Datasets:

Hendrik-a

/

arxiv

like 0

we study the hypothesis that observed x - ray / extreme ultraviolet emission from coronae of magnetically active stars is entirely ( or to a large part ) due to the superposition of flares , using an analytic approach to determine the amplitude distribution of flares in light curves . the flare - heating hypothesis is motivated by time series that show continuous variability suggesting the presence of a large number of superimposed flares with similar rise and decay time scales . we rigorously relate the amplitude distribution of stellar flares to the observed histograms of binned counts and photon waiting times , under the assumption that the flares occur at random and have similar shapes . our main results are : ( 1 ) the characteristic function ( fourier transform of the probability density ) of the expected counts in time bins is = \}$ ] , where is the mean flaring interval , is the characteristic function of the flare amplitudes , and is the flare shape convolved with the observational time bin . ( 2 ) the probability of finding counts in time bins is = . ( 3 ) the probability density of photon waiting times is = with = the mean count rate . an additive independent background is readily included . applying these results to euve / ds observations of the flaring star ad leo , we find that the flare amplitude distribution can be represented by a truncated power law with a power law index of 2.3 0.1 . our analytical results agree with existing monte carlo results of and . the method is applicable to a wide range of further stochastically bursting astrophysical sources such as cataclysmic variables , gamma ray burst substructures , x - ray binaries , and spatially resolved observations of solar flares .

astro-ph0312405

since the late 1930s the solar corona has been known to be much hotter than the photosphere , although the latter delineates the surface through which energy enters the corona . the corona is thus not in a state of simple heat flow equilibrium , and its high temperature must be supplied with some form of mechanical , electrical or radiative work which is dissipated at coronal heights . several heating mechanisms have been proposed , such as alfvn @xcite or ion cyclotron @xcite waves , or electric currents @xcite and magnetic reconnection @xcite converting magnetic shear into jets , and , ultimately , into heat . some of these processes operate in a continuous mode , whereas others invoke transients . comprehensive reviews may be found , e.g. , in @xcite , @xcite , @xcite , @xcite , and @xcite . with increasing sensitivity of the observations it was recognized that the ` quiet ' solar corona contains in fact ` microflares ' and ` nanoflares ' down to the actual resolution limit . clearly , these events are the signature of non - equilibrium processes , and therefore represent promising candidates for the primary heating mechanism of the solar corona @xcite . stars at high magnetic activity levels show properties that are reminiscent of continuous flaring processes , such as persistent coronal temperatures of 530 mk , high coronal electron densities , or continuous non - thermal radio emission . see @xcite for a summary of observational results . foremostly , x - ray and extreme ultraviolet ( euv ) time series of such stars reveal a high level of flux variability on time scales similar to flares @xcite . limited detector sensitivity could give the wrong impression that the emission is due to steadily emitting coronal sources , while new , highly sensitive observations in at least some cases show little indication of a base level emission that could be defined as being genuinely ` quiescent ' @xcite . a coronal heating mechanism that should successfully explain x - ray and euv emission from magnetically active stars inevitably has to address the question of continuous variability . in the context of microflare heating one is interested in the frequency distribution of flare amplitudes . in particular , it is essential to have reliable estimators of the small but frequent flares , since these presumably carry a major part of the total energy . however , these small flares substantially overlap , so that their amplitudes are hard to determine from the noisy observations . one way out is to use flare identification algorithms that also include flare superposition @xcite . another way out is to resort to a forward model , which constrains the flare amplitude distribution by model assumptions . within a monte carlo framework , this approach has been followed by @xcite and in @xcite . in the present article we explore _ analytical _ forward methods . we describe a statistical model of superimposed flares which provides an exact relation between the ( univariate ) probability density of the flare amplitudes and the ( univariate ) probability densities of the observed binned count rates and photon waiting times . as the method involves the histograms of binned count rates and photon waiting times only , it is insensitive to permutation of the time bins and to the presence of long data gaps . the price for this robustness is the loss of temporal information , which is replaced by the assumption that all flares have similar shape and occur at random . see @xcite and @xcite for a complementary approach focusing on the temporal autocorrelation of the light curves ( and thus on the first two moments of the univariate count rate distribution ) . in what follows we basically work with the observables ` counts ' and ` count rates ' , with emphasis on the effect of flare superposition and counting noise . the conversion into luminosity and flare energy is briefly discussed in section [ discussion_sect ] . while the energy release and relaxation processes are not known in detail , there is observational support for the similarity of flares on a global scale at least for ` simple ' flares . stellar flares of the type investigated in this paper can usually well be represented by a sharp rise and an exponential decay . examples supporting this view are found in @xcite , @xcite , @xcite , and @xcite . the observed durations of flares , as investigated in the solar case , are almost independent of the flare amplitude or radiated energy even if 3 - 5 observed orders of magnitude are included . ( @xcite , see also summary in @xcite ) . the decay times of flares studied below were found to scatter around a constant value in @xcite , independent of peak flux . since in stellar observations 3 orders of magnitude of flare energy are usually sufficient to explain the observed emission , we shall assume that a constant decay time is a good approximation for our case as well . for completeness , the case of a weak dependence of the decay time scale @xmath15 on radiated total energy @xmath16 , @xmath17 as suggested by one stellar investigation , was studied in both @xcite and @xcite , with the result that the main conclusions on the importance of small flares were reinforced . observations also suggest that individual flares occur at random and independently of each other ( for the solar case , see @xcite ) . this independence might suggest that flares trigger an exhaustive ` reset ' where mechanical stress ( and thus memory ) is completely destroyed . we shall thus adopt the following model for the instantaneous count rate in a fixed energy band : @xmath18 } \label{model}\ ] ] where @xmath19 represents the ( dimensionless ) flare shape , @xmath20 [ ct s@xmath21 are the flare amplitudes , and @xmath22 [ ct s@xmath21 is a background intensity . we shall agree that @xmath23 , and define @xmath24 to be the flare duration . all flare amplitudes @xmath20 are positive and drawn from the same probability density @xmath25 . ( from here on , @xmath9 denotes a generic real positive variable . ) the flare times @xmath26 occur at random with mean flaring interval @xmath3 [ s ] , and it is assumed that all @xmath20 , @xmath26 , and the background are statistically independent of each other . the observed counts are supposed to be a poisson process with time - varying intensity @xmath27 . such a non - uniform poisson process is not unambiguously defined , and we use here the construction shown in fig . [ poissonprocess_fig ] : the photon arrival times are distributed like the @xmath28-components of uniformly distributed points @xmath29 in the @xmath30-plane satisfying @xmath31 . note that the background is included in @xmath27 . when binned in time bins of duration @xmath0 , the counts are poisson distributed with parameter ( = expected number of counts in @xmath0 ) @xmath32 alternatively , we may work with the photon event list . in this case we consider the time intervals between consecutive photon arrivals ( ` photon waiting times ' ) . dimensional analysis shows that the flare - related part of the signal ( [ model ] ) is governed by the single dimensionless parameter @xmath33 . the conversion into discrete time bins involves a second dimensionless parameter , @xmath34 ( or , alternatively , @xmath35 ) . if the peaks are well - separated ( @xmath36 ) and counting noise is low , one can directly estimate the flare amplitudes from the observed peaks in the light curve . if flares overlap ( @xmath37 ) or counting noise is relevant , this is no longer possible . we resort then to the model ( [ model ] ) and ask for its predictions on the observed probability densities of binned counts @xmath7 and photon waiting times @xmath10 , which are available with high accuracy under stationary conditions . the goal is thus to establish the relation between @xmath25 on the one hand , and @xmath7 and @xmath10 on the other hand . the linearity of equation ( [ model ] ) , together with the statistical independence of its ingredients , suggests that a rigorous mapping from @xmath25 to @xmath7 and @xmath10 should exist . this is in fact the case , and the individual steps of this mapping are discussed in sections 3.1 - 3.4 . see table 1 for an overview on our notation . .notation@xmath38 [ cols="^,<,^ " , ] @xmath38the ( poisson- ) ` intensity ' is the instantaneous photon count rate ; the ( poisson- ) ` parameter ' is the expected number of photons in time bins @xmath0 . ` total ' means the sum of background and flares . we first ask for the probability that the flare signal @xmath39 assumes values between @xmath9 and @xmath40 . by statistical homogeneity , all times are equivalent . consider thus the time @xmath41 , and let it be centered in an auxiliary interval @xmath42 of length @xmath43 . neglecting flares outside @xmath42 as well as the background , the intensity at @xmath28 = 0 is @xmath44 . since @xmath45 is a sum of independent random variates , it is conveniently @xcite represented in a way which diagonalizes convolution , such as by its laplace- or fourier transform . we use here the latter , called the characteristic function @xcite and defined by @xmath46 where @xmath47 is the probability that @xmath45 is between @xmath9 and @xmath48 . the probability density is obtained from the characteristic function by fourier back transform , @xmath49 . we continue now with the calculation of the characteristic function @xmath50 . the essential point to notice is that @xmath50 can be partitioned according to the number of flares which occur in the interval @xmath42 . one can thus write @xmath51 , where @xmath52 is the probability of finding exactly @xmath53 flares in @xmath42 , and @xmath54 is the corresponding photon intensity from these @xmath53 superimposed flares in @xmath42 . it is understood that @xmath55 , because the case @xmath53 = 0 flare does not contribute to the count rate . since the times @xmath26 are independent and uniformly distributed in @xmath42 , the characteristic function of @xmath56 is @xmath57 the step from equation ( [ phi_q1 ] ) to equation ( [ phi_q2 ] ) also invokes the fact that all @xmath20 are independently @xmath58-distributed , and the definition of the characteristic function of the flare amplitudes , @xmath4 = @xmath59 . the step from equation ( [ phi_q2 ] ) to equation ( [ phi_q3 ] ) makes use of the symmetry of @xmath42 about @xmath28 = 0 . performing the sum over all possible numbers of flares and using equation ( [ phi_q3 ] ) , one finds that @xmath60 \big\ } \label{phi_ft } \ , .\ ] ] we now want to liberate ourselves from the artificial time interval @xmath42 , by extending it to infinity and thereby collecting all flare contributions . to this end we have to investigate whether the integral in equation ( [ phi_ft ] ) remains finite as @xmath61 . this turns out to be the case under rather weak conditions , one of which is as follows ( see appendix [ limit_appendix ] ) : if for some @xmath62 the flare shape @xmath63 decays faster than @xmath64 as @xmath65 , and if the probability density of the flare amplitudes decays no slower than @xmath66 as @xmath67 , then equation ( [ phi_ft ] ) remains well - behaved as @xmath61 . note that the above condition does not require the expectation value of the flare amplitudes to be finite ; this freedom will allow us to consider every power law index @xmath68 if the flare amplitudes have a power law distribution truncated at some lower cutoff . ( @xmath68 must be required to ensure that @xmath25 can be normalized . ) letting thus @xmath42 go to infinity , equation ( [ phi_ft ] ) becomes @xmath69 \big\ } \ , . \label{phi_f}\ ] ] the argumentation leading to equation ( [ phi_f ] ) can be repeated , with minor modifications , for the poisson parameter ( expected counts ) of time bins of duration @xmath0 . neglecting the background , the poisson parameter of time bin @xmath70 $ ] is @xmath71 , so that we merely have to replace @xmath19 by @xmath72 in order to obtain the characteristic function of @xmath73 , @xmath74 \big\ } \ , . \label{phi_f}\ ] ] equations ( [ phi_f],[phi_f ] ) are our central result ; they establish the connection between the probability density of flare amplitudes @xmath58 via @xmath75 and the distributions of photon intensities and expected counts , @xmath76 , @xmath77 via @xmath78 , @xmath79 . equation ( [ phi_f ] ) holds under the same conditions as equation ( [ phi_f ] ) . the function @xmath80 is a blurred version of the flare shape with resolution degraded to the observational time bin . it is easily seen that probabilistic normalization is preserved ( @xmath81=1 and @xmath82=1 if @xmath83=1 ) . it can also be shown that non - negative definiteness @xcite of @xmath4 is inherited under the transformations ( [ phi_f ] ) and ( [ phi_f ] ) , so that the @xcite theorem ensures that the fourier inverses of @xmath50 and @xmath1 represent , in fact , probability densities . the transformation @xmath84 redistributes amplitudes @xmath9 both towards higher ( flare overlap ) and lower ( flare tails ) values . it also broadens the probability density function by finite time binning , unless @xmath85 , in which case @xmath86 and @xmath87 . in the limit @xmath88 and if @xmath89 , normal behavior is approached , with @xmath90 concentrated about @xmath91 with relative spread of order @xmath92 . this expected result may be proven by formal expansion of equation ( [ phi_f ] ) about @xmath93 . so far the background has been neglected . when included , the total ( flare plus background ) photon poisson intensity at time @xmath28 = 0 is @xmath94 , and the total poisson parameter of the time bin @xmath95 $ ] is @xmath96 with @xmath97 . since the background is independent of the flares , the characteristic functions of the total photon poisson intensity and -parameter are @xmath98 with @xmath99 and @xmath100 the characteristic functions of @xmath101 and @xmath102 , respectively . if the background is constant then @xmath103 . the counts in the time bin [ @xmath104 are poisson distributed with parameter @xmath105 , where @xmath105 itself is @xmath106-distributed . the binned counts have thus the characteristic function @xmath107 which is @xmath108-periodic because the binned counts are integers . the probability density of the binned counts is obtained from fourier back transform of @xmath109 ; when expressed in terms of the original probability density @xmath25 , the probability of finding @xmath6 counts in bins of duration @xmath0 is @xmath110 if a gaussian approximation was used , with the binned counts modeled by @xmath111 with @xmath112 standard normal , then @xmath113 could assume any real number and the characteristic function is @xmath114 . this approximation becomes exact at @xmath115 , corresponding to large count rates . instead of time - binning the observed counts one may directly work with the photon event list , and consider the histogram of waiting times @xmath116 between consecutive photon arrivals . this has the advantage that no ad - hoc choice of time bins is needed , and that the waiting time bins can be adapted for optimum ( similar ) probability content . in addition , long data gaps ( @xmath117 ) and detector saturation ( @xmath118 ) are easily segregated , and the construction of the photon waiting time histogram amounts to a direct determination of the moment generating function of @xmath27 , which is ( almost ) the quantity provided by theory . the photon waiting times represent therefore a natural interface between theory and observation . to see why this is so , consider the ` waiting ' event that there is ( i ) one photon between @xmath119 and @xmath120 , ( ii ) no photon between @xmath120 and @xmath121 , and ( iii ) one photon between @xmath121 and @xmath122 ( the waiting time is thus @xmath123 ) . by the binomial theorem , the probabilities of ( i ) and ( iii ) are @xmath124 and @xmath125 , while from figure [ poissonprocess_fig ] we see that the probability of ( ii ) is @xmath126 . since ( i ) , ( ii ) and ( iii ) are independent of each other , the waiting event occurs with probability density @xmath127 . denoting the average over realizations of @xmath27 by angular brackets , and using the translation invariance @xmath128 , one finds that the probability density of photon waiting times @xmath129 is @xmath130 where the normalization constant @xmath131 is determined by the constraint @xmath132 = 1 , implying . ] that @xmath133 . while equation ( [ p_d ] ) is exact , it is generally difficult to calculate @xmath134 for arbitrary @xmath9 . this is , however , not needed because the photon waiting time is much shorter than the flare duration in any practically relevant case . one may thus resort to the approximation @xmath135 and @xmath136 the cumulative waiting time distribution is , accordingly , @xmath137 if the background is time - varying on time scale @xmath138 , the conditions @xmath139 and @xmath140 must be added to the right hand side of equations ( [ p_d0 ] , [ p_d1 ] ) . the range of validity of equation ( [ p_d0 ] ) was explored analytically ( appendix [ explicit_appendix ] ) and by monte carlo simulations ( section [ simulations ] ) . it was found to hold even if the assumptions on the right hand side are not strictly valid . the result ( [ p_d0 ] ) can also be derived in the following , perhaps more intuitive , way : let @xmath141 denote the probability that @xmath27 is between @xmath142 and @xmath143 in a random time interval of fixed duration . then , @xmath144 is the probability that a given photon has been emitted at poisson parameter @xmath142 . ( this probability is proportional to the total area of strips in the @xmath29 plane with @xmath145 see fig . [ poissonprocess_fig ] ) . assume now that @xmath27 varies only weakly during the local waiting time @xmath146 . thus @xmath144 is also the probability that a waiting time ` belongs ' to intensity @xmath142 . given the intensity @xmath142 , the photon waiting time has probability density @xmath147 , so that the ( marginal ) probability density of photon waiting times @xmath9 is @xmath148 , which is equation ( [ p_d0 ] ) . for special flare shapes and amplitude distributions , equations ( [ phi_f ] ) - ( [ p_d1 ] ) can be evaluated rigorously . this section summarizes results which are relevant for application to observational data ( section [ obs_sect ] below ) . exponential decay is a generic feature of many relaxation processes , and is a natural model of flares . observations of late - type main sequence stars often reveal a rather abrupt onset followed by an approximately exponential decay ( for stellar examples , see @xcite ) , with decay time almost independent of flare size ( e.g. , @xcite ; see also section ii ) . we shall thus use an exponential flare shape with constant decay time @xmath15 as a proxy for exact calculations . if the flare shape is @xmath149 with @xmath150 the heaviside step function , then @xmath151 ( see appendix [ explicit_appendix ] ) . the abrupt and artificial onset of the one - sided exponential flare shape may be softened by an exponential increase , thus creating a two - sided exponential shape : @xmath152 . since flare contributions from @xmath153 and @xmath154 to @xmath155 are statistically independent of each other , the characteristic function of their combined flare photon intensity is @xmath156 where @xmath157 is given by equation ( [ phi_f1 ] ) with @xmath15 replaced by @xmath158 . equations ( [ phi_f1 ] ) and ( [ phi_f_pm ] ) show that two - sided exponential flares have the same characteristic function @xmath78 as one - sided exponential flares with effective decay time @xmath159 . a similar result holds for @xmath79 if @xmath85 . from the observational point of view , this implies that one- and two - sided exponential flares can not be distinguished based on their short - term photon waiting time ( or binned count ) distribution . we therefore assume in the sequel that the flare shape is one - sided exponential , with the understanding that our results on @xmath25 are robust against the generalization of a one - sided to two - sided exponential shapes . for several forms of @xmath25 the characteristic function @xmath50 can be explicitly obtained . solvable cases include exponentials , power laws , and lvy distributions . details are given in appendix [ explicit_appendix ] ; here we cite the results for power laws , which are a traditional and observationally well confirmed model of the distributions of solar @xcite and stellar @xcite flares . if @xmath25 is a power law with exponent @xmath160 , truncated at a lower cutoff @xmath161 , @xmath162 and if the flare shape is exponential , then @xmath50 can be exactly expressed in terms of special functions ( appendix [ explicit_appendix ] ) . for practical purposes , these may be approximated by elementary functions . of particular interest is the cumulative photon waiting time distribution ( eq . [ p_d1 ] ) which predicts the probability content of arbitrary photon waiting time bins : @xmath163 with @xmath164^{-1}$ ] . the relative accuracy of the approximation ( [ pd_approx ] ) was found to be better than 0.001 if @xmath165 and @xmath160 @xmath166 2.01 . for @xmath160 @xmath166 2 the expectation value of the flare amplitudes exists and is given by @xmath167 . in the limit @xmath160 @xmath168 2 the right hand side of equation ( [ pd_approx ] ) converges to the unit step function @xmath169 ; this is also true if the exact equations ( [ phi_f_pl1 ] , [ phi_f_pl2 ] ) are used . the limit @xmath170 @xmath168 @xmath171 therefore predicts , theoretically , arbitrarily small photon waiting times . before application to real data , the results of section [ gensolution ] have been verified by several independent methods . the introduction of the characteristic function in section [ flare_contrib_sect ] may appear somewhat arbitrary , and the question arises whether a more direct approach exists . the answer is yes , but the calculation is tedious and difficult to make rigorous . we shall therefore only sketch it in appendix [ moment_appendix ] . here , we discuss validation based on an entirely independent analytical argument ( section [ peak_sect ] ) and on numerical simulations ( section [ simulations ] ) . for one - sided exponential flares and only for these the signal @xmath172 is a continuous - time markov process whose stationary measure can be obtained from a renewal argument . this provides an independent verification of equation ( [ phi_f1 ] ) , and has some interest on its own as it allows to relax the poisson assumption on the flare times ( see sect . [ discussion_sect ] below).the argument is as follows . discarding the background , the amplitudes at consecutive peaks are related by @xmath173 with @xmath174 and @xmath175 the flare waiting times . now , in the stationary case , the probability densities of @xmath176 and @xmath177 must agree . therefore their characteristic function @xmath178 must satisfy @xmath179 where @xmath180 is the flare waiting time distribution , and we have used the definition of the characteristic function and the fact that @xmath181 , @xmath176 and @xmath182 are statistically independent of each other . for uniformly distributed flares , the flare waiting times are exponentially distributed , @xmath183 . upon setting @xmath184 = @xmath185 , equation ( [ phi_eq ] ) simplifies thus to @xmath186 equation ( [ int_eq ] ) can be solved for @xmath187 by taking derivatives with respect to @xmath188 , yielding @xmath189 together with the initial ( normalization ) condition @xmath190 , equation ( [ diff_eq ] ) has the unique solution @xmath191 the exponential factor on the right hand side of equation ( [ phi_r_n ] ) acts as a low - pass filter , which broadens @xmath192 with respect to @xmath25 by an amount proportional to @xmath33 . as it must be , isolated flares reproduce the true flare amplitudes , @xmath193 . there is an apparent similarity between equations ( [ phi_f1 ] ) and ( [ phi_r_n ] ) , and the exact relation is as follows . by construction , the peaks @xmath176 are the sum of two independent random variates : the value of the continuous signal just before the jump , @xmath194 , plus the flare amplitude @xmath181 . from equation ( [ phi_r_n ] ) we see that @xmath194 has the characteristic function @xmath195 . since @xmath194 occurs with equal probability than any @xmath27 , the above expression must agree with the result of equation ( [ phi_f1 ] ) . this is in fact the case . equations ( [ phi_f ] , [ phi_f ] , [ p_c ] , [ p_d ] ) have also been verified by monte carlo simulations . in these simulations , a realization of @xmath27 is generated , and the binned poisson parameters @xmath196 = @xmath197 are numerically calculated . the binned counts are then simulated using a conventional poisson number generator , and their histogram is created . for the photon waiting times we proceed as indicated in figure [ poissonprocess_fig ] : a large number @xmath198 of points @xmath29 are uniformly distributed in the rectangle @xmath199 \times [ 0,y]$ ] , with @xmath200 and @xmath201 . the @xmath28-coordinates of points with @xmath31 constitute our photon list . note that the background is included in @xmath27 . the simulations are compared to numerical evaluations of eqns . ( [ p_c],[p_d],[p_d0 ] ) and to exact results ( appendix [ explicit_appendix ] ) . figure [ binned_counts_fig ] ( left column ) shows the theoretical ( gray ) and simulated ( black ) probability densities which occur in the transformation from the flare amplitudes into the binned counts . the flare amplitudes are uniformly distributed between 1 and 3 ct @xmath202 , the flare profile is a one - sided exponential with decay time @xmath15 = 8 s , the mean flaring interval is @xmath3 = 10 s , and the simulation involves @xmath203 time bins of duration @xmath0 = 3 s. these parameters represent marginally resolved , partially overlapping flares . accordingly , @xmath204 ( fig . [ binned_counts_fig ] line 2 left ) is smoothed with respect to @xmath25 ( line 1 left ) but has also a peak at @xmath9=0 which is due to low - count sequences between different flares . the binned background @xmath102 is modeled as chi square distributed with 3 degrees of freedom ( fig . [ binned_counts_fig ] line 3 left ) ; this ad - hoc choice yields a particularly simple characteristic function and is for illustration purposes only . the conversion of the binned poisson parameter ( fig . [ binned_counts_fig ] line 3 right ) into discrete counts renders the characteristic function @xmath108-periodic ( fig . [ binned_counts_fig ] line 4 right ) . for benchmarking purposes , the theoretical probability densities are obtained from numerical fourier inversion of the associated characteristic functions ( fig . [ binned_counts_fig ] right column , solid line - real part , dashed line - imaginary part ) . the transformation from the flare amplitudes into the photon waiting times is illustrated in figure [ waitingtime_fig ] . the flare amplitudes are exponentially distributed ( mean amplitude @xmath205 = 0.15 ct @xmath202 , mean flaring interval @xmath3 = 20 s ) and the flare shape is one - sided exponential ( @xmath15 = 40 s ) . the exponential flare amplitudes are chosen because they admit simple analytical results for @xmath10 ( see appendix [ explicit_appendix ] ) , which can be compared to the simulation . a constant background @xmath206 = 0.1 ct @xmath202 is present , and the simulated photon list includes @xmath207 photons . the theoretical @xmath208 ( fig . [ waitingtime_fig ] middle left gray ) is a chi square distribution with @xmath33 degrees of freedom ( eq . [ phi_f_exp ] ) , and @xmath10 ( fig . [ waitingtime_fig ] bottom left gray ) is given by the exact equation ( [ p_d_exp_bg ] ) . the characteristic functions of the flare amplitudes , flare photon intensity , and total ( flare plus background ) photon intensity are shown in the right column of figure [ waitingtime_fig ] . solid line denotes the real part , and dashed line denotes the imaginary part . a broad probability density corresponds to narrow characteristic functions , and vice versa . the constant background amounts to a modulation by @xmath209 . both binned counts and photon waiting times have also been simulated for power law flare amplitudes , which are used as a fitting template for the euve / ds observations below . figure [ sim_pl_fig ] shows the result of a simulation of one - sided exponential flares ( @xmath15 = 3000 s , @xmath3 = 333 s ) with lower cutoff @xmath210 = 0.007 ct @xmath202 and power law index @xmath160 = 2.36 . the simulation involves 3@xmath211 photons in 10@xmath212 time bins of duration @xmath0 = 100 s. again , the black line represents simulation , and the gray line represents theory , with @xmath10 being the short - term form ( eq . [ p_d0 ] ) computed by differentiation of the exact expression ( [ phi_f_pl1 ] ) for @xmath50 . a constant background @xmath206 = 0.09 ct @xmath202 is present , and an upper cutoff @xmath213 = 7 ct @xmath202 has been applied in the simulations and when evaluating @xmath7 from equation ( [ p_c ] ) . we have applied our theoretical results to a 3700 ks ( elapsed time ) observation of the active star ad leo acquired by the extreme ultraviolet explorer ( euve , e.g. @xcite ) deep survey ( ds ) instrument between april 2 and may 15 1999 . this time series has previously been analyzed by @xcite and in @xcite by monte - carlo simulations of the model ( eq . [ model ] ) , so that we can compare our findings with existing results . the deep survey instrument records photons between 52 and 246 with maximum sensitivity at 91 . figure [ euve_lc_fig ] shows an overview of the light curve , time - integrated over spacecraft orbits ( 5663 s each ) . during one spacecraft orbit , ad leo was in the euve field of view for typically 1500 s , followed by a @xmath214 4200 s interval without observations . furthermore , there are observation interrupts where the light curve of figure [ euve_lc_fig ] drops to zero . the period from 05/06 16:42 to 05/16 05:11 carries large contamination from energetic particles and is excluded from our analysis . figure [ euve_pd_fig ] shows the histograms of photon waiting times . the ` flares+background ' are collected across a spatial window centered at ad leo , while the ` background ' is determined from a larger disjoint area . the resolution of our photon arrival times is 0.1 s. this is not the full euve / ds time resolution , but has been considered sufficient with regard to the actual count rates . only photon waiting times from 0.25 s and 30 s are taken into account . the lower limit eliminates poor flare statistics and telemetry saturation at rare and extremely bright flares , and the upper limit eliminates low count rates and artifacts from ( much longer ) unobserved intervals . our selection comprises 118,637 photon waiting times , and the photon waiting time bin size is 0.1 s. the background waiting times ( fig . [ euve_pd_fig ] ` background ' black line ) are found to closely follow an exponential distribution ( fig . [ euve_pd_fig ] ` background ' gray line ) ; accordingly , we model the background as constant . when scaled to the ` flares+background ' collecting area , the best - fit background intensity is ( cf . @xcite ) @xmath215 once the background is known , we determine the flare amplitude distribution @xmath25 under the assumption that @xmath25 is a power law with exponent @xmath160 and lower cutoff @xmath210 ( eq . [ p_a_pl ] ) , and that the flare shape is one - sided exponential ( fig . [ euve_lc_fig ] inlet ) . these assumptions correspond up to a little relevant upper cutoff to those made by @xcite and in @xcite . we also assert that the short - term forms ( eqns . [ p_d0 ] , [ p_d1 ] ) of the photon waiting time distribution apply . our model parameters are thus ( @xmath216 ) . they are determined by minimizing @xmath217 where @xmath218 ( @xmath219 ) is the observed ( predicted ) number of waiting times in the @xmath6-th waiting time bin . the prediction is given by @xmath220 , and we use the exact expression ( [ phi_f_pl1 ] ) when computing the cumulative photon waiting time distribution ( eq . [ p_d1 ] ) . the sum in equation ( [ chi2 ] ) runs over @xmath221 = @xmath222 = 298 waiting time bins , and there are @xmath223 = @xmath221 @xmath224 3 = 295 degrees of freedom . the minimum reduced chi square is 0.989 , and the best - fit parameters are @xmath225 the resulting theoretical waiting time distribution is shown in figure [ euve_pd_fig ] ( ` flares+background ' gray line ) together with the observation ( ` flares+background ' black line ) . the errors of ( @xmath226 ) are not independent of each other , and the values given in equation ( [ results_d ] ) are projections of the 90% confidence domain ( see fig . [ euve_chi2_fig ] below ) . we also estimated the parameters ( @xmath226 ) from the observed histogram of binned counts ( fig . [ euve_pc_fig ] black line ) . the ( disjoint , adjacent ) time bins have duration @xmath0 = 100 s and contain , on average , 22 counts . time bins which would intersect the transition from or to an unobserved time interval are rejected . in order to abate the arbitrariness of choosing a binning offset , the starting time of the first time bin has been varied at random within @xmath0 , and the histogram of figure [ euve_pc_fig ] is an average over 50 random offsets . time bins with less than 2 counts and more than 80 counts are excluded , because they are affected by the binning offset and have poor statistics . the best - fit solution is again found by minimizing the chi square expression of equation ( [ chi2 ] ) , where @xmath218 and @xmath219 are now the observed and theoretical frequencies of @xmath6 counts in time bins of duration @xmath0 . the theoretical prediction is @xmath227 with @xmath131 = @xmath228 = 4959 and @xmath6 running over @xmath221 = @xmath229 = 78 photon number bins . since @xmath0 @xmath230 @xmath15 , we may use the short - term form of the probability density of binned counts , @xmath231 with @xmath232 and @xmath50 given by equation ( [ phi_f_pl1 ] ) . the minimum reduced chi square ( per @xmath223 = @xmath221 @xmath224 3 = 75 degrees of freedom ) is 1.003 , which is attained at @xmath233 the best - fit binned count distribution is superimposed in figure [ euve_pc_fig ] ( gray line ) . the errors given in equation ( [ results_c ] ) are again projections of the 90% confidence domain . the confidence domains of ( @xmath226 ) for both the photon waiting time and binned count methods are visualized in figure [ euve_chi2_fig ] . the contours represent chi square sections through the best fit solution ( crosses ) at levels @xmath234 , corresponding to ( 0.75 , 0.9 , 0.999 ) confidence levels ) was true then the probability that the observed @xmath235 ( eq . [ chi2 ] ) deviates from its expectation value @xmath223 by more than @xmath236 is about @xmath237 . the cited ` confidence levels ' represent the probabilities that this does not happen . ] . panels above the diagonal refer to the binned count method ; panels below the diagonal refer to the photon waiting time method . the formal 90% confidence level is marked boldface . we find that the relative error of @xmath160 is smaller than the errors of @xmath210 and @xmath33 for both the binned count and waiting time methods . the power law index @xmath160 can thus be estimated more reliably than the lower cutoff . the shape of the confidence domains indicates that the parameters are not independent of each other . the strongest interference is between the lower cutoff @xmath210 and the separation of individual flares , @xmath238 , which reflects the fact that the average flare photon flux , @xmath239 = @xmath240 , is sensitive to the product @xmath241 only . likewise , @xmath160 is positively coupled to both @xmath210 and @xmath33 because the function @xmath242 decreases with increasing @xmath160 ( @xmath160 @xmath166 2 ) . the results of the photon waiting time ( eq . [ results_d ] ) and binned count ( eq . [ results_c ] ) methods agree within the 90% errors , which is considered as satisfactory , although the binned count method systematically yields lower @xmath210 and larger @xmath33 than the photon waiting time method , whereas the product @xmath241 is similar in both cases ( 0.0541 for the waiting time method ; 0.0527 for the binned count method ) . the mean flaring interval @xmath3 and flare duration @xmath15 enter the short - term forms of binned count- and photon waiting time distributions only in the dimensionless combination @xmath33 . the decay constant @xmath15 of flares with amplitudes @xmath243 @xmath210 has been investigated by @xcite and in @xcite and found to be about 3000 s ; adopting this value , we find a mean flaring interval of @xmath244 ( average over both methods ) . the use of the short - term forms of the binned count- and photon waiting time distributions is justified , a posteriori , since @xmath0 @xmath230 @xmath15 , and since @xmath116 @xmath245 30 s @xmath230 @xmath15 and @xmath246 @xmath214 20 @xmath247 1 . in fact , direct numerical evaluations of the exact forms ( eqns . [ p_c ] , [ p_d ] ) at best - fit parameters ( eqns . [ results_d ] , [ results_c ] , [ flare_rate ] ) deviate from their short - term approximations by less than 2% . we have applied an analytical approach to the inversion of light curves to derive the amplitude distribution of constituent elementary flare light curves . the method uses either binned count light curves or the distribution of waiting times derived from photon lists . we have in particular considered flare shapes described by a one - sided exponential decay or a two - sided exponential profile , and we have studied flare amplitude distributions described by an exponential function and a truncated power law . the latter form has been favored in previous studies because solar flare statistics has revealed power - law distributions over a wide range in energy @xcite ; explicit stellar studies indicate power laws as well @xcite . we have thus used a power law to model euve / ds observations of ad leo , where we find a power law index of @xmath160 = 2.3 @xmath14 0.1 , a lower cutoff of @xmath210 = 0.004 @xmath14 0.002 ct @xmath202 , and that there are @xmath33 = 15 @xmath14 8 flares per flare duration above @xmath210 ; these values are compilations of the binned count- and photon waiting time results . our results agree with complementary numerical methods recently applied to similar data sets @xcite indicating power - law indices for the flare amplitude distribution of typically 2.0 2.4 , with most measurements concentrating around @xmath160 = 2.2 2.3 . while such values are necessary for power - law flare amplitude distributions to ` explain ' coronal heating on active stars , they are not sufficient . the emitted power observed from magnetically active coronae can only be due to the ensemble of flares _ if _ the power - law distribution extends to sufficiently low energies ; this conjecture is of course difficult to prove due to the strong superposition of near - simultaneous flares that can not be separated spatially , thus forming a pseudo - continuous emission level . when expressed in terms of flare energy , our numerical results ( eqns . [ results_d ] , [ results_c ] ) correspond to a flare peak luminosity and minimum flare energy of @xmath248 where it is assumed that one euve / ds count corresponds to @xmath249 erg emitted at ad leo ( see @xcite ) . alternatively , a fraction of the observed emission may be truly steady ( or variable on time scales much longer than flare time scales ) , but this level should then be below the lowest level seen in light curves between individual detected flares ; @xcite estimate that a mere 30% ( or less ) of the low - level emission of the dme ( binary ) star uv ceti could be ascribed to quasi - steady emission , the rest being recorded as variable on time scales of tens of minutes with shapes indicating continuous flaring . additional information contained in the differential emission measure distribution could help interpreting the physical processes participating in the heating of the coronal plasma @xcite . as mentioned in @xcite , a few words of caution are in order . a number of features can not be recognized in the data per se and would require additional information . specifically , the observations collect emission from across the visible hemisphere of the star , although heating mechanisms or , in our picture , flare amplitude distributions , may vary across various coronal structures ; further , some of the variability may be introduced by intrinsic changes in the large - scale magnetic field distributions ; emerging or disappearing active regions will certainly contribute to variability , as do changes in the visibility of individual regions on the star as they move into or out of sight due to the star s rotation . nevertheless , the light curve characteristics do not appear to be dominated by any of these longer - term effects , while variability on flare time scales is ubiquitous ; similar conclusions were drawn by @xcite , @xcite , @xcite , and @xcite . we therefore conclude that the present study further supports the view that stochastic flaring is of principal importance to coronae of magnetically active stars , and that coronae of such stars may significantly , if not entirely , be heated by flares . a practical benefit of the analytical method lies in its efficient assessment of the compatibility of model and observations . the computation of @xmath235 for a given set of parameters takes milliseconds on a medium - size work station , while a monte carlo simulation of comparable accuracy takes seconds or minutes due to slow ( @xmath250 ) convergence at rare large flares . the efficiency of the analytical approach allows to systematically explore the whole parameter space for the global chi square minimum . for instance , figure [ euve_chi2_fig ] has been obtained by a brute - force calculation of @xmath251 on a cubic lattice of dimension 90@xmath25295@xmath252100 , taking about one hour computation time , whereas a similar monte carlo simulation would take many weeks . the possibility for systematic exploration of the parameter space ( at low resolution ) is important because the chi square function has generally several local minima , so that traditional minimization routines such as conjugate gradients may fail . note that the simple contours of figure [ euve_chi2_fig ] depict only the local behavior around the global minimum . when comparing the binned counts- and photon waiting time methods it was found that the binned counts method has larger confidence domains than the photon waiting time method , but is also more robust in finding the global minimum . a good strategy is thus to start with the binned counts , using a low - resolution exploration of the parameter space . as soon as a the approximate global chi square minimum is found , this may be refined by the photon waiting time method . in any case the final solutions should be cross - checked for both methods , to ensure consistency and to obtain an error estimate for the parameters which goes beyond the errors of the separate methods . we would finally stress that our analytical method is in no way restricted to the particular time series studied in this paper . the method applies whenever the process can be described in the form of equation ( [ model ] ) , which may include models for so divers phenomena as cataclysmic variables , gamma ray burst substructures , and x ray binaries . there is also no need to restrict the process ( [ model ] ) a the real line ( see below ) . a drawback of the analytical method , compared to monte carlo simulations , is its limited capability to incorporate systematic instrumental artifacts . we close by mentioning a few straight forward extensions of the model ( eq . [ model ] ) . * flare clustering . * the assumption of independent flare times might be questionable in the light of more advanced avalanche models @xcite where flares may trigger each other . flare clustering can be accounted for by a flare waiting time distribution which does not create uniform flare times ( @xmath253 ) , but has a long - range tail permitting rare but large jumps . the ` clusters ' are between the jumps . if the flare amplitudes are still independently @xmath58-distributed , rise immediately and decay exponentially , then the functional equation ( [ phi_eq ] ) for the peak values still holds , but the separation into the form ( [ int_eq ] ) is no longer possible . however , equation ( [ phi_eq ] ) can be iterated : the sequence @xmath254 converges uniformly to the characteristic function of @xmath255 of clustered flares . all iterates are characteristic functions , and the characteristic function of the flare photon intensity is given by @xmath256 ( see section [ peak_sect ] ) . * ccd movies . * the arguments leading to equations ( [ phi_f ] ) and ( [ phi_f ] ) are not restricted to one - dimensional light curves . a useful generalization is to a sequence of spatially resolved ccd images od solar flares , where binning is over the space - time pixel @xmath257 , and the auxiliary interval @xmath42 of section 3.1 becomes @xmath258 . provided that the flares occur independently and uniformly in @xmath42 , the derivation of equation ( [ phi_f ] ) goes through without modifications if @xmath0 is replaced by @xmath259 and @xmath260 is replaced by @xmath261 with @xmath262 and @xmath263 the space - time flare shape . we thank marc audard for discussions on the euve data set . by definition , @xmath4 satisfies @xmath264 ( normalization ) . let us assume that @xmath265 as @xmath266 with @xmath62 and @xmath267 ; this is a valid form of @xmath4 which is realized , for @xmath268 , in lvy distributions @xcite . since @xmath269 as @xmath65 , the integral in equation ( [ phi_ft ] ) remains well - behaved for all @xmath188 when @xmath61 if @xmath270 , requiring that @xmath63 decays faster than @xmath64 as @xmath65 . so far about the decay of the flare shape . the decay of the probability density @xmath25 at large flare amplitudes @xmath9 is related by tauberian theorems @xcite to the small-@xmath188 limit of the characteristic function . for our assumption @xmath271 , one has that @xmath272 as @xmath67 . these are the conditions mentioned in section [ flare_contrib_sect ] . for special flare shapes and amplitude distributions , equations ( [ phi_f ] ) and ( [ phi_f ] ) can explicitly be evaluated . to this end , it is advantageous to cast them in the volterra form @xmath273 with kernels @xmath274 where @xmath275 is the @xmath53-th solution of @xmath276 ( kernel @xmath277 ) or @xmath278 ( kernel @xmath279 ) , respectively , and prime denotes derivative with respect to @xmath28 . the success of an explicit evaluation of equations ( [ volterra_f ] , [ volterra_f ] ) depends on the simplicity of @xmath4 and on the ability to invert the functions @xmath19 and @xmath80 for the kernels @xmath277 and @xmath279 . @xmath1 is generally harder to obtain than @xmath50 . if the flare shape is @xmath280 then the equation @xmath278 has the solutions @xmath119 = @xmath281 ( valid for the rise phase @xmath282 ) and @xmath121 = @xmath283 ( valid for the decay phase @xmath284 ) . the corresponding kernel contributions ( eq . [ kernel ] ) are @xmath285 and @xmath286 , and equation ( [ volterra_f ] ) becomes @xmath287 the mean of @xmath73 is @xmath288 , and its variance is @xmath289 . for short time bins ( @xmath85 ) one has that @xmath290 , and equation ( [ expdecay1 ] ) reduces to @xmath291 with @xmath50 given by equation ( [ phi_f1 ] ) . we present now several exactly solvable models for the flare amplitude distribution . for @xmath292 and one - sided exponential flare shape , all characteristic functions and the photon waiting time distribution can be rigorously calculated . the results are @xmath293 equation ( [ phi_f_exp ] ) indicates that @xmath208 is a chi square distribution with @xmath33 ( non - integer ) degrees of freedom . the combination of exponential flare amplitudes and exponential flare profiles has the peculiar property that @xmath294 ( and thus @xmath295 ) if @xmath296 . if a constant background @xmath206 [ ct @xmath202 ] is present , the exact ( eq . [ p_d ] ) and short - term ( eqns . [ p_d0 ] , [ p_d1 ] ) forms of the photon waiting time distributions are @xmath298 ^ 2 \hspace{-1 mm } + \frac{a^2 \tau e^{x/\tau}}{t } \big\ } \label{p_d_exp_bg } \\[2 mm ] p_\delta(x ) & \simeq & \frac{e^{-bx}(1+ax)^{-\tau / t}}{b+a\tau t^{-1 } } \big\ { \big ( b + \frac{a\tau t^{-1}}{1+ax } \big)^2 + \frac{a^2 \tau t^{-1 } } { ( 1+ax)^2 } \big\ } \ , .\label{p_d_exp_bg0}\end{aligned}\ ] ] equations ( [ p_d_exp_bg ] ) and ( [ p_d_exp_bg0 ] ) allow to benchmark ( and to be benchmarked by ) our monte carlo simulations ( section [ simulations ] ) . they also provide a measure of validity of the short - term approximation ( eqns . [ p_d0 ] , [ p_d1 ] ) : for example , for @xmath9 = @xmath299 , @xmath300 = 25 photons per flare , @xmath33 = 1 ( moderate flare overlap ) and @xmath206 = 0.2@xmath301 ( 20% background ) , the error of the short - term approximation is about 0.6% . if @xmath25 is a truncated power law ( eq . [ p_a_pl ] ) and if the flare shape is exponential , then @xmath50 may be expressed in terms of special functions . possible representations are @xmath302,[2,2,3-\nu],ia_0s ) \big\ } \label{phi_f_pl1 } \\ \phi_f(s ) & = & e^{\frac{\tau}{t}(\gamma - \frac{i\pi}{2 } + \frac{1}{\nu-1 } ) } \ , ( a_0s)^{-\tau / t } e^{-\frac{\tau}{t}\big\ { { \rm e_1}(-ia_0s ) + \nu ( -ia_0s)^{\nu-1 } \gamma(-\nu ,- ia_0s ) + \frac{e^{ia_0s}}{ia_0s } \big\ } } \label{phi_f_pl2}\end{aligned}\ ] ] where @xmath303 denotes the generalized hypergeometric function @xcite , @xmath304 is euler s constant , @xmath305 is the incomplete gamma function @xcite , @xmath306 is the exponential integral , and the principal branches of the @xmath307 and @xmath308 functions are to be used . equations ( [ phi_f_pl1 ] , [ phi_f_pl2 ] ) have been obtained by the computer algebra program maple . in numerical computations we use the form ( [ phi_f_pl1 ] ) with a taylor ( small - argument ) expansion of the hypergeometric function . for values of @xmath309 @xmath230 1 , this expansion converges rapidly , and machine precision is reached after @xmath310 6 orders if @xmath309 @xmath245 0.25 . in fact , already order two yields @xmath310 0.0005 rms relative precision for @xmath10 over the whole waiting time range 0 @xmath245 @xmath116 @xmath245 30s used in the euve data . this fact is exploited in the approximation ( [ pd_approx ] ) , which is based on an order - two taylor expansion of the hypergeometric function . if @xmath25 is a levy distribution @xcite with characteristic function @xmath311 , @xmath267 , and 0 @xmath310 @xmath312 @xmath310 1 , then @xmath25 vanishes at negative @xmath9 and therefore can model flare amplitudes . large flare amplitudes ( @xmath9 @xmath247 @xmath113 ) decay like @xmath25 @xmath214 @xmath313 , and superimposed flares have @xmath50 = @xmath314 with @xmath315 and @xmath316 is euler s constant ( see above ) . such lvy distributions might be a convenient model for small solar flares where power law indices @xmath310 2 have been reported @xcite . we ask for an elementary derivation of the results ( [ phi_f ] - [ p_d ] ) . the key observation is that equations ( [ phi_f ] ) and ( [ phi_f ] ) imply a simple relation between the moments of the flare amplitudes and the cumulants are defined by @xmath317 . cumulants of order @xmath318 can be expressed in terms of moments of order @xmath318 , and vice versa ; see , e.g. , @xcite . ] of the instantaneous photon intensity and the binned poisson parameter . this fact may be exploited to establish a relation between the moments of the flare amplitudes and the moments of the binned counts and photon waiting times , without recurrence to characteristic functions . we illustrate the procedure starting with the expectation value @xmath319 . interpreting the angular brackets as time average ( which is justified by the assumption of stationarity ) , the central limit theorem ensures that @xmath320 where @xmath321 is the number of flares in the interval @xmath42 ( see section [ flare_contrib_sect ] ) . when restricting the first sum in eq . ( [ time_mean ] ) to flares inside @xmath42 it has been supposed that contributions from outside @xmath42 decay sufficiently fast to be neglected , and we have used the fact that all @xmath20 and @xmath26 are statistically independent of each other . similarly , the second moment of @xmath45 is given by @xmath322 ^ 2 \ , dt + o\big({\cal n}_i^{1/2}\big ) \nonumber \\ & \to & \langle a \rangle^2 t^{-2 } \big ( \int \hspace{-1 mm } \xi(t ) \ , dt \big)^2 + \langle a^2 \rangle t^{-1 } \int [ \xi(t)]^2 \ , dt \label{time_var } \ , .\end{aligned}\ ] ] from equations ( [ time_mean ] ) and ( [ time_var ] ) we recognize that @xmath323 ^ 2 \ , dt$ ] : the variance of the instantaneous photon intensity equals the second moment of the flare amplitudes times a weight @xmath324 ^ 2 \ , dt$ ] . this finding can be generalized to higher cumulants if the division into @xmath325=@xmath53 and @xmath325@xmath326@xmath53 contributions in equation ( [ kl_split ] ) is replaced by a sum over partitions into equal and non - equal indices . we shall not discuss here the ( complicated ) combinatorial weights , but note that they correspond to those of the cumulants expressed in terms of moments ( e.g. , @xcite ) . as a consequence , @xmath327^n \ , dt \ \label{cumulants_f}\ ] ] equation ( [ cumulants_f ] ) is a weak form of equation ( [ phi_f ] ) , since it represents its taylor expansion around @xmath188=0 . it is easily seen that equation ( [ cumulants_f ] ) also applies to the binned poisson parameter @xmath73 if @xmath63 is replaced by @xmath328 . therefore the cumulants ( and thus the moments ) of both @xmath45 and @xmath73 can be expressed in terms of the moments of the flare amplitudes . the probability density of the binned counts @xmath6 is then given by @xmath329 , while the probability density of short ( section [ waitingtimesect ] ) photon waiting times @xmath9 is @xmath330 . although these series may be evaluated in special cases , it is generally much more convenient to work with the characteristic functions , which collect all the moments in a compact way . abramowitz , m. , & stegun , i. , 1970 , _ handbook of mathematical functions _ ( dover : dover publ . inc . ) ambruster , c. w. , sciortino , s. , & golub , l. 1987 , apj , 65 , 273 aschwanden , m. j. , tarbell , t. d. , nightingale , r. w. , schrijver , c. j. , title , a. , kankelborg , c. c. , martens , p. c. h. , & warren , h. p. , 2000 , apj , 535 , 1047 audard , m. , gdel , m. , & guinan , e. f. 1999 , , 513 , l53 audard , m. , gdel , m. , drake , j. j. , & kashyap , v. 2000 , apj , 541 , 396 audard , m. , gdel , m. , & skinner , s. l. 2003 , apj , 589 , 983 axford , w. i. , & mckenzie , j. f. , 1996 , _ astroph . space sci _ , 243 , 1 benz , a. o. , 1995 , in _ coronal magnetic energy release _ benz and a. krger ( berlin : springer ) benz , a. o. , & krucker , s. , 1999 , a&a , 341 , 286 bochner , s. , 1959 , _ lectures on fourier integrals _ ( princeton : univ . press ) butler , c. j. , rodon , m. , foing , b. h. , & haisch , b. m. 1986 , nature , 321 , 679 collura , a. , pasquini , l. , & schmitt , j. h. m. m. 1988 , , 205 , 197 crosby , n. b. , aschwanden , m. , & dennis , b. , 1993 , _ solar phys . _ , 143 , 275 datlowe , d. w. , elcan , m. j. , & hudson , h. s. 1974 , , 39 , 155 de jager , c. , heise , j. , van genderen , a. m. , foing , b. h. , ilyin , i. v. , kilkenny , d. s. , marvridis , l. , cutispoto , g. , rodon , m. , seeds , m. a. , et al . 1989 , a&a , 211 , 157 eadie , w. t. , drojard , d. , james , f. e. , roos , m. , & sadoulet , b. , _ statistical methods in experimental physics _ , 1971 ( amsterdam : north - holland ) feldman , u. , doschek , g.a . , & klimchuk 1997 , j.a . , _ apj _ , 474 , 511 gnedenko , b. v. , & kolmogorov , a. n. , 1954 , _ limit distributions for sums of independent random variables _ ( cambridge : addison - wesley ) gdel , m. , audard , m. , skinner , s. l. , & horvath , m. i. 2002 , apj , 580 , l73 gdel , m. , audard , m. , kashyap , v. , drake , j. , & guinan , e. , 2003 , apj , 582 , 423 gdel , m. , audard , m. , reale , f. , skinner , s. l. , & linsky , j. l. 2004 , a&a , in press gradshteyn , i. s. , & ryzhik , i. m. , 1980 , _ table of integrals , series and products _ ( new york : academic press ) heyvaerts , j. , & priest , e. , a&a , 1983 , 117 , 220 heyvaerts , j. , & priest , e. , a&a , 1984 , 137 , 63 hollweg , j. v. , coronal heating by waves , 1983 , in _ solar wind 5 _ , ed . m. neugebauer ( nasa cp-2280 , washington dc ) , 1 isliker , h. , a&a , 1996 , 310 , 672 kashyap , v. , & drake , j. j. 1999 , , 524 , 988 kashyap v. l. , drake , j. j. , gdel , m. , & audard , m. , 2002 , apj , 580 , 1118 kraev , u. , & benz , a. o. , 2001 , a&a , 373 , 318 krucker , s. , & benz , a. o. 1998 , , 501 , l213 lvy , p. , 1937 , _ thorie de laddition des variables alatoires _ ( paris : gauthiers - villars ) lin , r. p. , schwartz , r. a. , kane , s. r. , pelling , r. m. , & hurley , k. 1984 , apj , 283 , 421 lukacs , e. , 1970 , _ characteristic functions _ ( london : charles griffin ) malina , r. f. , & bowyer , s. , 1991 , in _ extreme - ultraviolet astronomy _ , ed . r. f. malina and s. bowyer ( new york : pergamon ) marsh , e. , & tu , c. y. , 1997 , a&a , 316 , l17 narain , u. & ulmschneider , p. 1990 , _ space science rev . _ , 54 , 377 narain , u. & ulmschneider , p. 1996 , _ space science rev . _ , 75 , 453 osten , r. a. , & brown , a. 1999 , apj , 515 , 746 parker , e. 1972 , apj , 174 , 499 pearce , g. , & harrison , r. a. 1988 , a&a , 206 , 121 priest , e. , & forbes , t. , 2000 , _ magnetic reconnection - mhd theory and applications _ ( cambridge : cambridge univ . press ) rota , g. , & shen , j. ( 2000 ) , _ j. of combinatorial theory _ , a 91 , 283 shimizu , t. 1995 , _ publ . soc . japan _ , 47 , 251 stein , m. l. 1999 , _ interpolation of spatial data _ , springer ulmschneider p. , rosner , r. , & priest , e. , 1991 , _ mechanisms of chromospheric and coronal heating _ ( berlin : springer ) vlahos , l. , georgoulis , m. , kluving , r. , & paschos , p. , 1995 , apj , 299 , 897 wheatland , m. s. 2000 , _ solar phys . _ , 191 , 381 wheatland , m. s. & craig , i. j. d. 2003 , apj , 595 , 458

electron spin relaxation due to the dyakonov - perel mechanism is investigated in bilayer graphene with only the lowest conduction band being relevant . the spin - orbit coupling is constructed from the symmetry group analysis with the parameters obtained by fitting to the numerical calculation according to the latest report by konschuh _ et al . _ [ phys . rev . b * 85 * , 115423 ( 2012 ) ] from first principles . in contrast to single - layer graphene , the leading term of the out - of - plane component of the spin - orbit coupling in bilayer graphene shows a zeeman - like term with opposite effective magnetic fields in the two valleys . this zeeman - like term opens a spin relaxation channel in the presence of intervalley scattering . it is shown that the intervalley electron - phonon scattering , which has not been reported in the previous literature , strongly suppresses the in - plane spin relaxation time at high temperature whereas the intervalley short - range scattering plays an important role in the in - plane spin relaxation especially at low temperature . a marked nonmonotonic dependence of the in - plane spin relaxation time on temperature with a minimum of several hundred picoseconds is predicted in the absence of the short - range scatterers . this minimum is comparable to the experimental data . moreover , a peak in the electron density dependence of the in - plane spin relaxation time at low temperature , which is very different from the one in semiconductors , is predicted . we also find a rapid decrease in the in - plane spin relaxation time with increasing initial spin polarization at low temperature , which is opposite to the situation in both semiconductors and single - layer graphene . a strong anisotropy between the out - of- and in - plane spin relaxations at high temperature is also revealed with the out - of - plane spin relaxation time being about two orders of magnitude larger than the in - plane one . detailed comparisons of the temperature and electron density dependences of the spin relaxation with the existing experiments [ phys . rev . lett . * 107 * , 047206 ( 2011 ) , phys . rev . lett . * 107 * , 047207 ( 2011 ) and nano lett . * 11 * , 2363 ( 2011 ) ] are reported .

1302.5880

in recent years , single - layer graphene has received much attention due to its two dimensionality , dirac - like energy spectrum and outstanding spin - coherence properties.@xcite specifically , the hyperfine interaction and spin - orbit coupling ( soc ) are considerably weak,@xcite suggesting a long spin relaxation time ( srt ) in single - layer graphene . this long srt makes single - layer graphene a promising candidate for spintronic application . so far , a number of experiments on spin relaxation have been carried out , with the srts being orders of magnitude shorter than theoretical prediction.@xcite this pronounced discrepancy has provoked many investigations on the extrinsic effects of spin relaxation such as adatoms,@xcite curvature,@xcite substrate effects@xcite and contacts.@xcite consisting of two layer graphene , bilayer graphene ( blg ) also possesses good spin - coherence properties.@xcite additionally , in contrast to single - layer graphene , blg has a tunable gap caused by an out - of - plane electric field,@xcite which can be used to turn on or off the electronic transport . in addition , in the absence of space inversion symmetry , the intrinsic soc in the pseudospin space of blg provides an out - of - plane effective magnetic field near the dirac points whereas the intrinsic soc of single - layer graphene only induces a shift of the energy spectrum.@xcite moreover , the electron - phonon coupling becomes more complex in blg due to the existence of more phonon modes.@xcite so far , some efforts have been made to investigate the intravalley electron - phonon coupling.@xcite specifically , the intravalley electron - acoustic ( ac ) phonon coupling has been calculated by viljas and heikkil@xcite with the continuum theory and also by borysenko _ et al._@xcite from first principles . to explore the intravalley electron - optical ( op ) phonon coupling , the tight - binding model , density functional theory and first principles approach are employed by viljas and heikkil,@xcite cappelluti _ et al._@xcite and borysenko _ et al._,@xcite respectively . however , the intervalley electron - phonon coupling , which has been shown to play an important role in spin relaxation in rippled single - layer graphene,@xcite has not been reported in the previous literature . all these intriguing features make blg an attractive material for investigation . very recently , some attention has been devoted to the spin relaxation in blg.@xcite experimentally , the srts of blg on sio@xmath0 substrate@xcite or in freely suspended blg@xcite are reported to be of the order of @xmath1-@xmath2ns . both yang _ et al._@xcite and han and kawakami@xcite measured the temperature and carrier dependences of the srt . et al._@xcite showed that the srt is weakly dependent on temperature . however , han and kawakami@xcite reported a marked decrease of the srt with the increase of the temperature . as for the carrier density dependence , the decrease of the srt with increasing carrier density at low temperature is reported in both experiments . however , at room temperature , yang _ et al._@xcite observed an increase of the srt with increasing density whereas the srt measured by han and kawakami@xcite exhibits a marginal density dependence . in addition , yang _ et al._@xcite also showed that the srt scales inversely with the mobility at both room and low temperatures . in contrast to the above two experiments based on the micromechanically exfoliated blg samples , avsar _ et al._@xcite carried out the experiment in chemical vapor deposition grown blg . the carrier density and temperature dependences of the srt measured are comparable to the ones by yang _ et al._@xcite both quantitatively and qualitatively . motivated by these experiments , diez and burkard@xcite calculated the electron spin relaxation due to the dyakonov - perel ( dp ) mechanism@xcite with only the long - range electron - impurity scattering included in blg . they claimed to obtain a good agreement with the experimental data of yang _ et al._.@xcite it is noted that the effective soc ( referred to as the spin - orbit field in ref . ) used in their calculation is induced by an interlayer bias voltage together with the intrinsic soc in the pseudospin space.@xcite however , the intrinsic soc in the pseudospin space becomes incomplete in the presence of the interlayer bias since the interlayer bias breaks the space inversion symmetry and leads to additional extrinsic soc terms in the pseudospin space.@xcite also even with only the intrinsic soc in the pseudospin space in their calculation , they overlooked the out - of - plane component of the effective soc . as will be shown later , the leading term of this component leads to a zeeman - like term in the two valleys and makes a dominant contribution to the in - plane spin relaxation . moreover , they neglected the electron - phonon and electron - electron coulomb scatterings . in fact , the electron - electron coulomb scattering has been demonstrated to be very important for the spin relaxation in semiconductor systems@xcite and also single - layer graphene at high temperature.@xcite the electron - phonon scattering has also been shown to play an important role in spin relaxation in both semiconductors@xcite and single - layer graphene@xcite at high temperature . therefore , both have to be included in studying the spin relaxation . in the present work , with the electron - electron coulomb , ( both the intra- and inter - valley ) electron - phonon , possible short - range@xcite as well as long - range electron - impurity scatterings explicitly included , we investigate the electron spin relaxation due to the dp mechanism in blg with only the lowest conduction band being relevant by the kinetic spin bloch equation ( ksbe ) approach.@xcite the soc of the lowest conduction band near the dirac points @xmath3 is constructed from the symmetry group @xmath4 in the presence of an out - of - plane electric field.@xcite its three components @xmath5 @xmath6 can be written as ( see appendix [ appa ] ) @xmath7,\label{eq1}\\ \omega_y^{\mu}({\bf k})&=&-\alpha_1(k)\cos\theta_{\bf k}+ \mu[\alpha_2(k)\cos 2\theta_{\bf k}\nonumber\\ & & \mbox{}-\alpha_3(k)\cos 4\theta_{\bf k}],\label{eq2}\\ \omega_z^{\mu}({\bf k})&=&\mu\beta_1(k)+\beta_2(k)\cos 3\theta_{\bf k},\label{eq3}\end{aligned}\ ] ] with @xmath8 standing for k(k@xmath9 ) valleys and @xmath10 and @xmath11 representing the magnitude and the polar angle of the momentum @xmath12 relative to k(k@xmath9 ) points , respectively . the coefficients @xmath13 ( @xmath14-@xmath15 ) and @xmath16 are given in appendix [ appa ] by fitting to the numerical results calculated according to the latest report by konschuh _ et al._@xcite from first principles with both the intrinsic and extrinsic soc terms in the pseudospin space included . the out - of - plane component @xmath17 [ eq . ( [ eq3 ] ) ] , which has not been reported in the previous literature , is induced by the intrinsic soc in the pseudospin space in the absence of space inversion symmetry , whereas the in - plane component is contributed by both the intrinsic and extrinsic socs in the pseudospin space . this is very different from the case in single - layer graphene where the intrinsic soc in the pseudospin space only induces a shift of the energy spectrum in the conduction band.@xcite we find that the magnitudes of both the out - of- and in - plane components decrease with the increase of @xmath10 at large momentum , indicating a suppression of the inhomogeneous broadening@xcite by increasing momentum.@xcite this is very different from the case in both semiconductors and single - layer graphene.@xcite additionally , the leading term of the out - of - plane component , i.e. , @xmath18 , serves as a zeeman - like term with opposite effective magnetic fields perpendicular to blg plane in the two valleys , similar to the case in rippled single - layer graphene.@xcite this zeeman - like term , together with the intervalley scattering , opens a spin relaxation channel . this spin relaxation channel suppresses the in - plane srt _ significantly _ at high temperature by the intervalley electron - phonon scattering and also at low temperature by the short - range scattering.@xcite in addition , we find that although such zeeman - like terms also exist in the in - plane components of the soc , their contributions to the out - of - plane spin relaxation are marginal since these terms are of the high order of the momentum . this leads to a strong anisotropy between the in - plane and out - of - plane spin relaxations . we also find that the in - plane srt shows a marked nonmonotonic dependence on temperature with a minimum down to several hundred picoseconds in the absence of short - range scatterers . this minimum is comparable to the experimental data . the nonmonotonic behavior results from the crossover between the weak and strong intervalley electron - phonon scattering . moreover , we predict a peak in the density dependence at low temperature , which is very different from the one in semiconductors.@xcite as for high temperature , the srt shows a monotonic increase with increasing density . we also find that the in - plane srt decreases rapidly with increasing initial spin polarization at low temperature , which is very different from the previous studies in both semiconductors@xcite and single - layer graphene.@xcite finally , we report detailed comparisons of the temperature and electron density dependences with the existing experiments.@xcite this paper is organized as follows . in sec . ii , we present our model and construct the ksbes . the main results are given in sec . iii . in sec . iiia , we investigate the temperature , electron density and also initial spin polarization dependences of the electron spin relaxation in the absence of short - range scatterers . the anisotropy of the spin relaxation is also addressed in this part . in sec . iiib , we show detailed comparisons of the temperature and electron density dependences of the spin relaxation with the existing experiments with the inclusion of short - range scatterers . we summarize in sec . iv . we start our investigation from the ab - stacked blg on sio@xmath0 substrate . the spinless @xmath19-band structure can be described by the effective @xmath20 hamiltonian:@xcite @xmath21 in the on - site orbital bloch basis @xmath22 , @xmath23 , @xmath24 and @xmath25 , where a@xmath26 refers to the a sublattice in the lower layer , b@xmath0 to the b sublattice in the upper layer , etc . , and @xmath27 represents the two - dimensional wave vector counted from the @xmath28 point . here , @xmath29 and @xmath30 describe the nearest - neighbor intralayer and interlayer hoppings whereas @xmath31 and @xmath32 stand for the indirect hoppings between two layers . @xmath33 represents the asymmetry in the energy shift of the on - site energy due to the interlayer hopping . the potential @xmath34 with @xmath35 , @xmath36 and @xmath37 representing the electron charge @xmath38 , the out - of - plane electric field and the effective electrostatic bilayer distance , respectively . @xmath39 $ ] is the nearest - neighbor structural function of the graphene hexagonal lattice with @xmath40 denoting the lattice constant . near the dirac points , to the first order of the momentum @xmath41 originated from the k(k@xmath9 ) points , @xmath42 . then the effective @xmath20 hamiltonian near the dirac points can be written as @xmath43 by exactly diagonalizing this hamiltonian , one obtains the eigenvalue @xmath44 and the eigenfunction @xmath45 with @xmath46 being the band index including two conduction bands and two valence ones . in our calculation , we concentrate on the heavily @xmath47-doped case with only the lowest conduction band being relevant whereas other bands are separated by large energy intervals . for convenience , the eigenvalue and the corresponding eigenfunction of the lowest conduction band at @xmath12 are denoted by @xmath48 and @xmath49 , respectively . by incorporating the spin degree of freedom , the effective hamiltonian of the lowest conduction band is given by @xmath50 here , @xmath51 are the pauli matrices . @xmath52 stands for the effective magnetic field of the soc , which is constructed from the symmetry analysis with the parameters obtained approximately by fitting to numerical results based on the report by konschuh _ et al._.@xcite the analytical form of the soc with three components @xmath5 @xmath6 is shown in eqs . ( [ eq1]-[eq3 ] ) and the explicit numerical calculation is given in appendix [ appa ] . it is noted that at large momentum , the analytical result agrees with the numerical one fairly well ( see appendix [ appa ] ) . we then construct the microscopic ksbes@xcite to study the electron spin relaxation in blg . the ksbes read@xcite @xmath53 where @xmath54 represent the density matrices of electrons with the diagonal terms @xmath55 describing the distribution functions and the off - diagonal ones @xmath56 denoting the spin coherence . the coherent term is given by @xmath57 , \label{eq8}\end{aligned}\ ] ] where @xmath58\equiv ab - ba$ ] is the commutator ; @xmath59 represents the coulomb hartree - fock ( hf ) term.@xcite @xmath60 denotes the screened coulomb potential with its form given in appendix [ appb ] . the form factor @xmath61 . @xmath62 are the scattering terms including the electron - electron coulomb ( @xmath63 ) , long - range electron - impurity ( @xmath64 ) , intravalley electron - ac phonon ( @xmath65 ) , electron - op phonon ( @xmath66 ) , electron - remote - interfacial ( ri ) phonon ( @xmath67 ) and especially the intervalley electron - phonon scattering including the electron-@xmath68 phonon ( @xmath69 ) , electron-@xmath70 phonon ( @xmath71 ) , electron-@xmath72 phonon ( @xmath73 ) and electron-@xmath74 phonon ( @xmath75 ) scatterings . their detailed expressions can be found in ref . .@xcite here , @xmath68 , @xmath70 and @xmath72 are the in - plane phonon modes corresponding to the representations @xmath76 , @xmath77 and @xmath78 of group @xmath79 , respectively whereas @xmath74 stands for the out - of - plane phonon mode.@xcite it is noted that the intravalley electron - phonon scattering has been reported in the previous literature as mentioned in the introduction.@xcite however , to the best of our knowledge , there is no specific report to the intervalley electron - phonon scattering . in this work , we derive the intervalley electron - phonon scattering matrix elements using the tight - binding model according to the arxiv version of ref . . the scattering matrix elements are given by @xmath80,\label{eq9}\\ & & \hspace{-0.4cm}|m^{\rm ka_2^{\prime}}_{{\mu{\bf k}},{\mu^{\prime}{\bf k}^{\prime}}}|^2=\frac{3a^2{\gamma_4^{\prime}}^2}{2\rho l_4 ^ 2 } \frac{\hbar^2}{2\omega_{\rm ka_2^{\prime}}}\delta_{\mu^{\prime},-\mu } \big[|{\psi^{\mu}_{{\bf k}}}^{\dagger}\gamma^3_{\rm d}\gamma^5_{\rm d}\psi^{\mu^{\prime}}_{{\bf k}^{\prime}}|^2\nonumber\\ & & \mbox{}\hspace{0.1cm}+|{\psi^{\mu}_{{\bf k}}}^{\dagger}\gamma^0_{\rm d}\psi^{\mu^{\prime}}_{{\bf k}^{\prime}}|^2\big],\label{eq10}\\ & & \hspace{-0.4cm}|m^{\rm ke^{\prime}}_{{\mu{\bf k}},{\mu^{\prime}{\bf k}^{\prime}}}|^2=\frac{3a^2}{\rho l_4 ^ 2 } \frac{\hbar^2}{4\omega_{\rm ke^{\prime}}}\delta_{\mu^{\prime},-\mu } \big[|{\psi^{\mu}_{{\bf k}}}^{\dagger}(\gamma_3^{\prime}\gamma^1_{\rm d}\gamma^5_{\rm d}-i\gamma_3^{\prime}\gamma^2_{\rm d}\nonumber\\ & & \mbox{}\hspace{0.1cm}-\gamma_4^{\prime}\gamma^0_{\rm d})\psi^{\mu^{\prime}}_{{\bf k}^{\prime}}|^2+|{\psi^{\mu}_{{\bf k}}}^{\dagger } ( i\gamma_3^{\prime}\gamma^2_{\rm d}-\gamma_3^{\prime}\gamma^1_{\rm d}\gamma^5_{\rm d}-\gamma_4^{\prime}\gamma^0_{\rm d } ) \psi^{\mu^{\prime}}_{{\bf k}^{\prime}}|^2\nonumber\\ & & \mbox{}\hspace{0.1cm}+4|{\psi^{\mu}_{{\bf k}}}^{\dagger}\gamma_4^{\prime}\gamma^3_{\rm d}\gamma^5_{\rm d } \psi^{\mu^{\prime}}_{{\bf k}^{\prime}}|^2\big],\label{eq11}\\ & & \hspace{-0.4cm}|m^{\rm kz}_{{\mu{\bf k}},{\mu^{\prime}{\bf k}^{\prime}}}|^2=\frac{{\gamma_1^{\prime}}^2}{\rho } \frac{\hbar^2}{2\omega_{\rm kz}}\delta_{\mu^{\prime},-\mu } \label{eq12}\end{aligned}\ ] ] where @xmath81 is the mass density of the blg;@xcite @xmath82 , @xmath83 , @xmath84 and @xmath85 are the energy spectra of @xmath68 , @xmath70 , @xmath72 , and @xmath74 phonon modes , respectively;@xcite @xmath86 ( @xmath87 ) is the derivative of @xmath88 with respect to the corresponding hopping bond length;@xcite @xmath89 is the bond length of the interlayer hopping @xmath32 ; @xmath90 ( @xmath91-@xmath15 , @xmath92 ) , @xmath93 and @xmath94 are @xmath20 dirac matrices given in appendix [ appc].@xcite we will show that the above intervalley electron - phonon scattering plays a significant role in the in - plane spin relaxation in our investigation . the remaining scattering matrix elements such as @xmath63 , @xmath64 , and @xmath95 ( @xmath96 ) are given in appendix [ appb ] . it is noted that we also include the short - range scattering in our calculation with the scattering term laid out later in this paper . .parameters used in the calculation ( from ref . unless otherwise specified ) . [ cols= " < , < , < , < " , ] by numerically solving the ksbes , one obtains the time evolution of spin polarization along direction @xmath97 @xmath98/n_e$ ] with @xmath99 being the electron density . then the srt can be determined from the slope of the envelope of the spin polarization @xmath100.@xcite the initial spin polarization @xmath101 is set to be @xmath102 % and the spin - polarization direction @xmath97 is along the @xmath103-axis unless otherwise specified . it is noted that all the parameters used in our calculation are listed in table i. it is also noted that the soc is calculated both fully numerically following the approach by konschuh _ et al._@xcite and analytically by using eqs . ( [ eq1]-[eq3 ] ) . the srt calculated with the analytical form of the soc and with the explicit numerical one agrees fairly well with each other . in our calculation , the out - of - plane electric field is chosen to be a typical value @xmath104v / nm@xcite and the long - range impurity density is taken to be @xmath105@xmath106 . the corresponding mobility in our investigation is of the order of @xmath107@xmath108/(vs ) , which is about @xmath109-@xmath110 orders of magnitude larger than those reported in the existing experiments in blg on sio@xmath0 substrate@xcite and comparable to the ones in the experiments in freely suspended blg.@xcite it is noted that the short - range scattering is not included in this subsection . -axis @xmath111 ( @xmath112 ) and the calculation without the long - range electron - impurity ( @xmath113 ) , electron - ac - phonon ( @xmath114 ) , electron - ri - phonon ( @xmath115 ) , electron - electron coulomb ( @xmath116 ) or intervalley electron - phonon ( @xmath117 ) scattering as function of temperature @xmath118 . it is noted that when @xmath119 k , only the total srt and the calculation without the intervalley electron - phonon scattering are shown since the remaining ones almost coincide with the total srt . in the calculation , the electron density @xmath120@xmath106.,width=321 ] we first investigate the temperature dependence of the spin relaxation . in fig . [ fig1 ] , the srt along the @xmath103-axis @xmath111 is plotted as function of temperature @xmath118 . it is seen that @xmath111 changes little with the temperature at low temperature ( i.e. , @xmath121k ) . however , as @xmath118 becomes larger , the srt decreases dramatically with increasing temperature . in particular , the srt reaches down to several hundred picoseconds at @xmath122k , which is comparable to the experimental values.@xcite a detailed comparison with the existing experiments will be discussed in sec . when @xmath118 further increases , the srt begins to increase with @xmath118 . to understand the behavior of the spin relaxation , we show in fig . [ fig1 ] the temperature dependence of the srt calculated with the electron - electron coulomb , electron - ri - phonon , electron - ac - phonon , long - range electron - impurity , or intervalley electron - phonon scattering removed , separately . we find that at low temperature , the intervalley electron - phonon scattering is negligible . therefore , spins relax independently in the two valleys and the srt is solely determined by the intravalley spin relaxation channel . specifically , the long - range electron - impurity scattering is dominant , which leads to a weak temperature dependence of the srt.@xcite with the increase of the temperature , the contribution of the long - range electron - impurity scattering becomes marginal whereas the intervalley electron - phonon scattering becomes important . this intervalley scattering opens another spin relaxation channel together with the zeeman - like term with opposite effective magnetic fields in the two valleys , which is similar to the case in rippled single - layer graphene.@xcite this zeeman - like term is given by @xmath18 [ eq . ( [ eq3 ] ) ] approximately . then the spin relaxation due to the above intervalley spin relaxation channel can be approximated by the rate equations@xcite @xmath123 where @xmath124 , @xmath125\hat{{\bf z}}$ ] and @xmath126 represent the in - plane spin vectors in each valley , spin precession vector along the @xmath127-axis and the intervally electron - phonon scattering time , respectively . according to the report by zhang _ et al._,@xcite the in - plane srt is given by @xmath128 here , @xmath129 stands for the srt along the @xmath130-axis . since the system we investigate is always in the degenerate regime , we take @xmath131 approximately with @xmath132 representing the fermi wave vector . in our calculation , the srt along the @xmath103-axis is solely determined by the intervalley electron - phonon scattering at @xmath133-@xmath134k . at @xmath135k , @xmath136 , i.e. , the intervalley electron - phonon scattering is in the weak scattering limit . as a result , the srt @xmath137 according to eq . ( [ eq14 ] ) , which decreases with the enhancement of the intervalley scattering as the temperature increases . nevertheless , at @xmath138k , @xmath139 , i.e. , the intervalley scattering is in the strong scattering limit . then the srt @xmath140 $ ] from eq . ( [ eq14 ] ) , which increases with increasing of @xmath118 . therefore , the srt first decreases then increases with increasing temperature when the intervalley scattering changes from the weak to strong scattering limit . the crossover between the weak and strong intervalley scattering limit is determined by @xmath141.@xcite at this crossover point , @xmath142 here @xmath143ps , very close to the value @xmath144ps shown in the figure . it is noted that the electron - electron coulomb scattering is unimportant in the whole temperature regime . then we turn to study the electron - density dependence . in fig . [ fig2 ] , the srt along the @xmath103-axis is plotted against the electron density at @xmath145 @xmath146k . one observes a peak in the electron - density dependence of the srt at @xmath135k whereas the srt increases with the increase of the electron density when @xmath147k . we first focus on the case at @xmath135k and show the srt with the intervalley electron - phonon scattering excluded in fig . it is seen that when the electron density is low , the intervalley electron - phonon scattering is marginal . the srt is then determined by the intravalley spin relaxation channel and hence increases with the suppression of the inhomogeneous broadening@xcite as the electron density increases.@xcite the suppression of the inhomogeneous broadening can be understood from the decrease of the spin splitting with increasing momentum at high momentum [ see eqs . ( [ eq1]-[eq3 ] ) ] , as shown in fig . [ fig9 ] in appendix [ appa ] . this behavior is very different from the case in both semiconductors and single - layer graphene.@xcite when the electron density further increases , the intervalley electron - phonon scattering becomes important . this intervalley scattering is always in the weak scattering limit with @xmath148 and therefore @xmath137 according to eq . ( [ eq14 ] ) . as a result , the srt decreases with the increase of the intervalley scattering and hence the increase of the electron density.@xcite therefore , in the whole electron - density regime , a peak is observed . it is noted that this peak is very different from the one predicted by jiang and wu@xcite in semiconductors where the peak is attributed to the crossover from the nondegenerate - to - degenerate limit . as for the case at @xmath147k , the srt is always dominated by the intervalley electron - phonon scattering as shown in fig . [ fig2 ] by comparing the srt calculated with and without the intervalley electron - phonon scattering . since this intervalley scattering is in the strong scattering limit with @xmath149 , the srt @xmath140 $ ] from eq . ( [ eq14 ] ) . this srt increases with increasing electron density since both @xmath150 and the intervalley scattering time @xmath151 decrease with the increase of the electron density.@xcite -axis @xmath111 as function of the electron density @xmath99 with @xmath112 ( @xmath152 ) and without @xmath115 ( @xmath114 ) the intervalley electron - phonon scattering at @xmath153k.,width=321 ] -axis @xmath111 @xmath114 ( @xmath154 ) and the calculation without the coulomb hf term @xmath152 ( @xmath112 ) or the intervalley electron - phonon scattering @xmath113 ( @xmath115 ) as function of the initial spin polarization @xmath155 at @xmath153k . in the calculation , the electron density @xmath156@xmath106.,width=321 ] the initial spin polarization dependences of the srt along the @xmath103-axis at @xmath145 and @xmath157k are shown in fig . it is seen that the srt decreases rapidly with the increase of initial spin polarization @xmath155 at @xmath135k whereas the srt shows a slight increase with increasing @xmath155 at @xmath147k . the rapid decrease of the srt at @xmath135k is very different from the previous studies in both semiconductors@xcite and single - layer graphene@xcite where the srt increases significantly with increasing initial spin polarization . in the previous studies , the increase of the srt originates from the contribution of the coulomb hf term.@xcite this term serves as an effective magnetic field along the direction of the spin polarization , which blocks the spin precession induced by the dresselhaus@xcite and/or rashba@xcite socs and slows down the spin relaxation . however , the contribution of the coulomb hf term is marginal to the spin relaxation in our investigation , which is shown in fig . [ fig3 ] by comparing the the srt calculated with and without the coulomb hf term . here , the srt is determined by the intervalley electron - phonon scattering , as shown in fig . [ fig3 ] by comparing the calculation with and without the intervalley electron - phonon scattering . the intervalley electron - phonon scattering transfers electrons between the two valleys . the transferred electrons experience opposite effective magnetic fields induced by the zeeman - like term in the two valleys , which are not affected by the effective magnetic field induced by the coulomb hf term . moreover , the intervalley electron - phonon scattering is in the weak scattering limit at low temperature and hence the srt @xmath137 [ see eq . ( [ eq14 ] ) ] . the intervalley scattering time @xmath151 decreases significantly with the increase of the initial spin polarization as the density of states increases with the polarization . therefore , the srt shows a rapid decrease with the initial spin polarization at @xmath135k . as for the case of @xmath147k , the srt is also determined by the intervalley electron - phonon scattering in comparison with the case of @xmath135k . however , in contrast to the case of @xmath135k , the intervalley scattering is in the strong scattering limit . this makes the srt present an opposite trend to the case of @xmath135k . a mild increase of the srt results from the insensitivity of @xmath151 to the initial spin polarization at @xmath147k . @xmath158-axis @xmath111 @xmath159 as function of temperature @xmath118 . in the calculation , the electron density @xmath120@xmath106.,width=321 ] we also address the anisotropy of the spin relaxation with respect to the spin polarization direction . due to the existence of the out - of - plane effective magnetic field [ see eq . ( [ eq3 ] ) ] , the in - plane srts are identical.@xcite for comparison , we plot the temperature dependence of the srt along the @xmath103 ( @xmath127)-axis as the red curve with crosses ( blue curve with open squares ) in fig . it is seen that in comparison with the srt along the @xmath127-axis , the one along the @xmath103-axis is comparable at low temperature and orders of magnitude smaller at high temperature . as mentioned previously , the srt along the @xmath103-axis is strongly suppressed by the intervalley spin relaxation channel induced by the intervalley electron - phonon scattering and the zeeman - like term in the two valleys at high temperature . however , the contribution of this spin relaxation channel to the srt along the @xmath127-axis is marginal [ see eqs . ( [ eq1]-[eq2 ] ) ] . therefore , a strong anisotropy between the out - of- and in - plane spin relaxations is observed at high temperature . as function of temperature @xmath118 . @xmath116 ( @xmath114 ) corresponds to the case of the gate voltage v@xmath160v in the experiment of han and kawakami;@xcite @xmath112 ( @xmath115 ) stands for the case of electron density @xmath161@xmath106 in the experiment of avsar _ et al._;@xcite @xmath154 ( @xmath117 ) represents the case of @xmath162@xmath106 in the experiment of yang _ et al._;@xcite @xmath163 denotes the case of @xmath164@xmath106 in the experiment of neumann _ et al._.@xcite ( a ) red solid ( orange chain ) curve stands for our calculation corresponding to the case of v@xmath160v in the absence of the short - range scatterers . in the inset , we show the temperature dependence of the long - range impurity density @xmath165 corresponding to the case of v@xmath166v . ( b ) grey dashed and skyblue double - dotted chain curves are obtained from our calculations corresponding to the case of v@xmath166v by including the short - range scatterers , with a fixed short - range impurity density @xmath167@xmath106 and a tunable @xmath168 ( shown in the inset ) , respectively . we also show the dependence of @xmath169 on @xmath168 ( with the scale on the top of the frame ) at @xmath170k with other conditions the same as the case of v@xmath166v . the short - range potential strength @xmath171mevm@xmath172.,title="fig:",width=332 ] as function of temperature @xmath118 . @xmath116 ( @xmath114 ) corresponds to the case of the gate voltage v@xmath160v in the experiment of han and kawakami;@xcite @xmath112 ( @xmath115 ) stands for the case of electron density @xmath161@xmath106 in the experiment of avsar _ et al._;@xcite @xmath154 ( @xmath117 ) represents the case of @xmath162@xmath106 in the experiment of yang _ et al._;@xcite @xmath163 denotes the case of @xmath164@xmath106 in the experiment of neumann _ et al._.@xcite ( a ) red solid ( orange chain ) curve stands for our calculation corresponding to the case of v@xmath160v in the absence of the short - range scatterers . in the inset , we show the temperature dependence of the long - range impurity density @xmath165 corresponding to the case of v@xmath166v . ( b ) grey dashed and skyblue double - dotted chain curves are obtained from our calculations corresponding to the case of v@xmath166v by including the short - range scatterers , with a fixed short - range impurity density @xmath167@xmath106 and a tunable @xmath168 ( shown in the inset ) , respectively . we also show the dependence of @xmath169 on @xmath168 ( with the scale on the top of the frame ) at @xmath170k with other conditions the same as the case of v@xmath166v . the short - range potential strength @xmath171mevm@xmath172.,title="fig:",width=359 ] as mentioned in the introduction , experiments on spin relaxation have been carried out in blg on sio@xmath0 substrate@xcite or in freely suspended blg@xcite very recently by different groups . the in - plane srts in these experiments are of the order of @xmath1-@xmath2ns , which are comparable to our theoretical results at high temperature ( several hundred picoseconds ) . this motivates us to carry out a detailed comparison with the existing experiments . in these experiments , the gate voltage , temperature and electron density@xcite ( tuned by the gate voltage ) are given explicitly . however , some parameters such as the out - of - plane electric field and the impurity density are not available in the experiments . in our calculation , the out - of - plane electric field is taken to be the ratio of the gate voltage to the thickness of the substrate approximately.@xcite the impurity density is obtained by fitting to the mobility or the spin diffusion coefficient given in the experiments.@xcite in addition , we assume the initial spin polarization along the @xmath103-axis without loss of generality since the in - plane spin relaxations are identical as pointed out previously . we first compare the temperature dependence of the srts obtained from the experiments and calculation from our model . in these experiments , the mobilities are reported of the order of @xmath173-@xmath107@xmath108/(vs ) . the srts from different groups are comparable as shown in fig . however , the srt from the experiment of han and kawakami@xcite decreases with increasing temperature whereas the ones from the experiments of yang _ et al._@xcite and avsar _ et al._@xcite show a marginal temperature dependence . in addition , han and kawakami@xcite reported the longest srt among these experiments , indicating that their experimental data may contain less extrinsic effects . motivated by this , we show our comparison with the experiment of han and kawakami@xcite in fig . [ fig5](a ) . the experimental data with the gate voltage v@xmath160v corresponding to the electron density @xmath174@xmath106 , is labeled as @xmath175 . by fitting to the spin diffusion coefficient given in the experiment , we obtain the long - range impurity density @xmath165 shown in the inset as the curve with @xmath113 . in contrast to the long - range impurity scattering , the contribution of other scatterings to the spin diffusion coefficient ( mobility ) is marginal . then we calculate the srt corresponding to v@xmath160v shown as the red solid ( orange chain ) curve . a crossover is observed in the temperature dependence of the srt , similar to the one shown in sec . it is seen that the srt from our calculation is comparable to the experimental data at high temperature . this suggests that the intervalley spin relaxation channel induced by the intervalley electron - phonon scattering and the zeeman - like term in the two valleys addressed in sec . iiia , plays an important role at high temperature . nevertheless , the srts are still larger than the experimental data with an observable difference at high temperature . additionally , at low temperature , our results are orders of magnitude larger than the experimental ones since the intervalley electron - phonon scattering is marginal at low temperature and the srt is determined by the long - range electron - impurity scattering in our calculation . due to the existence of all these discrepancies , one has to take into account extrinsic effects such as adatoms,@xcite curvature,@xcite substrate effects@xcite and also contacts.@xcite in the above experiments on blg , the spin polarized electrons are all injected from ferromagnetic contacts.@xcite as reported , the ferromagnetic electrodes generate stray fields , which cause spin precession around the contact.@xcite as a result , the stray fields may have a strong influence on the observed experimental results . additionally , electron scattering by the tunneling contacts may also be responsible for the available experimental data.@xcite in addition to the above extrinsic effects , the short - range scatterers may also be an extrinsic source.@xcite here , we discuss the contribution of short - range scatterers@xcite in detail . the scattering term induced by the short - range scatterers is given by @xmath176 with @xmath168 and @xmath177 denoting the short - range impurity density and the constant short - range potential strength , respectively.@xcite this scattering term contributes to both the intra- and inter - valley scatterings . the intervalley scattering opens an intervalley spin relaxation channel together with the zeeman - like term in the two valleys [ see eq . ( [ eq3 ] ) ] , similar to the intervalley electron - phonon scattering . however , in contrast to the intervalley electron - phonon scattering , the intervalley scattering induced by the short - range scatterers is _ insensitive _ to the temperature and may play an important role in the in - plane spin relaxation especially at low temperature . here , we choose a fixed short - range impurity density @xmath167@xmath106 and potential strength @xmath171mevm@xmath172 . it is noted that although with this additional short - range scattering included , the mobility is still dominated by the long - range electron - impurity scattering since the contribution of the long - range electron - impurity scattering is about four orders of magnitude larger than that of the short - range one . we calculate the temperature dependence of the srt corresponding to the case of v@xmath166v in the experiment of han and kawakami@xcite by including the additional short - range scattering . the result is plotted as the grey dashed curve in fig . [ fig5](b ) . a crossover is also observed at high temperature due to the contribution of the intervalley electron - phonon scattering as pointed out previously . by comparing our calculation with the experimental data labeled as @xmath116 , we find that our result becomes comparable at low temperature . furthermore , by fitting to the experimental data with tunable short - range impurity density shown as the curve with @xmath152 in the inset , our result labeled as the skyblue double - dotted chain curve agrees fairly well with the experimental data in the temperature regime lower than the crossover . as for the temperature higher than the crossover , our result becomes even larger than the experimental one once the short - range scattering is included . this is because that the intervalley electron - phonon scattering is in the strong scattering limit . with the inclusion of the short - range scattering , the intervalley scattering becomes stronger and the srt becomes larger . this discrepancy may be attributed to other extrinsic effects . as mentioned above , the srt can be tuned by the short - range impurity density . we show the short - range impurity density dependence of the in - plane srt at @xmath170k as the purple dotted curve in fig . [ fig5](b ) , with all other conditions remaining the same as the case of v@xmath166v . a valley is also observed , similar to the case of the temperature dependence discussed above . at low temperature , the intervalley electron - phonon scattering is negligible . the srt is then determined by the intervalley short - range scattering . the valley originates from the crossover between the weak and strong intervalley short - range scattering limit by tuning the short - range impurity density . at the crossover point , the srt @xmath178ps according to eq . ( [ eq15 ] ) , close to the value @xmath179ps shown in the figure . this crossover point gives the minimum of the srt and can be taken as a criterion for determining the possible contribution of the short - range scattering . specifically , the crossover point @xmath180 , @xmath181 , @xmath182 , @xmath183 , and @xmath184ps corresponding to the case of the experimental data shown as @xmath114 , @xmath112 , @xmath115 , @xmath154 and @xmath117 , respectively . we find that both the experimental data of han and kawakami@xcite and avsar _ et al._@xcite are larger than the crossover point whereas the one of yang _ et al._@xcite is smaller . this suggests the experimental data of han and kawakami@xcite and avsar _ et al._@xcite can be explained by the short - range scatterers . as for the experiment of yang _ et al._,@xcite some other extrinsic effects have to be considered . as function of electron density . the dotted curves with symbols correspond to the experimental data of han and kawakami ( hk)@xcite at @xmath185k and avsar _ et al._@xcite and yang _ et al._@xcite at @xmath186k . the solid curves with symbols correspond to the experimental data at room temperature of han and kawakami,@xcite avsar _ et al._@xcite and yang _ et al._.@xcite red solid ( blue chain ) curve is from our calculation corresponding to the case of room ( low ) temperature in the experiment of han and kawakami.@xcite it is noted that the blue chain curve is calculated with the short - range scatterers with the short - range impurity density @xmath187@xmath106 and potential strength @xmath171mevm@xmath172.,width=321 ] then we turn to compare the electron density dependence of the srts from the experiments and our calculation . among the experiments , han and kawakami@xcite measured the longest srt at low temperature ( up to several nanoseconds ) shown as the dotted curve with @xmath114 in fig . this srt is about one order of magnitude larger than the ones from other experimental cases . in addition , the srts from the experiments show different dependences on electron density . specifically , at low temperature , both avsar _ et al._@xcite and yang _ et al._@xcite show a decrease of the srt with the increase of the electron density whereas a mild peak is observed by han and kawakami.@xcite at room temperature , the srts from avsar _ et al._@xcite and yang _ et al._@xcite increase with increasing electron density whereas han and kawakami@xcite reported an insensitive density dependence of the srt . we first show our comparison with the experimental data at low temperature . as shown in fig . [ fig5](a ) , our result is orders of magnitude larger than the experimental data of han and kawakami@xcite at low temperature without short - range impurities . however , with the inclusion of the short - range scatterers , our result shows good agreement with the experimental data . here , with the short - range scatterers explicitly included , we also show our comparison with the experimental data of han and kawakami.@xcite with a fixed short - range impurity density @xmath187@xmath106 and potential strength @xmath171mevm@xmath172 , the electron density dependence of the srt from our calculation is shown as blue chain curve in fig . we find that our result marginally agrees with the experimental data shown as the dotted curve with @xmath114 . better agreement can be obtained with tunable short - range impurity density . in addition , we also find that the experimental data of avsar _ et al._@xcite can be explained by the short - range scatterers . however , to account for the experimental data of yang _ et al._,@xcite one has to include other extrinsic effects . this is because the srt of yang _ et al._@xcite is smaller than the crossover point in the short - range impurity density dependence of the srt ( @xmath188-@xmath189ps ) , which is taken as a criterion for the possible contribution of the short - range scattering as mentioned previously . we then address our comparison with the experimental data at room temperature . as shown in fig . [ fig5](a ) , our result is larger than the experimental data of han and kawakami@xcite with an observable difference at room temperature without short - range impurities . with the short - range scatterers included , the srt from our calculation becomes even larger than the experimental data as mentioned previously , indicating that other extrinsic effects have to be taken into account . here , we show our comparison with the experiment of han and kawakami@xcite without extrinsic effects . the experimental data and our result are shown as the solid curve with @xmath113 and the red solid curve in fig . [ fig6 ] , respectively . we find that the srt in our calculation is about three times as large as the experimental one in low electron density regime and one order of magnitude larger than the experimental one in high density regime . in summary , we have investigated the electron spin relaxation due to the dp mechanism in blg with only the lowest conduction band being relevant . the soc of the lowest conduction band is constructed from the symmetry group analysis with the parameters obtained by fitting to the numerical calculation according to the latest report by konschuh _ et al._@xcite from first principles with both the intrinsic and extrinsic soc terms in the pseudospin space included . we find that the magnitudes of both the out - of- and in - plane components of the soc decrease with increasing momentum at large momentum , indicating a suppression of the inhomogeneous broadening with the increase of the momentum . this is different from the case in both semiconductors and single - layer graphene . additionally , the leading term of the out - of - plane component of the soc serves as a zeeman - like term with opposite effective magnetic fields in the two valleys , which is similar to the case in rippled single - layer graphene . this zeeman - like term , together with the intervalley electron - phonon and/or possible intervalley short - range scatterings , opens an intervalley spin relaxation channel , which has not been reported in the literature in blg . the intervalley electron - phonon scattering is derived by using the tight - binding model . in addition to the intervalley electron - phonon and short - range scatterings , we also include the long - range electron - impurity , electron - electron coulomb and intravalley electron - phonon scatterings to calculate the srt . we find that the in - plane srt is strongly suppressed by the intervalley electron - phonon scattering at high temperature . in contrast to the intervalley electron - phonon scattering , the intervalley short - range scattering is insensitive to the temperature and plays an important role in the in - plane spin relaxation especially at low temperature . in the absence of short - range scatterers , a marked nonmonotonic temperature dependence of the in - plane srt is predicted with a minimum srt down to several hundred picoseconds . this nonmonotonic behavior originates from the crossover between the weak and strong intervalley electron - phonon scattering . moreover , we predict a peak in the electron density dependence of the in - plane srt at low temperature , which is very different from the one in semiconductors . at high temperature , the in - plane srt increases monotonically with increasing density . we also find that the in - plane srt decreases rapidly with the increase of the initial spin polarization at low temperature . this is very different from the previous studies in both semiconductors and single - layer graphene where the srt increases significantly with increasing initial spin polarization . the physics is understood that the spin relaxation time is determined by the intervalley electron - phonon scattering , which transfers electrons between the two valleys . the transferred electrons experience opposite effective magnetic fields in the two valleys , which are not affected by the coulomb hf term . in addition , a strong anisotropy between the out - of- and in - plane spin relaxations is also addressed at high temperature where the out - of - plane srt is about two orders of magnitude larger than the in - plane one . as for low temperature , the out - of- and in - plane spin relaxation times are comparable . we also show our comparison with the existing experiments of han and kawakami,@xcite avsar _ et al._@xcite and yang _ et al._.@xcite we find that without intervalley scattering , the srt is orders of magnitude larger than the experimental data in the whole temperature regime . with the intervalley electron - phonon scattering explicitly included , the srt from our calculation becomes comparable to the experimental data at high temperature but still orders of magnitude larger than the experimental data at low temperature . in addition , a crossover in the temperature dependence of the in - plane srt is shown , which also results from the crossover between the weak and strong intervalley electron - phonon scattering . with the inclusion of the short - range scatterers , our result agrees fairly well with the experimental data in the temperature regime lower than the crossover . as for the temperature higher than the crossover , other extrinsic effects have to be included . moreover , a crossover point is also shown in the short - range impurity density dependence of the in - plane srt , which gives a minimum for determining the possible contribution of the short - range scattering . the experimental srt larger than the minimum can be explained by the possible short - range scatterers whereas other extrinsic effects have to be considered for experimental data smaller than the minimum . this work was supported by the national basic research program of china under grant no . 2012cb922002 and the strategic priority research program of the chinese academy of sciences under grant no . one of the authors ( l.w . ) would like to thank b. y. sun for double checking the derivation of the electron - phonon scattering matrix elements . , ( b ) @xmath190 and ( c ) @xmath191 as function of the out - of - plane electric field @xmath36 with @xmath192-@xmath92 . it is noted that @xmath193 , @xmath194 , @xmath195 , @xmath196 and @xmath197 ( @xmath14-@xmath15 ) are dimensionless.,title="fig:",width=257 ] , ( b ) @xmath190 and ( c ) @xmath191 as function of the out - of - plane electric field @xmath36 with @xmath192-@xmath92 . it is noted that @xmath193 , @xmath194 , @xmath195 , @xmath196 and @xmath197 ( @xmath14-@xmath15 ) are dimensionless.,title="fig:",width=257 ] , ( b ) @xmath190 and ( c ) @xmath191 as function of the out - of - plane electric field @xmath36 with @xmath192-@xmath92 . it is noted that @xmath193 , @xmath194 , @xmath195 , @xmath196 and @xmath197 ( @xmath14-@xmath15 ) are dimensionless.,title="fig:",width=257 ] the soc term of the lowest conduction band near the dirac points can be described by a @xmath198 matrix @xmath199 in the basis @xmath200 where @xmath201 and @xmath202 ( @xmath203 ) . the coefficients @xmath204 can be obtained by exactly diagonalizing the @xmath20 effective hamiltonian given in eq . ( [ eq5 ] ) . here , @xmath205 with @xmath206 ( @xmath207 ) being the spin - orbit matrix elements near the dirac points in the pseudospin space given by konschuh _ et al._.@xcite specifically , @xmath208+\frac{\mu-1}{2}[{c_3^{\mu}}^*c_1^{\mu}(i\lambda_4 ) \nonumber\\ & & \mbox{}\hspace{-0.4cm}+{c_1^{\mu}}^*c_2^{\mu}(-i\lambda_0)+{c_4^{\mu}}^*c_2^{\mu}(-i\lambda_4^{\prime})+{c_2^{\mu}}^*c_3^{\mu}(-i\lambda_3)\nonumber\\ & & \mbox{}\hspace{-0.4cm}+{c_3^{\mu}}^*c_4^{\mu}(i\lambda_0^{\prime})].\label{a7}\end{aligned}\ ] ] @xmath209 and @xmath210 are the strengths of single - layer - like intrinsic soc ; @xmath211 and @xmath212 represent single - layer - like extrinsic spin - orbit strengths ; @xmath213 , @xmath214 and @xmath215 stand for interlayer spin - orbit parameters.@xcite then the effective magnetic field of the soc @xmath216 in eq . ( [ eq6 ] ) is given by @xmath217 , @xmath218 and @xmath219 . it is noted that the contribution of other energy bands to the soc of the lowest conduction band is marginal in our calculation . and ( b ) @xmath220 as function of the out - of - plane electric field @xmath36 with @xmath192-@xmath221 . it is noted that @xmath222 , @xmath223 , @xmath224 and @xmath225 ( @xmath226 ) are dimensionless.,title="fig:",width=283 ] and ( b ) @xmath220 as function of the out - of - plane electric field @xmath36 with @xmath192-@xmath221 . it is noted that @xmath222 , @xmath223 , @xmath224 and @xmath225 ( @xmath226 ) are dimensionless.,title="fig:",width=283 ] as reported by konschuh _ et al._,@xcite the small group of the dirac points is @xmath4 in the presence of an out - of - plane electric field . following from the symmetry group @xmath4,@xcite we derive an analytical form of the soc near the dirac points [ see eqs . ( [ eq1]-[eq3 ] ) ] . in these equations , the coefficients @xmath13 ( @xmath14-@xmath15 ) and @xmath227 ( @xmath14-@xmath110 ) are obtained by fitting to the numerical calculation from eqs . ( [ a5 ] ) and ( [ a7 ] ) using the pad approximation.@xcite specifically , the coefficients @xmath13 ( @xmath14-@xmath15 ) in eqs . ( [ eq1]-[eq2 ] ) read @xmath228 the coefficients @xmath227 ( @xmath14-@xmath110 ) in eq . ( [ eq3 ] ) are given by @xmath229 here , the coefficient @xmath230ev , which is independent of the applied electric field . the electric field dependences of @xmath231 ( @xmath14-@xmath15 , @xmath192-@xmath92 ) and @xmath232 ( @xmath14-@xmath110 , @xmath192-@xmath221 ) are plotted in figs . [ fig7 ] and [ fig8 ] , respectively . it is noted that the pad approximation@xcite can give a precise description of @xmath13 ( @xmath14-@xmath15 ) and @xmath227 ( @xmath14-@xmath110 ) at large momenta near the dirac points , which can be populated and play an important role in spin relaxation according to the experimental conditions ( heavily doped).@xcite with only a simple linear @xmath10-order approximation , @xmath13 and @xmath227 agree with the numerical results only at very small momentum near the dirac points . we also show in fig . [ fig9 ] the spin splitting of the lowest conduction band calculated with analytical form of the soc ( blue dashed curve ) and with the explicit numerical one ( red solid curve ) . we find that the analytical result agrees fairly well with the numerical one at large momentum . in addition , the spin splitting decreases with increasing momentum at large momentum , which is very different from the case in both semiconductors@xcite and single - layer graphene.@xcite near the k point as function of the momentum @xmath233 under a typical electric field @xmath104v / nm . solid ( dashed ) curve represents the numerical ( analytical ) result . the electron - electron coulomb scattering matrix element is given by @xmath235 . the screened coulomb potential @xmath236 with @xmath237 being the screening under the random phase approximation.@xcite @xmath238 is the two - dimensional bare coulomb potential with @xmath239 and @xmath240 being the fermi velocity in single - layer graphene@xcite and the dimensionless wigner - seitz radius,@xcite respectively . @xmath241 is given by@xcite @xmath242 where @xmath243 . the long - range electron - impurity scattering matrix element @xmath244 , in which @xmath245 and @xmath246 stand for the impurity charge number and effective distance of the impurity layer to the blg sheet.@xcite it is noted that when we calculate the electron - electron coulomb and long - range electron - impurity scatterings , we take the distance between the graphene layers to be zero approximately.@xcite it is further noted that the coulomb potential in the hf term is in the static screening limit , i.e. , @xmath247 . then we turn to the electron - phonon scattering matrix elements . for the intravalley electron - ac - phonon scattering , the scattering matrix element @xmath248 where @xmath249 ; @xmath250 and @xmath251 represent the deformation potential@xcite and acoustic phonon velocity,@xcite respectively . for the intravalley electron - ri - phonon scattering , @xmath252 with @xmath253 being the thomas - fermi screening length.@xcite for sio@xmath0 substrate , the energy spectra of the two remote phonon modes are denoted as @xmath254 and @xmath255 , respectively ; @xmath256 and @xmath257 are the corresponding dimensionless coupling parameters.@xcite the matrix elements for the intravalley electron - op - phonon scattering ( @xmath66 ) are derived using the tight - binding model according to the arxiv version of ref . . the intravalley electron - op - phonon scattering includes both the electron in - plane ( @xmath258 ) and out - of - plane ( @xmath259 ) op phonon scatterings . the electron in - plane op phonon scattering matrix element is given by @xmath260\psi^{\mu}_{{\bf k}^{\prime}}|^2+|{\psi^{\mu}_{{\bf k}}}^{\dagger } [ i\gamma_0^{\prime}(\sin\theta_{\bf q}\sigma^{01}_{\rm d}-\cos\theta_{\bf q}\nonumber\\ & & \mbox{}\hspace{-1.9cm}\times\sigma^{02}_{\rm d})+a\gamma_4^{\prime}(i\cos\theta_{\bf q}\gamma^3_{\rm d}- \sin\theta_{\bf q}\gamma^3_{\rm d}\gamma^5_{\rm d})/(\sqrt{3}l_4)]\psi^{\mu}_{{\bf k}^{\prime}}|^2\nonumber\\ & & \mbox{}\hspace{-1.9cm}+|{\psi^{\mu}_{{\bf k}}}^{\dagger}[\gamma_0^{\prime}(\cos\theta_{\bf q}\sigma^{23}_{\rm d } -\sin\theta_{\bf q}\sigma^{13}_{\rm d})-a\gamma_3^{\prime}(\cos\theta_{\bf q}\gamma^1_{\rm d}\gamma^5_{\rm d}\nonumber\\ & & \mbox{}\hspace{-1.9cm}-i\cos\theta_{\bf q}\gamma^2_{\rm d}-i\sin\theta_{\bf q}\gamma^1_{\rm d}-\sin\theta_{\bf q}\gamma^2_{\rm d } \gamma^5_{\rm d})/(2\sqrt{3}l_3)]\psi^{\mu}_{{\bf k}^{\prime}}|^2\nonumber\\ & & \mbox{}\hspace{-1.9cm}+|{\psi^{\mu}_{{\bf k}}}^{\dagger}[i\gamma_0^{\prime}(\cos\theta_{\bf q}\sigma^{01}_{\rm d } + \sin\theta_{\bf q}\sigma^{02}_{\rm d})-a\gamma_4^{\prime}\gamma^3_{\rm d}(i\sin\theta_{\bf q}\nonumber\\ & & \mbox{}\hspace{-1.9cm}+\cos\theta_{\bf q}\gamma^5_{\rm d})/(\sqrt{3}l_4)]\psi^{\mu}_{{\bf k}^{\prime}}|^2\big\},\end{aligned}\ ] ] where @xmath261 is the energy of the in - plane op phonon modes;@xcite @xmath262 is the bond length corresponding to the interlayer hopping @xmath31 ; and @xmath263 and @xmath264 are @xmath20 dirac matrices given in appendix [ appc].@xcite here , @xmath265 is the polar angle of the momentum @xmath266 . the electron out - of - plane op phonon scattering matrix element reads @xmath267 being the energy of the out - of - plane op phonon mode.@xcite the relevant dirac matrices@xcite used in the electron - phonon scattering are @xmath268 @xmath269 @xmath270 with @xmath271 being @xmath198 unit matrix . yang , j. balakrishnan , f. volmer , a. avsar , m. jaiswal , j. samm , s. r. ali , a. pachoud , m. zeng , m. popinciuc , g. gntherodt , b. beschoten , and b. zyilmaz , phys . lett . * 107 * , 047206 ( 2011 ) . it is noted that due to the complicated energy spectrum of blg , we employ the single - layer graphene approximation to calculate the spin diffusion coefficient in blg according to the report by zhang and wu.@xcite

we study the x ray spectral variability of the narrow line seyfert 1 galaxy ngc 4051 as observed during two _ xmm newton _ observations . to gain insight on the general behaviour , we first apply model independent techniques such as rms spectra and flux - flux plots . we then perform time resolved spectral analysis by splitting the observations into 68 spectra ( 2 ks each ) . the data show evidence for a neutral and constant reflection component and for constant emission from photoionized gas , which are included in all spectral models . the nuclear emission can be modelled both in terms of a `` standard model '' ( pivoting power law plus a black body component for the soft excess ) and of a two component one ( power law plus ionized reflection from the accretion disc ) . both models reproduce the source spectral variability and can not be distinguished on a statistical ground . the distinction has thus to be made on a physical basis . the standard model results indicate that the soft excess does not follow the standard black body law ( l ) despite a variation in luminosity by about one order of magnitude . the resulting temperature is consistent with being constant and has the same value as observed in pg quasars . moreover , although the spectral slope is correlated with flux , which is consistent with spectral pivoting , the hardest photon indexes are so flat (.31.4 ) as to require rather unusual scenarios . furthermore , the very low flux states exhibit an invertedflux behaviour which disagrees with a simple pivoting interpretation . these problems can be solved in terms of the two component model in which the soft excess is not thermal , but due to the ionized reflection component . in this context , the power law has constant slope ( about 2.2 ) and the slope flux correlation is explained in terms of the relative contribution of the power law and reflection components which also explains the shape of the flux flux plot relationship . the variability of the reflection component from the inner disc closely follows the predictions of the light bending model , suggesting that most of the primary nuclear emission is produced in the very innermost regions , only a few gravitational radii from the central black hole . [ firstpage ] galaxies : individual : ngc 4051 galaxies : active galaxies : seyfert x - rays : galaxies

astro-ph0602191

ngc 4051 is a nearby ( z=0.0023 ) low luminosity narrow line seyfert 1 ( nls1 ) galaxy which exhibits extreme x ray variability in flux and spectral shape on both long and short timescales . the source sometimes enters relatively long and unusual low flux states in which the hard spectrum becomes very flat ( @xmath3 ) while the soft band is dominated by a much steeper component ( @xmath4 , or blackbody with @xmath5 kev ) . most remarkable is the 1998 _ bepposax _ observation reported by guainazzi et al ( 1998 ) in which the source reached its minimum historical flux state , the spectral slope was @xmath6 , and the overall spectrum was best explained by assuming that the nuclear emission had switched off leaving only a reflection component from distant material ( clearly visible in the hard spectrum ) which still responded to previous higher flux levels . a different interpretation assumes that the nuclear emission in low flux states originates so close to the central black hole that only a tiny fraction can escape the gravitational pull of the hole so that the nuclear continuum is virtually undetectable at infinity ( e.g. miniutti & fabian 2004 ) . in general , the 210 kev spectral slope appears to be well correlated with flux . however , the correlation is not linear with the slope hardening rapidly at low fluxes and reaching an asymptotic value at high fluxes ( see e.g. lamer et al 2003 , based on _ rxte _ monitoring data over about three years ) . this behaviour is relatively common in other sources as well , such as mcg6 - 30 - 15 , 1h 0419577 , ngc 3516 ( among many others ) and might be due to flux correlated variations of the power law slope produced in a corona above an accretion disc as originally proposed by haardt & maraschi ( 1991 ; 1993 ) and haardt et al . ( 1994 ) and related to changes in the input soft seed photons ( e.g. maraschi & haardt 1997 , haardt , maraschi & ghisellini 1997 , poutanen & fabian 1999 , zdziarski et al 2003 ) . on the other hand , such slope flux behaviour can be explained in terms of a two component model ( mchardy et al 1998 ; shih , iwasawa & fabian 2002 ) in which a constant slope power law varies in normalisation only , while a harder component remains approximately constant hardening the spectral slope at low flux levels only , when it becomes prominent in the hard band . in the well studied case of mcg6 - 30 - 15 , detailed spectral and variability analysis ( e.g. fabian & vaughan 2003 ; vaughan & fabian 2004 ) allowed the presence of two main spectral components to be disentangled , namely a variable power law component ( plc ) with constant slope , and a reflection dominated component ( rdc ) from the inner accretion disc which does not follow the plc variation , confirming for this particular case the validity of the two component model and ruling out spectral pivoting as the main source of spectral variability . the contribution from a distant reflector is limited to 10 per cent or less ( see e.g. the long _ chandra _ observation reported by young et al 2005 ) i.e. the rdc is completely dominated by lowly ionized reflection from the inner disc , including the broad relativistic fe line . from a theoretical point of view , the lack of response of the disc rdc to the plc variations has been interpreted in terms of strong gravitational light bending by miniutti et al ( 2003 ) and miniutti & fabian ( 2004 ) , which predicts this behaviour during normal flux periods and gives a correlation at low flux levels only . it should be stressed that , in the framework of the two component model , the 210 kev slope flux correlation in ngc 4051 can not entirely be explained by the contribution of a distant reflector in the hard band ( the presence of which is required by the detection of a narrow component to the iron line ) . indeed , as shown by lamer et al ( 2003 ) the correlation is still present even if the distant reflector contribution is taken into account . even at low flux levels , soft and hard components are significantly variable and well correlated , which excludes any dominant extended emission ( torus and/or extended scattering region ) . extended emission is undetected by _ chandra _ on @xmath7 100 pc scales ( collinge et al 2001 ; uttley et al 2003 ) . therefore , the residual slope flux correlation can be due to intrinsic pivoting or to an additional , almost constant , spectral component ( i.e. rdc from the disc , as for mcg6 - 30 - 15 ) . throughout the paper for `` two component model '' we mean a parametrization of the nuclear emission in terms of two components . other spectral components such as absorption , emission from photoionized gas , and reflection from distant matter can be present but do not represent nuclear emission . other important clues on the nature of the spectral variability in ngc 4051 ( and other sources as well ) comes from the so called flux flux plot analysis , first presented by churazov , gilfanov & revnivtsev ( 2001 ) to demonstrate the stability of the disc emission in the high / soft state of cyg taylor , uttley & mchardy ( 2003 ) have applied the same technique to a few seyfert galaxies by using _ rxte _ data . in the case of ngc 4051 , uttley et al ( 2004 ) have performed such an analysis by using the same _ xmm newton _ data we are presenting here and showed that the distribution of high , intermediate , and low flux state data points are smoothly joined together , indicating that the same process causing the spectral variability in high and intermediate flux states continues to operate even at the very low flux levels probed by the _ xmm newton _ data ( uttley et al 2004 ) . this evidence seems to challenge the interpretation of the spectral variability as due to variable absorption by a substantial column of photoionized gas ( pounds et al . 2004 ) . here we report results based on the two _ xmm newton _ observations of ngc 4051 . the data have been previously analysed by uttley et al ( 2004 ) and pounds et al ( 2004 ) who reach different conclusions on the nature of the spectral variability in ngc 4051 and on the main spectral components that come into play . our analysis is complementary to previous studies and offers a novel point of view that may also be relevant for other similar sources . ngc 4051 has been observed twice by _ xmm newton_. the first observation took place during revolution 263 ( 2001 may ) for @xmath7117 ks , the second during revolution 541 ( 2002 november ) for @xmath752 ks . these data are public and have been obtained from the data archive . the analysis has been made with the latest version of the sas software , starting from the odf files . during the first observation the epic pn and the mos2 were operated in small window mode with the medium filter . no significant pile - up is present in the pn data , so single and double pixel events were selected . the mos2 data show @xmath71% of pile - up when the count rate is higher than 5 counts / s ( during the first observation it sometimes reaches 15 counts / s ) . this could invalidate the results of time resolved spectral variability , because , even if pile - up is not strong , its effect is increasing with flux . thus , for the time resolved spectral analysis , we use the mos2 data selecting only single events and extracting the source photons within a circular region of 30@xmath8 radius when the count rate is lower than 5 counts / s , and within an annular region with 7@xmath8 and 30@xmath8 radii , during the higher count rate intervals . the mos1 data were in fast uncompressed mode and , due to cross calibration uncertainty , we did not used those . the second observation was timed to coincide with an extended period of low x ray emission from . during this observation the pile - up is negligible in all the epic cameras and single plus double pixel events are used for mos1 , mos2 and pn cameras . the background spectrum has been taken from source free regions in the same chip as the source and ancillary and response files have been created . after filtering out periods of high background , we obtain net exposures of about 97 ks ( @xmath769 ks for the pn small window mode ) and 39 ks for the first and second observation respectively . at the moment the pn camera is not well calibrated below 0.5 kev . for the data we are presenting here , for example , the ratio between the mos and pn camera below 0.5 kev is of about 20 per cent . thus , the data from the pn camera are used in the 0.510 kev band only , while mos data are used down to 0.2 kev . all spectral fits include absorption due to the line - of - sight galactic column of @xmath9=1.32@xmath1010@xmath11 @xmath12 ( elvis et al . in the subsequent spectral analysis , errors are quoted at the 90 per cent confidence level ( = 2.7 for one interesting parameter ) . the _ xmm newton _ broadband light curve of ngc 4051 exhibits large amplitude count rate variations in both observations , typical for this source and for nls1 galaxies in general . fast large amplitude variability is superimposed on longer trends which are mainly characterized by persistent low flux periods in which variability is suppressed . the _ xmm newton _ light curves shown in fig . [ lcs ] are a good representation of this behaviour which can be described in more general terms by the rms flux relation in which the absolute rms variability is proportional to the source flux ( uttley & m@xmath13hardy 2001 ) . light curves such as those shown in fig . [ lcs ] are just a realization of the rms flux relation which naturally produces large amplitude spikes at high fluxes and periods of relative quiescence at low fluxes ( e.g. uttley , m@xmath13hardy & vaughan 2005 ) . the general shape of the broadband spectrum of ngc 4051 in both observations can be roughly described by the presence of a power law continuum , reflection from distant matter including a narrow 6.4 kev fe line , and a prominent soft excess below about 1 kev . the fe line is unresolved and fluxes are consistent with each other in the two observations . a warm absorber imprints its presence in the spectrum particularly through two edge like features at 0.74 and 0.87 kev ( o and o ) . even modelling the absorption , some structures persist between 0.8 kev and 1 kev , particularly evident in the low flux state . this further structure can be interpreted as a broad absorption structure ( as suggested by pounds et al 2004 ) , emission around 0.9 kev ( see collinge et al 2001 , uttley et al 2004 ) , or a combination of the two . spectral variability is clearly present between the two observations with the source being much harder in the second , lower flux observation . moreover , when the source is faint and hard , the spectrum above @xmath7 2 kev appears to be more curved than in the high state . to illustrate the differences in spectral slope and hard curvature between the two observations we show the two 310 kev time averaged spectra in the top panel of fig . [ test ] . the best fitting power law slope is indicated , the fit including a narrow 6.4 kev fe line as well . the spectral curvature can be interpreted as the effect of absorption by a relatively high column of ionized gas or could be the signature of a relativistically blurred reflection component from the accretion disc . two main mechanisms have been proposed so far to explain the spectral variability of ngc 4051 . pounds et al ( 2004 ) performed a comparative analysis of the two _ xmm newton _ observations based on the time averaged spectra and on the simultaneous high resolution rgs data in the soft band . their main conclusions is that the spectral variability between the high flux ( rev . 263 ) and low flux ( rev . 541 ) observations is mainly due to variable absorption by a substantial column of photoionized gas in the line of sight . the authors suggest that in the low state ( observed about 20 days after the source entered one of its characteristic low state periods ) part of the gas had recombined in response to the extended period of lower x ray flux so providing a partial coverer ( with @xmath14 @xmath12 ) . the variable opacity of the absorber could then explain in a natural way both the flat spectrum and the hard curvature that is observed in the second low flux observation . this interpretation is challenged by the analysis by uttley et al ( 2004 ) who proposed spectral pivoting as the main driver of the spectral variability in ngc 4051 and pointed out that a variation in the absorber opacity in the line of sight during the low state is not consistent with the nature of the spectral variability , as revealed by flux flux plot analysis , in which no sharp transition is seen between the two observations . the flux flux relationship rather indicates that the spectral variability is flux dependent and the low flux states of the first observation are virtually identical in spectral shape to the high flux state of the second . this would rule out that the low flux spectral shape during the second observation depends on the fact that data were obtained following an extended low flux state allowing the absorbing gas to recombine . it is clear that the pounds et al interpretation implies that the spectral shape is different in the two observations . for instance , the low flux states of the first ( high flux ) observation should not be as flat and curved as the spectrum in the second observation . this idea can be easily checked by comparing two spectra : the first one is representative of the very low flux state during the first observation ( around 60 ks , see fig . [ lcs ] ) and the second one is extracted from the second observation . we choose the two spectra to have the same exposure and approximately the same count rate . the two spectra are shown in the bottom panel of fig . [ test ] in the hard band . the best fitting power law slope is also indicated . the fit also includes fe emission at 6.4 kev . firstly , the spectral slope is exactly the same in the two spectra . therefore the slope is count rate dependent and the time averaged flat spectrum during the low flux observation has thus little to do with the extended period of low flux preceding the second observation . secondly , the two spectra also seem to have approximately the same hard curvature . when fitted with a partial covering model , both spectra require a large column density of @xmath15@xmath16 @xmath12 and a covering fraction of 6070 per cent . this demonstrates that not only the apparent spectral slope is flux dependent , but also that the hard curvature depends on flux only . from the above analysis , it seems likely that the spectral variability in ngc 4051 is simply flux dependent and that the second low flux observation is not peculiar but indeed identical to low flux states during the first observation . as already mentioned by uttley et al ( 2004 ) one possible explanation for the flux dependent spectral variability is spectral pivoting . in this paper we shall also explore an alternative explanation invoking a reflection component from the accretion disc . before considering time resolved spectral analysis , we perform some model independent tests with the aim of identifying the main spectral components and their relative contribution to the spectrum and variability , in order to have a guide for our subsequent analysis . we start our analysis of the spectral variability of by calculating the rms spectrum , which measures the total amount of variability as a function of energy ( edelson et al . 2002 ; vaughan et al . 2004 ; ponti et al . [ rms ] shows the rms spectra during rev . 263 and 541 . the rms spectra have been calculated with time bins of @xmath72 ks ( the same as the subsequent time resolved spectral analysis ) and with energy bins chosen in order to have at least 300 counts per energy time bin . in this way the poissonian noise is negligible compared to the rms value and thus the use of the uncertainty given in ponti et al . ( 2004 ) is valid . during the high flux observation of rev . 263 ( fig . [ rms ] ) the variability rapidly increases toward softer energies , i.e. the broadband emission tends to steepen as it brightens . some reduction in the variability is seen below about 800 ev followed by a plateau below 500 ev . a similar rms shape has been observed in other sources ( ponti et al . 2005 ; vaughan et al . 2003 ; vaughan et al . 2004 ) and the two simplest explanations for the broadband trend invoke either spectral pivoting of the variable component , or the two component model ( see e.g markowitz , edelson & vaughan 2003 ) . another important feature of the rms spectrum is a drop of variability at 6.4 kev , the energy of the narrow fe line which is seen in the time averaged spectrum . such a drop shows that the narrow fe line ( and therefore the associated reflection continuum ) is much less variable than the continuum ( if variable at all ) indicating an origin in distant material . some structure is also present around 0.9 kev , with the possible presence of either a drop at that energy or two peaks . in the bottom panel of the same figure , we show the rms spectrum obtained for the low flux observation during rev 541 . the variability below 3 kev is strongly suppressed with respect to the high flux observation . moreover , the shape of the rms spectrum is totally different and rather unusual . the trend of increasing variability toward softer energies breaks down completely and the most striking feature is the marked drop of variability around 0.9 kev . this feature , as well as the other drop of variability at @xmath70.55 kev , could have the same nature as the structure seen in the high flux rms spectrum , but is much more significant and prominent here . as a sanity check , we show in fig . [ lcsoft ] the source light curve around 0.9 kev and in two bands below and above for comparison . the source appears to be constant around 0.9 kev ( @xmath17=144.2 for 140 degrees of freedom ) . the constant hypothesis is unacceptable fits in the other two bands . as in the high flux observation , the variability is also suppressed around 6.4 kev . the drop at 67 kev is here even more dramatic due to the reduced continuum level and therefore increased visibility of the constant narrow fe line . the rms spectrum analysis given above indicate the presence of two constant or weakly variable components . the first is related to x ray reflection and fe emission and is seen in both observations . the second is associated with the drop in variability around 0.9 kev which is clearer in the low flux data where the contribution of the continuum is lower . [ rms ] shows a clear minimum at 6.4 kev indicating that the narrow fe emission line is less variable than the continuum in both observations . the fe line is clearly detected in the time averaged spectra of both observations . in the low flux one , the line is unresolved , has an energy of @xmath18 kev , a flux of @xmath1910@xmath20 ph s@xmath21 @xmath12 , and an equivalent width of about 260 ev . the line energy and flux are consistent with measurements obtained in the high flux observation and also with previous data from _ chandra _ ( collinge et al 2001 ) and _ bepposax _ ( guainazzi et al 1998 ) . as mentioned , a constant and narrow fe emission line is also required by the rms spectra . since the fe line must be associated with a reflection continuum , in all subsequent fits we always include a constant and neutral reflection model from magdziarz & zdziarski ( 1995 ) and a narrow fe line . the reflection continuum normalization is chosen such that the fe line equivalent width ( with respect to such continuum only ) is consistent with the _ bepposax _ observation ( ew@xmath22700 ev ) by guainazzi et al ( 1998 ) . this is because that observation caught ngc 4051 in a particularly low flux state in which the primary continuum had switched off leaving only reflection from distant matter in the hard band , thereby providing a most useful way to constrain that component ( under the reasonable assumption that it remains constant ) . the second constant component is clearer in the low flux state rms spectrum ( bottom panel of fig . [ rms ] ) and appears as a deep minimum , that reach almost null variability , at around 0.9 kev and with a lower variability at @xmath70.55 kev ( see also fig . [ lcsoft ] ) . these prominent drops indicate the presence of constant emission around @xmath70.9 and @xmath70.5 kev , as the drop at 6.4 kev is easily interpreted as constant fe emission . moreover , an emission like feature is seen in the pn spectrum of the low flux observation and in previous _ low flux spectra as well ( collinge et al 2001 ; uttley et al 2003 ) . the energy of the feature in the pn spectrum is @xmath7 0.9 kev , consistent with ne ( and possibly fe emission plus o recombination continuum , or rrc ) . a second feature is also detected around 0.50.6 kev , where emission from o and o is expected . as an illustration , in the top panel of fig.[r12 ] , we show the ratio of the low flux pn and mos data to a simple continuum model comprising galactic absorption power law and blackbody in the relevant energy band . the energies of the emission features and their intensities over the continuum correspond with the two drops seen in the rms spectrum . the larger strength of these features during rev . 541 is due to the lower continuum . in fact , the same drop of variability at about 0.9 kev observed during the rev . 263 can be reproduced by the same constant emission lines with a higher continuum . to understand the nature of these constant emission components , we then inspected the simultaneous high resolution rgs data to better search for possible narrow emission lines that could explain the features seen at ccd resolution . as already shown by pounds et al ( 2004 ) the high resolution rgs data during the low flux _ xmm newton _ observation indeed show an emission line spectrum which is very similar to that of typical seyfert 2 galaxies ( e.g.kinkhabwala et al 2002 ; bianchi et al 2005 ) . we take the same continuum as in the ccd data and add a o edge at 0.74 kev to account for the absorption seen there . the most prominent emission lines are due to the o triplet ( dominated by the forbidden and intercombination lines and with no clear resonance ) , the o k@xmath23 line and the n k@xmath23 line ( see fig . [ rgs ] ) . the ne ( forbidden ) line is also clearly detected at 0.905 kev with a flux of @xmath24 ph s@xmath21 @xmath12 , not enough to account for the pn feature around 0.9 kev . however , the ne line sits on top of a broad feature most likely due to o rrc and possibly unresolved fe emission lines . when fitted with a crude gaussian model in the rgs data , such a feature has an energy of @xmath25 kev and a flux of @xmath26 ph s@xmath21 @xmath12 . by combining the ne line with this broad feature ( @xmath27 ev ) we obtain a flux of ( 0.7 - 1.1)@xmath28 ph s@xmath21 @xmath12 around 0.9 kev . we then included ( with all parameters fixed to the rgs results ) the n and o lines in the mos / pn model , together with a o edge at 0.74 kev ( fixed energy ) and a gaussian emission line with free energy , width , and normalization to account for the blend of ne line plus broad feature around 0.9 kev . the resulting fit is acceptable and is shown as a ratio plot in the bottom panel of fig . as for the emission feature at 0.9 kev , we measure an energy of @xmath29 kev and a flux of @xmath30 ph s@xmath21 @xmath12 , consistent with the rgs upper limit . as mentioned in pounds et al ( 2004 ) excess absorption might be present around 0.76 kev ( possibly related to a m shell unresolved transition array from fe ) . if included in both ccd and rgs data , this has the effect of reducing the broad 0.9 kev feature intensity , making so much more difficult to reproduce the variability drop in the rms spectra . the above analysis demonstrates that emission from photoionized gas is present in the low flux data and detected both in the rgs and in the lower resolution ccd detectors . such components , as shown by the rms spectra , are constant over time , thus then are likely to come from relatively extended gas ( see pounds et al 2004 ) . a constant emission component , indeed , does explain the unusual drop of variability around 0.9 kev in the low flux rms spectrum . the line intensities are also consistent ( though not detected ) with the higher flux observation data . they are not detected because of the much higher continuum level . we then included , in addition to the constant neutral reflection from distant material , such lines with fixed energies and intensities ( as obtained from the low flux rgs data ) in all subsequent fits . another model independent approach that provides information on the dominant driver of the spectral variations is the flux flux technique ( see taylor , uttley & m@xmath13hardy 2003 ; uttley et al 2004 ) . we recall here that a linear flux flux relationship reveals that the data can be described by a simple two component model where one component varies in flux but not in spectral shape , while the other remains constant both in flux and spectral shape . an offset on the hard or soft axis allows one to infer in which of the two bands the constant component contributes the most . on the other hand , if the flux flux relationship has a power law functional form , then the data are consistent with spectral pivoting of the variable component . uttley et al ( 2004 ) plotted the 210 kev band count rate versus the 0.10.5 kev one and found a very accurate power law flux flux relationship which was interpreted as evidence for spectral pivoting of the variable component . the choice of the soft and hard energy ranges by uttley et al ( 2004 ) is driven by the requirement of having well separated bands to best reveal the spectral variability . however the 0.10.5 kev band does not seem to have the same variability properties of the broadband continuum . this is clearly seen in the top panel of fig . [ rms ] in which the soft plateau indicates the presence of a component with different variability properties than the continuum above 1 kev . such a component is likely to be that producing the soft excess . therefore we consider as the reference soft band the 11.4 kev range to avoid possible contamination from undesired components . in all previously reported analysis of the x ray spectrum of ngc 4051 , the intermediate 11.4 kev band is dominated by a power law continuum which is not strongly affected by absorption , soft excess , or reflection . as for the hard band , we consider the 4 - 10 kev range , simply to allow enough leverage for spectral variability to show up clearly . in the top panel of fig . [ ff1 ] , we show the flux flux relationship in the chosen hard and soft bands . the two _ xmm newton _ observations are shown with different symbols . as already pointed out by uttley et al ( 2004 ) the low flux state data join gently on to the normal / high flux state ones , showing that the spectral variability is likely to be due to the same mechanism at all flux levels . there is no dramatic discontinuity between the two observations suggesting , once again , that the spectral changes are not due to a dramatic change in one of the physical parameters ( such a substantial change in the absorbing column , as suggested by pounds et al 2004 ) . the flux flux relation is clearly non linear with some curvature showing up especially at low fluxes . in the bottom panel , we show a binned version of the same plot ( minimum of 10 points per bin ) . in the same figure , the solid line is the best fit for a power law model representing a situation in which the main driver of the spectral variability is spectral pivoting ( see e.g. taylor , uttley & m@xmath13hardy 2003 for details ) . the fit is performed as in uttley et al ( 2004 ) with a power law relationship including constant components which are required in both the hard and the soft band . the power law relationship does not provide a good fit to the data ( @xmath31 for 22 degrees of freedom ) , mainly because the model has too much curvature at high fluxes and not enough at low fluxes . this does not exclude spectral pivoting as the main driver for the spectral variability , but suggests that we need to look for an alternative explanation . in fact , if the lowest flux data points are excluded , the flux flux relationship is linear with high accuracy . this is shown as a dotted line in the bottom panel of fig . [ ff1 ] which is the best fit linear relationship obtained when the lowest flux data points are ignored . we obtain a very good fit ( @xmath32 for 15 degrees of freedom ) with large hard offset of @xmath33 . as mentioned , a linear relationship suggests that the spectral variability is due to a two component model , i.e. to the relative contribution of a variable component whose spectral shape is flux independent ( e.g. a constant @xmath2 power law varying in normalization only ) and a constant or weakly variable harder component . if the flux flux relationship at normal / high fluxes is indeed linear , one can compute the contribution of the constant component in the 410 kev from the y axis offset @xmath34 . since the mean 4 - 10 kev rate for the normal / high flux data points is @xmath35 , this means that the putative constant component contribution in the 410 kev band is as large as 50 per cent in the normal / high flux states . however , if indeed the two component model is a fair representation of the spectral variability at normal / high fluxes , the low flux data imply that the spectral variability must smoothly change regime at low fluxes producing the gentle low flux bending in the relationship . such a transition could be reproduced if , below a given flux level , the constant component ( or part of it ) is not constant anymore but rather follows the changes of the variable one . in fact , any two component model in which the less variable component varies together with the variable one at low fluxes and saturates to an approximately constant flux at higher fluxes would produce the same kind of relationship as observed here . in order to check if the source spectral variability could be due to the two component model we try to reproduce the rms spectrum during rev . 263 with a simple phenomenological parametrization ( the rms spectrum during rev . 541 is not considered because its shape is mainly dictated by the constant components coming from distant material ) . the model is composed of a direct power law component ( plc ) and an ionized reflection dominated component ( rdc , from ross & fabian 2005 ) the spectrum of which is convolved with a kernel with fixed inner and outer disc radius ( to @xmath36 and @xmath37 respectively ) , and fixed disc inclination ( @xmath38 ) . in the rms simulation the contribution from neutral distant reflection is also considered . we assume that all the source variability is due to the variation of the normalisation of the direct plc that has a spectral index fixed at @xmath2=2.4 ( lamer et al . 2003 ) and that the reflection component is completely constant not only in normalisation , but also in spectral shape ( i.e. ionisation and relativistic blurring parameters ) . to simulate the rms we vary the plc normalization from the lowest to the highest value observed . for the rdc we have chosen the best fit values of the spectrum during the low flux state of rev . 263 ( @xmath3950 and disc emissivity index of @xmath406 ) . figure 8 shows this simulated rms ( dotted line , green in the colour version ) and , for comparison , the simulated rms with the reflection parameters of @xmath41=5 and @xmath23=4 ( solid line , blue in the colour version ) , that also roughly reproduce the low flux spectrum . the interplay between the plc and the rdc can roughly reproduce the shape of the rms ( within @xmath420.02 ) . this means that the two component model can explain the bulk of the spectral variability requiring no further variable component at low energy . the variability in the soft band can be explained by a variable power law and a constant soft excess ( which is in this case due to constant rdc ) clearly , a completely constant reflection component is an over approximation . in fact , although , it is possible to reproduce the bulk of the variability , some variations of the normalization of the rdc are required in order to match with more details the rms spectra and to reproduce the constant spectral shape below 0.5 kev ( see section 6.3.2 ) . in sources such as ngc 4051 where large amplitude fast variability is associated with spectral variability , time resolved or flux resolved spectroscopy provides much more detailed information on the nature of the source than time averaged spectroscopy . in the particular case of ngc 4051 , flux resolved spectroscopy could be a risky procedure to adopt because of the small central black hole mass ( and therefore the short dynamical timescale associated with it ) . as an example , if we were to consider the very high flux state of the first _ xmm newton _ observation ( see left panel of fig . [ lcs ] ) and make a selection say above 40 counts / s , we would accumulate spectra from intervals sometimes spaced by 70 ks . this timescale is about 140 times the dynamical timescale at 10 @xmath43 ( where @xmath44 ) for a black hole mass of @xmath45 ( e.g. shemmer et al 2003 ) which is , in our opinion , too large a factor to extract any truly significant information on the nature of the spectral variability of the source . we therefore explore time resolved spectroscopy on the shortest possible timescale ( set by requiring good quality time resolved individual spectra ) . in the following , we present results obtained by performing such an analysis on a 2 ks timescale ( about four dynamical timescales at 10 @xmath43 ) . the 2 ks slices that have been used in the analysis are shown in fig . [ lcs ] and have been numbered for reference . it is well known that the 210 kev spectral slope in ngc 4051 ( and other sources ) is correlated with the source flux . in the top panel of fig . [ gf1 ] , we show such a correlation when all the 2 ks spectra are fitted with a simple power law model in the 210 kev band . data for the two observations are shown with different symbols . the photon index increases with flux and seems to saturate around @xmath46 with a @xmath47 between the low and high flux states . however , the power law model is clearly inadequate to fit the 210 kev spectra in ngc 4051 . this is because of the presence of a narrow 6.4 kev fe k@xmath23 line and the associated reflection continuum in the hard band . as shown above , such a component is constant and likely originates in some distant material ( such as the torus ) . we then fit again the 2 ks data by including the constant reflection component and show in the bottom panel of fig . [ gf1 ] the resulting @xmath2flux relationship . the @xmath48 between the low and high flux states is now reduced to about 0.70.8 . again the photon index seems to saturate at high fluxes to @xmath46 . this value is slightly different from the one ( @xmath49 ) observed during a long rxte observation campaign ( lamer et al . the difference could be due to long term spectral variability . moreover , the photon index seems to saturate not only at high fluxes but also at low fluxes ( to @xmath50@xmath51 ) . since the two asymptotes are not extremely well defined , such behaviour could still be consistent with spectral pivoting , even though it should be stressed that the low flux photon index is still somewhat hard to be explained in terms of standard comptonization models and would suggest some peculiar physical scenario ( such as a photon starved corona ) . on the other hand , the @xmath2flux relationship seen in the bottom panel of fig . [ gf1 ] can be explained if an additional and weakly variable component is present in the 210 kev band . if so , such a component dominates the low flux states and the hard photon index measured there is just a measure of its intrinsic spectral shape in the 210 kev band , while it is overwhelmed by the power law at high fluxes where @xmath46 is then the intrinsic power law photon index . in other words , as an alternative to spectral pivoting , the @xmath2flux relationship can be explained by assuming the presence of i ) a constant reflection component from distant matter ; ii ) a variable power law with constant or weakly variable slope @xmath46 ; iii ) an additional component with approximate spectral shape of @xmath52@xmath51 in the 210 kev band which must be almost constant at normal / high fluxes and is allowed to vary significantly only at low fluxes . we point out here that our flux flux plot analysis showed that the two component model might indeed be appropriate in the normal / high flux states . if so , the constant component contributes by about 50 per cent in the 410 kev band ( from the y axis offset in the linear fit ) and this is much too large a fraction to be accounted for by constant reflection from distant material ( and associated narrow fe line ) which contributes in that band by only 10 per cent . the two component model interpretation of the flux flux relationship at normal / high fluxes implies that an additional constant component is present in the normal / high flux states and contributes by about 40 per cent to the 410 kev band . such a component naturally produces the observed @xmath2flux relationship because as the flux drops the spectrum becomes more and more dominated by it . as a first step , we consider a simple continuum model comprising a power law plus black body ( bb ) emission to model the prominent soft excess . this is , in many respects , the `` standard model '' to fit agn x ray spectra and , though often crude and phenomenological , provides useful indications that can guide further analysis . the power law slope is free to vary to account for possible spectral pivoting . we add to the model neutral photoelectric absorption with column density fixed to the galactic value . to account for the presence of absorbing ionized gas in the line of sight we include two edges ( o and o ) . we searched for possible variations of their energies without finding any evidence , so we fixed the absorption energies to the rest frame ones . this is clearly a crude approximation for the effects of the warm absorber in ngc 4051 , but the quality of the 2 ks spectra is not such to suggest the use of more sophisticated models . in addition , the constant components discussed above ( reflection from distant matter and emission from photoionized gas ) are also included in the spectral model . the overall model has six free parameters only . the resulting fits are very good with a reduced @xmath17 ranging between 0.9 and 1.2 with a mean close to unity . therefore , from a statistical point of view , a single model is able to describe the spectral variability of the source in a very satisfactory way . below , we discuss the main results of the fitting procedure and their main physical consequences . in fig . [ bbpo ] ( top panel ) we show the relation between the measured bb temperature and its intensity , which is proportional to the luminosity ( in the present case a@xmath53=10@xmath54 correspond to l@xmath53=9.7@xmath1010@xmath55 ergs s@xmath21 , assuming h@xmath56=71 mpc km@xmath21 s@xmath21 ) . the bb luminosity spans about one order of magnitude , while the bb temperature ranges from about 90 ev to 140 ev with an average of 110120 ev . such a temperature is somewhat high , but can still be consistent with that expected from a black hole with mass @xmath57 ( shemmer et al 2003 ; m@xmath58hardy et al 2004 ) accreting at high rate ( for the given black hole mass , the expected disc temperature is of the order of 70 ev for an object accreting at the eddington rate ) . we point out that even in the lowest flux states the temperature ( about 100 ev ) is larger than predicted and , in the framework of standard disc models , would already require the system to be accreting at the eddington rate . if the soft excess in ngc 4051 is really due to bb emission from a constant area , we should be able to detect the bb law ( l@xmath59 ) in data spanning one order of magnitude in bb luminosity . however , such a relation is ruled out . a fit of the l@xmath53 versus temperature data ( top panel of fig . [ bbpo ] ) with the bb law ( leaving only the normalization as a free parameter ) produces a very bad fit with @xmath60 for 67 degrees of freedom . the relation appears to be much steeper ( with an index of 6.36.4 rather than 4 ) , but even so , a power law fit is unacceptable . the measured temperature appears to be much more constant ( the fit with a constant gives a @xmath61 for 67 degrees of freedom ) than predicted by standard disc models . the properties of the soft excess in ngc 4051 are remarkably similar to those of 26 bright radio quiet quasars ( considered as a population ) studied e.g. by gierlinski & done ( 2004 ) and selected from the _ xmm newton _ archive by their strong uv continuum ( such that they have strong accretion disc emission ) . the quasar sample spans a wide range of black hole masses and luminosities and should therefore exhibit a wide range of disc temperatures ( between a few ev to about 70 ev ) . however , when fitted with a cool comptonized component , the measured temperature of the soft excess is remarkably constant throughout the sample with a mean of 120 ev ( and very small variance ) . it is a rather surprising coincidence that this is exactly the average temperature we measure for the soft excess in ngc 4051 , considering also that the black hole mass in ngc 4051 is about 3 orders of magnitude smaller than the typical black hole mass in the quasar sample . moreover , no correlation between the measured soft excess temperature of quasars and the expected disc temperature was found , in line with the results presented above . such discrepancies naturally raises questions on the real nature of the x ray soft excess . as pointed out by gierlinski & done ( 2004 ) , the remarkable constancy of the `` soft excess temperature '' might indicate an origin in atomic rather than thermal physical conditions . possible candidates seems to be absorption ( e.g. from a disc wind , as suggested by gierlinski & done 2004 ) and/or ionized reflection ( e.g. from the accretion disc itself , as suggested e.g. in ross & fabian 2005 ) . the `` standard model '' of the source emission seems to be inadequate to describe the corona parameters as well . in the bottom panel of fig . [ bbpo ] we show the power law slope as a function of the 0.510 kev flux . in the normal / high flux observation the photon index steepens with flux , as already pointed out in the 210 kev analysis . this behaviour is consistent with spectral pivoting . once again , the lowest @xmath2 are of the order of 1.4 which is a hard spectral shape to be accounted for by standard comptonization models and might require a photon starved corona scenario . the main problems arise at very low fluxes , where the @xmath2flux relation is inverted in a manner that is not consistent with simple spectral pivoting . such behaviour could indicate the presence of an additional soft ( steep ) component which becomes prominent at low fluxes and is not properly accounted for by the black body component . in fact , the power law is trying to fit the soft data by steepening the index at low fluxes and leaves significant residuals in the hard band where a slope of about 1.31.4 would be more appropriate even at very low fluxes ( see fig . [ gf1 ] ) . as a final comment , the mean optical depth of the ionized absorber ( here modelled crudely by the o and o edges ) is of the order of 0.10.3 , consistent with the value obtained from the rgs analysis ( pounds et al . these values are left free to vary and they seem to suggest some variations during the two observations . nevertheless , the uncertainty associated with these is so big that a fit to the optical depth with a constant during the first observation results in a @xmath17 of 26 ( o ) and 47 ( o ) for 47 degrees of freedom , while during the second low flux observation the @xmath17 are 29 and 7 for 19 degrees of freedom . the results of the fits with the `` standard model '' presented above suggest that the soft excess is not ( or not completely ) truly disc black body emission . moreover , if the spectral variability is due to spectral pivoting , the power law slope at low fluxes is somewhat too hard to be accounted for by standard comptonization ( haardt et al . in addition , a steeper component appears in the soft band at the lowest flux levels . we then explore here an alternative model to describe the spectral variability of ngc 4051 . in order to reproduce the flux flux plot and the @xmath2flux relationship without imposing spectral pivoting , we need two main continuum components . the first one is variable in normalization only retaining an approximately constant spectral shape at all flux levels . we make here the assumption that this can be represented by a power law component ( plc ) . its slope should be similar to the high flux asymptote of the @xmath2flux relationship , i.e. @xmath62 2.2 . the other component must be almost constant in normal / high flux states where its contribution to the 410 kev band is of the order of 40 per cent . in the lower flux states it has instead to correlate with the plc in order to reproduce the flux flux plot . from the 210 kev @xmath2flux relationship we have also some indication that its spectral shape in the hard band can be approximated by a slope of @xmath631.31.4 . the obvious candidate is a relativistically blurred reflection dominated component ( rdc ) from a ionized accretion disc . such a component is characterised by an x ray spectrum which is hard in the hard band ( @xmath631.31.6 depending on the parameters ) and steep in the soft band therefore naturally producing a soft excess . as pointed out e.g. by ross & fabian ( 2005 ) when fitted with a thermal model in the _ xmm newton _ band , such a soft excess has a temperature of the order of 150 ev , very similar to the constant temperature of radio quiet bright quasars and to the one we have measured in ngc 4051 . since the rdc is produced in the inner disc by irradiation by the plc , the two components should always be well correlated while , in the present case , the flux flux plot would require that the rdc is well correlated with the plc at low fluxes but is then only weakly variable at higher fluxes despite large changes in the plc flux . this is however precisely the main feature of the gravitational light bending model proposed by miniutti et al ( 2003 ) and miniutti & fabian ( 2004 ) . in this model the variability is dominated by a power law component ( plc ) which changes in flux but not in spectral shape . the plc is ( naively ) assumed to have constant intrinsic luminosity and all the plc variability is due to light bending effects as the primary plc sources have different positions at different times with respect to the central black hole . the plc illuminates the accretion disc giving rise to a reflection component which is affected by the relativistic effects arising in the inner disc . such a reflection dominated component ( rdc ) does not respond trivially to the plc variability because of strong gravity effects ( see miniutti & fabian 2004 for more details ) . at low fluxes the rdc is indeed well correlated with the plc , but it gradually varies less and less as the flux increases reaching a regime in which it becomes almost constant in a range in which the plc varies up to a factor 5 . therefore , the model successfully reproduces the two component model ( variable plc plus constant rdc ) at mean / high fluxes and since it predicts a correlation between the two components at low fluxes , it has the potential of explaining the curved flux flux plot relationship without invoking spectral pivoting . in this framework , low flux states correspond to situations in which the plc originates few @xmath43 from the black hole strongly illuminating the innermost disc . light bending reduces at the same time the observed plc flux at infinity . as the plc source is further away from the black hole , more photons can escape to infinity ( so that the observed plc flux increases ) and the illuminating flux on the disc is reduced . the model therefore predicts also a relation between the emissivity profile of the rdc on the disc and flux . the lower the flux , the more centrally concentrated the disc irradiation and the steeper the emissivity . the two component model we have tested comprises then i ) a plc with fixed slope at @xmath64 ( consistent with the high flux asymptote of the @xmath2flux relationship ) and variable normalization and ii ) an ionized rdc ( from ross & fabian 2005 ) the spectrum of which is convolved with a kernel with fixed inner and outer disc radius ( to @xmath36 and @xmath37 respectively ) , and fixed disc inclination ( @xmath38 ) . the inner disc radius corresponds to the innermost stable circular orbit around a kerr black hole . the only free parameters of the relativistic kernel is then the index @xmath65 of the disc emissivity profile ( @xmath66 ) . the ionized reflection model is appropriate for solar abundances and has ionization parameter and normalization as free parameters , while the photon index of the illuminating power law in the rdc model is tied to that of the plc and therefore fixed to @xmath64 . as for the `` standard model '' discussed in the previous section , the overall spectral model also includes galactic absorption , constant emission from photoionized gas and from a distant reflector and the o and o edges with fixed energies . the number of free parameters in the model is just the same as in the `` standard model '' case ( 6 free parameters ) . the model reproduces very well the data at all flux levels with a reduced @xmath17 between 0.8 and 1.2 . given that the standard model and the two component one have the same number of free parameters , we can directly compare the resulting averaged reduced @xmath17 which are @xmath67 for the standard model and @xmath68 for the two component one . therefore , the two component model is statistically indistinguishable from the standard one and has to be considered as a possible alternative to be accepted or rejected on a physical rather than statistical basis . as mentioned , the plc used to model the data has a fixed slope of @xmath64 at all flux levels . in fact , we also tested a variable @xmath2 fit to the data and found that all 68 spectra are consistent with @xmath64 with only three exceptions in which the statistics is significantly better with a different power law slope ( with small @xmath69 ) . therefore , we demonstrate here by direct spectral fitting that the curvature in both the flux flux plot and the @xmath2flux relationship does not imply spectral pivoting but is only consistent with it . the second characteristic of the model is that the soft excess is not due to a thermal component anymore ( see fig.[refl_emod ] ) . it is in fact the result of including ionized reflection from the disc , which naturally produces a soft excess . the `` constant temperature '' problem can then be naturally solved because of the very non thermal nature of that component . the same kind of model successfully reproduces the soft excesses in other sources as well ( e.g. fabian et al 2004 ; 2005 ) . the ionization parameter of the reflection spectrum is not very well constrained by the data and most spectra are consistent with @xmath41 between 50 and 300 erg cm s@xmath21 with only few exceptions below and above ( and no clear trend with flux ) . the reason for the non accurate measurement of @xmath41 is probably twofold . firstly , the spectra lack sharp emission lines which would help in constraining the ionization state ( if present , the emission lines are broadened and skewed because emitted from the inner disc ) . secondly , we are using a reflection model with a single uniform ionization parameter over all the disc surface from the innermost to the outermost regions . this is clearly a rough approximation for a real case in which the ionization parameter has a structured radial profile . such approximation is likely to play a role in the non accurate measurement of @xmath41 . on the other hand , the disc reflection emissivity index of the relativistic blurring model is better constrained . in fig . [ blur ] we show the emissivity index @xmath65 as a function of the plc flux . some trend can be seen in the figure with low flux states generally corresponding to steeper emissivity profiles ( with @xmath65 of the order of 5 ) than high flux ones . as mentioned above , this result is in line with the predictions of the light bending model ( miniutti & fabian 2004 ) . the other main prediction of the light bending model is the rdc versus plc relation . as mentioned , the model predicts a correlation at low flux levels and an almost constant rdc at higher flux levels , despite large variation in the plc . in fig . [ gflux ] we show the 0.510 kev flux of the rdc versus the plc flux in the same band . the rdc is well correlated with the plc at low flux levels ( the solid line represent perfect correlation between the two components ) . however , as the plc flux increases , the correlation clearly breaks down and the rdc is much less variable ( about a factor 2.5 ) than the plc ( about a factor 7 ) in very good agreement with the qualitative predictions of the model . the main difference between the predicted and observed behaviour is that some residual correlation is seen at normal / high flux ( while the model predicts an almost constant rdc ) . such residual correlation can be easily interpreted by intrinsic luminosity variations of the plc source which are not accounted for in the model . we finally remark that , if the light bending interpretation is correct , the rdc versus plc behaviour requires the primary source of plc to be centrally concentrated above the accretion disc within 15 gravitational radii at most from the central black hole ( see miniutti & fabian 2004 ) . in particular , the correlation between the two component is predicted to occur only if the primary source of the plc is within @xmath7 5 gravitational radii from the hole . therefore , one consequence of this interpretation is that low flux states are characterized by an extremely compact region of primary emission , well inside the relativistic region around the black hole . we tested both spectral models further by comparing the rms spectra with synthetic ones , obtained from the best fitting models discussed above . obviously , the better the fits to the 2 ks spectra are , the more closely the rms spectra will be matched , but the comparison is not trivial because the @xmath17 obtained from the time resolved spectroscopy is a global quantity in the energy space ( in fact it depend over all the energies considered ) while the rms is a local quantity ( it does not depend on what is the model at other energies ) . it is possible , for example , that the spectral model slightly fails always in the same energy band . if this is the case , a comparison between observed and simulated rms spectra is useful . in fig . [ rms2 ] we show the observed and synthetic rms spectra for both observations . in back ( blue in the colour version ) we plot the rms spectra obtained from the two component model best fitting parameters and in grey ( green ) that from the standard bb plus power law model . both models reproduce well the broadband shape of the rms spectra in the two observations with residuals of the order of few per cent only . this is possible , allowing the three parameters of the rdc component ( ionization , emissivity index and normalization ) to vary ( see sect . we do not exclude the presence of further soft component with a possibly different physical origin , but we show that that is not required when variations of the parameters of the rdc are considered . in particular , the sharp features in the rms spectra have been reproduced thanks to the inclusion of the two constant components due to distant neutral reflection and photoionized gas emission . this reinforces our previous qualitatively interpretation indicating that the emission line intensities and their constancy ( not only during each observation but also between the two one year apart ) are needed in order to reproduce the rms spectra . the variability spectrum during the rev . 263 is less affected by the two constant components and carries more informations about the nuclear variability . its shape can be explained equivalently well by the two models investigated . in the `` standard scenario '' the gradual drop of variability with energy is mainly due to the steepening of the power law slope with flux , while the lower variability at low energy is due to a less variable bb component . in the two component model scenario it is expected that the source has lower variability both a high and low energy , because these are the regions where the almost constant relativistic reflection component is more important . in order to estimate the contribution , to the overall spectral variability , due to the variations of the warm absorber , we have simulated the rms spectra both with constant and varying optical depth edges . the inserts of figure [ rms2 ] show the enlargement of the rms in the soft energy band , where the effect of the warm absorber are stronger . the grey ( green in the colour version ) line shows the synthetic rms obtained starting from the best fitting parameters with the `` standard model '' and with varying optical depth edges , while the black ( orange in the colour version ) line shows the rms expected from refitting the data with the imposition of a constant warm absorber . we show here only the results obtained considering the standard model , but similar results are obtained with the two component model . the inserts of figure [ rms2 ] show that , when the constancy of the optical depth is imposed , some deviations in the reproduction of the rms are present . the intensity of these deviations are of the order of few percent suggesting that a variation of the warm absorber is possible but the bulk of the spectral variability is due to the interplay of the back body and power law components ( power law and reflection , in the two component model ) if the continuum is left completely free to vary . the analysis we presented above shows that rms spectra , flux - flux plots and individual 2 ks spectra can be explained either in terms of a standard back body plus power law model or of a two component model . in both cases , the presence of constant emission from photoionized gas and from a distant reflector are required . the standard model comprises a blackbody component which provides the soft excess and a power law with variable slope ( and normalization ) . the blackbody temperature is surprisingly constant given that its luminosity spans one order of magnitude and does not seem to follow the blackbody law ( @xmath70 ) . moreover the average temperature we measure ( about 110 - 120 ev ) is remarkably similar to the average temperature measured in a sample of 26 radio quiet quasars with typical black hole masses three orders of magnitude larger than in ngc 4051 . these surprising results suggest that a pure thermal component is not appropriate to account for all the soft excess in ngc 4051 . in fact , as shown in crummy et al . ( in prep . ) the soft excess in type 1 active galactic nuclei is consistent , like here , with being produced by ionized reflection . the power law slope in correlates with flux except at very low flux levels where the relation expected from simple spectral pivoting breaks down . moreover the hardest photon index we measure ( about 1.31.4 ) seem too hard to be explained in terms of standard comptonization models . it should be stressed also that while spectral pivoting describes well the flux flux plot relationship between the 0.10.5 kev and 210 kev bands ( see uttley et al 2004 ) , it fails ( or at least reduces its accuracy ) when the reference soft band is chosen to be between 1 kev and 1.4 kev . this is somewhat surprising because the latter band is dominated ( whatever spectral model is preferred ) by the power law component , while at softer energies a contribution from the prominent soft excess is unavoidable . in the context of the two component model , the soft excess is instead accounted for by ionized reflection from the accretion disc which simultaneously hardens the hard spectral shape as the flux drops . the apparent constant temperature of the soft excess is then explained , because the soft excess spectral shape is dictated by atomic rather than thermal mechanisms . the two component model is consistent with the variability being dominated by a constant slope power law which varies in normalization only . the @xmath2flux relationship is then interpreted as spurious and due to the relative flux of the power law and reflection components . the rdc is in fact correlated with the plc at low fluxes and varies with much smaller amplitude at normal / high fluxes . such behaviour explains not only the spurious @xmath2flux relationship , but is also consistent with the flux flux plot analysis . we investigated the x ray spectral variability of the narrow line seyfert 1 galaxy with model independent techniques ( rms spectra and flux flux plots ) and with time resolved spectral variability . the main features of the x ray emission of are here summarized . 1 . the data of show evidence for a distant neutral and constant reflection component contributing by about 10 per cent in the 410 kev band at mean fluxes . 2 . during the low flux observation a photoionized gas imprints its presence in the soft x ray band . this emission is consistent with being constant not only during , but also between the two observations one year apart . it is not clearly detected during the high flux observation because of a much stronger continuum . + the constancy of this component , as well as the previous one , is primarily shown thanks to the powerful rms spectra . 3 . the nuclear emission has been interpreted both in a `` standard scenario '' ( consisting of bb plus power law emission ) and by a two component ( plc plus ionized rdc from the disc ) . both models reproduce the rms spectra , and describe the time resolved spectra in a comparable way from a statistical point of view . our results are consistent with the warm absorber playing a minor role in the spectral variability . the standard model results indicate that the bb emission does not follow the l@xmath0 relation , even if more than an order of magnitude in luminosity is spanned . the best fit bb temperature is almost constant , clustering its value between 0.1 and 0.12 kev . the power law slope is consistent with spectral pivoting and steepens from 1.31.4 to about 2.2 with flux . the hardest photon indexes are so flat to require rather unusual scenarios such as a photon starved corona . moreover , the very low flux states exhibit an inverted @xmath2flux behaviour which disagree with a simple pivoting interpretation . it is then possible that further unmodelled soft / steep component manifest itself at very low fluxes . 5 . in the framework of the two component model the soft excess is interpreted as the soft part of ionized reflection from the accretion disc . the constant temperature problem is then solved in terms of atomic rather than truly thermal processes . moreover , the rdc explains the @xmath2flux relationship at all flux levels in terms of the relative contribution of a plc with constant slope @xmath64 and the rdc . the variability properties of the rdc with respect to the plc explain also the reason why the flux flux plot is curved without the need for invoking spectral pivoting . the rdc is well correlated with the plc at low fluxes and varies with much smaller amplitude than it at normal / high fluxes , in line with the predictions of the light bending model in which the variability is mainly due to the fact that primary sources close to the black hole appear fainter than sources far from it because of light bending , which reduces the net observed flux at infinity as the source ( of the plc ) is closer and closer to the hole . we then propose the light bending model as a possible alternative explanation for the spectral variability of ngc 4051 in which most of the nuclear emission is emitted from the very inner regions of the flow , only few gravitational radii from the central black hole . from a physical point of view , the model does not require extreme parameters . the only true requirement is that the plc originates in some compact region above the accretion disc and that the active region changes location in time ( or disappears and a new one appears at a different location ) but always within about 15 gravitational radii from the center . we stress here that if , as it is generally believed for high accretion rate sources ( such as ngc 4051 ) , the bulk of the x ray emission comes from the inner regions of the accretion flow , the effects we are talking about are simply unavoidable . no radiation can escape from the inner regions of the flow without being affected by gravitational light bending and other relativistic effects . this is true not only for disc emission ( and reflection ) but also for the primary radiation . the authors would like to thank mauro dadina , giorgio palumbo , giuseppe malaguti , pierre olivier petrucci , roderick johnstone michael mayer and the anonymous referee for many useful comments , suggestions and technical help . based on observations obtained with xmm - newton , an esa science mission with instruments and contributions directly funded by esa member states and nasa . gp thanks the european commission under the marie curie early stage research training programme for support . gm and ki thank the pparc for support . acf thanks the royal society for support .

we describe a new numerical 3d relativistic hydrodynamical code and present the results of three validation tests . a comparison of an axisymmetric jet simulation using the 3d code , with corresponding results from an earlier 2d code , reveal that a ) the enforcement of axisymmetry in the 2d case had no significant influence on the global morphology and dynamics ; b ) although 3d studies typically have lower resolution than those using 2d , limiting their ability to fully capture internal jet structure , such 3d studies can provide a reliable model of global morphology and dynamics . the 3d code has been used to study the deflection and precession of relativistic flows . we find that even quite fast jets ( ) can be significantly influenced by impinging on an oblique density gradient , exhibiting a rotation of the mach disk in the jet s head . the flow is bent via a potentially strong , oblique internal shock that arises due to asymmetric perturbation of the flow by its cocoon . in extreme cases this cocoon can form a marginally relativistic flow _ orthogonal _ to the jet , leading to large scale dynamics quite unlike that normally associated with astrophysical jets . exploration of a flow subject to a large amplitude precession ( semi - angle ) shows that it retains its integrity , with modest reduction in lorentz factor and momentum flux , for almost jet - radii , but thereafter , the collimated flow is disrupted . the flow is approximately ballistic , with velocity vectors _ not _ aligned with the local jet ` wall ' . however , sufficiently large changes in flow direction take place within the jet that for observers close to the jet axis , significant changes in doppler boost would be evident along the flow . we consider simple estimators of the flow emissivity in each case and conclude that a ) while the oblique internal shocks which mediate a small change in the direction of the deflected flows have little impact on the global dynamics , significantly enhanced flow emission ( by a factor of ) may be associated with such regions ; and b ) the convolution of rest frame emissivity and doppler boost in the case of the precessed jet invariably leads to a core - jet - like structure , but that intensity fluctuations in the jet can not be uniquely associated with either change in internal conditions or doppler boost alone , but in general are a combination of both factors .

astro-ph0202402

collimated extragalactic flows , exhibiting a complex pattern of internal structures , often with high brightness temperature and/or superluminal speed , have been explored with increasing spatial and temporal resolution over the last two decades @xcite . remarkably , it has become evident over the last five years that the highly energetic flows associated with active galactic nuclei are also found in galactic objects with stellar mass ` engines ' . in the galactic superluminals grs 1915 + 105 @xcite and gro j165540 @xcite the observed motions indicate jet flow speeds up to @xmath5 . further , there is compelling evidence from the observation of optical afterglow that gamma - ray bursts are of cosmological origin , and whether produced by the mergers of compact objects , or accretion induced collapse ( aic ) , simple relativistic fireball models seem ruled out , the data strongly favoring highly relativistic jets @xcite . thus a detailed understanding of the dynamics of collimated relativistic flows has wide application in astrophysics . a particularly important facet of these flows has been revealed over the last decade as the quantity and quality of images of extragalactic relativistic jets have increased : the curvature of parsec - scale flows is the norm , not the exception ; see e.g. , @xcite . some of this curvature may be ` apparent ' , resulting from a more modestly curved flow seen close to the line of sight . nevertheless , such flows must posses intrinsic curvature , and we are forced to ask how such highly relativistic flows can exhibit significant curvature , and yet retain their integrity , transporting energy and momentum efficiently to the kiloparsec scale and beyond . indeed , the structure ( large - scale curvature , sheaths and filaments , and ` knots ' of emission ) of such flows provides a powerful diagnostic tool , that may elucidate the intrinsic instability , precession , deflecting obstacles and cross - winds to which the flow is subjected ( although it is important to note that the mere presence of structure can not uniquely determine its origin : at a minimum we must explore the velocity and magnetic fields of such features , and their evolution ) . numerous lines of evidence point convincingly to the occurrence of both transverse and oblique shocks , e.g. , @xcite in such flows ; how do they form and evolve ? specifically , how do transverse shocks propagate along a curved flow ; will the shock plane rotate with respect to the flow axis ; will the shock strengthen or weaken ? what role does flow curvature have to play in explaining phenomena such as stationary knots between which superluminal components propagate @xcite , knots which brighten after an initial fading @xcite , and changing component speed @xcite . all these issues are potentially important in both the extragalactic and stellar jet contexts , and are amenable to study through hydrodynamic simulation but all demand that such simulations be 3d . numerical hydrodynamical studies of relativistic astrophysical jets have been available for barely five years . @xcite and @xcite explored _ steady _ relativistic jets , but it was only with the advent of robust , shock - capturing 2d schemes @xcite that a full exploration of relativistic flows began . @xcite , @xcite and @xcite have pioneered 3d mhd studies where fluid is injected into a strong , ordered magnetic field , but the detailed structure and evolution of 3d flows remains largely unexplored @xcite . the current paper describes extension of our original 2d code @xcite to 3d , and its first application to axisymmetric , deflected and precessed flows . our immediate goals are to study : the internal structures that arise from the development of kelvin - helmholtz instability , which may explain features such as those seen in the jet of m 87 ( e.g. , components e and a @xcite ) ; the formation of oblique shocks which may mediate a change in flow direction ( e.g. , 1127@xmath6 - 145 and cta 102 @xcite ) ; and precession which may underly the _ evolving _ , approximately helical trajectories exhibited by a number of parsec - scale flows ( e.g. , bl lac @xcite ) . we assume an inviscid and compressible gas , and an ideal equation of state with constant adiabatic index . we use a godunov - type solver which is a relativistic generalization of the method due to @xcite , and @xcite , in which the full solution to the riemann problem is approximated by two waves separated by a piecewise constant state . we evolve mass density @xmath7 , the three components of the momentum density @xmath8 , @xmath9 and @xmath10 , and the total energy density @xmath11 relative to the laboratory frame . defining the vector ( in terms of its transpose for compactness ) @xmath12 and the three flux vectors @xmath13 @xmath14 @xmath15 the conservative form of the relativistic euler equation is @xmath16 the pressure is given by the ideal gas equation of state @xmath17 the godunov - type solvers are well known for their capability as robust , conservative flow solvers with excellent shock capturing features . in this family of solvers one reduces the problem of updating the components of the vector @xmath18 , averaged over a cell , to the computation of fluxes at the cell interfaces . in one spatial dimension the part of the update due to advection of the vector @xmath18 may be written as @xmath19 in the scheme originally devised by @xcite , a fundamental emphasis is placed on the strategy of decomposing the problem into many local riemann problems , one for each pair of values of @xmath20 and @xmath21 , to yield values which allow the computation of the local interface fluxes @xmath22 . in general , an initial discontinuity at @xmath23 due to @xmath20 and @xmath21 will evolve into four piecewise constant states separated by three waves . the left - most and right - most waves may be either shocks or rarefaction waves , while the middle wave is always a contact discontinuity . the determination of these four piecewise constant states can , in general , be achieved only by iteratively solving nonlinear equations . thus the computation of the fluxes necessitates a step which can be computationally expensive . for this reason much attention has been given to approximate , but sufficiently accurate , techniques . one notable method is that due to ( * ? ? ? * hll ) , in which the middle wave , and the two constant states that it separates , are replaced by a single piecewise constant state . one benefit of this approximation , which smears the contact discontinuity somewhat , is to eliminate the iterative step , thus significantly improving efficiency . however , the hll method requires accurate estimates of the wave speeds for the left- and right - moving waves . @xcite analyzed the hll method and found good estimates for the wave speeds . the resulting method combining the original hll method with einfeldt s improvements ( the hlle method ) , has been taken as a starting point for our simulations . in our implementation we use wave speed estimates based on a simple application of the relativistic addition of velocities formula for the individual components of the velocities , and the relativistic sound speed @xmath24 , assuming that the waves can be decomposed into components moving perpendicular to the three coordinate directions . in order to compute the pressure @xmath25 and sound speed @xmath24 we need the rest frame mass density @xmath26 and energy density @xmath27 . however , these quantities are nonlinearly coupled to the components of the velocity as well as to the laboratory frame variables via the lorentz transformation : @xmath28 @xmath29 @xmath30 @xmath31 @xmath32 where @xmath33 is the lorentz factor and @xmath34 . when the adiabatic index is constant it is possible to reduce the computation of @xmath26 , @xmath27 , @xmath35 , @xmath36 and @xmath37 to the solution of the following quartic equation : @xmath38 where @xmath39 . this quartic is solved at each cell several times during the update of a given mesh using newton - raphson iteration . our scheme is generally of second order accuracy , which is achieved by taking the state variables as piecewise linear in each cell , and computing fluxes at the half - time step . however , in estimating the laboratory frame values on each cell boundary , it is possible that through discretization , the laboratory frame quantities are unphysical they correspond to rest frame values @xmath40 or @xmath41 . at each point where a transformation is needed , we check that the conditions @xmath42 and @xmath43^{1/2}$ ] are satisfied , and if not , recompute cell interface values in the piecewise constant approximation . we find that such ` fall back to first order ' rarely occurs . the relativistic hlle ( rhlle ) method constitutes the basic flow integration scheme on a single mesh . we use adaptive mesh refinement ( amr ) in order to gain spatial and temporal resolution for given computer resources . the amr algorithm used is a general purpose mesh refinement scheme which is an outgrowth of original work by @xcite and @xcite . the amr method uses a hierarchical collection of grids consisting of embedded meshes to discretize the flow domain . we have used a scheme which subdivides the domain into logically rectangular meshes with uniform spacing in the three coordinate directions , and a fixed refinement ratio of @xmath44 . the amr algorithm orchestrates i ) the flagging of cells which need further refinement , assembling collections of such cells into meshes ; ii ) the construction of boundary zones so that a given mesh is a self - contained entity consisting of the interior cells and the needed boundary information ; iii ) mechanisms for sweeping over all levels of refinement and over each mesh in a given level to update the physical variables on each such mesh ; and iv ) the transfer of data between various meshes in the hierarchy , with the eventual completed update of all variables on all meshes to the same final time level . the adaption process is dynamic so that the amr algorithm places further resolution where and when it is needed , as well as removing resolution when it is no longer required . adaption occurs in time , as well as in space : the time step on a refined grid is less than that on the coarser grid , by the refinement factor for the spatial dimension . more time steps are taken on finer grids , and the advance of the flow solution is synchronized by interleaving the integrations at different levels . this helps prevent any interlevel mismatches that could adversely affect the accuracy of the simulation . the time step value is computed by applying the cfl condition to every cell within the computational domain , using the relativistic sum of the local flow velocity and sound speed , and picking the globally minimum value of @xmath45 . the corresponding time step on the coarsest level is computed , and multiplied by a cfl number , typically @xmath46 . in order for the amr method to sense where further refinement is needed , some monitoring function is required . we have used a combination of the gradient of the laboratory frame mass density , a test that recognizes the presence of contact surfaces , and a measure of the cell - to - cell shear in the flow , the choice of which functions to use being determined by which part of the flow is of most significance in a given study . since the tracking of shock waves is of paramount importance , a buffer ensures the flagging of extra cells at the edge of meshes , ensuring that important flow structures do not ` leak ' out of refined meshes during the update of the hydrodynamic solution . the combined effect of using the rhlle single mesh solver and the amr algorithm results in a very efficient scheme . where the rhlle method is unable to give adequate resolution on a single coarse mesh the amr algorithm places more cells , resulting in an excellent overall coverage of the computational domain . the testing of codes for both 2d and 3d numerical hydrodynamics , and the appraisal of the quality of the ` solver ' , in terms of its ability to accurately capture both the structure and location of shocks and contact surfaces is always challenging , because of the sparsity of problems with analytic solutions ; this is particularly true in the case of relativistic flows . we present the results of three common tests below , but also note that the solver employed in the current code is a direct extension to 3d ( with a recast from fortran 77 to fortran 90 ) of the solver described by @xcite . evidence for the accuracy and robustness of that code comes from , in addition to its application to test problems : a ) the general agreement between studies performed with that code and with independently constructed codes , e.g. , that of @xcite ; b ) the agreement between simulations performed with that code and analytic estimates of the expected morphology and dynamics , e.g. , @xcite ; and c ) the agreement between internal jet structures found in simulations performed with that code and the predictions of linear stability analyses @xcite . the primary goal of the tests presented here is to validate the extension and translation of the original 2d code , and its inclusion in a newly constructed amr ` harness ' . the initial condition for the 1d relativistic shock tube problem comprises two piecewise constant states separated by a discontinuity at the origin . the left and right states have pressure and rest density : @xmath47 , @xmath48 , @xmath49 , and @xmath50 . the fluid is initially at rest and the adiabatic index is taken to be @xmath51 . for time @xmath52 a rarefaction wave propagates to the left and a contact surface and shock wave , which bound a narrow region of shocked flow , propagate to the right ; the analytic solution has been provided by @xcite . the computational domain is taken to be @xmath53 $ ] . we have run the shock tube problem with the shock propagating along the @xmath54 , @xmath55 and @xmath56 axes , in each case in ` unigrid ' mode ( a single level , high resolution grid ) and with refinement by two levels of @xmath44 each , with the base grid resolution chosen so that the resolution of the finest grid matches that of the unigrid runs . the code produces comparable results independent of the coordinate direction of the shock propagation , and within refined regions an adaptive solution that is as good in quality of fit to the analytic solution as is the unigrid case . ( in general the unigrid and refined solutions are not identical to machine accuracy , because it is possible that in the refined case depending on the choice of parameters that control where refinement occurs there are sections of the computational domain that remain unrefined ; such domains can lead to very small changes in the location of refined structures . ) figure [ rst ] shows an example of the numerical and analytic solutions for the 1d shock tube tests . the left panels show , from top to bottom , the rest frame density , pressure and lorentz factor for the whole computational domain ; the right panels show in detail the structure of the flow in the vicinity of the contact / shock . the dashed line is the analytic solution . the initial condition for the 3d relativistic shock reflection problem comprises two piecewise constant states separated by a discontinuity at an arbitrary radius of 0.5 . the inner and outer states have pressure and rest density : @xmath57 , @xmath58 , @xmath59 , and @xmath60 . the fluid is initially at rest and the adiabatic index is taken to be @xmath51 . the values were chosen to permit comparison with the test run results of @xcite for a 1d shock - patching code . for time @xmath52 a shock propagates towards the origin , reflects , and moves outwards to interact with the contact surface between the initial states . the computational domain is taken to be @xmath61 $ ] , @xmath62 $ ] , and @xmath63 $ ] . the base grid ( amr level @xmath64 ) was chosen to contain 15 points , yielding a base grid discretization scale of @xmath65 . note that the discretization scale of the base grid is larger than the radius of the initial discontinuity ! figure [ sph_refl ] shows the evolution of the laboratory frame density @xmath66 , the laboratory frame internal energy density @xmath67 , and the @xmath68component of the fluid velocity @xmath69 along the @xmath68axis . the left panel shows the evolution of the fields with three additional levels of refinement ( @xmath70 ) , whereas the right hand panel shows the evolution of the fluid with four additional levels of refinement ( @xmath71 ) . the output times , @xmath72 , and @xmath73 , were chosen so as to easily compare with results from the @xmath74d code written by wen , panaitescu , and laguna ( figure 4 , right panel of @xcite ) . although the discretization scale for the @xmath74d code ( @xmath75 ) is almost an order of magnitude smaller than the discretization scale of the finest grid of our 3d amr code ( @xmath76 ) , we find good agreement between the two . note that the effective resolution of this simulation is 1135 x 1135 x 1135 , which would require over 200 gb ( gigabytes ) of memory . in contrast , our amr code requires only 1.9 gb to run this simulation . in this case , amr techniques have decreased the memory footprint of the simulation by two orders of magnitude ! the 3d relativistic shock reflection problem does not exhibit highly relativistic flow speeds during the evolution covered by the simulation , and the major structures are interior to the location of the initial discontinuity at @xmath77 . in order to validate the code for flows @xmath78 in 3d and to explore the symmetry preserving properties of the code , we have performed a 3d blastwave simulation . in this case a shell of high density and pressure propagates outward from the initial discontinuity . coding errors and discretization effects rapidly become apparent as an asymmetric flow . ideally we would have used the blandford - mckee blastwave solution @xcite as initial data , confirming that while the shock speed remained highly relativistic the computed solution matched the analytic one . however , when using only a wedge - shaped portion of the entire sphere , of opening angle @xmath79rad , and initial energy and density that yield a shock lorentz factor @xmath80 , and peak fluid lorentz factor @xmath81 , the leading structure is so narrow that even five levels of refinement ( requiring @xmath82 gb of memory ) did not adequately resolve it : the highest lorentz factor in the initial data being @xmath83 . it was concluded that the blandford - mckee blastwave constitutes a problem that is too demanding to be a useful validation test unless very significant computer resources are employed . we have thus opted to simulate the 3d - equivalent of the shock tube problem : the initial condition for the 3d relativistic blastwave problem comprises two piecewise constant states separated by a discontinuity at an arbitrary radius @xmath84 . a numerical solution with spherical symmetry , and initial conditions that produced marginally relativistic flow speed , was presented by @xcite ; here we adopt an initial state the leads to higher lorentz factors of order those encountered in the applications discussed in the next sections . the inner and outer states have pressure and rest density : @xmath85 , @xmath86 , @xmath59 , and @xmath87 . the fluid is initially at rest and the adiabatic index is taken to be @xmath51 . for time @xmath52 a spherical shock wave propagates to larger radius . substantial lorentz factors ( @xmath88 ) are generated , and by the end of the simulation the propagating structure encompasses @xmath89 times the volume of the initial high pressure sphere , which is ample development for isolating asymmetries that might arise . some asymmetry is to be expected , because the criteria used for flagging cells to be refined is applied automatically , and a combination of differences due to rounding , coupled with the clustering algorithm , leads to a non - uniform distribution of refined patches . taking the computational domain to be @xmath90 $ ] , @xmath91 $ ] , and @xmath92 $ ] , figure [ blast ] shows the final evolution ( at @xmath93 ) of a run with @xmath94 levels of refinement , yielding @xmath95 finest level cells across the blastwave diameter at the end of the computation . the two panels show laboratory frame density ( @xmath7 ) and lorentz factor ( @xmath96 ) for cuts along the two orthogonal coordinate directions in the plane @xmath97 , plus additional cuts that bisect these two axes . the peak laboratory frame density along the diagonal cuts is @xmath98 higher than along the coordinate directions . this is associated with an oscillation in the lorentz factor near to its peak value @xmath88 , in the cuts that bisect the coordinate directions . the oscillation in the lorentz factor cuts is a result of the fact that even with an effective resolution of @xmath99 cells encompassing the blastwave , the scale of the leading edge of the evolving structure is so narrow that the lorentz factor varies from @xmath88 to @xmath100 over a scale of a couple of the finest cells . off axis , the chance , and changing , location of cells with respect to the thin shell of high lorentz factor leads to a small amplitude irregularity in the peak value , evident as fluctuation along the radial cut . notice however that an inability to fully capture the extreme values encountered on the finest scale , does not in general influence the global location of major flow structures or their values away from these extremes . in particular , there is no global asymmetry . it is well - known from nonrelativistic hydrodynamic studies that the most rapidly growing modes of the kelvin - helmholtz instability are suppressed in 2d simulations ( cf . @xcite ) ; only a fully 3d treatment can capture features likely to play a major role in determining internal jet structure . however , as internal flow structures thermalize only a small fraction of the flow s bulk energy , axisymmetric simulations should provide a reliable picture of the gross morphology and dynamics of an initially axisymmetric flow . in order to test this expectation , and to learn what internal jet structures arise in the absence of externally driven perturbations in 3d , we have rerun case b of @xcite , using the 3d code . if the 2d and 3d results are similar in terms of gross morphology and dynamics , we can conclude that these attributes are insensitive to the details of small scale flow structure , and thus that a ) earlier 2d results provided a valid picture of these relativistic flows on the large scale ; and b ) relatively low resolution 3d simulations should provide a valid picture of the overall morphology and dynamics of deflected and precessed jets . the parameters for this simulation are ( lorentz factor ) @xmath101 , ( relativistic mach number ) @xmath102 , and ( adiabatic index ) @xmath103 . the ambient density is @xmath104 that of the inflowing jet rest density , and the jet and ambient medium are initially in pressure balance . we have achieved the same resolution within the jet ( @xmath105 fine cells across the jet radius ) by using @xmath94 levels of refinement , of @xmath44 each . however , to avoid prohibitive memory requirements we have followed the jet for only @xmath106 jet - radii , as compared with @xmath107 jet - radii in the original 2d run , and in the 3d simulation we have used the finest grid only within @xmath108 jet - radii of the axis thus the cocoon and bow are under - resolved . the computation was terminated as the bow shock reached the far edge of the domain . figure [ axicomp ] shows a schlieren ( gradient ) rendition of the laboratory frame density after 400 computational cycles . for comparison , we show also a 2d simulation performed with the same maximum resolution as that achieved in 3d , having propagated @xmath106 jet - radii . in the absence of an externally applied periodic perturbation to the flow , the development of structure within the jet depends wholly on the driving of available modes by naturally occurring perturbations in particular , pressure perturbations arising from kelvin - helmholtz ` fingers ' growing at the contact surface between shocked jet and shocked ambient medium , and a pressure wave at the inflow @xcite . in the original 2d simulation the contact surface exhibited instability only as the bow approached the edge of the computational domain . the poor resolution with which the contact is captured in 3d , and the shorter duration of the simulation , ensure that no manifestation of this slowly growing instability is seen , and thus that little jet structure develops as a result . however , the 3d simulation does exhibit a much stronger pressure wave at inflow than is seen in the corresponding 2d case . this is due to the radically different boundary conditions necessary in 3d : inflow is imposed on the plane @xmath109constant , and involves cells cut by the jet boundary for which state variables must be established through a volume - weighted average of the internal and external values . to avoid a ` leakage ' of jet momentum into the ambient material , fixed , initial values are used across the entire boundary plane at every time step . in the 2d simulation above the jet inflow conditions were chosen to be ` extrapolated ' or ` rigorous outflow ' , but the initial conditions were not enforced for the duration of the computation . the ambient flow thus evolved to smear out the shear layer near the inflow in the 2d case , reducing the shear - induced wave in the jet . thus for 2d and 3d simulations _ of equal resolution _ there is stronger driving of available modes by a pressure wave at the inflow @xcite and over most of its length the 3d simulation exhibits more internal structure ; the 2d simulation manifests internal structure of significant amplitude only just upstream of the mach disk , where the development of instability in the better - resolved contact surface drives a weak pressure wave into the jet . from a measurement of the well - defined locations of pressure minima along the flow axis we estimate a wavelength for internal structure in the 3d case of @xmath110 jet radii . as noted in @xcite , the actual jet radius is somewhat larger than the inflow radius , and thus analytic predictions based on the inflow scale slightly underestimate the wavelength of excited modes . bearing this in mind , it is plausible to identify the mode excited in the 3d simulation as the second body mode , with @xmath111 . this is consistent with the fact that the lower order modes have longer wavelength , and are oblique to the jet axis , and the oblique inlet pressure wave more easily couples to these than to higher order ones . in global terms the 2d and 3d simulations are similar : the inclination of the leading bow shock , a well - defined mach disk standing back from the contact surface , and a cocoon that ( near to the head ) spans at most @xmath112 jet - radii . this similarity is quantified by comparing the variation in density , pressure and lorentz factor for cuts along the flow axis . figure [ axiquant ] shows that the global morphology differs little between the 2d and 3d cases . the only significant difference is seen in the off - axis cuts just upstream of the head , and reflects the more extensive , but lower amplitude , pressure wave - driven flow structure in the 3d case as discussed above . this has minimal influence on the global morphology because the internal structures are associated with only slight dissipation of the flow energy : the global dynamics is determined primarily by the energy and momentum flux along the jet and that is captured as well in 3d as in 2d . the similarity is even more striking in the dynamics : the average speed of advance of the bow shock is @xmath113 in the 2d case , and @xmath114 in the 3d case . these results have an important implication . our 3d simulations encompassing the global dynamics ( jet , bow and contact ) will not be able to achieve a spatial resolution necessary to capture the flow contact surface well - enough to reveal it s instability , with the consequent driving of certain normal modes of the jet . however , in as much as it is the global dynamics of the source that is of interest , even the modest resolution 3d simulations that are currently viable do provide a valid picture . we are thus in a position to address the influence on global morphology of jet deflection through an encounter with ambient inhomogeneities , or large amplitude precession of the inflow . note that as regards exploring the details of internal jet structure , we may study a _ pre - existing _ flow , established across the computational domain in pressure balance with a _ low density _ ambient medium , representing the cocoon established after the passage of a bow shock . within available resources we have performed such simulations with @xmath115 cells across the jet diameter , confirmed that almost pure normal modes may be excited through precession - induced driving at frequencies suggested by linear stability theory , and demonstrated that linear stability theory correctly predicts jet structures even for highly nonlinear development ( @xmath116 ) @xcite . as noted in 1 , oblique shocks are likely to play a major role in the understanding of evolving parsec - scale jets , and in the detailed morphology of kiloparsec - scale flows such as m 87 . while models based on transverse propagating shocks have been extremely successful in explaining both the radio polarization and broadband behavior of bl lac objects and qsos @xcite , it has become evident recently that only oblique structures are capable of providing a quantitative explanation for many features seen on a range of length scales @xcite . oblique shocks have been widely studied in the context of recollimated flows , e.g. , @xcite and for their role in mediating bends in supersonic flows , e.g. , @xcite . such structures are stationary , but disturbances to the flow are likely to generate propagating oblique structures also . here we explore both stationary and propagating oblique structures that arise due to a change in the jet direction . curvature may result from the nonlinear development of flow instability , as may internal structure , where , ultimately , modes driven by perturbations to the jet steepen to form shocks . we shall address this issue in future studies of the development of instability . here we are concerned with flow curvature induced by an external influence , and the associated internal structure that mediates the change in flow direction . a jet flow may be bent by a cross - wind , an ambient pressure gradient , an ambient density gradient , or interaction with a discrete ambient inhomogeneity ( cloud ) . a cross - wind will cause a large scale curvature such as seen in nat and wat sources @xcite , and is unlikely to be the origin of subparsec - scale curvature , particularly given the implausibility of sustaining a cross wind deep within a galactic potential . an ambient pressure gradient will also cause large scale curvature and will relax unless sustained by a gravitational potential . in the latter case , the gradient is likely to be in the same sense as that in which the jet propagates from the central engine , leading to little or no curvature . we have thus opted to study the interaction of a relativistic flow with an ambient density gradient , as this seems the external influence most likely to cause flow curvature , and which can be regarded as an idealization of the interaction with a discrete cloud . an issue with this type of study , known from the simulation of nonrelativistic jet interaction with clouds @xcite , is that a nontrivial interaction occurs only for a narrow range of jet parameters . on either side of this range the momentum flux of the jet is either large enough that the cloud provides negligible hindrance to the jet flow , or so low that the jet is effectively stopped by a rigid obstacle . as the thrust of a relativistic jet increases rapidly with lorentz factor , we explore in detail only flows with modest lorentz factor : @xmath117 . the ambient density gradient has been modeled as a smooth ramp running from low density @xmath118 to high density @xmath119 away from the inflow , at angle @xmath120 to the inflow plane , and with ramp width @xmath121 measured in units of the jet inflow radius . the ambient medium is assumed to have uniform pressure distribution , implying a lower temperature in the high density region ; that is consistent with the high density region modeling a condensation that has achieved pressure balance with the surrounding material after experiencing thermal instability . indeed , observations of the ism , e.g. , @xcite , show that within each phase , the dispersion in density variations is of order the mean density , and the range of densities encountered within any phase spans typically less than an order of magnitude , with even smaller pressure variations . to assess the maximum likely consequence of such density variations , we consider a jump in density by one order of magnitude , over a scale of order the jet radius . as anticipated , for ambient gradients more - or - less parallel to the inflow plane the jet material is either deflected to form a bubble of hot gas encompassed by the slightly distorted interface to the high density domain , or rapidly penetrates that interface . the relativistic nature of the flow does not lead to the formation of distinctive structures . only interfaces that are highly inclined to a flow allow the flow to retain its integrity , and lead to the formation of distinct internal structures . in the following subsections we present results for a range of flow speeds and density gradient orientations . we have limited these studies to @xmath122 , as the jet will respond only weakly to more gradual gradients . figure [ defl25a ] illustrates the case @xmath125 , @xmath126 , @xmath124 and @xmath127 for a jet with @xmath128 and @xmath123 . note that a region of extremely low density and pressure and high shear ( a change in @xmath129 by @xmath130 over a scale @xmath131 jet - radii ) develops on the low density side of the jet immediately to the left of the inflowing jet in the left panel . the jet exhibits a number of oblique structures , primarily driven by the impact of the cocoon , the dynamics of which are clearly seen in the third panel : the cocoon is one - sided , and constitutes a structure orthogonal to the jet flow , with marginally relativistic motion partially directed towards the base of the jet . a large fraction of the low density cocoon material is moving orthogonal to , and away from , the jet with a speed @xmath132 , fast enough to exhibit ( if seen face on ) modest but significant doppler boosting ( by @xmath133 ) . that part of the cocoon that impacts the ` base ' of the jet does so with a speed @xmath134 , but the low density and small volume of this portion of the cocoon mean that the influence of this flow is only to slightly perturb the jet . further evolution would have the jet continue to propagate within the higher density material , and the head would no longer be aware of the ambient density gradient . however , until the head is many more jet - radii further on , the shocked jet material will preferentially follow the channel that takes it back on one side of the jet , maintaining a source of driving for structures internal to the jet . the fourth panel in figure [ defl25a ] shows a grey scale version of the first panel , colored blue , added to a reddened version of a similar snapshot 4% of the final time earlier . unchanged regions are thus rendered neutral , while the evolved portions are evident as red or blue ( or as ` ghost ' images if seen in monochrome ) , and enable us to set limits on the speed of propagation of the internal jet structures ; these we find to be propagating at a speed @xmath135 of the marginally relativistic bow shock to a good approximation they are stationary . these structures mediate a change in jet direction ( by @xmath136 ) midway between inflow and mach disk , which in view of the approximate stationarity of the structures must persist as the bow moves well beyond the density interface . above the ` bend ' the velocity component @xmath137 has an average sense and magnitude ( @xmath138 ) consistent with the entire body of the jet following the deviation in flow direction , while several jet - radii upstream of the mach disk an oblique structure may be seen in the left - most panel of figure [ defl25a ] ( crossing the jet from lower left to upper right ) which is a shock in which the pressure jumps by @xmath139 and which mediates a change in the flow direction so that close to mach disk / bow shock the flow has resumed its original direction of propagation . by rendering the laboratory frame density gradient in slices parallel to the inflow plane , figure [ defl25b ] shows the extent of the cocoon , also delimited by the twin - peaks in the pressure and density plots of figure [ defl25c ] which shows plots of lorentz factor , pressure , rest frame density and momentum flux along a stack of cuts , such as the one shown by the horizontal white line in the lower - left panel of figure [ defl25b ] . the upper - left panel showing the lorentz factor makes clear the slowing of the jet just upstream of the mach - disk , but also that the higher pressure there leads to a higher momentum flux as seen in the lower right panel . in fact , the momentum flux increases with @xmath56 , up to and inclusive of the 11th slice , and manifests significant variation across the cross - section of the flow at different @xmath56 , as a consequence of the internal structure . along a locus of peak lorentz factor or momentum flux spanning planes @xmath109constant ( the result is insensitive to the choice of variable , and the locus closely follows the inflow axis ) a ) the lorentz factor retains its inflow value ( @xmath140 ) for @xmath141 jet radii , slowly dipping to @xmath82 upstream of the bow shock at @xmath142 jet radii ; b ) the momentum flux varies slowly , by less than @xmath3% , along the first @xmath142 jet radii , with a value within @xmath143% of its inflow value just upstream of the bow . interaction with the ambient medium has lead to the development of discernible internal structure not seen in the axisymmetric case , but the global dynamics of the jet ( if not the cocoon ! ) are minimally influenced by this interaction . we conclude that inhomogeneity in the ambient medium may lead to significant , oblique internal jet structure albeit essentially stationary with little flow energy dissipation , or change in the overall direction of the flow . stationary vlbi structures , which may be a manifestation of such oblique shocks , are now known to be very common @xcite . figure [ defl50a ] shows the evolution of a faster jet ( @xmath1 ) with the same inclined density gradient . the behavior is not dramatically different from that exhibited by the @xmath123 flow , but a number of important features may be seen more clearly here . as the flow meets the density gradient , a clockwise rotation of the mach disk occurs ( upper panels in figure [ defl50a ] ) , while after the initial interaction , the mach disk rotates counter - clockwise ( bottom panels in figure [ defl50a ] ) to lie orthogonal to the density incline by the last cycle of computation . the maximum clockwise rotation of the disk is @xmath144 , and the maximum counter - clockwise rotation is @xmath145 : highly polarized emission from the downstream flow of the mach disk may well be the dominant feature on vlb maps , but remain unresolved , the direction of the electric vector being used to establish the likely flow direction ; deviations from the transverse , by tens of degrees , could lead to a significant misestimation of the local flow direction . the jet deflection is mediated by an internal shock whose development starts to become evident in the fifth panel as a structure crossing the mach disk . by the last cycle , this has developed to form an internal structure extending upstream from the mach disk , with an inclination comparable to that of the density gradient . the higher momentum flux of the jet in this case ensures that deflection occurs only within the extent of this oblique shock , and the evolution is not followed far enough to see realignment of the jet along its original direction of propagation . the internal shock that moderates the jet deflection is quite strong , the rest frame density and pressure jumps being @xmath88 and @xmath146 respectively . figure [ defl10a ] shows the evolution of a @xmath148 jet with the same inclined density gradient . the speed and jet - to - ambient density contrast are comparable to those of cases c and d studied by @xcite wherein it was found that the ambient medium presented little obstruction to the flow , and little internal jet structure arose . in the present case the jet head moves with high efficiency ( @xmath149 ; see @xcite ) through even the densest ambient gas at relativistic speed , and all sections of the head cross the density incline before a significant rotation of the mach disk can occur . a rotation of the mach disk , deflection of the jet , and generation of oblique internal shocks occurs only if the jet is of low enough speed that the densest ambient regions lead to an inefficient , slow head advance there . figure [ defl50sa ] shows the evolution of a @xmath1 jet with a density gradient inclined at a shallower angle @xmath151 than explored previously . as would be anticipated the clockwise / counter - clockwise rotation of the mach disk as the jet crosses the interface is slight . however , this shallower angle case sheds light on the origin of the jet deflection . a conical pressure wave which steepens to form a weak conical shock is driven _ asymmetrically _ into the jet flow at the location of the ambient gradient , and the asymmetric nature of the flow perturbation leads to its deflection . some jet - radii beyond this point an oblique internal shock that cuts the mach disk returns the flow to its original direction . the inclination of the density gradient allows us to follow the jet further into the denser ambient medium in this case . by the end of the simulation both the morphology of the cocoon and the velocity field within it show that the shocked jet flow is similar albeit still somewhat asymmetric to that seen for propagation in a uniform environment . however , without the low density region to act as a sink for the backflowing shocked jet material , the cocoon at left , just behind the head , starts to inflate . that induces a curvature to the contact surface between shocked jet and shocked ambient material , enhanced by the kelvin - helmholtz instability , and this in turn drives a pressure wave into the jet just upstream of the mach disk . a normal mode of the jet is excited ( cf . @xcite ) which steepens to form a conical shock upstream of the mach disk . the influence on the flow is quite significant : constriction of the flow accelerates the jet upstream of the conical shock to @xmath152 , but the shock decelerates the flow to @xmath153 ; the corresponding pressure jump is @xmath154 . in as much as the types of interaction just discussed are uniquely associated with the leading edge of jets , the structures that result are localized , and will not form a web of propagating shocks along the length of a flow . however , this study highlights that the mach disk associated with the hotspots of kiloparsec ( and larger ) scale jets may display orientations revealed by their polarization structure that bears no simple relation to the jet orientation and accompanied by emission from associated oblique shocks , if the head experiences a significantly inhomogeneous ambient medium . further , on any scale , and in particular on the unresolved sub - parsec scale , a jet that restarts after a period of quiescence that allows a previously formed channel to relax , or that restarts with a new orientation , will display polarized emission that is a complex sum of that from the rotating mach disk and developing internal shocks , and which on such scales will evolve within observable time frames . symmetries seen in twin - jet sources , in particular the s - symmetric sources , e.g. , @xcite , and the intriguing success of binary black hole models for the optical - radio waveband behavior of oj 287 , e.g. , @xcite , suggest that jets are subject to a precessional motion , and provide a quantitative explanation for the possible origin of that motion , respectively . even if precessional motion associated with a massive binary system is not the cause of a time - dependent curvature seen in some parsec - scale flows ( e.g. , bl lac ; mutel et al . , private communication ) , an exploration of the impact of precession on flow structure and evolution will give insight into the internal structures to be found in evolving , curved flows . as a first step in exploring the role of precession , we have applied precession to the inflow of the axisymmetric jet discussed above : a jet with inflow lorentz factor of @xmath112 . the inflow precesses on a cone of semi - angle @xmath2 with a frequency @xmath155 rad measured in time units set by the inflow radius and speed . the relatively large computational domain , @xmath156 jet - radii , meant that we could employ only three grid levels , with refinement by @xmath44 , providing @xmath157 fine cells across the inflowing jet diameter . in the approximation that both the inflow speed and bow propagation speed are close to unity , the precession rate implies @xmath158 revolutions of the jet during evolution across the computational domain . the computation was terminated when the bow was @xmath159% of the way to the domain s far edge , as by then significant dissipation and disruption of the jet had occurred . renders in planes parallel to the inflow plane reveal a small influence of the boundary on structures close to the edge of the computational domain . however , there is no evidence that the limited lateral extent of the computational domain influences the spine of the flow , which is that structure of most significance to us . figure [ preca ] shows a sequence of slices in the plane @xmath97 at equally - spaced intervals during the evolution , using schlieren plots of laboratory frame density . superficial inspection of the final slice suggests that to a significant extent , the jet has retained its integrity , and is driving a bow not dissimilar to that seen in the unprecessed case . in particular , on average the bow is advancing at @xmath160 , which is @xmath161% the speed with which it advanced in the unprecessed case , suggesting a relatively undiminished jet thrust . however , figure [ precb ] shows from left to right , a schlieren render of the pressure there are dramatic variations of pressure along the jet s path , suggestive of significant dissipation ; a linear render of the pressure ( dominated by the leading bow ) with 3-velocity vectors superposed evidently the jet s momentum has been shared with a broad sheath of material ; and finally , the lorentz factor showing that despite precession , the jet does retain its integrity for almost @xmath3 jet - radii , but thereafter the flow speed drops dramatically . the asymmetric bow is being driven forward by a flow in which the pressure maintains a high enthalpy , and which is to some extent focussed onto a small area at the bow s apex . figure [ precc ] quantifies this , showing the run of rest frame density , pressure , lorentz factor and momentum flux ( or ` discharge ' ) , @xmath162 , along the spine of the flow . this spine was defined by computing weighted , average @xmath54 and @xmath55 values in a series of planes @xmath109constant , within a cylinder of diameter @xmath163 jet - radii aligned with the jet inflow ( to exclude complex structures near to the edge of the computational domain ) . using the lorentz factor as a weighting function produced a locus barely distinguishable from that found using the momentum flux as a weighting function , leading us to the conclusion that this approach indeed picks out a physically significant core flow . the quantities plotted in figure [ precc ] are spatial averages within cylindrical pills of one jet radius , aligned locally with the spine . increasing the radius of these sampling volumes decreased the average lorentz factor and momentum flux , but did not change the character of the variation of these quantities along the spine . beyond @xmath3 jet - radii the lorentz factor falls well below its initial value and the local flow direction becomes more chaotic . indeed , the orientation of the velocity vectors within the jet , relative to the local jet ` direction ' defined by the locus of the spine is a quantity of fundamental importance . is the flow locally parallel to the jet boundary , with little momentum imparted to the jet s wall , leading to the expectation that flow disturbances ( e.g. , shocks ) would follow the jet channel , or do fluid elements move more ballistically ? the bottom panel of figure [ precc ] shows the angle between the spine of the flow and the inflow direction ( i.e. , the @xmath56-direction ) as a dashed line , and the angle between the spine of the flow and the velocity as a solid line . the former changes quite smoothly along the flow , increasing to a value well above that imposed at inflow , in conjunction with a decline in the @xmath56 component of momentum flux . by @xmath164 jet - radii large fluctuations associated with jet disruption become evident . if the velocity vectors tracked the flow , the solid line would remain close to zero . however , until significant disruption of the jet is evident , the angle between the spine of the flow and the velocity closely follows , but lies a little below , the angle of the spine . this is because the velocity vectors are almost entirely along the @xmath56-axis , with a slight offset ( typically @xmath165 ) towards the local channel direction ( with the potential to change the doppler boost of emitted radiation by more than @xmath106% ) . thus while there is a weak tendency for velocities to follow the jet channel , the motion is substantially ballistic , with significant momentum imparted to the jet wall , displacing that surface as the flow evolves . figure [ precd ] shows the evolution of the location of the spine in 9 equally separated planes @xmath56=constant to a little beyond @xmath3 jet - radii . ( the jet spends too little time further from the inflow to study its evolution there . ) the complex pattern of dashed lines that join these points arises because the spine is captured at various phases of precession in the 28 epochs sampled , but the stable , conical path of the spine is evident for tens of jet radii from the inflow . interaction between the precessing flow and its cocoon of shocked jet material has imparted a significant ` forward ' velocity to the latter . within the broad region delimited by the contact surface between shocked jet and shocked ambient material the entire flow moves forward at a marginally relativistic speed , forming a plateau about a spine of highly relativistic flow . the corresponding velocity distribution for the unprecessed case shows significant forward flowing cocoon material only within a jet s radius of the jet itself . this contrasts even more dramatically with the results seen in highly underdense jet simulations , e.g. , @xcite , where the cocoon material shows significant _ backflow _ in the laboratory frame ( albeit away from the jet , near to the contact surface ) , with the potential for significant doppler enfeeblement of its radiation flux . the broad forward flow seen in the current simulation is suggestive of the core - sheath morphology suggested to explain observations implying both relativistic and nonrelativistic motions in the same source , e.g. , @xcite , and the recent observation of a source with a ` spine - sheath ' morphology , evident in the polarization map @xcite . whether jets that propagate from the sub - parsec scale to the tens of kiloparsec scale can retain their integrity under the influence of large amplitude precession can be answered only through larger scale computations involving a diverging flow , wherein the local scale length to jet radius ratio is bounded . we have shown that jets impacting an inclined ambient density gradient are subject to the development of internal waves and oblique shocks . it is impractical to run the simulations to a point where the leading bow shock leaves the computational domain , but we have shown that these internal structures are slow moving , and argued that their driving flows persist long after the bow has passed the deflecting structure ; thus an interesting question is the extent to which these internal structures would be observable in maps not dominated by the bow shock ( which even if in proximity to the structures of interest , might contribute little to the synchrotron emissivity e.g. , @xcite ) . in the simulation with @xmath123 , @xmath124 ( see figure [ defl25a ] ) the oblique , high pressure structure upstream of the flow s head spans the jet in the plane parallel to the ambient density gradient , and in a plane orthogonal to that , constitutes a wedge of extent @xmath166 jet radii , and maximum width @xmath167 jet radii . it thus fills @xmath168% of the volume of the perturbed jet segment . the peak pressure in this structure is @xmath169 , while in the adjacent cocoon @xmath170 , and in the initial state @xmath171 . if the emission is optically thin , with spectrum @xmath172 , there is equipartition between radiating particles and random magnetic field ( @xmath173 ) , and the number of radiating particles is proportional to the internal energy and thus the pressure , the emissivity is @xmath174 @xcite . for @xmath175 the emissivity of the oblique structure is @xmath89 times that of the quiescent jet . allowing for the @xmath168% filling factor , this structure would be @xmath176 times brighter than a corresponding segment of unperturbed jet . the simulation with @xmath1 and @xmath150 ( see figure [ defl50sa ] ) also shows a prominent oblique internal structure in the final time frame ; again , the structure spans the jet in a plane parallel to the ambient gradient and its section orthogonal to this direction approximates an ellipse with major axis tilted slightly with respect to the jet s inflow direction . the filling factor is @xmath164% . the corresponding pressure values are @xmath177 , @xmath170 and @xmath178 ( the initial pressure being higher than in the slow jet case for fixed mach number ) . following the same arguments as used above to estimate the component flux , it follows that this structure would be @xmath179 times brighter than a corresponding segment of unperturbed jet . we conclude that where jets are disturbed by ambient density structure , jet emission may be enhanced by a significant factor ( @xmath4 ) locally , with little impact on the jet s global dynamics . a simple prescription for the appearance of the precessed jet ( see figure [ preca ] ) can not be readily given , as the structure is highly asymmetric , and while the local jet boundary evolves at subrelativistic speed , significant doppler boosting is associated with the instantaneous flow velocity , and varies with observer location , even for observers with the same inclination to the inflow axis . maps are being computed by c. m. swift , following the approach described by @xcite and these results will be presented elsewhere . for the purpose of characterizing brightness variations that might be associated with such flows , here we consider the variation of a simple estimator of the source intensity along the flow s spine . we have computed an effective emissivity @xmath180 as a function of distance along the spine , where @xmath181 , the doppler factor , and the boost is appropriate for stationary , bounded flows @xcite . the well - defined , high lorentz factor channel evident in figure [ precb ] illustrates that at least within a jet radius of the spine , the internal flow conditions are quite uniform , and that the significantly doppler boosted flow suffers little divergence before disruption just upstream of the bow . the observed intensity will be determined by integrating the emissivity across the doppler boosted flow for some line of sight , and given the cross - flow uniformity , and comparable line of sight depth for each flow segment , @xmath182 provides a reasonable estimator of the source brightness . we have adopted @xmath175 , and an angle of view with respect to the inflow axis @xmath183 , the latter being of interest as the critical cone of the inflowing jet ( with @xmath101 ) is @xmath184 . the detailed run of @xmath182 along the spine depends upon the azimuthal angle chosen . however , for all cases , a narrow peak in doppler factor occurs between @xmath143 and @xmath142 jet radii along the spine , and convolved with the broad pressure peak at that location ( see figure [ precc ] ) always leads to a substantial intensity enhancement : the peak ` intensity ' is typically @xmath163 to @xmath143 times that in the ` intensity ' minimum immediately downstream . beyond @xmath142 jet radii , significant ( by a factor of @xmath185 or more ) fluctuations in ` intensity ' occur , the details of which depend on observer orientation , but which are _ not _ uniquely tied to variations in the underlying pressure distribution or doppler factor although there is a trend for pressure fluctuations to underpin the ` intensity ' variations as the head is approached . the flow thus exhibits a core - jet morphology , in which individual features can move both towards and away from the jet s origin according to the phase of precession by a number of jet radii . in particular , if the core is associated with the inner - most region where emissivity and high doppler boost conspire to produce a high intensity component , the location of that component will not be stable to within a few jet - radii . as the precession time scale is likely to be long compared with the time scale for propagation of an individual component along the flow , this will have minimal effect on a sequence of vlbi measurements that define the trajectory of one component . however , it could lead to long - term changes in the relative position of stationary components , and change the inner jet orientation so as to influence the details of the interaction of propagating and stationary components , e.g. , as seen in @xmath186 @xcite . this precessed flow has a number of characteristics exhibited by bl lac objects : a core - jet morphology , with stationary , or slowly moving components , and a flow that appears to dissipate rather close to the core . jets associated with qsos may typically be followed further than those associated with bl lac objects , so perhaps one facet of the class distinction relates to the amplitude of precessional perturbation to which the jet is subjected . evidently a small helical perturbation ( transverse velocity of order 1% of longitudinal velocity ) does _ not _ lead to disruption of the flow @xcite . our immediate goal is to generate a well - sampled grid of simulations for a range of lorentz factor and precession cone semi - angle , to establish both the largest amplitude of precession for which flows of a given lorentz factor avoid macroscopic break - up , and the physical origin of that break up . the extent to which propagating shocked fluid , thought to explain moving vlbi components , will follow the helical channel , or follow approximately ballistic trajectories , is also a subject for future study . this work was supported in part by nsf grants ast 9617032 and phy 9979985 , by nrac allocation mca93s025 at ncsa , and by the ohio supercomputer center . we thank the referee for many useful comments that helped to improve the presentation of this work . alberdi , a. , gmez , j. l. , marcaide , j. m. , marscher , a. p. , & prez - torres , m. a. 1999 , in bl lac phenomenon , ed . l. o. takalo & a. sillanp ( san francisco : asp ) , 452 alberdi , a. , gmez , j. l. , marcaide , j. m. , marscher , a. p. , & prez - torres , m. a. 2000 , , 361 , 529 aller , h. d. , hughes , p. a. , freedman , i. , & aller , m. f. 1999,in bl lac phenomenon , ed . l. o. takalo & a. sillanp " ( san francisco : asp ) , 45 aloy , m. a. , ibez , j. m.@xmath187 , mart , j. m.@xmath187 , gmez , j .- l . , & mller , e. 1999 , , 523 , l125 attridge , j. m. , roberts , d. h. , & wardle , j. f. c. 1999 , , 518 , l87 berger , m. 1982 , dissertation , stanford university berger , m. , & colella , p. 1989 , j. comp . phys . , 82 , 67 bicknell , g. v. , & begelman , m. c. , 1996 , , 467 , 597 blandford , r. d. & mckee , c. f. 1976 , phys . fluids , 19 , 1130 bowman , m. 1994 , , 269 , 137 cawthorne , t. v. 1991 , in beams and jets in astrophysics , ed . a. hughes ( cambridge : cambridge univ . press ) , 187 daly , r. a. , & marscher , a. p. 1988 , , 334 , 539 dar , a. 1998 , , 500 , l93 denn , g. r. , mutel , r. l. , & marscher , a. p. 1999 , , 129 , 61 duncan , g. c. , & hughes , p. a. 1994 , , 436 , l119 einfeldt , b. 1988 , siam j. numerical analysis , 25 , 294 godunov , s. k. 1959 , mat . , 47 , 271 hanasz , m. , & sol , h. 1996 , , 315 , 355 hardee , p. e. , & clarke , d. a. 1992 , , 400 , l9 hardee , p. e. , hughes , p. a. , rosen , a. , & gomez , e. a. 2001 , , 555 , 744 hardee , p. e. , rosen , a. , hughes , p. a. , & duncan , g. c. 1998 , , 500 , 599 harten , a. , lax , p. , & van leer , b. 1983 , siam rev . , 25 , 35 heinz , s. , & begelman , m. c. 1997 , , 490 , 653 higgins , s. w. , obrien , t. j. , & dunlop , j. s. 1999 , , 309 , 273 hjellming , r. m. , & rupen , m. p. 1995 , , 375 , 464 hughes , p. a. , aller , h. d. , & aller , m. f. 1989a , , 341 , 54 hughes , p. a. , aller , h. d. , & aller , m. f. 1989b , , 341 , 68 icke , v. 1991 , in beams and jets in astrophysics , ed . a. hughes ( cambridge : cambridge univ . press ) , 232 jorstad , s. g. , et al . 2001 , , 134 , 181 koide , s. , sakai , j .- nichikawa , k .- & mutel , r. l. 1996 , , 464 , 724 komissarov , s. s. , & falle , s. a. e. g. 1996 , in energy transport in radio galaxies and quasars , ed . e. hardee , a. h. bridle , & j. a. zensus ( san francisco : asp ) , 173 komissarov , s. s. , & falle , s. a. e. g. 1998 , , 297 , 1087 leahy , p. j. 1991 , in beams and jets in astrophysics , ed . a. hughes ( cambridge : cambridge univ . press ) , 100 lister , m. l. , marscher , a. p. , & gear , w. k. 1998 , , 504 , 702 lobanov , a. p. , & zensus , j. a. 1996 , in energy transport in radio galaxies and quasars , ed . e. hardee , a. h. bridle , & j. a. zensus ( san francisco : asp ) , 109 marscher , a. p. & gear , w. k. 1985 , , 298 , 114 marscher , a. p. 1997 , , 190 , 5106 mart , j. m.@xmath187 , mller , e. , & ibez , j. m.@xmath187 1994 , , 281 , l9 mart , j. m.@xmath187 , mller , e. , font , j. a. , & ibez , j. m.@xmath187 1995 , , 448 , l105 mart , j. m.@xmath187 , mller , e. , font , j. a. , ibez , j. m. z. , & marquina , a. 1997 , , 479 , 151 mendoza , s. , & longair , m. s. 2001 , , 324 , 149 mendoza , s. , & longair , m. s. 2001 , astro - ph/0109125 mioduszewski , a. j. , hughes , p. a. , & duncan , g. c. 1997 , , 476 , 649 mirabel , i. f. , & rodrguez , l. f. 1994 , , 371 , 46 mirabel , i. f. , & rodrguez , l. f. 1995 , annals of the new york academy of sciences , eds . h. bhringer , g. e. morfil , & j. trmper , 759 , 21 mutel , r. l. , phillips , r. b. , su , b. , & bucciferro , r. r. 1990 , , 352 , 81 muxlow , t. w. b. , & garrington , s. t. 1991 , in beams and jets in astrophysics , ed . a. hughes ( cambridge : cambridge univ . press ) , 52 myers , p. c. 1978 , , 225 , 380 nichikawa , k .- i . , et al . 1997 , , 483 , l45 nichikawa , k .- i . , 1998 , , 498 , 166 perlman , e. s. , biretta , j. a. , zhou , f. , sparks , w. b. , & macchetto , f. d. 1999 , , 117 , 2185 polatidis , a. g. , & wilkinson , p. n. 1998 , , 294 , 327 rosen , a. , hughes , p. a. , duncan , g. c. , & hardee , p. e. 1999 , , 516 , 729 shaffer , d. b. , marscher , a. p. , marcaide , j. , & romney , j. d. 1987 , , 314 , l1 thompson , k. 1986 , j. fluid mech . , 171 , 365 tingay , s. j. , et al . 1995 , , 374 , 141 valtonen , m. j. , & lehto , h. j. 1997 , , 481 , l5 van putten , m. h. p. m. 1993 , , 408 , l21 van putten , m. h. p. m. 1994 , int . j. bif . & chaos , 4 , 57 van putten , m. h. p. m. 1996 , , 467 , l57 wardle , j. f. c. , cawthorne , t. v. , roberts , d. h. , & brown , l. f. 1994 , , 437 , 122 wen , l. , panaitescu , a. , & laguna , p. 1997 , , 486 , 919 wilson , m. j. 1987 , , 226 , 447 zensus , j. a. 1997 , , 35 , 607

we consider the exponential matrix representing the dynamics of the fermi - bose model in an undepleted bosonic field approximation . a recent application of this model is molecular dimers dissociating into its atomic compounds . the problem is solved in spatial dimensions by dividing the system matrix into blocks with generalizations of hankel matrices , here refered to as-block - hankel matrices . the method is practically useful for treating large systems , i.e. dense computational grids or higher spatial dimensions , either on a single standard computer or a cluster . in particular the results can be used for studies of three - dimensional physical systems of arbitrary geometry . we illustrate the generality of our approach by giving numerical results for the dynamics of glauber type atomic pair correlation functions for a non - isotropic three - dimensional harmonically trapped molecular bose - einstein condensate . department of mathematics , technical university of denmark , 2800 kgs . lyngby , denmark . +center for mathematical sciences , lund university , box 118 , 22100 lund , sweden .

1211.1806

the fermi - bose model under study here forms the underlying basis for a range of phenomena in condensed matter and ultra - cold atomic physics . it was proposed in the context of high - temperature superconductivity by friedberg and lee @xcite , and in ultracold gases it corresponds to the theory of resonance superfluidity with feshbach molecules @xcite . the latter forms the basis of a model for describing the physics of the bcs - bec crossover @xcite . more recently , the fermion - boson model has been used for analyzing the decay of double occupancies ( doublons ) @xcite in a driven fermi - hubbard system @xcite . the particular situation that we concentrate on in this article corresponds to spontaneous dissociation of a bose - einstein condensate of molecular dimers into fermionic atoms @xcite . this process represents a fermionic counterpart of parametric down - conversion in quantum optics . after recent experimental achivements of molecular dissociation @xcite , our aim here is to take the theory for numerical modeling of dissociation into fermionic atoms from the state of principally possible to the state of being useful in practice . let us here in brief present the three - wave interaction type hamiltonian of interest in this work @xcite . @xmath3 here all the quadratic terms are collected in @xmath4 and contains kinetic- and potential energy terms , @xmath5 stands for a bosonic field operator , whereas @xmath6 ( @xmath7 ) describe two particle fields that can be two fermions ( bosons ) in different spin states , finally @xmath8 is the strength of the fermion - boson ( boson - boson ) coupling term . in modern condensed matter physics the fermi - bose model have two major areas of applicability . first the so called `` s - channel '' model in high - temperature superconductivity @xcite . in this context it model the formation dynamics of bosonic cooper - pairs , ( 400,60 ) ( 185,32)(1,0)60 ( 135,32)(-1,0)60 ( 245,28)(-1,0)60 ( 75,28)(1,0)60 ( 60,30 ) ( 260,30 ) ( 149,30 ) ( 171,30 ) ( 157,24)~ ( 60,13)(0,1)34 ( 260,47)(0,-1)34 ( 149,13)(0,1)34 ( 171,47)(0,-1)34 ( 130,1)cooper - pair ( 55,1)@xmath9 ( 255,1)@xmath10 where the two atomic particles , the electrons , are fermions . in the field of ultra - cold atomic physics , it can model the dissociation of ultra - cold bosonic molecules @xcite , ( 400,60 ) ( 180,32)(1,0)65 ( 140,32)(-1,0)65 ( 245,28)(-1,0)65 ( 75,28)(1,0)65 ( 60,30 ) ( 260,30 ) ( 155,30 ) ( 165,30 ) ( 160,30 ) ( 60,13)(0,1)34 ( 260,47)(0,-1)34 ( 155,13)(0,1)34 ( 165,47)(0,-1)34 ( 135,1)molecule ( 42,1)atom 1 ( 242,1)atom 2 and hence we allow here for the atomic particles to be either two fermions or two bosons . various generalizations of time - dependent dmrg @xcite to higher dimensional systems is a topic of large present interest in the computational physics community @xcite , but have not yet reached a useful status for large higher dimensional systems as of interest here . methods for bosonic evolution based on stochastic differential equations ( sdes ) , with the ability for independent stochastic trajectories to be carried out on computer clusters for large systems , are succesful in many situations @xcite but are restricted to simulations of molecules dissociating into bosonic atoms . in this article we outline a method to study the fermionic time evolution for effective heisenberg equations that are linear in creation and annihilation operators . we apply the method to the evaluation of analytic short - time asymptots for the glauber s second order correlation functions @xcite , for a non - isotropic three - dimensional molecular condensate dissociating into fermionic atoms , against numerical data . we focus first on a general formulation in a @xmath0-dimensional cartesian momentum base , such that convenient numerical solutions can be directly obtained for an arbitrary shaped bosonic field . however , for a specific application the numerical performance may be further increased by using additional geometrical symmetries or by formulating the equations in a different basis . except for generating valuable results in certain physical regimes , the method presented here can also be useful as a reference for validation of approximate analytic results or to evaluate more advanced numerical methods , such as for example the gaussian fermionic phase - space representation ( gpsr ) that have recently been applied to the fermi - bose model with a uniform bosonic field @xcite and an implementation of gpsr for dissociation from a non - uniform molecular bec is in progress . turning off stochastic terms in the sdes of the gpsr , one obtain in effect the so called pairing mean - field theory ( pmft ) @xcite . furthermore , keeping the molecular variables in pmft undepleted will be equivalent to the formulation in the present article but is numerically less suitable . * * there is an obvious computational advantage to solve for the single operator dynamics , instead of solving for pairs of operators , the latter is done e.g. in the pmft ( and gpsr ) discussed above . to give one relevant example , to simulate a three - dimensional non - uniform field on a cartesian momentum grid of size @xmath11 requires in effect , as we will show in this paper , only to be able to store part of a ( sparse ) @xmath12 @xmath0-block - hankel matrix of the size @xmath13 and to multiply this @xmath0-block - hankel matrix with vectors only for each final time - point of interest . on the other hand c - number based mean - field methods for fermi - bose systems like pmft , where the basic variables represents pairs of operators with two indices , requires in this case to propagate @xmath14 variables in time through sufficiently many small time - steps up to the final time - point of interest . the methods for stochastic evolution that represents single bosonic field operators with complex stochastic fields mentioned above , see @xcite for a recent review , do not have any direct corresponding useful method for fermions . for example the gpsr that can treat the quantum dynamics of the fermi - bose model exact @xcite , involves basis elements that represents pairs of single operators @xcite , hence it is generally more restrictive in the size of the computational grid . due to the existence of several related methods for bosonic dynamics , we focus in the present article in particular on applications to systems with fermionic atomic operators . however , the specific analogue theory of two distinguishable bosonic atomic operators is also presented for comparison , since its formulation differ only with a sign . advantages with the method presented here are _ i ) _ to be able to study effects of quantum statistics [ by chosing @xmath15 ( @xmath16 ) for fermionic ( bosonic ) atomic particles ] ; _ ii ) _ to obtain accurate correlation functions for short times i.e. for a small number of atomic particles , where stochastic evolution methods have a low signal to noise ratio ; and _ iii ) _ to conviniently treat systems with a moderate size ( relative to the ram memory ) of the computational grid ( e.g. lower dimensional systems ) , where deterministic values of the observables are obtained fast even on a single standard pc . the article is organized as follows . section [ sec : the - momentum - space - operator ] provides the heisenberg equation of motion to be treated within the concept of molecular dissociation into fermionic ( bosonic ) atoms and also briefly mention dissociation into two indistinguishable bosonic atoms and the related problem of condensate collision . in section [ sec : non - uniform - molecular - field ] we relate blocks in the system matrix responsible for atom - molecule coupling to the so called @xmath0-block - hankel matrices . in section [ sub : a - theorem - for ] we formulate and prove results for the structure of the solution of heisenberg equation of motion . section [ sub : obtaining - the - observables ] provides the reader with the practical details of how to use the mathematical results in obtaining physical observables . we apply the general method to a three - dimensional non - isotropic harmonically trapped bose - einstein condensate in section [ sec : results ] and present numerical results for atomic correlation functions that we compare to recently derived analytic short - time asymptotes . finally the article is summarised in section [ sec : summary ] . to model the dissociations of a bose - einstein condensate of diatomic molecules into pairs of constituent atoms , we start with the following effective quantum field theory hamiltonian @xmath17 @xmath18 here we assume that the molecules are made of either two distinguishable bosonic atoms or two fermionic atoms in different spin states . in both cases , @xmath19 is a bosonic field operator for the molecules , satisfying the standard commutation relations @xmath20=\delta^{d}(\mathbf{x}-\mathbf{x}^{\prime})$ ] , with @xmath0 being the spatial dimension of the system @xcite . the atomic field operators , @xmath21 ( @xmath7 ) , satisfy either bosonic commutation or fermionic anti - commutation relations , @xmath22=\delta_{ij}\delta^{d}(\mathbf{x}-\mathbf{x}^{\prime})$ ] and @xmath23=[\widehat{\psi}_{i}(\mathbf{x},t),\widehat{\psi}_{j}(\mathbf{x}^{\prime},t)]=0 $ ] or @xmath24 and @xmath25 , depending on the underlying statistics @xcite . the first term in the hamiltonian ( [ hamiltonian ] ) describes the kinetic energy where the atomic masses are @xmath26 and @xmath27 , whereas the molecular mass is @xmath28 . for simplicity , we will consider the case of equal atomic masses ( same isotope atoms ) , with @xmath29 and @xmath30 . the coupling constant @xmath31 is responsible for coherent conversion of molecules into atom pairs , e.g. via optical raman transitions , an rf transition , or a feshbach resonance sweep and microscopic expressions for @xmath8 can be found in @xcite and references therein . the detuning @xmath32 is defined to give the overall energy mismatch @xmath33 between the free two - atom state in the dissociation threshold and the bound molecular state ( including the relative frequencies of the raman lasers or the frequency of the rf field , again see @xcite and references therein for details ) . unstable molecules , spontaneously dissociating into pairs of constituent atoms , correspond to @xmath34 , with @xmath35 being the total dissociation energy . in what follows we will treat the dissociation dynamics in the undepleted molecular condensate approximation in which the molecules are represented as a fixed classical field . the approximation is valid for short enough dissociation times during which the converted fraction of molecules does not exceed about @xmath36 @xcite . in this regime the dissociation typically produces low density atomic clouds for which the atom - atom @xmath37-wave scattering interactions are negligible . for dissociation into bosonic atoms , also the effects of atom - atom @xmath37-wave scattering have been investigated in @xcite , and the validity of the undepleted molecular condensate approximation was found to still hold for a converted fraction of molecules up to about 5% . even though 5%10% conversion efficiencies seem small , nevertheless they can produce mesoscopic ensembles of pair - correlated atoms with interesting quantum statistics and nontrivial many - body correlations if one starts with large - enough molecular condensates , such as containing at least @xmath38@xmath39 molecules . further on , for dissociation into fermionic atoms in the regime of pauli - blocking , where the atomic occupation numbers are strictly limited by the number of available states @xcite , the undepleted field approximation is expected to be accurate even for large times @xcite . in addition to the issue about temporal depletion , the atom - molecule interactions will initially appear as an effective spatially dependent detuning due to the mean - field interaction energy @xcite . to neglect this effect can be motivated by operating at relatively large absolute values of the dissociation detuning @xmath40 so that it dominates the mean - field energy shift @xcite . the trapping potential for preparing the initial molecular bec with any residual atoms being removed is omitted from the hamiltonian ( [ hamiltonian ] ) since we assume that once the dissociation is invoked , the trapping potential is switched off , so that the dynamics of dissociation is taking place in free @xmath0-dimensional space . we assume that the switching on of the atom - molecule coupling @xmath8 and switching off of the trapping potential is done in a sudden jump at time zero . accordingly the preparation stage is reduced to assuming a certain initial state of the molecular bec in a trap , after which the dynamics is governed by the hamiltonian ( [ hamiltonian ] ) . from the heisenberg equation of motion with the hamiltonian taken from ( [ hamiltonian ] ) , we have for the three field operators @xmath41,\ j=0,\:1,\:2.\label{eq : heisenberg_eq_for_j_1_2_3}\ ] ] in order to obtain linear operator equations the undepleted molecular field approximation is first invoked as follows . assuming that the molecules are in a coherent state initially , the density profile @xmath42 is in principal given by the ground state solution of the standard gross - pitaevskii equation . we then replace the molecular field operator by its coherent mean - field complex function @xcite , the so called condensate wave - function , @xmath43 from ( [ hamiltonian ] ) , ( [ eq : heisenberg_eq_for_j_1_2_3 ] ) and ( [ apa ] ) we then write down the heisenberg equations for the remaining two coupled atomic field operators as follows @xmath44\widehat{\psi}_{1}\left(\mathbf{x},t\right)+q\chi\sqrt{\rho_{0}\left(\mathbf{x}\right)}\exp\left(i\theta\left(\mathbf{x}\right)\right)\widehat{\psi}_{2}^{\dag}\left(\mathbf{x},t\right),\label{eq : heisenberg - eqs}\ ] ] @xmath45\widehat{\psi}_{2}^{\dagger}\left(\mathbf{x},t\right)+\chi\sqrt{\rho_{0}\left(\mathbf{x}\right)}\exp\left(-i\theta\left(\mathbf{x}\right)\right)\widehat{\psi}_{1}\left(\mathbf{x},t\right).\label{eq : heisenberg - eqs_lower}\ ] ] the sign given by @xmath46 in the second term in ( [ eq : heisenberg - eqs ] ) is @xmath15 for fermionic and @xmath16 for bosonic atoms throughout the paper , as a consequence of different operator ( anti- ) commutator relations . multiplying ( [ eq : heisenberg - eqs ] ) and ( [ eq : heisenberg - eqs_lower ] ) with @xmath47 , where @xmath48 is the quantization volume and we assume for simplicity @xmath49 to be the spatial length of the system in any direction , followed by integration over @xmath50 , we can interpret the differential equations for a given @xmath51 in terms of the fourier operators in momentum space @xmath52 the operators @xmath53 satisfy the usual ( commutation- ) anti - commutation relations @xmath54_{-q}=\delta_{ij}\delta_{\mathbf{k},\mathbf{k}^{\prime}}$ ] and @xmath55_{-q}=[\widehat{a}_{\mathbf{k},i},\widehat{a}_{\mathbf{k}',j}]_{-q}=0 $ ] ( i.e. for @xmath15 , @xmath56_{+1}\equiv\left\ { \:,\:\right\ } $ ] ) . since the effective hamiltonian corresponding to ( [ eq : heisenberg - eqs])-([eq : heisenberg - eqs_lower ] ) is quadratic in the field operators , higher - order moments or expectation values of products of creation and annihilation operators will factorize according to wick s theorem into products of the normal and anomalous densities @xmath57@xmath58 and @xmath59 . upon applying the fourier transform , the equations ( [ eq : heisenberg - eqs ] ) and ( [ eq : heisenberg - eqs_lower ] ) become @xmath60 @xmath61 where @xmath62 denotes the complex conjugate . the kinetic part is @xmath63 . the effective atom - molecule coupling constant is @xmath64 @xcite . finally the complex fourier coefficients @xmath65 of the condensate wave function describing the molecular mean - field in momentum - space is defined analogue to ( [ def : fouireroperator ] ) @xmath66 for a real condensate wavefunction we have @xmath67 , while a non - zero phase - function @xmath68 can be used to represent for example an initial vortex state as in @xcite . we give the general theory for a static complex condensate wave - function @xmath69 in the following , while we for the numerical example in section [ sec : results ] choose @xmath70 to be real . for the case where the molecular density is uniform and constant @xmath71 , there is only one non - zero fourier coefficient and only operators with index @xmath51 and @xmath72 couples in ( [ eq : matrixapphe_first_eq])-([eq : matrixapphe ] ) . finally , if the molecular density is non - uniform and time dependent @xmath73 , we need to determine the fourier coefficients ( [ deffouriercoefficients ] ) for each time - step in an iterative process where the dynamics of the molecular mean - field is taken into account . this is a topic for future work , however , the major effects to be taken into account are : expansion of the condensate @xcite ; temporal depletion of the number of molecules @xcite ; and finally the spatial dependences , such as a larger local dissociation rate in the points of highest initial molecular density @xcite . defining such an iterative process , the results of the present article will be applicable in each time - step of the evolution . when the molecular condensate consists of pairs of indistinguishable bosonic atoms of a single spin - state , for example @xmath74rb@xmath75 @xcite , heisenberg equation for one bosonic operator describe the dissociation dynamics @xcite . a similar set of heisenberg equations for bosonic operators of one spin - state can also be formulated for the dynamics of condensate collisions within the time - dependent bogoliubov approach @xcite . hence , also for this problem the method discussed in the present article can be applied @xcite . for a reference we start with presenting the relevant results for a size - matched uniform system . size - matched here means that the spatial dimensions of the uniform bosonic field @xmath76 , with the same ( central- ) particle density @xmath71 , are chosen such that the initial number of molecules are the same as for the non - uniform system of interest , hence @xmath77 . under the condition of a uniform molecular field , i.e. that do not depend on the spatial coordinates @xmath78 , equations ( [ eq : matrixapphe_first_eq])-([eq : matrixapphe ] ) with initial vacuum states for the atoms have analytical solutions for the normal- and anomalous atomic moments @xmath79 respectively @xmath80 . the related ( pmft ) complex differential equations with initial conditions @xmath81 are @xcite @xmath82 @xmath83 where @xmath84 . the corresponding solutions to ( [ eq : pmfteqsys1])-([eq : pmfteqsys2 ] ) , calculated explicitly in section [ sub : comparison - with - the ] , are @xmath85 @xmath86 @xmath87 note that for bosons ( @xmath16 ) , e.g. , the resonance mode ( @xmath88 ) leads to a bose - enhancement effect in the atomic occupations , described by @xmath89 which grows exponentially with time , consequently this illustrate that the undepleted molecular field approximation ( [ apa ] ) is only realistic for a short time . in contrast to this , for fermionic atoms , the atomic occupations undergo sinusoidal oscillations and can be kept to a small fraction of the number of molecules also for large times . as noted in @xcite the moments ( [ eq : uniform_nk ] ) and ( [ eq : uniform_mk ] ) fulfills the equality @xmath90 for a uniform molecular field . we here give the theoretical framework needed for the main analytic results of the article , which is given in the next section , for how to efficiently solve the fermi - bose model for a non - uniform molecular field . in particular we show how the system matrix for the dynamics of the fermi - bose model of a physical system in @xmath0 spatial dimensions can be classified in terms of generalizations of hankel matrices . in cases where the shape of the molecular condensate @xmath91 posseses certain geometrical symmetrices , such as spherical symmetry , a reduction of the number of atomic creation and annihilation operators can be implemented . alternatively , a base with atomic operators that are directly defined e.g. in a spherical coordinate system can be used @xcite . in this article , however , we treat the case of a general dense cubic lattice . in practice , the set of indices @xmath51 ( and @xmath92 ) lie on a finite grid in @xmath93 of the form @xmath94 , where @xmath95 and @xmath96 for all @xmath97 . by abuse of notation , we will often suppress the factor @xmath98 and identify @xmath51 with @xmath99 , i.e. we will write e.g. @xmath100 in place of @xmath101 wherever convenient . whenever @xmath99 or @xmath102 appears it will be implicitly understood that it stays within the above limitations . to set the scene , we first discuss a system in one spatial dimension . for a non - uniform system with @xmath103 , set @xmath104 and identify the systems of annihilation operators @xmath105 and creation operators @xmath106 with the @xmath107 dimensional column vector @xmath108^{t}$ ] . under this identification , heisenberg equations ( [ eq : matrixapphe_first_eq])-([eq : matrixapphe ] ) can then be visualized in terms of a @xmath109-system - matrix @xmath110 composed of four blocks of @xmath111-matrices @xmath112=\left[\begin{array}{cc } a_{11 } & a_{12}\\ a_{21 } & a_{22}\end{array}\right]\left[\begin{array}{l } \widehat{a}_{-k,1}\\ \vdots\\ \widehat{a}_{k,1}\\ \widehat{a}_{-k,2}^{\dagger}\\ \vdots\\ \widehat{a}_{k,2}^{\dagger}\end{array}\right].\label{odeinmatrixform}\ ] ] it then follows directly from ( [ eq : matrixapphe ] ) that the coupling matrices @xmath113 and @xmath114 become hankel matrices . a hankel matrix have the following structure @xmath115 , i.e. the elements are identical when the sum of the row and column indices are constant @xcite . for @xmath116 , the summation operator that takes the system of creation operators @xmath117 to the system of annihilation operators @xmath118 is defined via @xmath119 we call this a @xmath0-dimensional finite hankel operator . these have been studied e.g. in @xcite for @xmath120 . obviously , there are multiple ways to represent the two systems @xmath121 and @xmath118 as a @xmath122dimensional vector . however , with this done , the hankel structure of the corresponding coupling matrix @xmath113 is lost . in the next section we will construct such a concrete representation in which @xmath113 turns out to be a @xmath0-block - hankel matrix . recall that @xmath104 . the following function @xmath123 is a one - to - one mapping from the @xmath0-dimensional lattice @xmath124 to @xmath125 . the inverse is @xmath126 note that the equations ( [ hindexmapd ] ) and ( [ hinversindexmapd ] ) are really nothing else than a change of base from @xmath127 to @xmath128 for integer numbers . we will later also use the following property of the map above @xmath129 we now identify the two lattices of operators @xmath130 and @xmath131 with the corresponding column vector @xmath132^{t}$ ] , where we use the simplified indices @xmath133 instead of @xmath134 . then the @xmath0-dimensional system ( [ eq : matrixapphe_first_eq])-([eq : matrixapphe ] ) can be represented in matrix form analogue to ( [ odeinmatrixform ] ) as @xmath135=\left[\begin{array}{cc } a_{11 } & a_{12}\\ a_{21 } & a_{22}\end{array}\right]\left[\begin{array}{l } \widehat{a}_{1,1}\\ \vdots\\ \widehat{a}_{b^{d},1}\\ \widehat{a}_{1,2}^{\dagger}\\ \vdots\\ \widehat{a}_{b^{d},2}^{\dagger}\end{array}\right].\label{odeinmatrixform1}\ ] ] clearly , @xmath136 and @xmath137 are diagonal matrices whereas @xmath113 corresponds to the @xmath0-dimensional hankel operator ( [ eq : defhankelsuperoperator ] ) and @xmath114 to the lower counterpart in ( [ eq : matrixapphe ] ) . as the matrices , @xmath113 and @xmath114 are not hankel matrices for @xmath116 , but rather exhibit a block - hankel structure @xcite , we will refer to them as @xmath0-block - hankel matrices . in the coming two sections we discuss properties of the four block matrices @xmath138 and @xmath137 . we will use the following notation ; @xmath139 denotes transpose , @xmath62 is complex conjugation , and @xmath140 represents both the two previous operations combined . moreover , the operation of transposing in the skew - diagonal will be denoted @xmath141 , i.e. @xmath142 or equivalently @xmath143 where @xmath144 denotes the skew - diagonal identity , i.e. the matrix obtained by reversing the order of the columns of the identity matrix @xmath145 . from ( [ eq : neweq54 ] ) and the property @xmath146 , it also follow that _ _ the skew - diagonal transpose of the product of two general matrices @xmath147 and @xmath148 fulfills _ _ @xmath149 which will be used later . finally , the skew - diagonal transpose combined with complex conjugation will be denoted @xmath150 . inspection of ( [ eq : matrixapphe_first_eq])-([eq : matrixapphe ] ) and ( [ odeinmatrixform1 ] ) shows that @xmath113 is given elementwise by @xmath151 if we denote the coordinates of @xmath152 by a sup - index @xmath153 , those of @xmath154 with a @xmath155 and the coordinates for the fourier coefficients by @xmath156 , then the coordinates for the rows and columns of @xmath113 fulfills @xmath157 which considerably simplify any practical implementation of @xmath113 . note also that when @xmath65 is defined on the same @xmath51-lattice as the atomic operators , we necessarily have @xmath158 when @xmath159 , see figure [ fig : d - block_hankel_matrix ] for an illustration . similarly to ( [ eq45 ] ) , we have @xmath160 such that ( @xmath161 ) @xmath162 due to this identity we are satisfied with discussing properties of @xmath113 in the remainder . first we note that @xmath163 which is immediate by ( [ eq45 ] ) . to get additional structure , we impose extra conditions on the condensate wave - function @xmath164 . for many physical applications , @xmath165 can be chosen real . in this case , we note that the fourier coefficients ( [ deffouriercoefficients ] ) have the following symmetry @xmath166 . therefore we get from ( [ eq : firstpropertyofmap ] ) @xmath167 and from ( [ eq : generalrelationfora21 ] ) that @xmath168 furthermore , for the physically important case of a condensate wave - function that is real and even , i.e. with @xmath169 , such as for example for a condensate in a harmonic trap , we have that @xmath65 is real so @xmath170 , which combined with ( [ eq : a12sdhequala12 ] ) implies that @xmath171 and from ( [ eq : generalrelationfora21 ] ) that @xmath172 finally , note that in the real uniform case , @xmath173 , we have non - zero entries only at positions where @xmath174 , i.e. @xmath175 , as in section [ sub : uniform dimensional - mbec ] . according to the heisenberg equations ( [ eq : matrixapphe_first_eq])-([eq : matrixapphe ] ) and the representation ( [ odeinmatrixform1 ] ) , the matrices @xmath136 and @xmath137 become diagonal ( kinetic energy in momentum space ) , where the @xmath176:th diagonal element of @xmath136 is given by @xmath177 and similarly @xmath178 finally , we note that @xmath179 which follows directly from ( [ eq : firstpropertyofmap ] ) , since @xmath180 . much of the structure of the system - matrix @xmath110 is preserved in matrix functions defined on @xmath110 , which can be used to reduce the computational complexity when evaluating @xmath181 for the system s evolution in time . naturally , the more conditions we impose on the condensate wave - function @xmath182 in ( [ deffouriercoefficients ] ) , the more structure is preserved . we present the corresponding identities in order of increasing symmetry on @xmath182 , starting with a general complex condensate wave - function . throughout we will let @xmath183 and prove the results for the cases of fermionic ( @xmath15 ) and bosonic ( @xmath16 ) atoms simultaneously . given an arbitrary square ( even sized ) @xmath184-matrix @xmath127 we decompose it into its four blocks @xmath185.\ ] ] below we list a number of useful matrix identities which all can be verified by direct computation @xmath186\left[\begin{array}{cc } b_{11 } & b_{12}\\ b_{21 } & b_{22}\end{array}\right]\left[\begin{array}{cc } 0 & \mathbb{i}\\ q\mathbb{i } & 0\end{array}\right]=\left[\begin{array}{cc } b_{22 } & qb_{21}\\ qb_{12 } & b_{11}\end{array}\right],\label{eq1}\ ] ] @xmath187^{\dagger}=\left[\begin{array}{cc } b_{11}^{\dagger } & b_{21}^{\dagger}\\ b_{12}^{\dagger } & b_{22}^{\dagger}\end{array}\right],\label{eq3}\ ] ] @xmath187^{sdt}=\left[\begin{array}{cc } b_{22}^{sdt } & b_{12}^{sdt}\\ b_{21}^{sdt } & b_{11}^{sdt}\end{array}\right].\label{eq4}\ ] ] with these three identities at hand , the following lemmas are all immediate . for example the first one is a direct application of ( [ eq1 ] ) alone . _ the matrix identities _ _ @xmath188 are equivalent to @xmath189b^{*}\left[\begin{array}{cc } 0 & \mathbb{i}\\ q\mathbb{i } & 0\end{array}\right].\label{eq6'}\ ] ] _ the second lemma follows by combining ( [ eq1 ] ) with ( [ eq4 ] ) . _ the matrix identities _ _ @xmath190 are equivalent to @xmath189b^{sdt}\left[\begin{array}{cc } 0 & \mathbb{i}\\ q\mathbb{i } & 0\end{array}\right].\label{eq5'}\ ] ] _ finally combining ( [ eq1 ] ) with ( [ eq3 ] ) we obtain the third lemma . _ the matrix identities @xmath191 are equivalent to @xmath192b^{\dagger}\left[\begin{array}{cc } 0 & \mathbb{i}\\ \mathbb{i } & 0\end{array}\right].\label{eq7'}\ ] ] _ _ let @xmath127 be the system - matrix @xmath110 constructed in section [ constrcoupling ] from the condensate wave - function @xmath182 . then the identities in ( [ eq6 ] ) of lemma [ l1 ] are always satisfied . moreover , if @xmath182 is real then the identities in ( [ eq5 ] ) of lemma [ l2 ] hold and if @xmath182 is real and even , the identities in ( [ eq7 ] ) of lemma [ l3 ] hold . _ _ proof . _ the first claim follows by combining ( [ eq : generalrelationfora21 ] ) and ( [ eq : a22equala11 * ] ) , and the second follows by combining ( [ eq : a21qa12sdt ] ) and ( [ eq : a11sdtequala11 ] ) . when @xmath182 is also even the elements of @xmath113 are real , and hence ( [ eq : a12transposea12 ] ) can be written @xmath193 ( analogously @xmath194 ) . since clearly @xmath195 , the last claim is established as well . @xmath196 we need some preparation for the main theorem and corollaries , which basically says that the above identities are preserved when forming @xmath181 . first we note that @xmath197\left[\begin{array}{cc } 0 & q\mathbb{i}\\ \mathbb{i } & 0\end{array}\right]=\left[\begin{array}{cc } \mathbb{i } & 0\\ 0 & q^{2}\mathbb{i}\end{array}\right]=\left[\begin{array}{cc } \mathbb{i } & 0\\ 0 & \mathbb{i}\end{array}\right].\label{eq2}\ ] ] we now recall that an analytic function @xmath198 defined on all of @xmath199 is called an entire function , and that for each convergence radius @xmath200 one can find a constant @xmath201 such that @xmath202 , @xcite . this allows us to define the function @xmath203 for any matrix @xmath127 ( with the corresponding matrix norm @xmath204 ) , since @xmath205 which ( picking any @xmath206 ) shows us that @xmath207 is a cauchy sequence in the set of matrices @xmath208 with the operator norm @xmath204 . the limit is thus a well defined matrix since this is a banach space . we recall that @xmath204 is defined as @xmath209 and in particular that for any given index @xmath210 we have @xmath211 we are now ready for the main result . _ let @xmath198 be an entire function and let @xmath110 be a matrix that satisfies either of the identities in ( [ eq6 ] ) , ( [ eq5 ] ) or ( [ eq7 ] ) . then @xmath212 satisfies the same identities . _ _ proof . _ let us suppose that @xmath110 satisfies the identities of ( [ eq5 ] ) . by lemma [ l2 ] we then have ( [ eq5 ] ) , which combined with ( [ eq8 ] ) and ( [ eq2 ] ) , yields that @xmath213(a^{k})^{sdt}\left[\begin{array}{cc } 0 & \mathbb{i}\\ q\mathbb{i } & 0\end{array}\right],\ ] ] for any @xmath214 . thus the matrix identities in ( [ eq5 ] ) are satisfied for the corresponding blocks of @xmath215 when @xmath216 is a monomial . since ( [ eq5 ] ) are also preserved upon taking linear combinations of matrices that all satisfy ( [ eq5 ] ) , we conclude that ( [ eq5 ] ) holds whenever @xmath216 is a polynomial . finally , since in the general case we have @xmath217 in the operator norm , and since the identities ( [ eq5 ] ) are preserved upon taking limits with respect to this norm [ see ( [ eq9 ] ) ] , the general result follows . the proofs related to ( [ eq6 ] ) and ( [ eq7 ] ) are analogue . @xmath196 * remark * : connecting to proposition [ p1 ] , we note that when @xmath127 arise from a general complex condensate wave - function , one still has more structure than what is expressed in ( [ eq6 ] ) . for example @xmath218 which is easily seen to be equivalent to @xmath219b^{t}\left[\begin{array}{cc } 0 & -\mathbb{i}\\ \mathbb{i } & 0\end{array}\right].\ ] ] however , since @xmath220\left[\begin{array}{cc } 0 & -\mathbb{i}\\ \mathbb{i } & 0\end{array}\right]=\left[\begin{array}{cc } -\mathbb{i } & 0\\ 0 & -\mathbb{i}\end{array}\right],\ ] ] these properties are _ not _ preserved in @xmath203 , except for in the special case when @xmath216 is an odd function . we now sum up our conclusions for the physically most interesting cases . _ given a real condensate wave - function @xmath182 and @xmath221 , set @xmath222 then @xmath223 has the structure _ @xmath224,\label{eq : corollary1relationsform}\ ] ] _ where in addition we have for the two blocks _ @xmath225 _ proof . _ we can see that @xmath223 has the above structure if and only if it satisfies the identities in ( [ eq6 ] ) and ( [ eq5 ] ) of lemmas [ l1 ] and [ l2 ] . moreover , @xmath110 satisfies these identities by proposition [ p1 ] . the desired conclusion thus follows by theorem [ t1 ] . @xmath196 it is clear from corollary [ c1 ] that we only need to calculate half of the matrices @xmath226 and @xmath227 in order to fully determine @xmath223 , which generally reduces the computational cost from @xmath228 to @xmath229 elements in this case . _ suppose that @xmath182 is a real and even condensate wave - function . then , in addition to the identities in corollary [ c1 ] , we have _ @xmath230 _ proof . _ by proposition [ p1 ] @xmath110 satisfies the identities in ( [ eq7 ] ) of lemma [ l3 ] , and hence so does @xmath223 by theorem [ t1 ] . it is easy to see that these identities combined with the structure in ( [ eq : corollary1relationsform ] ) and ( [ eq : corollary1relationsforblocks ] ) proven in corollary [ c1 ] implies that the above identities are satisfied for @xmath223 . @xmath196 from corollary [ c2 ] follows that we only need to calculate a quarter of the matrices @xmath226 and @xmath227 in order to fully determine @xmath223 , which further reduces the computational cost to @xmath231 elements in this case . in this section we show how to use the results of the previous section in obtaining physical observables for the atoms . we start from the following general block form of the solution @xmath232 to heisenberg equations ( [ eq : matrixapphe_first_eq])-([eq : matrixapphe ] ) in matrix form ( [ odeinmatrixform1 ] ) @xmath233=\left[\begin{array}{cc } m_{11 } & qm_{12}\\ m_{21 } & m_{22}\end{array}\right]\left[\begin{array}{l } \widehat{a}_{1,1}\left(0\right)\\ \vdots\\ \widehat{a}_{b^{d},1}\left(0\right)\\ \widehat{a}_{1,2}^{\dagger}\left(0\right)\\ \vdots\\ \widehat{a}_{b^{d},2}^{\dagger}\left(0\right)\end{array}\right],\label{matrixsolutiontoheisenberg}\ ] ] it is obvious that the results of the previous section will simplify the practical calculations of @xmath234 and hence the physical observables . however , first we show in the next section how to generally obtain first - order moments for pairs of atomic operators directly from ( [ matrixsolutiontoheisenberg ] ) . we now denote by @xmath235 the @xmath176-row - vector of the block matrix @xmath236 , where @xmath237 is mapped to @xmath238 according to section [ sub : ordering - the - lattice ] . as a first example , for an annihilation operator of the @xmath239 spin - state in row @xmath176 in the left hand side of ( [ matrixsolutiontoheisenberg ] ) we have @xmath240 where we for notational and computational convenience introduce the two operators @xmath241,\:\:\hat{v}\equiv\left[\begin{array}{l } \widehat{a}_{1,2}(0),\ldots,\widehat{a}_{b^{d},2}(0)\end{array}\right]^{\dagger}=\left[\begin{array}{l } \widehat{a}_{1,2}^{\dagger}\left(0\right)\\ \vdots\\ \widehat{a}_{b^{d},2}^{\dagger}\left(0\right)\end{array}\right],\ ] ] which are naturally constructed from ( [ matrixsolutiontoheisenberg ] ) . with @xmath242 and @xmath243 we have introduced in effect a calculus , where only terms in the expectation values containing the matrix @xmath244 or @xmath245 give a contribution , while all other terms are zero . this is due to the ( anti- ) commutator relations applied to an initial vacuum state for the atoms . see the examples leading to ( [ eq : matrixappoomkkprim ] ) and ( [ eq : matrixappoonkkprim1 ] ) below for further details . with these rules , different vacuum expectation values of atomic operator pairs can conveniently be written down in terms of standard matrix products between complex row- and column - vectors , where both the vectors are defined from certain rows in one of the four block matrices @xmath236 . we now give the corresponding expression to ( [ eq : matrixappaf1op ] ) for an annihilation operator of the @xmath246 spin - state @xmath247 note that analogously to ( [ eq : matrixappaf1op ] ) , we can also alternatively write ( [ eq : matrixappooak2 ] ) on the form @xmath248 , and for a specific moment one chose the option that allow the operators to meet in the middle of the operator pair . from ( [ eq : matrixappaf1op ] ) and ( [ eq : matrixappooak2 ] ) we can then demonstrate in detail how to calculate the anomalous moments , @xmath249 @xmath250 @xmath251 @xmath252 @xmath253 which is a time - dependent complex number as expected . in a similar way , using the hermitian conjugate of ( [ eq : matrixappaf1op ] ) , @xmath254 we have for the normal moments of the @xmath255 spin - state @xmath256 @xmath257 from ( [ eq : matrixappooak2 ] ) it follows that the corresponding result to ( [ eq : matrixappoonkkprim1 ] ) for the @xmath258 spin - state is @xmath259 @xmath260 we finally confirm by direct calculations from ( [ eq : matrixappooak2 ] ) and ( [ eq : adaggerk1 ] ) that @xmath261 as expected from ( [ eq : matrixappoomkkprim ] ) . any other first - order moment that can be formed by two atomic operators , such as for example @xmath262 , can be shown to be zero . we now note that with @xmath263 in ( [ eq : matrixappoonkkprim1 ] ) we have @xmath264 which is a real number as expected from the physical interpretation of the occupation of spin-@xmath265 atoms in state @xmath51 . motivated by ( [ eq : matrixappoonkkprim1 ] ) and ( [ eq : matrixappoonkkprim2 ] ) we see that the fact that the two spin - states are analogue in the fermi - bose model ( [ eq : hinspatialcoordinates ] ) , with initial atomic vacuum in both spin - states , physically supports the identity @xmath266 given in corollary [ c1 ] . hence , applying corollary [ c1 ] it follows from ( [ eq : matrixappoonkkprim1 ] ) and ( [ eq : matrixappoonkkprim2 ] ) that we can write for arbitrary spin @xmath267 similarly the anomalous moments from ( [ eq : matrixappoomkkprim ] ) then take the form @xmath268 in terms of only the blocks @xmath226 and @xmath227 . in this section we will deduce the solutions ( [ eq : uniform_nk ] ) and ( [ eq : uniform_mk ] ) direct from the exponentialmatrix in order to check the formalism presented . we start by rewriting ( [ eq : defexpat ] ) according to _ @xmath269 _ _ _ then starting from the system matrix for a uniform molecular field with _ _ @xmath84 @xmath270,\ ] ] we first note , using @xmath271 , that the even powers of @xmath110 are diagonal , hence we have @xmath272.\label{eq : firstpartofmsum}\ ] ] secondly , we see by writing the odd powers as @xmath273 that @xmath274 @xmath275.\label{eq : secondpartofmsum}\ ] ] finally , we can from ( [ eq : rewritingmin2sums ] ) , ( [ eq : firstpartofmsum ] ) and ( [ eq : secondpartofmsum ] ) write up @xmath223 on the form of ( [ matrixsolutiontoheisenberg ] ) @xmath276 from ( [ eq : analyticmforuniformcase ] ) we can then via ( [ eq : matrixappoonkkprimsigma ] ) and ( [ eq : mkkprimfinal ] ) retrieve ( [ eq : uniform_nk ] ) and ( [ eq : uniform_mk ] ) of section [ sub : uniform dimensional - mbec ] . however , it is important to stress that the ( skew- ) diagonal blockmatrices of @xmath223 lose this structure for a non - uniform condensate wave - function , and in the general case it follows by cauchy - schwarz that the equality ( [ eq : matrixappuniformsolutions ] ) changes into the following inequality @xmath277 the above inequality have been investigated numerically for the case of dissociation into bosonic atoms in @xcite and it implies limitations on the strength of various correlations for non - uniform systems @xcite . higher - order moments , such as in the simplest case , the combination of two pairs of operators , are factorized according to wick s theorem @xcite * * which is implicit from the decorrelation assumption in use within the undepleted molecular field approximation here . as an example we calculate glauber s correlation function for two atoms in the same spin - state @xcite @xmath278 for a numerical implementation of ( [ eq : matrixappoog2kkprim ] ) we have , according to ( [ eq : matrixappoonkkprimsigma ] ) , @xmath279 we here also give glauber s correlation function for two atoms in opposite spin - states @xmath280 for a numerical implementation of ( [ eq : defbbcorr ] ) we have , according to ( [ eq : matrixappoonkkprimsigma ] ) and ( [ eq : mkkprimfinal ] ) , @xmath281 up to this point , we have shown how to reduce the computational needs for obtaining physical observables to the calculation of only a fraction of the full exponential matrix @xmath232 . we now discuss how we perform the necessary numerical calculations for the physical observables in practice , while the results are presented in the next section . the matrix @xmath181 can be calculated numerically in many different ways , see for example @xcite for a review of methods . for example , the two most obvious ones are : _ i ) _ use the definition in ( [ eq : defexpat ] ) and calculate all powers until some cut - off in @xmath214 , @xmath282 , or @xmath283 if @xmath110 is nilpotent ; _ ii ) _ diagonalize @xmath110 and use all its eigenvalues ( @xmath284 ) and eigenvectors ( @xmath285 ) to change to the @xmath286-basis , such that @xmath287 , where @xmath288 . both the natural methods _ i ) _ and _ ii ) _ in practice needs many matrix - matrix operations , and have computationally expensive performance for large matrices . as explained in section [ sub : first - order - atomic - moments ] , all observables are obtained as matrix products between different row- and column - vectors defined from rows of the blocks of @xmath223 . it is obvious that row @xmath153 of @xmath223 is obtained by multiplying @xmath289 with the unit vector @xmath290^{t}$ ] from the right . for this task , powerful matrix - free algorithms exists for general matrices @xcite , i.e. that calculate the results of an exponential matrix acting on an arbitrary vector , without the need to perform any matrix - matrix operations . using the results of the corollaries in section [ sub : a - theorem - for ] , the fact that @xmath136 is diagonal and hence can be represented as a vector , and the powerful property of @xmath113 being a @xmath0-block - hankel matrix , we use our own modified version of the open access software expokit for sparse matrices @xcite . as will be reported elsewhere , we have optimized the algorithm from @xcite for the implementation of @xmath0-block - hankel matrices . this software optimization , combined with the use of the results in section [ sub : a - theorem - for ] , give us the crucial advantages necessary for implementating large fermi - bose systems , compared to any brute force calculation of the full exponential matrix . while the complex matrix @xmath223 , or any of its blocks , is generally not sparse , the matrix @xmath113 is sparse especially after truncation of the smallest fourier coefficients , see figure [ fig : d - block_hankel_matrix ] of the next session for an example . clearly any such actual truncation have to be evaluated with convergence tests . overall , our optimized exponentiation procedure in effect reduces the original problem of calculating the complex ( non - sparse ) @xmath184-matrix @xmath232 to the simpler problem of performing matrix - vector operations with the @xmath0-block - hankel matrix @xmath291 , which in addition obey further symmetries , see e.g. ( [ eq : a12transposea12 ] ) and ( [ eq : a12sdhequala12 ] ) . furthermore , @xmath113 is real for the physically important case of a condensate wave - function that is even , which reduces the information to store in @xmath113 by an additional factor of two in this case . finally , let us stress the important consequence of section [ sub : first - order - atomic - moments ] , that the calculation of each row in the blocks @xmath236 is independent , such that the work can conviniently be distributed across several computers in parallel to reduce computation time if needed . this is in principle a very crucial advantage of the presented formalism , when applied to large systems . we here present specific numerical results for atomic correlation functions for a non - isotropic three - dimensional system . we use a @xmath292 grid in momentum space , which corresponds to a system of linear differential operator equations with a @xmath184-system matrix where @xmath293 , i.e. in general @xmath294 matrix elements . the theorem presented in section [ sub : a - theorem - for ] in combination with the algorithm improvements briefly discussed in section [ sec : numerical - calculations - of ] allow systems of this size to be solved on one standard pc in the order of @xmath295 hour . as pointed out in section [ sec : numerical - calculations - of ] , substantially larger grids can be attacked with calculations on several computers in parallell . -block - hankel structure for the actual @xmath113 block - matrix used in the numerical example with @xmath296 and @xmath297 reported in this section . figure ( a ) illustrates the entire @xmath298 @xmath113 matrix , while ( b ) , ( c ) and ( d ) shows zoomed in regions corresponding to the circle of the previous subfigure . note that in ( b ) each dot represents a @xmath299-block - hankel matrix , while in ( c ) and ( d ) each dot represent a single non - zero matrix element . dashed ( dotted ) lines in ( a ) shows the skew - diagonal ( diagonal ) which corresponds to the skew - diagonal transpose ( sdt ) ( respectivelly sdh for complex matrix elements ) and transpose symmetries discussed in section [ sub : structure - of - the - d - block - hankel - matrices ] . however , let us stress that the symmetry of a @xmath0-block - hankel matrix is considerable higher , since the block - matrices in ( a ) and ( b ) [ elements in ( d ) ] are repeated downwards parallell to the respective skew - diagonals . grey triangular regions in ( a ) and ( d ) shows where the conditions @xmath159 for @xmath300 respectivelly @xmath301 are fulfilled . as discussed in section [ sub : structure - of - the - d - block - hankel - matrices ] , these regions does not contain any non - zero matrix elements due to the restricted fourier coefficients lattice size . the fact that the regions of non - zero elements do not reach out to the grey fields are due to the truncation implemented , see text for details . [ fig : d - block_hankel_matrix ] ] in order to use a realistic set of parameters for our numerical example , we chose an harmonic trap with frequences such that the size of the molecular field along the @xmath302-direction @xmath303 m is two times the size along the @xmath304-direction @xmath305 m , while the size along the @xmath306-direction @xmath307 m is set to an intermediate value . with a central molecular peak density of @xmath308m@xmath309 , this corresponds to @xmath310 molecules . choosing @xmath311 dimers @xcite we have an atomic mass of @xmath312 kg . the molecule - atom dissociation parameter is @xmath313m@xmath314/s here @xcite . we set the dissociation detuning to @xmath315s@xmath316 which is large enough to ensure that the dissociation energy is larger than thermal excitations at nk temperatures ( i.e. @xmath317 ) . with a characteristic time of @xmath318ms this corresponds to a dimensionless detuning of @xmath319 @xcite . the momentum lattice in use have a spacing @xmath320m@xmath316 which is smaller than the smallest width of the molecular momentum distribution @xmath321 @xcite , and have been confirmed numerically to resolve the dynamics of the relevant structures in the atomic momentum distribution . the corresponding resonance momenta is then @xmath322 . for small times we can , for the purpose of validating the physical parameters , estimate the number of atoms by the following linear expression in time @xmath323 from the above formula it is clear that the number of atoms increase with @xmath324 for a three - dimensional system . this have earlier been studied explicitly for a uniform system , see fig . 1 of @xcite . hence for cases with large detuning the validity of the undepleted field approximation is limited to short dissociation times @xmath325 . for the parameters of section [ sub : parameters - of - the ] we have @xmath326 which results in @xmath327 atoms at @xmath328 , and hence a conversion ratio of less than 1% , ensuring the validity of the results from the undepleted field approximation @xcite . the presented estimate of atom numbers from the fermi s golden rule ( [ eq : fermi - golden - rule ] ) was later confirmed by the numerical calculations for the parameter values in use here . , at times @xmath329 ( @xmath318ms ) , calculated along different directions in 3d . in ( a ) we show numerical results ( solid thin curves ) from ( [ eq : matrixappoog2kkprimnum ] ) along the direction @xmath330 , i.e. with @xmath331 and @xmath332 ( @xmath333m@xmath316 ) , while in ( b ) we show the corresponding result along the direction @xmath334 . the analytic short - time asymptotes of ( [ eq : g_jj^2_ana ] ) are represented by the fat curves plotted at @xmath335 only . in fact the short - time asymptotes for the collinear correlations are in qualitative agreement with the numerical results even up to @xmath336 . however , a zoom in reveals quantitative deviations seen as a narrowing of the width of the correlation signal with time , this is also in agreement with detailed 1d results reported in fig . 8 of @xcite . in general the fermionic collinear correlations are here showing a pauli - blocking dip at @xmath337 , while the characteristic width of the correlation signal have been confirmed to be inverselly proportional to the size of the molecular bec source along the corresponding direction , i.e. @xmath338m@xmath316 in ( a ) and @xmath339 in ( b ) . in addition we compared with the corresponding results for bosonic atoms showing a so called hanbury - brown and twiss peak at @xmath337 ( dashed curves ) , shown only for the largest time here.[fig : atom - atom - correlation - functions],title="fig : " ] , at times @xmath329 ( @xmath318ms ) , calculated along different directions in 3d . in ( a ) we show numerical results ( solid thin curves ) from ( [ eq : matrixappoog2kkprimnum ] ) along the direction @xmath330 , i.e. with @xmath331 and @xmath332 ( @xmath333m@xmath316 ) , while in ( b ) we show the corresponding result along the direction @xmath334 . the analytic short - time asymptotes of ( [ eq : g_jj^2_ana ] ) are represented by the fat curves plotted at @xmath335 only . in fact the short - time asymptotes for the collinear correlations are in qualitative agreement with the numerical results even up to @xmath336 . however , a zoom in reveals quantitative deviations seen as a narrowing of the width of the correlation signal with time , this is also in agreement with detailed 1d results reported in fig . 8 of @xcite . in general the fermionic collinear correlations are here showing a pauli - blocking dip at @xmath337 , while the characteristic width of the correlation signal have been confirmed to be inverselly proportional to the size of the molecular bec source along the corresponding direction , i.e. @xmath338m@xmath316 in ( a ) and @xmath339 in ( b ) . in addition we compared with the corresponding results for bosonic atoms showing a so called hanbury - brown and twiss peak at @xmath337 ( dashed curves ) , shown only for the largest time here.[fig : atom - atom - correlation - functions],title="fig : " ] we here explicitly illustrate the @xmath0-block - hankel matrix @xmath113 that is used in the numerical calculations of a physical system for @xmath296 here . hence , it is evident from the figures [ fig : d - block_hankel_matrix ] ( a ) , ( b ) and ( c ) that we can zoom in @xmath296 times on @xmath113 and reveal a repeating pattern . clearly after the last zoom in [ figure [ fig : d - block_hankel_matrix ] ( d ) ] , we are left with a structure of a usual hankel matrix . we have used a truncation of the molecular bec source such that fourier coefficients with a modulus less than 2% of the leading coefficient is neglected . this procedure have been evaluated by reconstruction of the bec by the inverse fourier transform . it was also found that the 2% level of truncation resulted in correlation functions ( see figure [ fig : atom - atom - correlation - functions ] ) that could not be distingushable by the eye from the correlation functions obtained with a 4% level of truncation . for @xmath296 the thomas - fermi ( tf ) density profile of the molcular bec is given by @xmath340 for @xmath341 [ and @xmath342 otherwise ] , which is underlying an analytic derivation of the asymptotes . here @xmath343 is the thomas - fermi radius along the spatial direction @xmath344 . we are here interested in collinear ( cl ) density correlations between two momentum components at @xmath51 and @xmath345 , for which the displacement @xmath346 is along one of the cartesian coordinates , @xmath347 . the detailed derivation of short - time asymptotes for the correlation functions in this case was reported in @xcite . the cl correlations following from this derivation is @xmath348^{2}}{\left[(k_{j}-k_{j}^{\prime})r_{\mathrm{tf},j}\right]^{5}},\label{eq : g_jj^2_ana}\ ] ] where @xmath349 denotes bessel functions of the first kind . the qualitative behavior of the cl correlation functions are similar as in lower dimensions @xcite , whereas the quantitative differences enter e.g. through the width and the peak values . the widths of ( [ eq : g_jj^2_ana ] ) is @xmath350 and the peak value of ( [ eq : g_jj^2_ana ] ) is static to leading order @xcite . in figure [ fig : atom - atom - correlation - functions ] we show results for the evaluation of the analytic short - time asymptote ( [ eq : g_jj^2_ana ] ) , strictly valid in the @xmath325 limit , against numerical results for times @xmath351 , where @xmath328 roughly corresponds to the first maximum in time of the oscillating fermionic atom numbers @xmath352 . we have described how to effectivelly calculate the dynamics of linear heisenberg operator equations for the fermi - bose model applied to the problem of molecular dissociation . we note that a similar framework have been used to obtain numerical results for a non - isotropic 2d system on a @xmath353 grid in @xcite . we have here generalized the approch to @xmath0 spatial dimensions with the use of @xmath0-block - hankel matrices . in particular we have explicitly explored a non - isotropic 3d system on a @xmath292 grid numerically on a standard pc . such a grid can resolve relevant atom dynamics in momentum space for realistic parameters @xcite , and naturally extends earlier studies of non - uniform 1d and 2d systems @xcite , and is more realistic than previous treatments of uniform 3d systems @xcite . we finally stress that the results presented can be used to handle a complex bosonic mean - field of any geometry in any spatial dimension . we thank karn kheruntsyan and roger sidje for valuable discussions at an early stage , and johnny kvistholm for artistic assistance with figure 1 .

a one - parameter generalisation of the symmetric macdonald polynomials and interpolations macdonald polynomials is studied from the point of view of branching rules . we establish a pieri formula , evaluation symmetry , principal specialisation formula and-difference equation for . we also prove a new multiple-gauss summation formula and several further results for basic hypergeometric series based on .

0903.3996

let @xmath3 be a partition , i.e. , @xmath4 is a weakly decreasing sequence of nonnegative integers such that @xmath5 is finite . let the length @xmath6 of @xmath3 be the number of nonzero @xmath7 . for @xmath8 and @xmath9 the schur function @xmath10 is defined as @xmath11 where @xmath12 is the vandermonde product . if @xmath13 then @xmath14 . from its definition it is clear that @xmath10 is a symmetric polynomial in @xmath15 of homogeneous degree @xmath16 , and that @xmath17 forms a basis of the ring of symmetric functions @xmath18^{{\mathfrak{s}}_n}$ ] . a classical result for schur functions is the combinatorial formula @xmath19 here the sum is over all semi - standard young tableau @xmath20 of shape @xmath3 , and @xmath21 is shorthand for the monomial @xmath22 with @xmath23 the number of squares of the tableau filled with the number @xmath24 . one of the remarkable facts of is that it actually yields a symmetric function . the conventional way to view a semi - standard young tableau of shape @xmath3 ( and length at most @xmath25 ) as a filling of a young diagram with the numbers @xmath26 such that squares are strictly increasing along columns and weakly increasing along rows . given two partitions ( or young diagrams ) @xmath27 write @xmath28 if @xmath29 and @xmath30 is a horizontal strip , i.e. , if the skew diagram @xmath30 contains at most one square in each column . then an alternative viewpoint is to consider a young tableau of shape @xmath3 as a sequence of partitions @xmath31 where @xmath32 denotes the empty partition . for example , for @xmath33 the tableau @xmath34 may be encoded as @xmath35 the above description implies that a recursive formulation of the schur functions , equivalent to the combinatorial formula , is given by the _ branching rule _ @xmath36 subject to the initial condition @xmath37 . if we let @xmath29 be a pair of partitions and define the skew schur function @xmath38 of a single variable @xmath39 as @xmath40 then the branching rule for schur functions takes the more familiar form @xmath41 the macdonald polynomials @xmath42 @xcite are an important @xmath43-generalisation of the schur functions , and the @xmath44 for @xmath9 form a basis of the ring @xmath45 , where @xmath46 . a classical result in the theory is that the macdonald polynomials satisfy a combinatorial formula not unlike that of the schur functions ; @xmath47 where @xmath48 is a function that admits an explicit combinatorial description . importantly , if @xmath20 has no more than @xmath25 rows it factorises as @xmath49 where , as before , @xmath50 is the sequence of partitions representing @xmath20 . probably the simplest ( albeit non - combinatorial ) expression for @xmath51 is ( * ? ? ? * page 342 ) @xmath52 where @xmath53 with @xmath54 . ( note that @xmath55 since @xmath28 . ) it follows from the above that the macdonald polynomials , like the schur functions , can be described by a simple branching rule . namely , @xmath56 subject to the initial condition @xmath57 . again we may define a single - variable skew polynomial @xmath58 for @xmath29 @xmath59 to turn the branching formula for the macdonald polynomials into @xmath60 in view of the above two examples of symmetric functions admitting a recursive description in the form of a branching formula , a natural question is _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ can one find more general branching - type formulas that lead to symmetric functions ? _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ to fully appreciate the question we should point out that it is not at all obvious that if one were to take ( or , equivalently , ) as the definition of the schur functions or ( or ) as the definition of the macdonald polynomials , that the polynomials in question are symmetric in @xmath15 . assuming throughout that @xmath61 let the ( generalised ) @xmath1-shifted factorials be defined as follows : @xmath62 and let @xmath63 and @xmath64 . then probably the best - known example of a branching rule generalising and resulting in symmetric polynomials is @xmath65 the @xmath66 are the interpolation macdonald polynomials of knop , okounkov and sahi @xcite , and is ( * ? ? ? * proposition 5.3 ) . for comparison with @xcite , we have @xmath67 from it is clear that the top - homogeneous component of @xmath68 is the macdonald polynomial @xmath69 so that @xmath70 forms an inhomogeneous basis of @xmath71 . for @xmath72 and @xmath73 such that @xmath30 is a horizontal strip and such that @xmath74 denote by @xmath75 and @xmath76 the partitions @xmath77 and @xmath78 . note that @xmath79 is again a horizontal strip . it follows from that @xmath80 for @xmath74 . it also follows from and that @xmath81 if we replace @xmath82 in and then change @xmath83 and @xmath84 the branching rule for the interpolation macdonald polynomials may thus be recast as @xmath85 in this paper we consider a generalisation @xmath86 of the macdonald polynomials and the macdonald interpolation polynomials defined recursively by the branching rule @xmath87 our interest in the functions @xmath88 is not merely that they provide another example of a class of symmetric functions defined by a simple branching formula . indeed , it may be shown that the more general @xmath89 is also symmetric . moreover @xmath90 is a limiting case ( reducing bc@xmath91 symmetry to @xmath92 symmetry and breaking ellipticity ) of rains bc@xmath91 symmetric abelian interpolation functions @xcite so that @xmath88 is not actually new . however , it turns out that the function @xmath88 has a number of interesting properties , not shared by @xmath90 or the more general bc@xmath91 abelian functions . for example , @xmath88 satisfies a pieri formula that not just implies the standard pieri formulas for macdonald polynomials , but also gives an @xmath2 generalisation of the famous @xmath1-gauss summation formula . specifically , with @xmath93 a suitable normalisation of the functions @xmath88 defined in equation , and @xmath94 an arbitrary finite alphabet , we prove that @xmath95 for @xmath96 this simplifies to the well - known @xmath1-gauss sum @xmath97 we begin with a remark about notation . if @xmath98 is a symmetric function we will often write @xmath99 with @xmath100 ( and refer to @xmath94 as an alphabet ) instead of @xmath101 with @xmath8 , the latter notation being reserved for function that are not ( a priori ) symmetric . following this notation we also use @xmath102 where @xmath103 denotes the ( disjoint ) union of the alphabets @xmath94 and @xmath104 , and @xmath105 where @xmath106 denotes the alphabet @xmath94 with the single letter @xmath39 added . in the following we review some of the basics of macdonald polynomial theory , most of which can be found in @xcite . let @xmath107 be the @xmath1-shift operator acting on the variable @xmath108 : @xmath109 then the macdonald polynomials @xmath110 for @xmath100 are the unique polynomial eigenfunctions of the macdonald operator @xmath111}(-c)^{{\lverti\rvert}}t^{\binom{{\lverti\rvert}}{2 } } \prod_{\substack{i\in i \\ j\not\in i } } \frac{tx_i - x_j}{x_i - x_j } \prod_{i\in i } t_{q , x_i},\ ] ] where @xmath112:=\{1,\dots , n\}$ ] . explicitly , @xmath113 for later reference we state the coefficient of @xmath114 of this equation separately ; if @xmath115 then @xmath116 for each square @xmath117 in the ( young ) diagram of a partition ( i.e. , for each @xmath118 and @xmath119 , the arm - length @xmath120 , arm - colength @xmath121 , leg - length @xmath122 and leg - colength @xmath123 are given by @xmath124 and @xmath125 where @xmath126 is the conjugate of @xmath3 , obtained by reflecting the diagram of @xmath3 in the main diagonal . note that the generalised @xmath1-shifted factorial can be expressed in terms of the colengths as @xmath127 with the above notation we define the further @xmath1-shifted factorials @xmath128 , @xmath129 and @xmath130 as @xmath131 and @xmath132 then the macdonald polynomials @xmath133 are defined as @xmath134 we also need the skew macdonald polynomials @xmath135 and @xmath136 defined for @xmath29 by [ pxya ] @xmath137 note that @xmath138 and @xmath139 , and that @xmath140 . to simplify some later equations it will be useful to extend the definitions of @xmath135 and @xmath136 to all partition pairs @xmath27 by setting @xmath141 if @xmath142 . from , and it follows that for @xmath143 a scalar , [ hom ] @xmath144 where @xmath145 . for subsequent purposes it will be convenient to also introduce normalised ( skew ) macdonald polynomials @xmath146 and @xmath147 as [ norm ] @xmath148 { \mathsf{q}}_{{\lambda}/\mu}(x)&=t^{n(\mu)-n({\lambda } ) } \frac{c'_{{\lambda}}}{c'_{\mu}}\ , q_{{\lambda}/\mu}(x)= t^{n(\mu)-n({\lambda } ) } \frac{c_{{\lambda}}}{c_{\mu}}\ , p_{{\lambda}/\mu}(x ) , \label{skewqp}\end{aligned}\ ] ] where @xmath149 note that no additional factors arise in the normalised form of : @xmath150 and that @xmath151 and @xmath152 . if we define the structure constants @xmath153 by @xmath154 then @xmath155 with @xmath156 the @xmath43-littlewood richardson coefficients . below we make use of some limited @xmath3-ring notation ( see @xcite for more details ) . let @xmath157 be the @xmath158th power - sum symmetric function @xmath159 and let @xmath160 . then the @xmath161 form a basis of @xmath162 where @xmath163 . given a symmetric function @xmath99 we define @xmath164:=\phi_{a , b}(f),\ ] ] where @xmath165 is the evaluation homomorphism given by @xmath166 in particular , @xmath167=f(1,t,\dots , t^{n-1})$ ] is known as the principal specialisation , which we will also denote as @xmath168 , and @xmath169=f(1,t , t^2,\dots)$ ] . from ( * equation ( 6.24 ) ) @xmath170=a^{{\lvert{\lambda}-\mu\rvert } } \sum_{\nu}(b / a)_{\nu}{\mathsf{f}}_{\mu\nu}^{{\lambda}},\ ] ] which , by @xmath171 , also implies that @xmath172=(a)_{{\lambda}}{\quad\text{and}\quad}{\mathsf{p}}_{{\lambda}}\biggl[\frac{1-a}{1-t}\biggr ] = t^{2n({\lambda } ) } \frac{(a)_{{\lambda}}}{c_{{\lambda}}c'_{{\lambda}}}.\ ] ] this last equation of yields the well - known principal specialisation formula @xmath173 the cauchy identity for ( skew ) macdonald polynomials is given by @xmath174 the product on the right - hand side may alternatively be expressed in terms of the power - sum symmetric functions as @xmath175 it thus follows from and , as well as some elementary manipulations , that application of @xmath176 ( acting on @xmath104 ) turns the cauchy identity into @xmath177 { \mathsf{p}}_{{\lambda}/\mu}(x)= \biggl(\:\prod_{x\in x } \frac{(cx)_{\infty}}{(ax)_{\infty } } \biggr ) \sum_{{\lambda}}{\mathsf{q}}_{\mu/{\lambda}}\biggl[\frac{a - c}{1-t}\biggr ] { \mathsf{p}}_{\nu/{\lambda}}(x).\ ] ] for @xmath178 ( followed by the substitution @xmath179 and then @xmath180 ) this is the @xmath1-binomial identity for macdonald polynomials @xcite @xmath181 for later reference we also state the more general @xmath182 instance of @xmath183{\mathsf{p}}_{{\lambda}}(x).\ ] ] for reasons outlined below we will refer to this as a pieri formula . let @xmath184 and @xmath185 be defined by @xmath186 ( for combinatorial expressions for all of @xmath187 and @xmath188 , see @xcite ) . further let @xmath189 and @xmath190 the @xmath158th elementary symmetric function . then the macdonald polynomials @xmath191 satisfy the pieri formulas [ pierir ] @xmath192 now observe that for @xmath193 yields @xmath194 whereas for @xmath195 it yields @xmath196{\mathsf{p}}_{{\lambda}}(x).\ ] ] since @xmath197 we therefore have @xmath198p_{{\lambda}}(x),\end{aligned}\ ] ] where we have also used . identifying @xmath199 0 & \text{otherwise } \end{cases}\ ] ] and @xmath200= \begin{cases}\displaystyle ( -1)^{{\lvert{\lambda}-\mu\rvert } } \psi'_{{\lambda}/\mu } & \text{if $ \mu'\preccurlyeq{\lambda}'$ } \\[1 mm ] 0 & \text{otherwise } \end{cases}\ ] ] these two formulas are equivalent to the pieri rules of . the skew polynomials can be used to define generalised @xmath1-binomial coefficients @xcite as @xmath201}{0pt}{}{{\lambda}}{\mu}}= { \genfrac{[}{]}{0pt}{}{{\lambda}}{\mu}}_{q , t}:= { \mathsf{q}}_{{\lambda}/\mu}\biggl[\frac{1}{1-t}\biggr].\ ] ] in particular has @xmath202}{0pt}{}{{\lambda}}{\mu}}=0 $ ] if @xmath142 and @xmath203}{0pt}{}{(m)}{(k)}}=\prod_{i=1}^k \frac{1-q^{i+m - k}}{1-q^i}={\genfrac{[}{]}{0pt}{}{m}{k}}\ ] ] with on the right the classical @xmath1-binomial coefficients @xmath202}{0pt}{}{m}{k}}= { \genfrac{[}{]}{0pt}{}{m}{k}}_q$ ] . if @xmath204 then @xcite @xmath205}{0pt}{}{{\lambda}}{{\lambda}_{(i)}}}=\frac{c'_{{\lambda}}}{c'_{{\lambda}_{(i)}}}\ , \psi'_{{\lambda}/{\lambda}_{(i)}}.\ ] ] this , together with ( * * thorme 9 , bis ) @xmath206}{0pt}{}{{\lambda}}{\mu } } = ( 1-q)\sum_{i=1}^n q^{-{\lambda}_i}t^{i - n } { \genfrac{[}{]}{0pt}{}{{\lambda}}{{\lambda}_{(i ) } } } { \genfrac{[}{]}{0pt}{}{{\lambda}_{(i)}}{\mu}},\ ] ] where @xmath207 provides a simple recursive method to compute the generalised @xmath1-binomial coefficients . [ lemlas ] assume that @xmath208 . then @xmath209}{0pt}{}{{\lambda}}{\mu}}&= \frac{s_{\mu}({\langle 0\rangle})}{s_{{\lambda}}({\langle 0\rangle } ) } \det_{1\leq i , j\leq n}\biggl ( { \genfrac{[}{]}{0pt}{}{{\lambda}_i+n - i}{\mu_j+n - j}}\biggr ) & & \text{if $ t = q$ } \\ \intertext{and } \label{tis1 } { \genfrac{[}{]}{0pt}{}{{\lambda}}{\mu}}&=\sum_{u^{+}=\mu } \prod_{i=1}^n { \genfrac{[}{]}{0pt}{}{{\lambda}_i}{u_i } } & & \text{if $ t=1$.}\end{aligned}\ ] ] in the above @xmath210 denotes a sum over compositions @xmath211 in the @xmath92 orbit of @xmath212 . assume that @xmath213 . then simplifies to @xmath214}{0pt}{}{{\lambda}}{\mu}}=q^{n(\mu)-n({\lambda})}\ , \frac{c_{{\lambda}}}{c_{\mu}}\ , s_{{\lambda}/\mu}\biggl[\frac{1}{1-q}\biggr],\ ] ] where we have also used and the fact that for @xmath213 the ( skew ) macdonald polynomials reduce to the ( skew ) schur functions . let @xmath215 be the @xmath158th complete symmetric function . by application of the jacobi trudi identity ( * ? ? ? * equation ( i.5.4 ) ) @xmath216 and the principal specialisation formula @xcite @xmath217 the generalised @xmath1-binomial coefficient can be expressed as a determinant @xmath203}{0pt}{}{{\lambda}}{\mu}}= \frac{s_{\mu}({\langle 0\rangle})}{s_{{\lambda}}({\langle 0\rangle})}\ , \det_{1\leq i , j\leq n } \biggl(\frac{(q;q)_{{\lambda}_i+n - i}}{(q , q)_{\mu_j+n - j}}\ , h_{{\lambda}_i-\mu_j - i+j}\biggl[\frac{1}{1-q}\biggr]\biggr).\ ] ] since @xmath218=\frac{1}{(q)_r}\ ] ] this establishes the first claim . the second claim follows in analogous manner . since making the substitution @xmath219 in the right - hand side of is somewhat problematic it is best to first use the symmetry ( * ? ? ? * equation ( 2.12 ) ) @xmath220}{0pt}{}{{\lambda}}{\mu}}_{q , t}={\genfrac{[}{]}{0pt}{}{{\lambda}'}{\mu'}}_{t^{-1},q^{-1}}.\ ] ] since @xmath221 we therefore get @xmath203}{0pt}{}{{\lambda}}{\mu}}_{q^{-1},1}= { \genfrac{[}{]}{0pt}{}{{\lambda}'}{\mu'}}_{1,q } = q^{n(\mu')-n({\lambda}')}\ , \frac{c_{{\lambda}'}(1,q)}{c_{\mu'}(1,q)}\ , e_{{\lambda}/\mu}\biggl[\frac{1}{1-q}\biggr].\ ] ] using @xmath222 and @xmath223=\frac{q^{\binom{r}{2}}}{(q;q)_r}\ ] ] as well as @xmath224 , it follows that @xmath203}{0pt}{}{{\lambda}}{\mu}}_{q^{-1},1}= \sum_{u^{+}=\mu } \prod_{i=1}^n q^{-{\lambda}_i u_i}{\genfrac{[}{]}{0pt}{}{{\lambda}_i}{u_i}}.\ ] ] finally replacing @xmath225 yields the second claim . in this section we consider the question posed in the introduction : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ can one find new ( ? ) branching - type formulas , similar to , and , that lead to symmetric functions ? _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ assume that @xmath226 is a fixed nonnegative integer , and let @xmath227 denote a finite sequence of parameters . then we are looking for branching coefficients @xmath228 such that @xmath229 subject to the initial condition @xmath230 defines a symmetric function . in the above @xmath231 . of course , for @xmath232 combined with implies that @xmath233 if one wishes to only consider symmetric functions with the standard property @xmath234 then the additional condition @xmath235 must be imposed . because we assume the branching coefficients to be independent of @xmath25 , it may perhaps seem we are excluding interesting classes of symmetric functions such as the interpolation macdonald polynomials . as will be shown shortly , assuming @xmath25-independence is not actually a restriction , and may easily be recovered as a special case of . now let us assume that yields a symmetric function @xmath236 for all @xmath237 . ( for @xmath238 and @xmath239 this is obviously not an assumption . ) then , @xmath240 is a symmetric function in @xmath241 ( for @xmath237 ) . for it to also be a symmetric function in @xmath242 we must have @xmath243 implying that for fixed @xmath3 @xmath244 where @xmath245 . hence a sufficient condition for to yield a symmetric function is @xmath246 for partitions @xmath247 such that @xmath248 . as a first example let us show how to recover the macdonald interpolation polynomials of the introduction . to this end we take @xmath249 , @xmath250 , and @xmath251 clearly , the resulting polynomials @xmath252 correspond to the interpolation polynomials after the specialisation @xmath253 . to see that is indeed satisfied we substitute the above choice for the branching coefficient ( recall the convention that @xmath254 if @xmath142 ) to obtain @xmath255 the identity is easily proved using rains sears transformation for skew macdonald polynomials ( * ? ? ? * corollary 4.9 ) @xmath256 p_{\mu/\nu}\biggl[\frac{aq / de - a^2q^2/bcde}{1-t}\biggr ] \\ = \sum_{\mu } \frac{(aq / d , aq / e)_{{\lambda}}(b , c)_{\mu}}{(aq / d , aq / e)_{\mu}(b , c)_{\nu } } \ , p_{{\lambda}/\mu}\biggl[\frac{1-aq / bc}{1-t}\biggr ] p_{\mu/\nu}\biggl[\frac{aq / bc - a^2q^2/bcde}{1-t}\biggr].\end{gathered}\ ] ] after simultaneously replacing @xmath257 and taking the @xmath258 limit we obtain . if , more generally , we let @xmath259 in and take the @xmath258 limit we find that @xmath260 p_{\mu/\nu}\biggl[\frac{y - by}{1-t}\biggr ] \\ = \sum_{\mu } \frac{(a / y)_{{\lambda}}(a / bz)_{\mu}}{(a / y)_{\mu}(a / bz)_{\nu}}\ , p_{{\lambda}/\mu}\biggl[\frac{y - by}{1-t}\biggr ] p_{\mu/\nu}\biggl[\frac{z - bz}{1-t}\biggr].\end{gathered}\ ] ] the macdonald interpolation polynomials may thus be generalised by taking @xmath261 , @xmath262 and @xmath263.\ ] ] the polynomials @xmath264 defined by @xmath265 m_{\mu}(x_1,\dots , x_{n-1};a / b , b)\ ] ] subject to @xmath266 are symmetric . moreover , the interpolation macdonald polynomials corresponds to @xmath267 the polynomials @xmath268 are an example of a class of symmetric functions for which @xmath13 does not imply vanishing . for example , @xmath269= t^{n({\lambda } ) } z^{{\lvert{\lambda}\rvert } } \frac{(a / z , b)_{{\lambda}}}{c_{{\lambda}}}.\ ] ] the next example corresponds to okounkov s bc@xmath91 symmetric interpolation polynomials @xcite ( see also @xcite ) . [ propbcn ] if we take take @xmath261 , @xmath270 and @xmath271 in then the resulting functions @xmath272 are symmetric . writing @xmath273 instead of @xmath274 , the ( laurent ) polynomials @xmath273 satisfy the symmetries @xmath275 ( this follows easily using that @xmath276 . ) moreover , okounkov s bc@xmath91 interpolation macdonald polynomials @xmath277 follow as @xmath278 ( since @xmath279 the @xmath273 are not more general than the @xmath277 . ) substituting the claim in and using gives @xmath280 this is with @xmath281 . our final example will ( in the limit ) lead to the functions studied in the remainder of the paper . [ proprab ] if we take @xmath261 , @xmath282 and @xmath283 in then the resulting functions @xmath272 are symmetric . substituting the claim in and using gives @xmath284 this is with @xmath285 . if we write @xmath90 instead of @xmath274 the symmetric functions of proposition [ proprab ] correspond to the functions described by the branching rule of the introduction . as already mentioned there , the @xmath90 are not new , and follow as a special limiting case of much more general functions studied by rains @xcite . more specifically , rains defined a family of abelian interpolation functions @xmath286 where @xmath8 . the @xmath287 are bc@xmath91 symmetric and , apart from parameters @xmath288 , depend on an elliptic nome @xmath289 . in ( * theorem 4.16 ) rains proved the branching rule @xmath290 where the branching coefficient @xmath291 is expressed in terms of the elliptic binomial coefficient @xmath292(v_1,\dots , v_k)}$ ] ( see ( * ? ? ? * equation ( 4.2 ) ) ) as @xmath293(azt^n , at^n / z , pqa / tb)}.\ ] ] if we define @xmath294 and compute the corresponding limit of we obtain the branching rule with @xmath295 . in the remainder of the paper we consider the symmetric function @xmath296 which , alternatively , is defined by the branching rule . because @xmath297 is a limiting case of the abelian interpolation function @xmath298 many properties of former follow by taking appropriate limits in the results of @xcite . for example , it follows from ( * ? ? ? * proposition 3.9 ) that the @xmath0 for @xmath100 satisfy a @xmath1-difference equation generalising . specifically , with @xmath299 the generalised macdonald operator @xmath300}(-1)^{{\lverti\rvert}}t^{\binom{{\lverti\rvert}}{2 } } \prod_{\substack{i\in i \\ j\not\in i } } \frac{tx_i - x_j}{x_i - x_j } \prod_{j\not\in i}(1-bx_j ) \prod_{i\in i } ( c - bt^{1-n}x_i)t_{q , x_i}\ ] ] we have @xmath301 below we will first prove a number of elementary properties of the functions @xmath0 using only the branching rule . like the previous result , most of these can also be obtained by taking appropriate limits in results of rains for the abelian interpolation functions @xmath302 . then we give several deeper results for @xmath0 ( such as theorem [ thmskewrc ] and corollaries [ pieri ] , [ corgauss ] and [ corktw ] ) that , to the best of our knowledge , have no analogues for @xmath302 or @xmath303 . first however we restate the branching rule in the equivalent form @xmath304 where @xmath305 . when @xmath306 we find from that @xmath307 from this it is clear that @xmath308 and that in the @xmath258 limit @xmath309 is given by @xmath310 . it also shows that @xmath311 all three statements easily generalise to arbitrary @xmath94 . [ lembindep ] for @xmath312 a scalar , @xmath313 [ lemclim ] for @xmath312 a scalar and @xmath314 , @xmath315 [ lemreduce ] let @xmath314 and @xmath3 a partition such that @xmath316 . define @xmath317 . then @xmath318 this last result allows the definition of @xmath0 to be extended to all weakly decreasing integer sequences @xmath72 . by all three statements are obviously true for @xmath94 a single - letter alphabet , and we proceed by induction on @xmath25 , the cardinality of @xmath94 . by , @xmath319 using and the appropriate induction hypothesis this yields @xmath320 establishing the first lemma . taking the @xmath258 limit on both sides of and then using induction we get @xmath321 where the last equality follows from . to prove the final lemma we consider with @xmath322 and , in accordance with the conditions of lemma [ lemreduce ] , with @xmath323 . since @xmath324 vanishes unless @xmath325 is a horizontal strip this implies that @xmath326 . the summand also vanishes if @xmath327 so that we we may assume that @xmath328 . defining @xmath329 and @xmath330 and using induction , as well as @xmath331 we get @xmath332 where in the final step we have used and @xmath305 . [ psp ] for @xmath3 such that @xmath9 , @xmath333 by lemma [ lembindep ] this may be stated slightly more generally as @xmath334 iterating using @xmath335 we obtain the generalised branching rule @xmath336 for @xmath337 . when @xmath338 this results in the claim . for @xmath3 such that @xmath9 set @xmath339 then @xmath340 we may view the evaluation symmetry as a rational function identity in @xmath341 . hence it suffices to give a proof for @xmath342 where @xmath343 runs over all integers such that @xmath344 . but ( see and ) @xmath345 and @xmath346 so that we need to prove that @xmath347 making the substitutions @xmath348 , @xmath349 and @xmath350 we get @xmath351 finally , by the principal specialisation formula for the interpolation macdonald polynomials @xcite , @xmath352 so that we end up with @xmath353 this is the known evaluation symmetry of the interpolation macdonald polynomials ( * ? ? ? * section 2 ) . it is clear from that @xmath354 with on the right a macdonald polynomial . the macdonald polynomials in turn generalise the jack polynomials @xmath355 , since @xmath356 . combining the last two equations it thus follows that @xmath357 curiously , there is an alternative path from @xmath0 to the jack polynomials as follows . for @xmath94 an alphabet let @xmath358 [ propjack ] we have @xmath359 let @xmath305 be a finite alphabet . replacing @xmath360 in and taking the @xmath361 limit yields @xmath362 where @xmath363 is a skew jack polynomial and @xmath364 using the homogeneity of @xmath363 the above can be rewritten as @xmath365 comparing this with @xmath366 the proposition follows . probably our most important new results for @xmath0 are generalisation of the skew cauchy identity , the pieri formula and the @xmath1-gauss formula . before we get to these result we first need a few more definitions . first of all , in analogy with , we set @xmath367 so that becomes @xmath368 furthermore , we also define the skew functions @xmath369 by @xmath370 and @xmath371 in other words , @xmath372 and @xmath373 . [ thmskewrc ] let @xmath374 and @xmath94 a finite alphabet . then @xmath375 { \mathsf{r}}_{{\lambda}/\mu}(x;b ) \\ = \biggl(\ : \prod_{x\in x } \frac{(cx , dx)_{\infty}}{(ax , bx)_{\infty } } \biggr ) \sum_{{\lambda } } \frac{(b / c)_{\mu}}{(b / c)_{{\lambda}}}\ , { \mathsf{q}}_{\mu/{\lambda}}\biggl[\frac{a - c}{1-t}\biggr ] { \mathsf{r}}_{\nu/{\lambda}}(x;d).\end{gathered}\ ] ] note that for @xmath376 the theorem simplifies to . we defer the proof of till the end of this section and first list a number of corollaries . [ pieri ] let @xmath374 and @xmath94 a finite alphabet . then @xmath377 { \mathsf{r}}_{{\lambda}}(x;b).\ ] ] this follows from the theorem by taking @xmath378 and then replacing @xmath379 by @xmath212 . when @xmath380 equation yields the pieri formula for macdonald polynomials . when @xmath381 and @xmath382 such that @xmath383 equation yields ( after replacing @xmath384 ) @xmath385}{0pt}{}{{\lambda}}{\mu } } { \mathsf{p}}_{{\lambda}}(x)= { \mathsf{r}}_{\mu}(x;b)\prod_{x\in x } \frac{(bx)_{\infty}}{(x)_{\infty}}.\ ] ] for @xmath378 this is the @xmath1-binomial formula for macdonald polynomials , and for @xmath376 it is lassalle s @xcite @xmath386}{0pt}{}{{\lambda}}{\mu } } { \mathsf{p}}_{{\lambda}}(x)= { \mathsf{p}}_{\mu}(x)\prod_{x\in x } \frac{1}{(x)_{\infty}}.\ ] ] the jack polynomial limit of is of particular interest . to concisely state this we need some more notation . let @xmath387}{0pt}{}{{\lambda}}{\mu}}_{q , q^{1/\alpha}}\ ] ] or , alternatively @xcite , @xmath388 where @xmath25 is any integer such that @xmath389 . further let @xmath390 with @xmath391 , @xmath392 and @xmath393 using all of the above , replacing @xmath394 in by @xmath395 and taking the ( formal ) limit @xmath361 with the aid of proposition [ propjack ] , we arrive at the following identity . for @xmath94 a finite alphabet @xmath396 another special case of worth stating is the following multivariable extension of the @xmath397 summation ( * ? ? ? * ii.5 ) , which follows straightforwardly by taking the @xmath398 limit , @xmath399 { \mathsf{r}}_{{\lambda}}(x;b)= { \mathsf{p}}_{\mu}(x)\prod_{x\in x } \frac{(cx)_{\infty}}{(bx)_{\infty}}.\ ] ] this provides an expansion of the right - hand side different from . if we let @xmath400 in theorem [ thmskewrc ] , use and then replace @xmath401 we obtain @xmath402 for @xmath378 we state this separately . [ corgauss ] for @xmath94 a finite alphabet @xmath403 as mentioned in the introduction , for @xmath96 this simplifies to the standard @xmath1-gauss sum thanks to . more generally , if we principally specialise @xmath404 and use the @xmath2 @xmath1-gauss sum simplifies to kaneko s @xmath1-gauss sum for macdonald polynomials ( * ? ? ? * proposition 5.4 ) @xmath405 as another consequence of the theorem we obtain an explicit expression for the taylor series of @xmath406 in @xmath341 . for @xmath29 let @xmath407 be defined as @xmath408 . that is @xmath409 [ corbtaylor ] we have @xmath410}{0pt}{}{{\lambda}}{\mu } } { \mathsf{p}}_{{\lambda}}(x),\ ] ] or , equivalently , @xmath411 { \mathsf{r}}_{\mu}(x;b)=\sum_{\substack{{\lambda}\supseteq\mu \\ { \lvert{\lambda}-\mu\rvert}=r } } ( 1)_{{\lambda}/\mu}{\genfrac{[}{]}{0pt}{}{{\lambda}}{\mu } } { \mathsf{p}}_{{\lambda}}(x).\ ] ] replacing @xmath412 in lemma [ lembindep ] and expressing the resulting identity in terms of the normalised function @xmath413 ( see ) we get @xmath414 combined with this implies that @xmath415}{0pt}{}{{\lambda}}{\mu } } { \mathsf{p}}_{{\lambda}}(x).\ ] ] the summand vanishes unless @xmath29 and so we may add this as a restriction in the sum over @xmath3 . then the limit @xmath382 limit may be taken without causing ambiguities , and the claim follows . corollary [ corbtaylor ] implies the following simple expressions for @xmath0 when @xmath213 ( schur - like case ) or @xmath219 ( monomial - like case ) . let @xmath100 . then @xmath416 since the two claims are proved in almost identical fashion we only present a proof of . the only significant difference is that the omitted proof of uses instead of . assume that @xmath213 . let @xmath417 and suppose that @xmath418 . since for any @xmath419 , one has the expansion @xmath420}{0pt}{}{k+r-1}{r}},\ ] ] the matrix @xmath421 factorises into the product of rectangular matrices @xmath422}{0pt}{}{k-1}{r - k } } \bigg)_{\substack{r=0,1,\ldots \\ k=\nu_1,\ldots,\nu_n}}.\ ] ] according to cauchy binet theorem , the determinant on the right - hand side of factorises into a sum of products of minors of these two matrices . on the other hand , by and , the expansion gives @xmath423 where we have also used and . using @xmath379 instead of @xmath3 , and @xmath424 , this becomes @xmath425}{0pt}{}{\eta_i-1}{\nu_j-1 } } b^{\eta_i-\nu_j}\biggr ) \det_{1\leq i , j\leq n}\bigl(x_i^{\eta_j}\bigr ) \ , \ ] ] which is precisely the cauchy binet expansion . the restriction @xmath418 is lifted using lemma [ lemreduce ] . recall that the macdonald polynomials are the eigenfunctions of the operator @xmath426 , see . because @xmath427 this can also be stated as @xmath428 where @xmath429 and @xmath430 is given in . a second consequence of corollary [ corbtaylor ] it a generalisation of this identity as follows . let @xmath431 and @xmath432 so that @xmath433 . we have @xmath434 define the operator @xmath435 as @xmath436 combining the @xmath437 instance of ( * ? ? ? * proposition 9 ) with gives @xmath438}{0pt}{}{{\lambda}^{(i)}}{{\lambda } } } { \mathsf{p}}_{{\lambda}^{(i)}}(x),\ ] ] where @xmath439 . since @xmath440 this implies that @xmath441}{0pt}{}{{\lambda}^{(i)}}{{\lambda } } } { \mathsf{p}}_{{\lambda}^{(i)}}(x).\ ] ] by corollary [ corbtaylor ] we can now compute the action of @xmath442 on @xmath413 : @xmath443}{0pt}{}{{\lambda}}{\mu } } { \mathcal{d}}_n^1(b ) { \mathsf{p}}_{{\lambda}}(x ) \\ & = \sum_{{\lambda}\supseteq\mu}b^{{\lvert{\lambda}-\mu\rvert } } ( 1)_{{\lambda}/\mu}{\genfrac{[}{]}{0pt}{}{{\lambda}}{\mu } } \omega_{{\lambda } } { \mathsf{p}}_{{\lambda}}(x ) \\ & \quad-(1-q ) \sum_{i=1}^n \sum_{{\lambda}\supseteq\mu } b^{{\lvert{\lambda}-\mu\rvert}+1 } q^{-{\lambda}_i-1}t^{i - n}(1)_{{\lambda}^{(i)}/\mu } { \genfrac{[}{]}{0pt}{}{{\lambda}^{(i)}}{{\lambda } } } { \genfrac{[}{]}{0pt}{}{{\lambda}}{\mu } } { \mathsf{p}}_{{\lambda}^{(i)}}(x ) \\ & = \sum_{{\lambda}\supseteq\mu}b^{{\lvert{\lambda}-\mu\rvert } } ( 1)_{{\lambda}/\mu}{\genfrac{[}{]}{0pt}{}{{\lambda}}{\mu } } \omega_{{\lambda } } { \mathsf{p}}_{{\lambda}}(x ) \\ & \quad-(1-q ) \sum_{i=1}^n \sum_{{\lambda}\supseteq\mu } b^{{\lvert{\lambda}-\mu\rvert } } q^{-{\lambda}_i}t^{i - n}(1)_{{\lambda}/\mu } { \genfrac{[}{]}{0pt}{}{{\lambda}}{{\lambda}_{(i ) } } } { \genfrac{[}{]}{0pt}{}{{\lambda}_{(i)}}{\mu } } { \mathsf{p}}_{{\lambda}}(x),\end{aligned}\ ] ] where we have also used that @xmath444 for @xmath445 . recalling the recursion , the sum over @xmath24 on the right can be performed to give @xmath446}{0pt}{}{{\lambda}}{\mu } } { \mathsf{p}}_{{\lambda}}(x).\ ] ] again using completes the proof . as a third and final application of corollary [ corbtaylor ] we derive a simple expression for @xmath447 , to be compared with ( obtained for @xmath376 ) or with . we have @xmath448 first we use and , as well as the fact that for @xmath30 a vertical strip ( i.e. , @xmath449 ) @xmath450 to put the proposition in the form @xmath451 { \mathsf{p}}_{{\lambda}}(x).\ ] ] since @xmath452}{0pt}{}{\mu}{\nu}}$ ] is lower - triangular ( with @xmath453 ) , it is invertible ( for the explicit inverse see @xcite ) . since @xmath444 for @xmath445 the above equation is thus equivalent to @xmath454}{0pt}{}{\mu}{\nu } } { \mathsf{p}}_{\mu}(x ) \\ = \biggl(\:\prod_{i=1}^n ( 1-cq^{{\lambda}_i}t^{n - i})\biggr ) \sum_{{\lambda}\supseteq\mu\supseteq\nu } b^{{\lvert{\lambda}-\nu\rvert } } \frac{(cqt^{n-1})_{\mu}}{(cqt^{n-1})_{{\lambda}}}\,(1)_{{\lambda}/\nu } { \genfrac{[}{]}{0pt}{}{\mu}{\nu } } { \mathsf{q}}_{{\lambda}/\mu}\biggl[\frac{q-1}{1-t}\biggr ] { \mathsf{p}}_{{\lambda}}(x)\end{gathered}\ ] ] for fixed @xmath379 . taking @xmath455 in ( * ? ? ? * corollary 4.9 ) in rains generalised @xmath1-pfaff saalschtz sum ( * ? ? ? * corollary 4.9 ) @xmath456 { \mathsf{q}}_{\mu/\nu}\biggl[\frac{b - c}{1-t}\biggr]= \frac{(a)_{\nu}(b)_{{\lambda}}}{(b)_{\nu}(c)_{{\lambda}}}\ , { \mathsf{q}}_{{\lambda}/\nu}\biggl[\frac{a - c}{1-t}\biggr]\ ] ] yields @xmath457}{0pt}{}{\mu}{\nu } } { \mathsf{q}}_{{\lambda}/\mu}\biggl[\frac{q-1}{1-t}\biggr ] = q^{{\lvert{\lambda}-\nu\rvert } } \frac{(c)_{{\lambda}}}{(c)_{\nu } } { \genfrac{[}{]}{0pt}{}{{\lambda}}{\nu}}.\ ] ] this allows the sum over @xmath212 on the right of to be carried out , leading to @xmath458}{0pt}{}{\mu}{\nu } } { \mathsf{p}}_{\mu}(x ) \\ & = \biggl(\:\prod_{i=1}^n ( 1-cq^{{\lambda}_i}t^{n - i})\biggr ) \sum_{{\lambda}\supseteq\nu } ( bq)^{{\lvert{\lambda}-\nu\rvert } } \frac{(ct^{n-1})_{{\lambda}}(cqt^{n-1})_{\nu } } { ( ct^{n-1})_{\nu}(cqt^{n-1})_{{\lambda}}}\,(1)_{{\lambda}/\nu } { \genfrac{[}{]}{0pt}{}{{\lambda}}{\nu } } { \mathsf{p}}_{{\lambda}}(x ) \\ & = \biggl(\:\prod_{i=1}^n ( 1-cq^{\nu_i}t^{n - i})\biggr ) \sum_{{\lambda}\supseteq\nu } ( bq)^{{\lvert{\lambda}-\nu\rvert } } ( 1)_{{\lambda}/\nu } { \genfrac{[}{]}{0pt}{}{{\lambda}}{\nu } } { \mathsf{p}}_{{\lambda}}(x).\end{aligned}\ ] ] by corollary [ corbtaylor ] this is the same as . to prove the theorem we first prepare the following result . [ propskew ] for @xmath459 partitions @xmath460 \\ = \frac{(a)_{\mu}}{(c / t)_{\mu}}\ , \frac{(a)_{\nu}}{(c / b)_{\nu}}\ , \frac{(c / a , c / b)_{\infty}}{(c , c / ab)_{\infty } } \sum_{{\lambda } } \frac{(c / bt)_{{\lambda}}}{(a)_{{\lambda}}}\ , { \mathsf{p}}_{\nu/{\lambda}}\bigl(\frac{c}{ab}\bigr ) { \mathsf{q}}_{\mu/{\lambda}}\biggl[\frac{1-b}{1-t}\biggr].\end{gathered}\ ] ] note that for @xmath400 this is @xmath461 which , for @xmath378 , simplifies to the @xmath1-gauss sum . key is the cauchy - type identity for skew macdonald polynomials due to rains ( * ? ? ? * corollary 3.8 ) : @xmath462 { \mathsf{q}}_{{\lambda}/\nu}\biggl[\frac{1-d}{1-t}\biggr ] \\ = \bigl(\frac{q}{c}\bigr)^{{\lvert\mu\rvert}}\frac{(a , b)_{\mu}}{(e / c , f / c)_{\mu}}\ , \bigl(\frac{q}{d}\bigr)^{{\lvert\nu\rvert } } \frac{(a , b)_{\nu}}{(e / d , f / d)_{\nu } } \qquad\qquad\qquad\qquad\qquad\\ \times \sum_{{\lambda } } \bigl(\frac{cd}{q}\bigr)^{{\lvert{\lambda}\rvert } } \frac{(e / cd , f / cd)_{{\lambda}}}{(a , b)_{{\lambda}}}\ , { \mathsf{p}}_{\nu/{\lambda}}\biggl[\frac{1-c}{1-t}\biggr ] { \mathsf{q}}_{\mu/{\lambda}}\biggl[\frac{1-d}{1-t}\biggr],\end{gathered}\ ] ] provided that the sum on the left terminates and the balancing condition @xmath463 holds . the prefactor @xmath464 refers to the sum on the left for @xmath178 , i.e. , to @xmath465 { \mathsf{q}}_{{\lambda}}\biggl[\frac{1-d}{1-t}\biggr ] \\ & = \sum_{{\lambda } } q^{{\lvert{\lambda}\rvert } } \frac{(a , b , c , d)_{{\lambda}}}{(e , f)_{{\lambda}}}\ , { \mathsf{p}}_{{\lambda}}\biggl[\frac{1}{1-t}\biggr].\end{aligned}\ ] ] since @xmath466 vanishes if @xmath467 , and since @xmath468=\frac{1}{(q)_k}\ ] ] if follows that for @xmath469 @xmath470 where the second equality follows from the @xmath1-pfaff saalschtz sum ( * ? ? ? * equation ( ii.12 ) ) . to make the same @xmath469 specialisation in the right - hand side of we first replace the sum over @xmath3 by @xmath471 ( using the fact that @xmath472 if @xmath473 ) . then @xmath474 is well - defined for @xmath475 . it thus follows that for for @xmath469 simplifies to @xmath476 \\ = \bigl(\frac{q}{t}\bigr)^{{\lvert\mu\rvert}}\frac{(a , q^{-n})_{\mu}}{(e / t , f / t)_{\mu}}\ , \bigl(\frac{q}{d}\bigr)^{{\lvert\nu\rvert}}\frac{(a , q^{-n})_{\nu}}{(e / d , f / d)_{\nu}}\ , \frac{(e / a , e / d)_n}{(e , e / ad)_n } \qquad\qquad\qquad\qquad\qquad\\ \times \sum_{{\lambda}\subseteq\nu } \bigl(\frac{dt}{q}\bigr)^{{\lvert{\lambda}\rvert } } \frac{(e / dt , f / dt)_{{\lambda}}}{(a , q^{-n})_{{\lambda}}}\ , { \mathsf{p}}_{\nu/{\lambda}}(1 ) { \mathsf{q}}_{\mu/{\lambda}}\biggl[\frac{1-d}{1-t}\biggr]\end{gathered}\ ] ] for @xmath477 . eliminating @xmath98 , taking the limit @xmath478 and using this results in the claim with @xmath479 . we are now prepared to prove theorem [ thmskewrc ] . we proceed by induction on @xmath314 . for @xmath480 we get the tautology @xmath481= \frac{(b / c)_{\mu}}{(b / c)_{\nu}}\ , { \mathsf{q}}_{\mu/\nu}\biggl[\frac{a - c}{1-t}\biggr].\ ] ] for @xmath482 we compute the sum on the right - hand side of , assuming the formula is true for the alphabet @xmath104 . ( recall that @xmath305 . ) using the branching rule and @xmath483 { \mathsf{p}}_{{\lambda}/\eta}(z ) = a^{{\lvert\eta-\nu\rvert } } { \mathsf{q}}_{{\lambda}/\nu}\biggl[\frac{1-c / a}{1-t}\biggr ] { \mathsf{p}}_{{\lambda}/\eta}(az)\ ] ] we obtain @xmath484 { \mathsf{r}}_{{\lambda}/\mu}(x;b ) \\ & \hphantom{:}=\sum_{{\lambda},\eta } a^{{\lvert\eta-\nu\rvert } } \frac{(b / c)_{{\lambda}}(bz / t)_{\eta}}{(b / c)_{\nu}(bz)_{{\lambda}}}\ , { \mathsf{q}}_{{\lambda}/\nu}\biggl[\frac{1-c / a}{1-t}\biggr ] { \mathsf{p}}_{{\lambda}/\eta}(az ) { \mathsf{r}}_{\eta/\mu}(y;b).\end{aligned}\ ] ] the sum over @xmath3 can be transformed by proposition [ propskew ] with @xmath485 and @xmath486 . as a result @xmath487 { \mathsf{p}}_{\nu/{\lambda}}(z ) { \mathsf{r}}_{\eta/\mu}(y;b ) \\ & = \frac{(cz , dz)_{\infty}}{(az , bz)_{\infty } } \sum_{{\lambda } } \frac{(dz / t)_{{\lambda}}}{(dz)_{\nu}}\ , { \mathsf{p}}_{\nu/{\lambda}}(z ) s_{\mu,{\lambda}}(y , a , b , c),\\end{aligned}\ ] ] where once again we have used , and where @xmath488 . by the induction hypothesis , @xmath489 may be replaced by the right - hand side of with @xmath490 , leading to @xmath491 { \mathsf{p}}_{\nu/{\lambda}}(z ) { \mathsf{r}}_{{\lambda}/\eta}(y;d).\ ] ] one final application of results in @xmath492 { \mathsf{r}}_{\nu/\eta}(x;d),\ ] ] which is the right - hand side of with @xmath493 . in this final section we prove a number of additional identities for basic hypergeometric series involving the function @xmath413 . for easy comparison with known results for one - variable basic hypergeometric series we define @xmath494 \\ : = \sum_{{\lambda } } \frac{(a_1,\dots , a_r)_{{\lambda}}}{(b_1,\dots , b_{s-1})_{{\lambda}}}\ , \bigl((-1)^{{\lvert{\lambda}\rvert}}q^{n({\lambda}')}t^{-n({\lambda})}\bigr)^{s - r+1}\ , z^{{\lvert{\lambda}\rvert } } { \mathsf{r}}_{{\lambda}}(x;b_s),\end{gathered}\ ] ] where @xmath94 is a finite alphabet . there is some redundancy in the above definition since by lemma [ lembindep ] @xmath495 . since @xmath413 vanishes if @xmath496 , and recalling @xmath497 ( see ) , it follows that @xmath498 & = \sum_{k=0}^{\infty } \frac{(a_1,\dots , a_r)_k}{(b_1,\dots , b_s)_k}\ , \bigl((-1)^k q^{\binom{k}{2}}\bigr)^{s - r+1}\ , z^k \\ & = { _ r\phi_s}\biggl[\genfrac{}{}{0pt}{}{a_1,\dots , a_r } { b_1,\dots , b_s};z\biggr],\end{aligned}\ ] ] with on the right the standard notation for one - variable basic hypergeometric series ( with dependence on the base @xmath1 suppressed ) . the reader is warned that only in this degenerate case do the `` lower - parameters '' @xmath499 enjoy full @xmath500 symmetry . using the above notation the @xmath1-gauss sum may be stated as @xmath501 = \prod_{x\in x } \frac{(cx / a , cx / b)_{\infty}}{(cx , cx / ab)_{\infty}}.\ ] ] to generalise this result we first prove the following transformation formula . [ gthmktw ] for @xmath502 and @xmath212 a partition , @xmath503 { \mathsf{r}}_{{\lambda}}(x;e ) \\ = \frac{(b)_{\mu}}{(d / c)_{\mu } } \biggl(\ : \prod_{x\in x } \frac{(fx , ex / a)_{\infty}}{(ex , fx / a)_{\infty } } \biggr ) \sum_{{\lambda } } \frac{(a , d / c)_{{\lambda}}}{(d)_{{\lambda}}}\ , { \mathsf{q}}_{{\lambda}/\mu}\biggl[\frac{e / a - cf / a}{1-t}\biggr ] { \mathsf{r}}_{{\lambda}}(x;f).\end{gathered}\ ] ] we replace @xmath505 and then rename the summation index @xmath3 on the right as @xmath212 . by the pieri formula of theorem [ pieri ] ( with @xmath506 ) the term @xmath507 on the right - hand side can be expanded as @xmath508 { \mathsf{r}}_{{\lambda}}(x;e),\ ] ] resulting in @xmath509 { \mathsf{r}}_{{\lambda}}(x;e ) \\ = \frac{(b)_{\nu}}{(d / c)_{\nu } } \sum_{{\lambda},\mu } \frac{(a)_{{\lambda}}(d / c)_{\mu}}{(d)_{\mu}}\ , { \mathsf{q}}_{{\lambda}/\mu}\biggl[\frac{f / a - e / a}{1-t}\biggr ] { \mathsf{q}}_{\mu/\nu}\biggl[\frac{e / a - cf / a}{1-t}\biggr ] { \mathsf{r}}_{{\lambda}}(x;e).\end{gathered}\ ] ] by with @xmath510 the sum over @xmath212 can be carried out , completing the proof . if we let @xmath513 in theorem [ gthmktw ] such that @xmath514 and then replace @xmath515 , we find @xmath516}{0pt}{}{{\lambda}}{\mu } } { \mathsf{r}}_{{\lambda}}(x;c ) \\ = \biggl(\ : \prod_{x\in x } \frac{(cx / a , azx)_{\infty}}{(cx , zx)_{\infty } } \biggr ) \sum_{{\lambda } } \bigl(\frac{c}{a}\bigr)^{{\lvert{\lambda}-\mu\rvert } } \frac{(a , abz / c)_{{\lambda}}}{(a , abz / c)_{\mu } } { \genfrac{[}{]}{0pt}{}{{\lambda}}{\mu } } { \mathsf{r}}_{{\lambda}}(x;az).\end{gathered}\ ] ] for @xmath378 this is a multiple analogue of heine s @xmath517 transformation ( * ? ? ? * equation ( iii.2 ) ) .

it is widely recognised that the effect of doping into a mott insulator is complicated and unpredictable , as can be seen by examining the hall coefficient in high cuprates . the doping effect , including the electron - hole doping asymmetry , may be more straightforward in doped organic mott insulators owing to their simple electronic structures . here we investigate the doping asymmetry of an organic mott insulator by carrying out electric - double - layer transistor measurements and using cluster perturbation theory . the calculations predict that strongly anisotropic suppression of the spectral weight results in the fermi arc state under hole doping , while a relatively uniform spectral weight results in the emergence of a non - interacting - like fermi surface in the electron - doped state . in accordance with the calculations , the experimentally observed hall coefficients and resistivity anisotropy correspond to the pocket formed by the fermi arcs under hole doping and to the non - interacting fermi surface under electron doping .

1606.09311

electron - hole doping asymmetry in mott insulators has been one of the key questions related to the origin of superconductivity in the proximity of the insulating state @xcite . to investigate this , one should prepare an exactly half - filled ( zero - doped ) mott insulator and then inject or extract electrons , preferably by an electrostatic method , in the same sample . however , it is difficult to perform such measurements on high-@xmath0 cuprates for the following three reasons . first , they are mostly either only hole - doped or electron - doped , and the crystallographic structures of the parent materials are often different . second , they require a strong electric field to tune their band filling owing to their high half - filled carrier density ( on the order of 10@xmath1 @xmath2 ) . third , since they are charge - transfer - type insulators , electrons and holes are doped into different electronic orbitals , which may obscure the pure doping asymmetry . in practice , the inverse of the hall coefficient monotonically increases with increasing doping concentration with the sign changing across the zero - doping point @xcite . this is a feature of band insulators rather than mott insulators . in contrast to the high-@xmath0 cuprates , the organic mott insulator @xmath3-(bedt - ttf)@xmath4cu[n(cn)@xmath4]cl @xcite ( abbreviated to @xmath3-cl hereafter ) , where bedt - ttf represents bis(ethylenedithio)tetrathiafulvalene , serves as an appropriate material for examining the electron - hole asymmetry of doped mott insulators because of its relatively low carrier density ( @xmath61.8@xmath710@xmath8 @xmath2 ) and its single electronic orbital nature . indeed , electrons and holes are doped into the same @xmath9 electron band and thus @xmath3-cl can be modeled as a single band hubbard model on an anisotropic triangular lattice ( fig . [ fig.1 ] ) . due to its high controllability of electronic states , @xmath3-cl has been studied intensively in terms of the bandwidth - controlled mott transition between mott insulating and superconducting states at half filling @xcite . the electrostatic method of carrier doping into organic mott insulators with field - effect transistor ( fet ) structure has been implemented by the authors @xcite . however , the limited doping concentration with the fet still forbids the observation of electron - hole asymmetry . here , by fabricating electric - double - layer transistors ( edlts ) based on thin single crystals of @xmath3-cl , we realised for the first time both electron and hole doping into the organic mott insulator . we measured the field effect on transport properties including the sheet resistivity , hall coefficient and resistivity anisotropy to elucidate the doping anisotropy of the electronic state in the doped organic mott insulator . to provide further insights into the fermi surface ( fs ) , we calculated the single - particle spectral functions for an effective model of @xmath3-cl by using cluster perturbation theory ( cpt ) @xcite . the calculations predict that strongly anisotropic suppression of the spectral weight results in the fermi arc state under hole doping , while a relatively uniform spectral weight results in the emergence of a non - interacting - like fs in the electron - doped state . in accordance with the calculations , the experimentally observed hall coefficients and resistivity anisotropy correspond to the pocket formed by the fermi arcs under hole doping and to the non - interacting fs under electron doping . we fabricated the edlt devices by mounting an ion gel on a hall - bar - shaped thin single crystal of @xmath3-cl and a au side gate electrode ( figs . 2a and 2b ) . first , we obtained the resistivity curve for the sheet resistivity @xmath10 at the temperature @xmath11 k ( fig . [ fig.2]c ) . as expected for a mott insulator , a clear ambipolar field effect with a resistance peak at the gate voltage @xmath12 v was observed . the hysteresis and the leakage current remained small without any signature of a chemical reaction between the electrolyte and @xmath3-cl . note that the field effect was about 3% at 220 k because thermally excited carriers in the bulk ( about 100 molecular layers in a typical sample ) dominated the electrical conduction owing to the low charge excitation gap ( ca . we also confirmed that the carrier tunability of the @xmath3-cl edlt greatly exceeded that of sio@xmath4-based fets by comparing these devices using the same @xmath3-cl crystal ( fig . [ fig.s1 ] ) . provided that the mobilities of the fet and edlt are equivalent , a gate voltage of + 1 v in the edlt corresponded to approximately 20% electron doping ( see supplementary note 1 ) . the resistivity curves at 220 k and the temperature dependences of the resistivity were reproducible after multiple temperature cycles ( see supplementary note 2 and fig . [ fig.s2 ] ) . with decreasing temperature , the ambipolar field effect became more distinct . figures [ fig.2]d and [ fig.2]e show the temperature dependence of the sheet resistivity under a negative gate voltage ( hole doping ) and positive gate voltage ( electron doping ) , respectively . under hole doping , the activation energy significantly decreased to less than 1 mev at @xmath13 = @xmath141.35 v. on the other hand , the doping of electrons tended to reduce the resistivity more effectively and metallic conductivity was observed at @xmath151 v. in a high - conductivity sample , we observed negative magnetoresistance accompanied with the upturn of resistivity at low temperature , which suggested a weak localisation effect ( see supplementary note 3 and fig . [ fig.s3 ] ) . analysis of the weak localisation indicated that the mean free path and dephasing length considerably exceeded the distance between bedt - ttf dimers indicating coherent transport . to observe a more significant difference between the electron- and hole - doped states , we measured the hall coefficient @xmath16 , which indicates the fs topology . the magnetic field @xmath17 dependence of the hall resistance @xmath18 at 30 k is shown in fig . [ fig.3]a ( @xmath16 is given by the slope ) . despite electron doping , the hall coefficients were clearly positive and a simple estimation of the carrier density , @xmath19 , gave about 0.8 holes per bedt - ttf dimer . since the injected carriers were electrons ( @xmath60.2 electrons / dimer ) , the concentration of which was much lower than that of the observed holes , it suggests that dense hole carriers are delocalised by electron doping . this is considered to be the doping - induced mott transition previously observed in fet devices @xcite . if we assume doping symmetry , a similar hall effect is expected for the hole - doped side . namely , hole doping immediately collapses the mott - hubbard gap and @xmath19 corresponding to 1@xmath20 ( @xmath21 : electron doping concentration , which is negative for hole doping ) holes per bedt - ttf dimer should be observed . however , as shown in fig . [ fig.3]a , the hall coefficient on the hole - doped side was about three times larger than that on the other side , and @xmath19 gave about 0.3 holes per bedt - ttf dimer . although @xmath19 increased with increasing absolute value of the gate voltage , the rate of increase was low ( only 7% difference between @xmath13 = @xmath140.8 v and -1.2 v ) . surprisingly , under hole doping , @xmath19 per dimer appears to be neither 1@xmath20 ( delocalised hole carriers + injected hole carriers ) nor @xmath20 ( only injected hole carriers ) , indicating that the localised carriers were partially delocalised by excess hole doping as shown in fig . [ fig.3]c . these results indicate that the fs topology is significantly different between the electron- and hole - doped states . for the gate voltage dependence of the hall mobility , see supplementary note 4 and fig . [ fig.s5 ] . the temperature dependence of @xmath16 ( fig . [ fig.3]b ) shows that the @xmath16 asymmetry was maintained up to about 50 k , above which the hall effect at the doped surface was obscured by the thermally excited carriers in the bulk . note that increase in @xmath16 at 30 k upon cooling is due to the decrease of the thermally excited carriers in the bulk ( see supplementary note 5 and fig . [ fig.s6 ] ) . the data below 25 k contain ambiguity owing to the non - ohmic behaviour ( see supplementary note 6 and fig . [ fig.s4 ] ) . what is the origin of @xmath16 under hole doping ? as shown in fig . [ fig.1 ] , @xmath3-cl forms essentially an anisotropic triangular lattice resulting in an ellipsoidal non - interacting fs which is folded by the zone boundary . in a fermi liquid with a single type of carrier , @xmath19 denotes the carrier density corresponding to the volume enclosed by the fs ( luttinger s theorem @xcite ) . under electron doping , @xmath19 is close to the carrier density corresponding to the non - interacting ellipsoidal fs ( known as the @xmath22-orbit in the study of shubnikov - de haas oscillation @xcite ) . by contrast , @xmath19 under hole doping somehow appears to correspond to the volume bounded by the lenslike closed portion of the fs ( known as the @xmath23-orbit ) . the doping dependences of the carrier density enclosed by the @xmath23- and @xmath22-orbits assigned for a @xmath3-bedt - ttf salt @xcite are plotted as reference in fig . [ fig.3]c . to verify this @xmath16 asymmetry , we calculated the single - particle spectral functions of the hubbard model on an anisotropic triangular lattice ( @xmath24 = @xmath140.8 , @xmath25 = 7 ) at 30 k using cpt and the results are shown in fig . [ fig.4 ] . first , the mott insulating state was reproduced at half filling , where the energy gap opened at all the k - points ( figs . 4b and 4e ) . when electrons were doped ( @xmath21 = + 0.17 ) , the fs appeared as shown in figs . 4c and 4f . the fs topology was the same as that of the non - interacting case , although the shape and the spectral weight were not exactly the same . in this state , the value of @xmath16 should be close to that of the fermi liquid state , as suggested by the experiments . on the other hand , the fs under hole doping ( @xmath21 = @xmath140.17 ) was very different from the non - interacting fs ( figs . 4a and 4d ) . the spectral weight near the z point was strongly suppressed . namely , arcs around the x point and pseudogaps near the z point appeared . the partial disappearance of fs by the pseudogap is similar to the fermi arc state observed in lightly hole - doped cuprates @xcite . in this state , luttinger s theorem is seemingly violated and the value of @xmath16 is no longer simply estimated . however , considering the first brillouin zone ( bz ) of @xmath3-cl , one can see that the fermi arcs are folded and form closed lenslike orbits . these orbits originate from the @xmath23-orbit mentioned above and correspond to the observed @xmath16 in the hole - doped states . these results imply that @xmath16 is predominantly governed by quasi - particles with relatively long lifetime ( bright points of the spectral function in the reciprocal space in fig . [ fig.4 ] ) . therefore , the asymmetry of the fs topology was clearly observed via @xmath16 . the partial disappearance of the fs can also be confirmed by the anisotropy in the carrier conduction . the strongly suppressed parts of the fs under hole doping should affect strongly the carrier conduction along the @xmath26-axis . indeed , our cpt calculations predict that the resistivity is more anisotropic in the hole - doped state than in the electron - doped state ( see supplementary note 7 and fig . [ fig.opt ] ) . we measured the resistivity anisotropy as shown in fig . [ fig.5 ] . the crystallographic axes were determined by the crystal shape and the sign of the seebeck coefficient at room temperature . at 220 k where the thermally excited carriers in the non - doped bulk dominates the conductivity , the @xmath26-axis resistivity @xmath27 was slightly higher than the @xmath28-axis resistivity @xmath29 in both electron and hole doping . with decreasing temperature , @xmath30 increased under hole doping while it decreased and remained near 1 in the non - doped and electron - doped states as shown in fig . [ fig.5]b . the contour plot of @xmath31 under various gate voltage and temperature ( fig . [ fig.5]c ) clearly exhibited that the resistivity in the hole - doped state was particularly anisotropic , in good qualitative agreement with our cpt calculations . the doping asymmetry is expected from the particle - hole asymmetric non - interacting band structure , as shown in fig . [ fig.1]c . the energy bands along the z - m axis are very flat and the van hove singularity ( vhs ) lies below the fs at half filling . with the doping of holes , the fs approaches the vhs and the effect of the interaction is expected to be enhanced . indeed , substantial suppression of the spectral weight of the fs along the z - m axis , on which the van hove critical points lie , is observed ( fig . [ fig.4]a ) . by contrast , the fs departs from the vhs with the doping of electrons , resulting in a weaker interaction effect and a more non - interacting - like fs . it is also indicated that the spin fluctuation is stronger in the hole - doped state because of the vhs . if a superconducting state could be induced by further hole doping , the transition temperature and pairing symmetry are of great interest since the hole - doped state is substantially different from the metallic state created by controlling the bandwidth at half filling . to summarise , without gate voltage , @xmath3-cl is a mott insulator due to electron interaction and high commensurability between the hole density and periodic potential ( 1 hole / 1 site ) . when holes or electrons are doped , the commensurability is reduced and the effect of electron interaction is weakened , resulting in carrier delocalisation . however , the effect of interaction remains strong at specific k - points where the energy dispersion is relatively flat . owing to the particle - hole asymmetric band structure , the pseudogaps appeared only near the van hove critical points under hole doping . in contrast to the calculations , our measurements did not observe the true metallic state ( @xmath32 ) at low temperature . this is probably due to the effect of localisation caused by finite random potential at the surface @xcite , as the negative magnetoresistance indicated . improvement of the surface roughness and cleanness may realise metallic phases under both dopings . thus , the doping asymmetry in a mott insulator was demonstrated by a simple process in an organic mott edlt . the above results show that organic mott insulators are good model materials for mott physics even in the case of band - filling control , and the electronic state in a simple mott insulator may be predictable in the framework of cpt . since molecular conductors consist of various molecules with different arrangements , the effect of doping into materials with different fs topologies is of interest in future experiments . furthermore , the presence / absence of the superconducting phase under doping and its doping asymmetry are also intriguing . * sample preparation . * the source , drain , and gate electrodes ( 18 nm - thick - au ) were patterned on a polyethylene naphthalate substrate ( teonex q65fa , teijin dupont films japan limited ) using photolithography . a thin ( @xmath6100 nm ) single crystal of @xmath3-(bedt - ttf)@xmath4cu[n(cn)@xmath4]cl ( abbreviated to @xmath3-cl hereafter ) was electrochemically synthesized by oxidizing bedt - ttf ( 20 mg ) dissolved in 50 ml of 1,1,2-trichloroethane ( 10% v / v ethanol ) in the presence of tpp[n(cn)@xmath4 ] ( tpp = tetraphenylphosphonium , 190 mg ) , cucl ( 60 mg ) and tpp - cl ( 100 mg ) . after applying current of 8 @xmath33a for 20 h , a thin crystal was transferred into 2-propanol with a pipette and was guided on top of the pen substrate . after the substrate was removed from the 2-propanol and dried , the @xmath3-cl crystal was shaped into a hall - bar using a pulsed laser beam with a wavelength of 532 nm . typical dimensions of the hall bar sample were approximately 15 @xmath33 m ( width ) @xmath7 40 @xmath33 m ( length ) @xmath7 100 nm ( thickness ) . the edlt devices were completed by mounting an ion gel ( or ionic liquid ) on the hall - bar - shaped thin single crystal of @xmath3-cl and a au side gate electrode . the basic techniques , such as the electrochemical synthesis of thin single crystals and the laser shaping , are common to our previous fet devices @xcite . as gate electrolytes , we employed poly(vinylidene fluoride - co - hexafluoropropylene ) [ pvdf - hfp ] with 58% w / w 1-butyl-3-methylimidazolium tetrafluoroborate [ bmim - bf@xmath34 ( samples # 1@xmath64 ) , n , n - diethyl - n - methyl - n-(2-methoxyethyl ) ammonium bis(trifluoromethanesulfonyl)imide [ deme - tfsi ] ( sample # 5 in fig . [ fig.s1 ] ) , polymethyl methacrylate [ pmma ] with 57% w / w deme - tfsi ( sample # 6 in fig . [ fig.s3 ] ) . the pvdf - hfp ion gel @xcite was spin - coated from an acetone solution in which the weight ratio between the polymer , ionic liquid , and solvent was 1:1.3:7.7 . although the ionic conductivity was the greatest in the liquid state , the large thermal stress usually broke the @xmath3-cl crystal at a low temperature ( @xmath35150 k ) . thinning of the gate electrolytes by gelation reduced the thermal stress at low temperatures , allowing the successful cooling of the samples down to 2 k. the area of the gate electrode was more than 100 times larger than that of the @xmath3-cl channel in order to apply the gate voltage effectively . * transport measurements . * the transport measurements were performed using a physical property measurement system ( quantum design ) . the temperature and magnetic field normal to the substrate were controlled in the range of 2 - 300 k ( cooling rate : 2 k / min ) and @xmath368 tesla , respectively . when the gate voltage was varied , the samples were warmed to 220 k. in the hall measurements , the magnetic field was swept in the range of @xmath368 tesla at a constant temperature and the forward and backward data were averaged to eliminate the small linear voltage drift . although it is difficult to experimentally determine the distribution of field - induced charge along the out - of - plane direction , we consider that the field - induced carriers predominantly exist in a single conducting molecular sheet @xcite . recently , the quantum oscillations in the doped surface of the dirac fermion system @xmath23-(bedt - ttf)@xmath4i@xmath37 @xcite indicated that only two conducting layers from the surface were doped , in which the carrier density in the second layer was only 1/7 of that in the first layer . the above assumption is also likely to apply to @xmath3-type bedt - ttf salts , where a shorter screening length is expected . * hubbard model on an anisotropic triangular lattice . * to examine the electron correlation effects in the organic doped mott insulator , we study the single - band hubbard model defined on an anisotropic triangular lattice , which is known as the simplest model for @xmath3-type bedt - ttf salts @xcite . the model is defined by the hamiltonian , @xmath38 where @xmath39 creates an electron on site @xmath40 with spin @xmath41 and @xmath42 . @xmath43 is the transfer integral between the neighboring sites @xmath40 and @xmath44 , @xmath45 is the on - site coulomb repulsion , and @xmath33 is the chemical potential . here , a single site consists of a single dimer and a unit cell contains two dimers with different orientation ( fig . [ fig.lattice]a ) . the transfer integral between the different ( same ) dimers is given as @xmath46 ( @xmath47 ) . the series of @xmath3-type bedt - ttf salts serves as an ideal realization of the hubbard model of eq . ( [ eq.ham ] ) . the anisotropy @xmath48 can be changed depending on substituted anions @xcite . the coulomb repulsion @xmath45 is estimated to be comparable to @xcite or larger than @xcite the band width , implying that the strongly correlated electronic state is realized at ambient pressure . applied pressure @xcite or bending strain on thin films @xcite can control the electron correlation parameter @xmath49 . moreover , the chemical potential @xmath33 can be tuned efficiently by applied gate voltage in the edlt , allowing for investigation of the electronic structure of electron- and hole - doped mott insulators on equal footing . the series of organic charge transfer salts is indeed the highly controllable counterpart of strongly correlated transition - metal compounds @xcite . in the main text , we have focused on a typical parameter set for @xmath3-type bedt - ttf salts of @xmath50 , @xmath51 , and @xmath52 mev . we have also considered the parameter set relevant for @xmath3-cl of @xmath53 , @xmath54 , and @xmath55 mev @xcite , and found that the fermi arc appears for hole - doped case also in this parameter set ( fig . [ fig.4another ] ) . the non - interacting tight - binding band structure , the density of states , and the fs at half filling for this parameter set are shown in figs . [ fig.lattice]c and [ fig.lattice]d . the strongly correlated hubbard model at half filling or the heisenberg model on the anisotropic triangular lattice have been investigated theoretically and the ground state phase diagrams with various phases including metal , antiferromagnetic insulator , spiral , and spin liquid ground states have been obtained @xcite . most of these previous calculations on the hubbard model at half filling have been carried out with @xmath56 . however , the sign of @xmath47 is crucial when it comes to the electron - hole asymmetry of carrier doping because the sign determines whether the vhs appears in hole - doped side or electron - doped side . the correct sign is @xmath57 @xcite and the vhs appears in the hole - doped side ( see fig . [ fig.1]c and fig . [ fig.lattice]c ) . * unit cell and brillouin zone . * in figs . [ fig.4 ] and [ fig.4another ] , we show the single - particle spectral functions and the fs in the bz of the unit cell consisting of two sites in the real space ( see blue shaded region in fig . [ fig.bz]a ) . the bz is relevant for the realistic structure of the @xmath3-type bedt - ttf salts and therefore the results are directly compared with those obtained by first - principles band - structure calculations @xcite . however , the primitive unit cell of the single - band hubbard model in eq . ( [ eq.ham ] ) on the anisotropic triangular lattice consists of one site ( see red shaded region in fig . [ fig.bz]a ) , unless a spontaneous symmetry breaking such as the nel order occurs . therefore the theoretical calculations for the hubbard model on the anisotropic triangular lattice are usually performed with this primitive unit cell @xcite . in order to avoid possible confusions caused by the different choice of the unit cell , we describe the relation between the two bzs of the one - site unit cell and the two - site unit cell . figure [ fig.bz]a shows the anisotropic triangular lattice with regular - triangular plaquettes . the primitive translational vectors for the one ( two)-site unit cell are given by @xmath58 ( @xmath59 ) , and the corresponding bz is diamond ( rectangular ) shaped , as shown in fig . [ fig.bz]c . we also show in fig . [ fig.bz]b the lattice which is topologically equivalent to the triangular lattice in fig . [ fig.bz]a but forms the square plaquettes with diagonal lines . notice that this square topology of the lattice is very often used for study of the hubbard model on the anisotropic triangular lattice @xcite . the primitive translational vectors for the one ( two)-site unit cell are again given by @xmath58 ( @xmath59 ) . the bz of the one ( two)-site unit cell is now square ( diamond ) shaped , as shown in fig . [ fig.bz]d . let us now consider the basis - transformation formula between the bases @xmath59 and @xmath58 , where the former is a set of the primitive translational vectors of the realistic structure of the @xmath3-type bedt - ttf salts , while the latter is that of the single - band hubbard model . it is easily found from fig . [ fig.bz]b that they are related by the @xmath60 rotation and the scale transformation . to be concrete , we consider a vector @xmath61 in the real space . it is easily shown that the coordinate @xmath62 is related to the coordinate @xmath63 as @xmath64 similarly , for a wave vector @xmath65 , where @xmath66 ( @xmath67 ) is the set of the reciprocal vectors of @xmath68 ( @xmath69 ) , ( @xmath70 is related to ( @xmath71 as @xmath72 for example , @xmath73 , and @xmath74 in the @xmath75 basis ( i.e. , @xmath76 , z , m , and x points in fig . [ fig.bz]c ) correspond to @xmath77 , and @xmath78 in the @xmath79 basis ( fig . [ fig.bz]d ) , respectively . finally , we note that the rectangular bz in fig . [ fig.bz]c corresponds to the antiferromagnetic bz of the square lattice in fig . [ fig.bz]d ( blue shaded regions ) . * cluster perturbation theory . * to examine the electron - hole asymmetry of the fs , the calculation of single - particle green s functions is required . for this purpose , we employ the cpt @xcite . the cpt is a theory of strong coupling expansion for the single - particle green s function and allows us to examine the single - particle excitations including quasi - particle excitations , mott gap , and even pseudogap phenomena in strongly correlated electron systems @xcite . as shown in the following , the cpt requires the fully interacting single - particle green s function ( or the self - energy @xcite ) of open - boundary clusters to obtain the single - particle green s function of the lattice . the basic notion of the scheme was introduced in the earlier work by gros and valent @xcite . let us outline the method of cpt . first , we divide the whole lattice on which the model hamiltonian @xmath80 is defined into identical finite - size clusters , each of which consists of @xmath81 sites . we shall denote the hamiltonian of a cluster as @xmath82 and the inter - cluster hopping as @xmath83 . here , @xmath82 is given in the right - hand side of eq . ( [ eq.ham ] ) but it is defined on a @xmath81-site cluster with open boundary conditions . next , the single - particle green s function @xmath84 of the cluster hamiltonian @xmath82 is calculated by the exact diagonalization method . applying the strong coupling expansion with respect to the inter - cluster hopping @xmath83 , the single - particle green s function is given as @xmath85 , where @xmath86 is the matrix representation of @xmath83 @xcite . by restoring the translational invariance , which is broken in @xmath87 due to the partitioning of the lattice , we obtain the single - particle green s function @xmath88 of the whole lattice for momentum @xmath89 and complex frequency @xmath90 as @xmath91_{ij } { \mathrm{e}}^{-{\mathrm{i}}{\mathbf}{k } \cdot ( { \mathbf}{r}_i - { \mathbf}{r}_j ) } , \label{eq.cpt}\ ] ] where @xmath92 is the position of the @xmath40-th site within the cluster @xcite . the cluster green s function matrix @xmath84 is given as @xmath93 with @xmath94 and @xmath95 where @xmath96 with @xmath97 being the temperature , @xmath98 is the @xmath99-th eigenstate of @xmath82 with its eigenvalue @xmath100 in ascending order , and @xmath101 is the partition function of the cluster . the sum over eigenstates is truncated by summing up to the @xmath102-th eigenstate . the upper bound @xmath102 of the sum is determined so as to satisfy that @xmath103 , where @xmath104 is the ground - state energy and @xmath105 is a small real number . for the finite temperature calculations at @xmath106 k , we set @xmath107 , in which for example the lowest @xmath108 eigenstates among @xmath109 eigenstates in total are included for the hole - doped case with @xmath110 . the inter - cluster hopping between different clusters in eq . ( [ eq.cpt ] ) is represented as @xmath111 } , \ ] ] where @xmath43 is given in eq . ( [ eq.ham ] ) , @xmath112 and @xmath113 are the translational vectors of the clusters , and @xmath114 ( @xmath115 and @xmath116 : integer ) indicates a relative position of two different clusters , site @xmath40 being located in one cluster and site @xmath44 in the other cluster . for the 12-site cluster , the translational vectors are @xmath117 and @xmath118 , and @xmath43 in eq . ( [ eq.tij ] ) is finite only when @xmath119 , @xmath120 , and @xmath121 , indicating the inter - cluster hopping between a cluster and its neighboring six clusters ( see fig . [ fig.lattice]a ) . for the 16-site cluster , the translational vectors are @xmath122 and @xmath123 , and @xmath43 in eq . ( [ eq.tij ] ) is finite only when @xmath119 , @xmath120 , and @xmath124 ( see fig . [ fig.lattice]b ) . note that in principle a cluster with a larger size gives better results in the cpt because the electron correlation effects including spatial fluctuations within a cluster are fully taken into account . we have shown in the main text and fig . [ fig.4another ] the single - particle spectral functions using the 12-site cluster at finite temperatures . we also show in figs . [ fig.akw1 ] and [ fig.akw2 ] the single - particle spectral functions for the 16-site cluster at zero temperature ( see also supplementary note 8) , where @xmath125 in eqs . ( [ eq.ge ] ) and ( [ eq.gh ] ) . the results clearly demonstrate that the electron - hole asymmetric reconstruction of the fs is observed irrespectively of the cluster sizes used . as a technical point , we have made use of the @xmath126 symmetry of the 12-site and 16-site clusters to reduce computational costs for solving the eigenvalue problem of @xmath82 and for calculating the cluster single - particle green s functions . * single - particle spectral function . * as shown in fig . [ fig.lattice]a , the partitioning of the triangular lattice into the 12-site clusters does not respect the @xmath126 symmetry ( for lack of the @xmath127 reflection symmetry ) and the two - site unit cell structure ( as @xmath112 and @xmath113 connect differently oriented dimers ) of the dimer model on the anisotropic triangular lattice . however , these can be restored by averaging the single - particle green s functions over four different momenta , i.e. , @xmath128 . \label{eq.gbar}\ ] ] this is because the average of the green s functions between @xmath129 and @xmath130 restores the @xmath131 reflection symmetry and the addition of the green s functions with wave vectors shifted by a primitive reciprocal vector , i.e. , @xmath132 , recovers the two - site unit cell structure in the bz . therefore , @xmath133 given in eq . ( [ eq.gbar ] ) provides the single - particle green s function relevant for the hubbard model defined on the anisotropic triangular lattice even with the 12-site cluster . on the other hand , partitioning of the lattice into the 16-site clusters respects both the @xmath126 symmetry and the two - site unit cell structure ( see fig . [ fig.lattice]b ) . therefore , the single - particle green s function with 16-site cluster is calculated in a standard way of the cpt for multi - band models described in @xcite . using @xmath133 , the single - particle spectral function is calculated as @xmath134 where a small imaginary part @xmath135 of the complex frequency gives the lorentzian broadening of the spectra . the fs is determined with the single - particle spectral function at zero energy , @xmath136 . we set the lorentzian broadening factor @xmath137 for the single - particle spectral functions and @xmath138 for the fs calculation . the single - particle spectral functions and fs are shown in figs . 4 , 16 and 17 ( @xmath24 = @xmath140.8 , @xmath25 = 7 and @xmath139 = 55 mev ) , and in fig . [ fig.4another ] ( @xmath24 = @xmath140.44 , @xmath25 = 5.5 and @xmath139 = 65 mev ) . the data that support the findings of this study are available from the corresponding authors upon request . we would like to acknowledge teijin dupont films japan limited for providing the pen films . computations have been done using hokusai facility of advanced center for computing and communication at riken . this work was supported by mext and jsps kakenhi ( grant nos . jp16h06346 , 15k17714 , 26102012 and 25000003 ) , jst erato , and mext nanotechnology platform program ( molecule and material synthesis ) . y.k . and k.s . contributed equally to this work . y.k . performed the planning , sample fabrication , cryogenic transport measurements and data analyses . k.s . and s.y . performed all cpt calculations and data analyses . j.p . and t.t . developed the ion gel for @xmath3-cl edlt , performed transfer curve analyses and improved the device performance . y.s . grew @xmath3-cl single crystals of high quality and performed magnetoresistance data analyses . , k.s . and h.m.y . wrote the manuscript . , s.y . , h.m.y . and r.k . supervised the investigation . all authors commented on the manuscript . supplementary information is available in the online version of the paper . reprints and permissions information is available online at www.nature.com/reprints . correspondence and requests for materials should be addressed to y.k . or h.m.y . the authors declare no competing financial interests . -cl without electron correlation . a , * molecular arrangement of the bedt - ttf layer and the effective anisotropic triangular lattice . @xmath139 is the nearest - neighbor hopping ( solid line ) and @xmath47 is the next - nearest - neighbor hopping ( dashed line ) between the bedt - ttf dimers . two differently oriented dimers are indicated by red and blue shades . * b , * schematic of fermi surface without electron interaction . the 1st bz of @xmath3-cl is indicated by a solid square . * c , * tight - binding band structure and density of states ( dos ) for @xmath24 = @xmath140.8 with @xmath139 = 55 mev . here , @xmath140 , z@xmath141 , m@xmath142 , and x@xmath143 . the fermi energy is denoted by dashed lines . ] -cl crystal is shaped into a hall - bar using a pulsed laser beam . * c , * gate voltage dependence of the sheet resistivity at 220 k. * d , e , * temperature dependence of the sheet resistivity under hole doping * ( d ) * and electron doping * ( e)*. the dashed lines denote @xmath144 25.8 k@xmath145 . the gate voltage was consecutively applied in the order of appearance , namely : 0 , @xmath140.6 , @xmath141 , @xmath141.2 , @xmath141.3 and @xmath141.35 v for hole doping and 0.4 , 0.8 , 0.9 , 1 and 1.3 v for electron doping . ] at @xmath13 = @xmath141.2 , 0 and + 1.2 v. the solid line represents the fit 8.78@xmath146/tesla with an activation energy @xmath147= 176 k to the data of the undoped bulk crystal . the error bars were calculated from the standard deviation of the hall resistance vs magnetic field plots . note that the data below 25 k contain further errors owing to the non - ohmic behaviour ( supplementary note 6 ) . * c , * gate voltage dependence of @xmath19 at 30 k. data from three different samples are shown . the solid lines denote the @xmath19 estimated from the volume bounded by the non - interacting fs ( blue ) and by the lenslike closed portion of the fs ( red ) . namely , the blue and red solid lines represent densities of the total carriers and the partial carriers enclosed by the lenslike closed portion , respectively . here , the electron doping concentration @xmath21 is related to the non - interacting hole density @xmath148 (= 1@xmath20 ) which is evaluated for the anisotropic hubbard model with @xmath45 = 0 . on the basis of the fet and edlt measurements , the top horizontal axis is set in that @xmath21 = 0.2 corresponds to @xmath13 = 1 v ( see supplementary note 1 ) . ] -cl for 17% hole doping * ( a ) * , half filling * ( b ) * and 17% electron doping * ( c ) * , determined by the largest spectral intensity at the fermi energy . * d , e , f , * single - particle spectral functions for 17% hole doping * ( d ) * , half filling * ( e ) * and 17% electron doping * ( f)*. the fermi energy is located at @xmath149 = 0 and the parameter set of this model is @xmath24 = @xmath140.8 , @xmath25 = 7 and @xmath139 = 55 mev . ] -axis ( @xmath26-axis ) shows positive ( negative ) seebeck coefficient at room temperature . * b , * temperature dependence of @xmath150 at @xmath13 = @xmath141.3 , 0 and + 1.3 v. here , @xmath151 ( @xmath152 ) is the resistivity along the @xmath26-axis ( @xmath28-axis ) . * c , * contour plot of @xmath150 as a function of @xmath97 and @xmath13 . the applied gate voltages were @xmath141.3 , @xmath141.2 , @xmath141 , @xmath140.8 , @xmath140.6 , @xmath140.4 , @xmath140.2 , 0 , 0.2 , 0.4 , 0.6 , 0.8 , 1 , 1.2 and 1.3 v. data are missing at low temperature and low doping ( white region ) due to the high resistance . ] -cl crystal . a , * schematic side view of the fet and edlt devices . * b * , transfer curves of sample # 5 at 220 k. the gate voltage dependence of the resistivity for the fet ( black square ) and edlt operations ( red circle ) are shown . blue triangle denotes the leakage current for the edlt operation . ] at @xmath154 and @xmath141.2 v , respectively . the measured @xmath16 and the errors are taken from fig . [ fig.3]b in the main text . green thick line is a simulation of @xmath16 at @xmath1551.2 v by supplementary eq . ( [ rhtotal ] ) . it is reproduced by the constant @xmath156 and thermally excited @xmath157 ( black thin line ) down to approximately 22 k. at lower temperature , the simulation is not applicable because @xmath158 is not measurable . ] [ [ supplementary - note-1-comparison - experiment - between - the - fet - and - edlt - devices - using - the - same - kappa - cl - crystal ] ] supplementary note 1 : comparison experiment between the fet and edlt devices using the same @xmath3-cl crystal ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ to check the carrier tunability of the electric double layer on @xmath3-cl , we compared the field effect of the edlt ( deme - tfsi gate ) and fet ( sio@xmath4 gate ) in the same @xmath3-cl crystal at 220 k ( sample # 5 , fig . [ fig.s1 ] ) . using the sio@xmath4 gate , we observed n - type fet behavior with a field - effect mobility of 2.6 @xmath160/vs . on the other hand , a clear ambipolar field effect with a resistance peak at approximately @xmath140.25 v was observed for the deme - tfsi gate . the hysteresis and the leakage current remained small without any signature of a chemical reaction between the electrolyte and @xmath3-cl . as shown in fig . [ fig.s1 ] , the resistivity curves for the fet and edlt almost coincide when we adjust the gate voltage axis . provided that the mobilities of the fet and edlt are equivalent , the capacitance of the electric double layer on the electron - doping side is about 500 times larger than that of the fet . as a result , gate voltage of + 1 v in the edlt corresponded to approximately 20% electron doping ( 3.6@xmath710@xmath161 @xmath2 ) . as shown in fig . [ fig.3]c , the hall coefficient at @xmath13 = + 1 v in sample # 2 ( which shows the smallest error ) gives a hole density of 1.47@xmath710@xmath8 @xmath2 at 30 k. using the half - filled hole density of @xmath3-cl , 1.86@xmath710@xmath8 @xmath2 , the density of injected electrons is estimated as 0.39@xmath710@xmath8 @xmath2 ( @xmath16210@xmath8 @xmath2@xmath141.47@xmath710@xmath8 @xmath2 ) , which corresponds to 21% electron doping . this value coincides the above estimate of 20% within 1% error . however , as stated in supplementary note 5 , the experimental carrier density contains influences of thermally excited carriers in the bulk , and the absolute value of the above estimate of 20% is not as accurate as it appears . despite these complications , the relative trend of the carrier density should be reliable enough to support the discussions in the main text , because similar amount of errors in the same direction ( or baseline shift ) should be included for all the measured data . [ [ supplementary - note-2-repeatability - after - low - temperature - measurements ] ] supplementary note 2 : repeatability after low temperature measurements ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ we checked the effect of multiple temperature cycles together with the gate voltage application , on the transfer curve at 220 k and the temperature dependence of the resistivity in sample # 4 . the applied gate voltage in each temperature cycle ( between 220 and 10 k ) was as follows : 0 , @xmath141.2 , @xmath141 , @xmath140.8 , @xmath140.6 , @xmath140.4 , @xmath140.2 , 0 , 0.2 , 0.4 , 0.6 , 0.8 , 1 , 1.2 , 1.3 1.4 , @xmath141.2 , @xmath141.3 , 0 , 1.3 , 0 , 1.3 and 1.35 v. as shown in fig . [ fig.s2]a , the transfer curve at 220 k was reproducible after 20 temperature cycles despite the slight shift of approximately 0.1 v to the hole - doped side . the temperature dependence of the resistivity ( fig . [ fig.s2]b ) was also reproducible after multiple temperature cycles with the gate voltage application of @xmath361.3 v. therefore , it is unlikely that the sample is mechanically damaged due to thermal stress or degraded by destructive chemical reaction between the sample and gate electrolyte . on the other hand , when we applied gate voltage at room temperature , the resistance irreversibly increased at @xmath163 v ( fig . [ fig.s2]c ) . moreover , when we dropped 1-butyl-3-methylimidazolium iodide ( bmim - i ) on @xmath3-cl , the crystal was immediately dissolved without gate voltage as shown in fig . [ fig.s2]d . thus , although our device is weak against gate voltage application at room temperature and/or ionic liquids with large oxidizability , appropriate choice of ionic liquid and temperature enables repeatable measurements . [ [ supplementary - note-3-negative - magnetoresistance - in - a - metallic - sample - under - electron - doping ] ] supplementary note 3 : negative magnetoresistance in a metallic sample under electron doping ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ in a high - conductivity sample ( sample # 6 ) , we observed negative magnetoresistance at 5 k which is the hallmark of the weak localisation effect . as shown in fig . [ fig.s3 ] , the negative magnetoresistance was fitted with the formula @xcite @xmath164\ ] ] where @xmath165 denotes the digamma function , and @xmath166 , @xmath17 , @xmath167 and @xmath168 are deviation of the sheet conductivity from that under zero magnetic field , the applied magnetic flux density , mean free path and dephasing length of the carriers , respectively . @xmath167 and @xmath168 were estimated to be 13.4 nm and 15.7 nm from the least - squares fitting . these values considerably exceed the distance between bedt - ttf dimers ( @xmath61 nm ) indicating coherent transport , in contrast to the mott insulating state . [ [ supplementary - note-4-gate - voltage - dependence - of - the - hall - mobility ] ] supplementary note 4 : gate voltage dependence of the hall mobility ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ the hall mobility @xmath169 is given by the product of the hall coefficient and the conductivity . figure [ fig.s5]a shows the gate voltage dependence of the hall mobility in sample # 1@xmath143 at 30 k. the results have two implications : first , the hall mobility is more sample - dependent than the hall coefficient ( fig . [ fig.3]c in the main text ) . namely , the mobility is highly sample dependent due to the surface conditions such as roughness , while the carrier density under gate voltage is more robust . second , the hall mobilities are comparable between the hole- and electron - doped states , although the data contain ambiguities owing to the conductivity anisotropy ( on the other hand , the hall effect is isotropic as shown in fig . [ fig.s5]b ) . this does not contradict to our calculations because the suppressed spectral function near the z - m line under hole doping do not predominantly contribute to the hall coefficient and conductivity . therefore , the difference of electron correlation effects between the electron- and hole - doped states can not be examined via the hall mobility in the present study . [ [ supplementary - note-5-influence - of - thermally - excited - carriers - on - the - hall - coefficient ] ] supplementary note 5 : influence of thermally excited carriers on the hall coefficient ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ since the samples consist of several tens of conducting bedt - ttf layers , the @xmath16 values more or less contain information of thermally excited carriers in the bulk . here , we show that the presence of thermally excited carriers in the bulk explains the temperature dependence of @xmath16 ( fig . [ fig.3]b ) . for sample @xmath170 , we can roughly estimate the value at the surface ( at @xmath97 = 30 k and @xmath1551.2 v ) on the assumption that @xmath16 at @xmath97 = 30 k and @xmath13 = 0 v purely originates from the bulk . when two types of hole carriers ( for bulk and surface ) coexist , the hall coefficient @xmath16 is expressed as @xmath171 where @xmath172 , @xmath173 , @xmath174 , and @xmath175 denote the mobility , carrier density , conductivity , and hall coefficient for the bulk / surface , respectively , and @xmath176 . the surface hall coefficient @xmath156 is therefore estimated from @xmath177 by substituting measured values ( @xmath16 @xmath178 , @xmath179 , @xmath180 , @xmath181 and @xmath182 ) at 30 k on the right - hand side of supplementary eq . ( [ rhsurface ] ) , we obtain @xmath183 at @xmath184 k and @xmath185 v. although this value contains ambiguity owing to the resistivity anisotropy , the value is close to that at 20 k ( 13.64 @xmath186 ) and within the error bars of @xmath16 at lower temperatures including @xmath187 20 , 15 and 10 k. the errors are calculated from the standard deviation of the @xmath18 vs magnetic field plots and the ambiguity originating from the non - ohmic behaviour at low temperature ( see supplementary note 6 ) . thus , the true value of @xmath156 at 30 k is probably close to the measured @xmath16 at 20 k , but the data below 30 k also contain non - ohmicity . we therefore used the data at 30 k in fig . [ fig.3 ] . although we have such multiple origins of the errors on @xmath16 value , the general trend of @xmath16 vs @xmath13 plot in the main text should not be as uncertain as the errors indicate , because those errors do not cause a random scattering but cause baseline shift which preserve the relative values between data points . thus we believe that the @xmath16 data are well representing the surface states and support our claims in the main text . having estimated the surface hall coefficient @xmath156 at 30 k , we then simulated the temperature dependence of the total hall coefficient @xmath16 at @xmath188 v with the two - carrier model in supplementary eq . ( [ rhtotal ] ) , by assuming constant surface hall coefficient @xmath189 and thermally excited bulk hall coefficient @xmath190 with @xmath191 k ( black thin line in fig . [ fig.s6 ] ) . the experimentally measured values are also used for the temperature dependence of @xmath192 , @xmath193 , and @xmath194 in supplementary eq . ( [ rhtotal ] ) . notice that the simulation is not applicable for the temperatures lower than approximately 22 k because @xmath193 is not measurable . as shown in fig . [ fig.s6 ] , the measured values of @xmath16 ( red circle ) are reproduced by this simple two - carrier model with constant surface hall coefficient @xmath156 and thermally excited bulk hall coefficient @xmath157 ( green thick line ) . the increase in @xmath16 upon cooling has been thus described by considering the influence of thermally excited carriers in the bulk . [ [ supplementary - note-6-current - voltage - characteristics - in - the - insulating - state - under - hole - doping ] ] supplementary note 6 : current - voltage characteristics in the insulating state under hole doping ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ since the hole - doped samples remained insulating , we measured the current - voltage characteristics at low temperature to test the ohmicity of the conduction . figures [ fig.s4]a and b show the current - voltage characteristics and the current dependence of the longitudinal resistance at @xmath13 = @xmath141.2 v and @xmath159 10 , 15 , 20 , 25 and 30 k. the current - voltage characteristics were non - ohmic below 20 k , whereas the longitudinal and hall resistances did not depend on the applied current at 30 k ( fig . [ fig.s4]b and c ) . the error bars in fig . [ fig.3]b are calculated by taking into account the standard deviation @xmath99 of the @xmath18 vs magnetic field plots and the ambiguity @xmath23 originating from the non - ohmic behaviour . since we applied 1 @xmath33a for the hall measurements , @xmath23 was defined as @xmath195 at each temperature , where @xmath196 and @xmath197 denote the resistances at 0 a and 1 @xmath33a in fig . [ fig.s5]b . the error bars in fig . [ fig.3]b were estimated from @xmath198 ( here , @xmath16 is the most probable value ) . as described in the main text , we have observed experimentally that the electric resistivity exhibits anisotropy in particular for the hole - doped case at low temperature ( @xmath199 k ) , and the anisotropy is gradually weakened with increasing the temperature . here , we calculate the optical conductivity along @xmath28- and @xmath26-axes and show that they indeed exhibits the anisotropy and the temperature dependence consistent with the experiments . the optical conductivity along @xmath28 ( @xmath26 ) direction is defined as @xmath200 where @xmath201 is the fermi distribution function , @xmath202 is the velocity of the non - interacting electrons along @xmath28-axis , and @xmath203 is that along @xmath26-axis . notice that the optical conductivity in supplementary eq . ( [ sigma ] ) does not include the vertex corrections . figures [ fig.opt]*a * , * b * , and * c * show the temperature dependence of the optical conductivity along the @xmath28- and @xmath26-axes for hole - doped , half - filled , and electron - doped cases , respectively . here we use the parameter set of @xmath53 , @xmath54 , and @xmath55 mev . at half filling , a broad peak appears at @xmath204 ev . this peak corresponds to the excitation from the lower hubbard band to the upper hubbard band . although the peak is slightly broadened when the temperature is increased , its position and overall structure are almost independent of the temperature since the energy scale of the electron interaction @xmath205 ev @xmath206 k is much larger than that of the temperature considered . in the doped cases , the peak at @xmath204 ev found at half filling disappears while a shoulder remains at @xmath207 ev . this spectral - weight transfer is the characteristic feature of the doped mott insulators @xcite also observed in the cuprate high - temperature superconductor @xcite . as shown in fig . [ fig.opt]*b * , the anisotropy is weak at half filling . this is because the high - energy excitations across the mott gap originates in the local electronic states , and therefore it is insensitive to the direction . when the carriers are doped , the zero - energy excitations become possible and the anisotropy is clearly observed in the low - energy region ( @xmath208 ev ) . for the hole doped case ( fig . [ fig.opt]*a * ) , @xmath209 is twice as large as @xmath210 at @xmath211 k in the low - energy limit , but the anisotropy becomes weaker when the temperature is increased to @xmath106 k. for the electron doped case at @xmath211 k , the anisotropy is smaller than that in the hole - doped case . however , as shown in fig . [ fig.opt]*c * , the anisotropy is reversed under the electron doping . this is in good accordance with the resistivity measurement under the electron doping at low temperatures . we have thus shown that the low - energy excitations in the optical conductivity exhibit significant spatial anisotropy and temperature dependence , which is in good qualitative agreement with the resistivity experiment . [ [ supplementary - note-8-additional - results - at - zero - temperature - with - the-16-site - cluster ] ] supplementary note 8 : additional results at zero temperature with the 16-site cluster ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ figure [ fig.akw1 ] summarizes the results of the 16-site cluster with @xmath50 , @xmath51 , and @xmath52 mev calculated at zero temperature . the electron - hole asymmetry of the fermi surface can be seen clearly in fig . [ fig.akw1]a-c . the single - particle spectral functions in fig . [ fig.akw1]d-f exhibit qualitatively the same results with those of the 12-site cluster shown in the main text . [ fig.akw1]g-i show the density of states @xmath212 where @xmath213 is the volume of the bz and the integration over @xmath89 is taken in the bz . the substantial suppression of the density of states at the fermi energy , i.e. , the pseudogap , is observed in the hole - doped case ( see fig . [ fig.akw1]g ) , consistent with the observation of the fermi arc . on the other hand , such suppression of the density of states at the fermi energy is hardly seen in the electron doped case ( see fig . [ fig.akw1]i ) . these results demonstrate that the electron - hole asymmetry is not affected by the clusters chosen for the cpt calculations . finally , we show the fermi surfaces in unfolded " bz in fig . [ fig.akw2 ] . these plots are easily compared with other theoretical calculations often performed on the triangular lattice with the single - site unit cell @xcite . [ fig.akw2]a-c show the fermi surface calculated with the 16-site cluster at zero temperature . here the unfolded bz is centred at @xmath76 point of the first bz of the triangular lattice with the two - site unit cell . [ fig.akw2]d-f are the same as panels a - c , but the unfolded bz is centered at @xmath76 point of one of the second bzs . in both plots , one can again observe clearly the electron - hole asymmetric reconstruction of the fermi surface . 99 dagotto , e. correlated electrons in high - temperature superconductors . phys . _ * 66 * , 763 - 840 ( 1994 ) . imada , m. , fujimori , a. & tokura , y. metal - insulator transitions . phys . _ * 70 * , 1039 - 1263 ( 1998 ) . lee , p. a. , nagaosa , n. & wen , x .- g . doping a mott insulator : physics of high - temperature superconductivity . _ _ * 78 * , 17 - 85 ( 2006 ) . segawa , k. , kofu , m. , lee , s .- h . , tsukada , i. , hiraka , h. , fujita , m. , chang , s. , yamada , k. & ando , y. zero - doping state and electron - hole asymmetry in an ambipolar cuprate . _ nature phys . _ * 6 * , 579 - 583 ( 2010 ) . williams , j. , kini , a. , wang , h. , carlson , k. d. , geiser , u. , montgomery , l. , pyrka , g. , watkins , d. & kommers , j. from semiconductor - semiconductor transition ( 42 k ) to the highest-@xmath0 organic superconductor @xmath3-(et)@xmath4cu[n(cn)@xmath4]cl ( @xmath0 = 12.5 k ) . _ * 29 * , 3272 - 3274 ( 1990 ) . yamamoto , h. m. , ueno , j. & kato , r. critical behavior of a filling - controlled mott - transition observed at an organic field - effect - transistor interface . _ phys . j. spec . top . _ * 222 * , 1057 - 1063 ( 2013 ) . suda , m. , kawasugi , y. , minari , t. , tsukagoshi , k. , kato , r. & yamamoto , h. m. strain - tunable superconducting field - effect transistor with an organic strongly - correlated electron system . * 26 * , 3490 - 3495 ( 2014 ) . kawasugi , y. , yamamoto , h. m. , tajima , n. , fukunaga , t. , tsukagoshi , k. & kato , r. field - induced carrier delocalization in the strain - induced mott insulating state of an organic superconductor . lett . _ * 103 * , 116801 ( 2009 ) . gros , c. & valent , r. cluster expansion for the self - energy : a simple many - body method for interpreting the photoemission spectra of correlated fermi systems . b _ * 48 * , 418 - 425 ( 1993 ) . snchal , d. , perez , d. & pioro - ladrire , m. spectral weight of the hubbard model through cluster perturbation theory . lett . _ * 84 * , 522 - 525 ( 2000 ) . kawasugi , y. , yamamoto , h. m. , tajima , n. , fukunaga , t. , tsukagoshi , k. & kato , r. electric - field - induced mott transition in an organic molecular crystal . b _ * 84 * , 125129 ( 2011 ) . luttinger , j. m. fermi surface and some simple equilibrium properties of a system of interacting fermions . _ * 119 * , 1153 - 1163 ( 1960 ) . mielke , c. h. , harrison , n. , rickel , d. g. , lacerda , a. h. , vestal , r. m. & montgomery , l. k. fermi - surface topology of @xmath3-bedt - ttf@xmath4cu[n(cn)@xmath4]br at ambient pressure . b _ * 56 * , r4309-r4312 ( 1997 ) . oshima , k. , mori , t. , inokuchi , h. , urayama , h. , yamochi , h. & saito , g. shubnikov - de haas effect and the fermi surface in an ambient - pressure organic superconductor [ bis(ethylenedithiolo)tetrathiafulvalene]@xmath4cu(ncs)@xmath4 . b _ * 38 * , 938 - 941 ( 1988 ) . marshall , d. s. , dessau , d. s. , loeser , a. g. , park , c .- h . , matsuura , a. y. , eckstein , j. n. , bozovic , i. , fournier , p. , kapitulnik , a. , spicer , w. e. & shen , z .- x . unconventional electronic structure evolution with hole doping in bi@xmath4sr@xmath4cacu@xmath4o@xmath214 : angle - resolved photoemission results . lett . _ * 76 * , 4841 - 4844 ( 1996 ) . norman , m. r. , ding , h. , randeria , m. , campuzano , j. c. , yokoya , t. , takeuchi , t. , takahashi , t. , mochiku , t. , kadowaki , k. , guptasarma , p. & hinks , d. g. destruction of the fermi surface in underdoped high-@xmath0 superconductors . _ nature _ * 392 * , 157 - 160 ( 1998 ) . yoshida , t. , zhou , x. j. , sasagawa , t. , yang , w. l. , bogdanov , p. v. , lanzara , a. , hussain , z. , mizokawa , t. , fujimori , a. , eisaki , h. , shen , z .- x . , kakeshita , t. & uchida , s. metallic behavior of lightly doped la@xmath215sr@xmath216cuo@xmath217 with a fermi surface forming an arc . lett . _ * 91 * 027001 ( 2003 ) . anderson , p. w. absence of diffusion in certain random lattices . * 109 * , 1492 - 1505 ( 1958 ) . abrahams , e. , anderson , p. w. , liccardello , d. c. & ramakrishnan , t. v. scaling theory of localization : absence of quantum diffusion in two dimensions . lett . _ * 42 * , 673 - 676 ( 1979 ) . lee , k. h. , kang , m. s. , zhang , s. , gu , y. , lodge , t. p. & frisbie , c. d. `` cut and stick '' rubbery ion gels as high capacitance gate dielectrics . _ adv . mater_. * 24 * , 4457 - 4462 ( 2012 ) . tajima , n. , yamauchi , t. , yamaguchi , t. , suda , m. , kawasugi , y. , yamamoto , h. m. , kato , r. , nishio , y. & kajita , k. quantum hall effect in multilayered massless dirac fermion systems with tilted cones . b _ * 88 * , 075315 ( 2013 ) . komatsu , t. , matsukawa , n. , inoue , t. & saito , g. realization of superconductivity at ambient pressure by band - filling control in @xmath3-(bedt - ttf)@xmath218cu@xmath218(cn)@xmath219 . _ j. phys . jpn . _ * 65 * , 1340 - 1354 ( 1996 ) . kurosaki , k. , shimizu , y. , miyagawa , k. , kanoda , k. & saito , g. mott transition from a spin liquid to a fermi liquid in the spin - frustrated organic conductor @xmath3-(et)@xmath218cu@xmath218(cn)@xmath219 . lett . _ * 95 * , 177001 ( 2005 ) . suda , m. , kawasugi , y. , minari , t. , tsukagoshi , k. , kato , r. & yamamoto , h. m. strain - tunable superconducting field - effect transistor with an organic strongly - correlated electron system . _ * 26 * , 3490 - 3495 ( 2014 ) . dahnken , c. aichhorn , m. , hanke , w. , arrigoni , e. & potthoff , m. variational cluster approach to spontaneous symmetry breaking : the itinerant antiferromagnet in two dimensions . b _ * 70 * , 245110 ( 2004 ) . uchida , s. ido , t. , takagi , h. , arima , t. , tokura , y. & tajima , s. optical spectra of la@xmath215sr@xmath220cuo@xmath221 : effect of carrier doping on the electronic structure of the cuo@xmath218 plane . b _ * 43 * , 7942 - 7954 ( 1991 ) .

current theories of structure formation predict specific density profiles of galaxy dark matter haloes , and with weak gravitational lensing we can probe these profiles on several scales . on small scales , higher - order shape distortions known as flexion add significant detail to the weak lensing measurements . we present here the first detection of a galaxy - galaxy flexion signal in space - based data , obtained using a new shapelets pipeline introduced here . we combine this higher - order lensing signal with shear to constrain the average density profile of the galaxy lenses in the hubble space telescope cosmos survey . we also show that light from nearby bright objects can significantly affect flexion measurements . after correcting for the influence of lens light , we show that the inclusion of flexion provides tighter constraints on density profiles than does shear alone . finally we find an average density profile consistent with an isothermal sphere . [ firstpage ] cosmology : observations gravitational lensing : weak galaxies : haloes dark matter

1011.3041

weak gravitational lensing is a powerful technique for studying the distribution of matter in the universe due to its ability to model the matter distribution in foreground structures , independent of the nature of the matter present . as the light from background sources is bent around foreground lenses , the galaxy images get distorted by the tidal gravitational field . the first - order distortion is known as shear and is essentially an elongation of the image causing the source galaxy to appear stretched in one direction . this type of distortion measurement has been used in a wide variety of cosmological studies ranging from modeling the large - scale structure using cosmic shear ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * for reviews ) to determining galaxy halo shapes using galaxy - galaxy lensing @xcite . first described by @xcite , the second - order distortion is a relatively new addition which has since been named flexion @xcite . there are two types of flexion relevant to weak lensing studies : the first flexion induces a skewness of the brightness profile whilst the second flexion is a three - pronged distortion . in combination with shear , these distortions cause the well - known banana shape of lensed source images . as flexion is effectively the gradient of shear , it is sensitive on small scales . this makes it an important complement to shear which is sensitive on relatively large scales only . by virtue of this , and of the orthogonality of the three measurements , flexion is highly beneficial to investigations of the inner profiles of dark matter haloes , where baryons become important , and to the detection of substructure in cluster haloes . indeed , it was recently shown @xcite that mass reconstructions profit from the use of flexions in combination with shear , and flexion has already been used to constrain the halo mass distribution and to detect substructure in clusters of galaxies @xcite . to provide more information on substructure and mass profiles , there are currently new statistical flexion tools being developed ( eg . * ; * ? ? ? * ; * ? ? ? another application , as discussed in @xcite , is to use both flexions in combination with shear to significantly tighten the constraints on galaxy halo ellipticities compared to using shear alone . the shape measurement technique known as shapelets @xcite works by decomposing a galaxy image into a series of 2d hermite polynomials . these provide a simple framework for describing the main galaxy image distortion operators , such as shear and flexion , and the convolution with the point - spread function ( psf ) . due to the flexible treatment of the psf , the shapelets formalism has an advantage over the currently most widely used shape measurement method , ksb ( from * ? ? ? * ) , since ksb uses an idealised model for the psf whilst shapelets is more versatile . the ksb equivalent for flexion is known as holics @xcite . since the field of weak lensing is relatively new , lensing measurements are continuously being improved in accuracy and applicability . being a statistical technique , however , the accuracy of the weak lensing results depends heavily on the amount of data available . galaxy - galaxy flexion has been tentatively observed @xcite using the ground - based deep lens survey ( dls ) , but to further investigate galaxy - size haloes more and better data is needed . with large surveys such as the canada - france - hawaii telescope legacy survey ( cfhtls ) and the red sequence cluster surveys ( rcs , rcs2 ) available , and new surveys like the 1500 square degree kilo - degree survey ( kids ) imminent , the future looks bright . however , a space - based data set provides better resolution and such a data set is already accessible to us : the hst cosmos survey . using this data we will in this paper improve on the galaxy - galaxy flexion measurements of @xcite . this paper is organised as follows : in section [ sec : flexion ] we review the formalism for shear and flexion , whilst we review the shapelets method in section [ sec : shapelets ] with a description of our implementation ( dubbed the mv pipeline ) in section [ sec : shapelets_mv ] . in section [ sec : tests ] we test the mv pipeline on simulations and in section [ sec : cosmos ] the pipeline is applied to data from the cosmos survey . we conclude in section [ sec : discussion ] . throughout this paper we assume the following cosmology ( wmap7 ; * ? ? ? * ) : @xmath0 we begin by briefly reviewing the weak lensing formalism . flexion is a second - order lensing effect first introduced by @xcite and further developed by @xcite . it arises from the fact that convergence and shear are not constant across a source image , and can be used to describe how these fields fluctuate . in the weak lensing regime , the lensed surface brightness of a source galaxy , @xmath1 , is related to the unlensed surface brightness , @xmath2 , via @xmath3\frac{\partial}{\partial x_i}\right\}f_0(\mathbf{x}).\ ] ] here @xmath4 is the identity matrix , @xmath5 denotes lensed coordinates , and @xmath6 with a second derivative of the lensing , where subscripts denote partial differentiation . and are the two components of the complex shear . the matrix @xmath7 describes how convergence and shear vary across a source image . we can re - express @xmath8 as the sum of two flexions : . the two flexions , the first flexion @xmath9 ( known as or one - flexion ) and the second flexion @xmath10 ( known as or three - flexion ) , are the derivatives of the convergence and shear fields . there are four flexion components , each of which may be written in terms of the third derivatives of the lensing potential @xcite : @xmath11 where and are the complex f and g flexions respectively . the full matrices @xmath12 and @xmath13 in terms of the four flexion components are written explicitly in . the shapelets basis function set was introduced by @xcite and is more fully described there . in summary , the surface brightness of an object @xmath1 can be expressed as a sum of orthogonal 2d functions @xmath14 where @xmath15 are the shapelets coefficients and the shapelets basis functions @xmath16 are defined as @xmath17 here @xmath18 is a normalization constant , @xmath19 is the shapelets scale radius , @xmath20 are coordinates on the image plane and @xmath21 is a hermite polynomial of order @xmath22 . the shapelets basis functions are easily recognised as the energy eigenstates of the 2d quantum harmonic oscillator ( qho ) . the formalism developed for the qho can also be applied to shapelets , providing analytical expressions for transformations such as shear and flexion . in theory , an object can be perfectly described through a decomposition into shapelets up to order but in practice the expansion has to be truncated . we truncate at combined order to avoid introducing a preferred direction . convolution with the point - spread function ( psf ) can also be done analytically in the shapelets formalism by simply multiplying the shapelets expansion by a psf matrix @xmath23 : @xmath24 where @xmath25 are the shapelets coefficients of the psf and @xmath26 , @xmath27 and @xmath28 are the scale radii of the psf , the object and the resulting psf convolved object respectively . @xmath29 is a convolution tensor which depends on the different scale radii and the full expression is given in @xcite . we introduce here an implementation of the shapelets method which builds on a previous implementation described in @xcite ( hereafter ) . this approach creates a shapelets representation of the brightness profile of a psf - convolved galaxy image . it also creates a model circular source and applies shear and flexion to it before convolving it with the point - spread function ( psf ) . finally it fits the galaxy image to this modeled source in order to find the amount by which it has been sheared and flexed . to first order in ellipticity @xmath30 and flexions @xmath31 and @xmath32 , the model object can be written as @xmath33\sum_{\mathrm{even}}^{n_c}c_nc^n\ ] ] where @xmath23 is the psf matrix , @xmath34 , @xmath35 , @xmath36 and @xmath37 are the translation , shear , f flexion and g flexion operators respectively and @xmath38 , @xmath39 , @xmath40 and @xmath41 are the corresponding coefficients . the translation terms here ensure that fits spoiled by undue centroid shifts are caught . the operators are acting on a circular source which can be expressed as a series of circular shapelets @xmath42 with coefficients @xmath43 where @xmath22 is even and the series is truncated at @xmath44 . the reason for truncating at @xmath45 rather than @xmath46 is to safeguard against psf structure at higher orders affecting the highest order shapelets used . to avoid introducing signal - to - noise ( s / n ) dependent biases , the @xmath46 is kept constant for all galaxies rather than being allowed to vary according to size or brightness . for faint sources , this means the higher - order coefficients will be noisy but unbiased . once we have a cartesian shapelets representation of both the sheared , flexed and psf convolved circular model and of the psf convolved object we want to fit , we convert them both into polar shapelets as described in . polar shapelets are simply cartesian shapelets of order @xmath47 expressed in polar coordinates , resulting in polar shapelets of order @xmath22 with angular order @xmath48 and @xmath49 even . the construction of these is discussed in @xcite and further investigated in @xcite and @xcite . in our implementation , the purpose of converting the model and object shapelets expansions into polar shapelets is to avoid truncation effects . , shear and operators acting on a polar shapelet of order @xmath50 generate terms at order @xmath51 , @xmath52 and @xmath53 respectively . by truncating the polar shapelets expansion in the diamond shape shown in figure [ fig : polars ] , i.e. only including terms up to order @xmath54 , @xmath55 , @xmath56 and @xmath57 in the fit , we minimise truncation effects from the mixing of orders . the model is fit to each source using least - squares , resulting in a simultaneous estimate for the ellipticity @xmath58 , the f flexion @xmath59 , and the g flexion @xmath60 . as explained in , the errors on the shapelet coefficients are derived from the photon noise and propagated through the @xmath61 function for this fit . by differentiating the @xmath61 at the best - fit , we obtain the covariances between the fit parameters , resulting in proper error estimates . in essence , the main development since is the addition of flexion to the model and the inclusion of higher order polar shapelets ( @xmath62 ) in the fit . several aspects of the pipeline , such as the choice of scale radius @xmath19 , the method of psf correction and the effect of noise on ellipticity estimates , have been thoroughly tested in as part of the development of the pipeline . in this section we will therefore focus on testing the recovery of shear and flexion . as participants in the gravitational lensing accuracy testing 2008 ( great08 ) challenge @xcite , we were able to contrast the shear measurement capability of the pipeline with that of the mv pipeline under different observing conditions . the challenge provided a large number of simulated sheared and pixelated galaxy images with added noise . the performance of the different shape measurement pipelines taking part was quoted in terms of a quality factor , or q - value , defined as @xmath63 where @xmath64 is a combination of the statistical spread in the simulations and the expected systematic errors . the superscripts @xmath65 and @xmath66 denote measured and true ( input ) values respectively and @xmath67 is the shear component @xmath68 for simulation image @xmath69 . the differences between the measured and true shears are averaged over different input shear sets @xmath70 and simulation branches @xmath71 . the whole expression is normalised by @xmath72 so that a method with a purely statistical spread in the measured shears will have a q - value of @xmath72 which is the level desirable for future surveys . in the case of great08 , @xmath73 and @xmath74 , giving a q - value nominator of @xmath75 . with this definition current methods , like those that took part in the earlier shear testing programme ( step ) @xcite , generally achieve @xmath76 . this is sufficient for current weak lensing surveys . for a more in - depth discussion on the q - value and its relation to the step parameters @xmath65 ( multiplicative bias ) and @xmath77 ( additive bias ) , we refer to @xcite . the overall q - value was similar for the and the mv pipelines , both in the lownoise_blind competition ( @xmath78 ) and in the realnoise_blind ( @xmath79 ) . when broken down into the separate observing condition branches some differences became apparent . in general the mv pipeline did exceptionally well under `` good '' observing conditions , e.g. for the high s / n branch or for well resolved galaxies . our own simulations described in the next section will further test the dependence of the mv performance on different observing conditions . .the different branches of flashes . four parameters are varied between the branches according to this table . [ cols="<,^,^,^,^",options="header " , ] as there is no flexion simulation set publicly available to date , we create our own flexion and shear simulations ( flashes ) . flashes are very similar to the great08 simulations in several respects . first , each galaxy is generated on a grid , ensuring that there is no overlap of objects , thus avoiding deblending issues . second , each simulation image consists of @xmath80 such objects . third , each galaxy is generated through the following sequence : ( i ) simulate a sheared and/or flexed ( elliptical ) galaxy model ( depending on simulation branch ) ; ( ii ) convolve with the psf ; ( iii ) apply the noise model . four parameters are varied between the different flashes branches ; the intrinsic galaxy shape , the light profile of the galaxies , the s / n of the galaxies and the shape of the psf . these parameters are detailed below and summarised in table [ tab : branches ] . all parameters except for the intrinsic ellipticities are kept constant in each simulation image , and all images are created using monte - carlo selection . this is very similar to the process described in and in @xcite , but with the photon trajectories being influenced by flexion as well as by shear if required . the galaxies are modeled with srsic intensity profiles @xcite with varying indices @xmath22 . a srsic index of @xmath81 is a gaussian profile whilst @xmath82 and @xmath83 are exponential and de vaucouleur profiles respectively . half of the flashes branches have intrinsically round galaxies whilst the other half consists of galaxies with intrinsic ellipticities picked randomly from the ellipticity distribution of objects in the cosmos survey . there is no intrinsic flexion included . the psfs applied to the simulations are nearly gaussian with a moffat profile @xmath84 of index @xmath85 . in half of the branches , the psf is round whilst in the other half it is elliptical in the horizontal direction with @xmath86 . to mimic the properties of the cosmos survey , we use a psf fwhm of 2.1 pixels and a psf convolved galaxy size of 5.8 pixels which is the typical size of the galaxies we use in our cosmos analysis . finally there are four s / n branches , with s / n being defined as . it is expected that shape measurements will be less accurate at low s / n . for this reason the mv pipeline applies a s / n cut at 10 in general . the low s / n branch of 8 is designed to test how biased measurements are below this cut . the high s / n branch of 100 tests biases under near - perfect noise conditions . the strength of the different distortions is picked randomly but with the following maximum values : @xmath87 , @xmath88 and @xmath89 . the value of each distortion component is kept constant across each image , but differs between the 30 images in each set , and between different sets . . for the results for @xmath90 , @xmath91 and @xmath92 please refer to appendix [ app : flashes].,width=317,height=317 ] to estimate the average distortion on each image we use two different techniques : a weighted average with weights inversely proportional to the measurement errors , and convex hull peeling ( chp ) . chp is an efficient way of eliminating outliers and is essentially a 2d median . a convex hull , in the context of a point cloud in e.g. the @xmath93 plane , is the minimal convex set of points containing that point cloud . thus if all the points in this convex set were connected , a polygon containing the entire point cloud would be produced . by peeling away convex hulls , outliers are removed from the point cloud and the remaining points may be averaged over to produce a mean unaffected by extreme results . this is the averaging technique we used in great08 where we peeled away 50% of the measurements before averaging . we employ the parameters @xmath65 and @xmath77 as used in step @xcite to quantify the performance of the software : @xmath94[eq : step_mc ] and similarly for the flexions , where @xmath95 represents the shear component . a negative multiplicative bias @xmath96 thus indicates that the distortion is generally underestimated . a systematic offset @xmath97 may be caused by e.g. insufficient psf correction . in figure [ fig : flashes_results ] we show the multiplicative bias of the first component for each of shear , f flexion and g flexion as a function of the different simulation branches ( please refer to appendix [ app : flashes ] for the remaining bias components ) . for these results we use a shapelets order of @xmath98 . we use sextractor @xcite to detect the objects in each simulation , which we then split into clean star and galaxy catalogues by matching to the input catalogue . we keep all properties apart from the one under investigation fixed at a fiducial value to allow for a fair comparison . the fiducial simulations in figure [ fig : flashes_results ] have intrinsically round , high s / n galaxies with gaussian light profiles and a circular psf . from the above figure it is clear that both flexions are likely to be underestimated , especially for higher srsic indices . the bias is also strongly s / n dependent , particularly for the f flexion . thus a s / n cut is essential to improve the performance of the mv pipeline , but a bias correction should also be implemented . investigating the dependence of @xmath65 on s / n further , we are able to fit the following power - law to our flashes results : @xmath99 where @xmath100 and @xmath101 are constants as follows : for shear ; for f flexion ; for g flexion . we will apply this bias correction to our shape measurements in cosmos , but since flashes have been tailored for this particular data the biases should be explored further before being applied to other surveys . at the core of weak galaxy - galaxy lensing is the averaging of the signal in rings centered on lenses consisting of single galaxies rather than a galaxy cluster . this type of analysis is robust as numerous systematics , induced by e.g. the psf , cancel out . different systematics may however be introduced , such as the light from the central , often bright , lens causing biases in the shape measurements as discussed in @xcite . to study this possible effect , we created simple simulations with sources placed in evenly spaced rings around a central lens . apart from source numbers and positions , the simulations were created in the same way as flashes . the s / n of the images was set to 200 to ensure minimum bias , and for the same reason the source galaxies had gaussian light profiles . the size and profile parameters of the lens were varied between images . the results for a lens with an exponential profile are shown in figure [ fig : galgal_results ] ( black stars ) , where we have used @xmath98 . we recover a near - perfect average signal in each source circle far from the lens . however , close to the lens the shear and g flexion are slightly affected , but , more strikingly , the f flexion is severely overestimated . the conclusion we draw from this is that _ bright objects can add significantly to the f flexion signal _ , due to light ` leaking ' into the shapelets fitting radius . this causes the pipeline to detect a source light profile that is skewed towards the lens , and interpreting it as extra f flexion . our solution is to remove any bright objects sufficiently close to the source being fit using a technique we introduce here as bright object removal ( bor ) . before decomposing a galaxy image into shapelets , we identify any bright objects that could conceivably intrude using selection criteria based on distance between the two objects , shapelets fitting radius of the source , and size and brightness of the intruding object . we then create srsic models of the intruding objects using galfit @xcite and subtract these models from the shapelets stamp before doing the fitting . it works well in these simulations , provided one is careful with the parameters given to galfit as input . the sky background value given to galfit is particularly important as a small error in this estimate results in postage stamp artifacts when the stamps are subtracted from the original image . in figure [ fig : galgal_results ] we also show the results if bor is switched on whilst the rest of the analysis is kept identical to the previous run ( green triangles ) . there is still some excess f flexion signal around 44 pixels , indicating that there may be some residual light remaining , but this excess is smaller than for the uncorrected measurements . this provides a confirmation that the measured signal reproduces the input signal well if bor is applied , and no new artifacts are introduced . we note , however , that the leaking light does not affect the cross component of the measurements , with the consequence that this effect can not be detected through the usual systematic checks . @xcite made a first detection of galaxy - galaxy flexion using the ground - based dls , proving that flexion can indeed be detected , but ultimately they were hampered by the small size of their sample , the lack of redshifts and the extra blurring caused by the atmosphere . therefore we choose the space - based cosmic evolution survey ( cosmos , @xcite ) as the first real dataset for the mv pipeline . thanks to the depth of this survey we will have access to more than a thousand times as many lens - source pairs as @xcite did . more than half of these have photometric redshifts meaning that the division of the sample into lenses and sources will be more accurate . the intention is to provide independent confirmation that galaxy - galaxy flexion has high enough s / n to be detected , and that the software presented in this paper is able to do it . we will also look closer to the lens than previous analyses and attempt to combine shear and flexion to give constraints on galaxy dark matter halo profiles . cosmos is to date the largest contiguous field imaged by the hubble space telescope ( hst ) with a total area of 1.64 deg@xmath102 . the 579 tiles were observed in f814w ( i - band ) by the advanced camera for surveys ( acs ) between october 2003 and november 2005 . each tile consisted of 4 dithered exposures of 507 seconds each ( 2028 seconds in total ) with about 95% of the survey area benefiting from the full 4 exposures . we use the images reduced by @xcite ( hereafter ) and also their catalogues for stars and galaxies , detected using sextractor . there are a total of 446 934 galaxies with @xmath103 in the mosaic catalogue , almost half of which have cosmos-30 photometric redshifts from @xcite . these redshifts are magnitude limited and cover the entire cosmos field to a depth of @xmath104 . galaxy - galaxy lensing is less affected by the problems plaguing cosmic shear analyses , since most systematic shape distortions induced by instruments cancel out when azimuthally averaged . still , we have to be careful not to introduce new systematic effects or biases , so correcting for the psf and the charge - transfer inefficiency ( cti ) ( e.g. * ? ? ? * ; * ? ? ? * ) is important . we use all galaxies with redshifts of @xmath105 as lenses . at higher redshifts the light from the lensing galaxies becomes difficult to account for due to the small angular separation on the sky , as explained further in appendix [ app : highz ] . furthermore , imposing a lens redshift cut will ensure that the vast majority of sources are truly background objects . our source catalogue is comprised of all objects with a shape measurement . we clean this catalogue using a series of conditions on size and measured shape , detailed in appendix [ app : cosmos_catalogues ] , the most important of which is to remove objects with @xmath106 . roughly two - thirds of the remaining sources have individual cosmos-30 photometric redshifts assigned to them . for the remaining third ( redshift bin 6 in ) we use the estimated redshift distribution employed by to assign mean angular diameter distance ratios ( @xmath107 ) to each lens - source pair . we are finally left with @xmath108 sources , corresponding to a source density of @xmath109 . for the pairs we use , the median lens redshift is @xmath110 and the median source redshift is @xmath111 . despite the excellent space - based resolution , we need to correct the galaxy shapes for the instrumental psf . the acs psf is known to fluctuate both spatially and temporally ( e.g. * ? ? ? * ; * ? ? ? * ) , a variation mostly driven by changes in telescope focus caused for example by the breathing of the telescope . we can map the psf using stars , but , in high - galactic latitude acs fields typically only @xmath112 stars are present . this number is too low for the standard approach of a polynomial interpolation . instead , we closely follow the analysis of , who conducted a principal component analysis ( pca ) of the acs psf variation as measured in dense stellar fields . details for the shapelets implementation of pca may be found in appendix [ app : cosmos_psf ] . a challenge with using ccd detectors in space is that they are not protected by the atmosphere . exposed , they continuously get bombarded by radiation , causing deterioration of the chip surface . the imperfections created in this way act as charge traps which causes inefficiency in the moving of electrons to read - out . this effect is known as cti ( e.g. * ? ? ? * ; * ? ? ? as the electrons get trapped and then released at a later point , charge trails following objects are created in the read - out direction , effectively causing a spurious shear signal in that direction . our correction for cti again closely follows , who derive parametric corrections for the change in polarization for both galaxies and stars . for more details on this correction , please refer to appendix [ app : cosmos_cti ] . once corrected , the galaxy - galaxy shear and flexion signals are weighted according to the geometric lensing efficiency of each lens - source pair . in the case of flexion there is an extra scale dependence of the signal . for the navarro - frenk - white ( nfw ) profile @xcite , the strength of the shear signal scales as @xmath113[eq : nfw_shear ] where @xmath114 , @xmath115 and @xmath116 are the angular diameter distances to the lens , to the source , and between lens and source respectively @xcite . the flexion signals scale as @xmath117[eq : nfw_flexion ] . we therefore weight the signals accordingly , scale them to a reference lens and source redshift and compute the weighted average in 25 logarithmic distance bins ( see appendix [ app : cosmos_signal ] for details ) . we use a reference lens redshift of @xmath118 since that is close to the effective median redshift of our lenses , and a reference source redshift @xmath119 . to estimate the errors on each bin and the covariances between them , we use 5000 bootstrap resamples of our source catalogue . the results from our galaxy - galaxy lensing analysis of the full cosmos lens and source sample is shown in figure [ fig : cosmos_results ] . in the left panel we plot the shear results as a function of physical distance from the lens . these results agree very well with those from ( see appendix [ app : ksb ] ) , providing an independent consistency check . to this we fit a singular isothermal sphere ( sis ) profile and a tentative nfw profile . due to the dependence on mass and redshift of the mean concentration parameter ( e.g. * ? ? ? * ) , the nfw profile is only an indication when the spread in lens masses and redshifts is as great as it is in the above sample . splitting the sample up into redshift and/or mass bins would increase the confidence in the fit , but decrease the s / n of the signals significantly . the middle and right panels show the f and g flexion results respectively , for the same lenses and sources . the profiles plotted here are identical to those plotted in the shear panel but _ translated _ into predictions for @xmath9 and @xmath10 , as opposed to fitted to the flexion data directly . the f flexion has a tendency to be overestimated compared to the predicted profile from the shear , and we investigate this discrepancy further in the following sections . we also note that we measure a g flexion that is very noisy and consistent with zero . this is most likely caused by lack of information in higher @xmath65-order shapelets for fainter sources , and we choose to use only shear and f flexion in the continued analysis . also shown in pink circles in figure [ fig : cosmos_results ] is the signal if we apply the multiplicative s / n - dependent bias correction implied by flashes . with this correction , the f flexion signal becomes slightly higher . this bias correction is only based on one specific set of simulations and is thus rather preliminary ; this is also indicated in the increased size of the error bars . correcting for the morphology - dependent bias requires accurate source morphology determination . using the photometric galaxy type estimates from @xcite as an indicator of morphology we find that @xmath120 of our source sample consists of likely de vaucouleur objects . this type estimate is not accurate enough to implement a morphology bias correction , but simply removing the de vaucouleur candidates we identified makes little difference to our results . it is clear , however , that an accurate bias calibration of the flexion amplitude , taking into account both source s / n and brightness profiles , requires further investigation . ) . black solid points represent the difference between the signals before and after correction , with the f flexion in units of @xmath121 , whilst green circles represent the cross term.,width=226 ] we now explore the tendency of the f flexion points to lie above the predicted profiles . as shown in section [ sec : tests_galgal ] , the shape measured may be affected by bright objects nearby . we implement bor in our cosmos analysis to see the effect on real data . for very well resolved objects , prominent spiral arms and other complications cause galfit to reject the single srsic profile fit . removing these objects , and the residual light from the wings of the profile ( figure [ fig : galgal_results ] ) , requires a more sophisticated model . for now we are only interested in a rough indication of the impact this light leakage has on a galaxy - galaxy signal so we will not correct for the few large objects in this paper . however , as shown in figure [ fig : cosmos_brightbin ] , the correction to the innermost f flexion bin is non - zero even without accounting for the very large objects . the shear is largely unaffected , but for flexion analyses in future deeper and larger surveys it will be important to correct for this effect . since flexion is more sensitive to the underlying mass distribution on small scales than shear is , we expect it to respond differently to the presence of substructure in galaxy haloes . to test whether this has any impact on our analysis we take a galaxy - size sis halo and populate it with subhaloes , allowing 20% of the mass to be in substructure . the total mass of the halo is @xmath122 and the galaxy is placed at @xmath123 with @xmath124 . we spread the substructure mass over 100 subhaloes , randomly distributed according to an sis density profile . finally we average the azimuthally averaged signal over 100 such galaxies . now , subhaloes are generally stripped . to approximate this we use a truncated sis ( tsis ) profile for the subhaloes ( see * ? ? ? * for constraints on parameters ) . the tsis convergence is given by @xmath125 where @xmath126 is the einstein radius and @xmath127 is a truncation scale where the profile steepens . on small scales ( @xmath128 ) the tsis behaves like an sis but at large scales ( @xmath129 ) the profile decreases as @xmath130 . the tsis shear is given in @xcite and the flexions are @xmath131 and @xmath132 where @xmath133 is the position angle of the background source . using the parameters above and a truncation scale for the subhaloes we get the results shown in figure [ fig : tsis ] . the shear profile fit is pulled down slightly compared to a smooth halo but the flexions are not similarly affected . due to the substructure the flexions are more scattered , but the overall trend is for the points to follow the smooth profile , or even slightly above in the f flexion case . thus the flexions seem overestimated compared to the shear fit . we stress however that the fraction of substructure used in this test ( 20% ) is high to exaggerate the effect . the test does show that substructure may affect the flexions differently to the shear , but its influence is likely less than the excess currently observed in cosmos . ( @xmath134 , @xmath135 and @xmath136 respectively ) . the white ( thick ) contour marks the joint confidence limits . the grey - scale is logarithmic in @xmath61.,title="fig:",width=283 ] ( @xmath134 , @xmath135 and @xmath136 respectively ) . the white ( thick ) contour marks the joint confidence limits . the grey - scale is logarithmic in @xmath61.,title="fig:",width=283 ] one of the most interesting potential uses of flexion is as an aid to shear in determining the inner density profiles of dark matter haloes . the two signals are sensitive to the underlying density profile on different scales , so combining the two will give us tighter constraints than either on their own . to combine the shear and flexion signals we have to take any correlation between them into account . assumed that the shear and flexion measurements would be uncorrelated . here we confirm this assumption through the correlation matrix between the shear and flexion bins , using 5000 bootstraps , shown in figure [ fig : cosmos_correlation ] . this implies that it is trivial to combine the shear and flexion information to find the profile of an average lens . we use the f flexion in conjunction with the shear to fit density profiles to the measured signal . for this purpose we try two different families of profiles : the power - law and the nfw . our general power - law is defined as @xmath137[eq : powerlaw_shear ] with @xmath138 the distance from the lens , and the amplitude @xmath139 and the index @xmath22 free parameters . an index of @xmath82 would be equivalent to an sis . the above expression is easily differentiated to give the f flexion @xmath140[eq : powerlaw_flexion ] the expressions for the nfw profiles are somewhat more complicated but they are given in full in @xcite and for shear and flexion respectively . here we leave the virial radius @xmath141 and the concentration @xmath77 as free and independent parameters . we fit the power - law and nfw profiles to the inner only as this is the region where becomes important and the shear profile is not affected by halo - halo contamination . the top panel in figure [ fig : cosmos_profiles ] shows that both the shear and the are consistent with an sis ( @xmath82 ) , although together they prefer a slightly lower power - law index of . the bottom panel shows that it is difficult to constrain the nfw concentration if it is left completely unrestricted . this analysis with two free and independent parameters is not completely representative , however , since simulations indicate a fixed mean mass - concentration relationship @xcite . it is also important to note that the average profile we constrain here is a composite of lenses in a large redshift range . detection at the high end of the redshift distribution tend to be biased towards intrinsically brighter objects than at the low end . we also combine measurements from lenses of different sizes and morphologies . nonetheless , combining shear and f flexion does provide tighter constraints than shear alone on the density profiles , and this is an important proof of concept . the resulting mass estimate for the average lens in cosmos from the combined nfw fit is with a concentration of . we have shown a significant detection of galaxy - galaxy f flexion for the first time with shapelets using the space - based cosmos data set . we used this flexion signal in conjunction with the shear to constrain the average density profile of the galaxy haloes in our lens sample . we found a power - law profile consistent with an sis . furthermore , we showed that the inclusion of f flexion provides tighter constraints on both power - law and nfw profiles , an important proof of concept . the galaxy - galaxy f flexion signal measured in cosmos is slightly higher than expected from the shear signal , especially if we apply the multiplicative bias correction . there is however no indication from the cross term that there are systematics present . the discrepancy could be partly due to insufficient nearby object light removal , but this is unlikely to explain the full offset . substructure in galaxy haloes may cause excess f flexion compared to what the shear measures . however , a large fraction of the galaxy halo mass has to be in substructure in order for the effect to become significant . we note that @xcite also find shear and f flexion signals that are inconsistent with each other ; the velocity resulting from an sis profile fit to their f flexion signal is nearly twice that found using shear . this is qualitatively consistent with our findings , which leads us to believe that there is something more fundamental affecting the signal . in the near future we would like to further investigate the dependence of these discrepancies on lens properties . we measure a galaxy - galaxy g flexion signal that is consistent with the predicted profile , but due to the large measurement errors it is also consistent with zero . this measurement is a lot noisier than the other two , an effect most likely caused by the fact that there is less information available in the higher @xmath65-order shapelets for fainter sources . to measure a g flexion signal we thus require many well - resolved sources , an extravagance not yet awarded us . future large space - based surveys such as euclid will enable us to investigate g flexion further , but for now f flexion is a promising tool in its own right . the software introduced in this paper , the mv pipeline , is able to detect these higher order lensing distortions . we have shown that in practice , the shapelets f flexion measure is affected by light from nearby bright objects and detailed a way to correct for this effect . this bor does require further sophistication to account for large , well resolved galaxies , galaxies which are not well described by the single srsic light profile employed here . from the flashes simulations it is clear that there is more work required in order to improve the accuracy of the f flexion measurements for future surveys . noise related biases are particularly significant for this type of shape measure , and we have modeled these biases in cosmos . in the future we hope to measure flexion on a larger survey , enabling us to reduce the noise so that we can investigate the trend with e.g. redshift and lens mass . a larger number of sources would also enable us to further tighten the profile constraints in the inner regions of dark matter haloes where baryons become important . it is not yet clear how well we can measure flexion on ground - based data , but surveys like kids , cfhtls and rcs2 should provide an excellent test - bed . we would like to thank our colleagues henk hoekstra , edo van uitert and elisabetta semboloni at leiden observatory , and peter schneider at bonn university , for useful discussions . gary bernstein drew our attention to the possibility of using convex hull peeling for our averages , and for this we would like to thank him . mv is supported by the european duel research - training network ( mrtn - ct-2006 - 036133 ) . ts acknowledges support from the netherlands organization for scientific research ( nwo ) . 44 natexlab#1#1 d. j. , amara a. , read j. i. , 2010 , mnras , 1249 d. j. , goldberg d. m. , rowe b. t. p. , taylor a. n. , 2006 , mnras , 365 , 414 e. , arnouts s. , 1996 , a&as , 117 , 393 s. , balan s. t. , bethge m. , gentile m. , harmeling s. , heymans c. , hirsch m. , hosseini r. , jarvis m. , kirk d. , kitching t. , kuijken k. , lewis a. , paulin - henriksson s. , schlkopf b. , velander m. , voigt l. , witherick d. , amara a. , bernstein g. , courbin f. , gill m. , heavens a. , mandelbaum r. , massey r. , moghaddam b. , rassat a. , rfrgier a. , rhodes j. , schrabback t. , shawe - taylor j. , shmakova m. , van waerbeke l. , wittman d. , 2010 , mnras , 405 , 2044 s. , shawe - taylor j. , amara a. , applegate d. , balan berge j. s. t. , bernstein g. , dahle h. , erben t. , gill m. , heavens a. , heymans c. , high f. w. , hoekstra h. , jarvis m. , kirk d. , kitching t. , kneib j. , kuijken k. , lagatutta d. , mandelbaum r. , massey r. , mellier y. , moghaddam b. , moudden y. , nakajima r. , paulin - henriksson s. , pires s. , rassat a. , refregier a. , rhodes j. , schrabback t. , semboloni e. , shmakova m. , van waerbeke l. , witherick d. , voigt l. , wittman d. , 2009 , annals of applied statistics , 3 , 6 a. r. , schaye j. , kay s. t. , dalla vecchia c. , 2008 , mnras , 390 , l64 x. , li g. , schneider p. , 2010 , a&a submitted ( arxiv : astro - ph/1008.3088 ) d. m. , bacon d. j. , 2005 , apj , 619 , 741 d. m. , natarajan p. , 2002 , apj , 564 , 65 a. j. , bridle s. l. , 2009 , mnras , 400 , 1132 c. , van waerbeke l. , bacon d. , berge j. , bernstein g. , bertin e. , bridle s. , brown m. l. , clowe d. , dahle h. , erben t. , gray m. , hetterscheidt m. , hoekstra h. , hudelot p. , jarvis m. , kuijken k. , margoniner v. , massey r. , mellier y. , nakajima r. , refregier a. , rhodes j. , schrabback t. , wittman d. , 2006 , mnras , 368 , 1323 h. , jain b. , 2008 , annual review of nuclear and particle science , 58 , 99 h. , yee h. k. c. , gladders m. d. , 2004 , apj , 606 , 67 o. , capak p. , salvato m. , aussel h. , mccracken h. j. , sanders d. b. , scoville n. , kartaltepe j. , arnouts s. , le floch e. , mobasher b. , taniguchi y. , lamareille f. , leauthaud a. , sasaki s. , thompson d. , zamojski m. , zamorani g. , bardelli s. , bolzonella m. , bongiorno a. , brusa m. , caputi k. i. , carollo c. m. , contini t. , cook r. , coppa g. , cucciati o. , de la torre s. , de ravel l. , franzetti p. , garilli b. , hasinger g. , iovino a. , kampczyk p. , kneib j. , knobel c. , kovac k. , le borgne j. f. , le brun v. , fvre o. l. , lilly s. , looper d. , maier c. , mainieri v. , mellier y. , mignoli m. , murayama t. , pell r. , peng y. , prez - montero e. , renzini a. , ricciardelli e. , schiminovich d. , scodeggio m. , shioya y. , silverman j. , surace j. , tanaka m. , tasca l. , tresse l. , vergani d. , zucca e. , 2009 , apj , 690 , 1236 n. , squires g. , broadhurst t. , 1995 , apj , 449 , 460 t. d. , miller l. , heymans c. e. , van waerbeke l. , heavens a. f. , 2008 , mnras , 390 , 149 e. , smith k. m. , dunkley j. , bennett c. l. , gold b. , hinshaw g. , jarosik n. , larson d. , nolta m. r. , page l. , spergel d. n. , halpern m. , hill r. s. , kogut a. , limon m. , meyer s. s. , odegard n. , tucker g. s. , weiland j. l. , wollack e. , wright e. l. , 2010 , arxiv astrophysics e - prints k. , 2006 , a&a , 456 , 827 a. , king l. j. , 2010 , mnras , 405 , 1854 a. , king l. j. , goldberg d. m. , 2010 , mnras submitted ( arxiv : astro - ph/1009.1018 ) a. , king l. j. , wilkins s. m. , 2009 , mnras , 395 , 1438 r. , hirata c. m. , broderick t. , seljak u. , brinkmann j. , 2006 , mnras , 370 , 1008 r. , seljak u. , kauffmann g. , hirata c. m. , brinkmann j. , 2006 , mnras , 368 , 715 r. , heymans c. , berg j. , bernstein g. , bridle s. , clowe d. , dahle h. , ellis r. , erben t. , hetterscheidt m. , high f. w. , hirata c. , hoekstra h. , hudelot p. , jarvis m. , johnston d. , kuijken k. , margoniner v. , mandelbaum r. , mellier y. , nakajima r. , paulin - henriksson s. , peeples m. , roat c. , refregier a. , rhodes j. , schrabback t. , schirmer m. , seljak u. , semboloni e. , van waerbeke l. , 2007 , mnras , 376 , 13 r. , refregier a. , 2005 , mnras , 363 , 197 r. , rowe b. , refregier a. , bacon d. j. , berg j. , 2007 , mnras , 380 , 229 r. , stoughton c. , leauthaud a. , rhodes j. , koekemoer a. , ellis r. , shaghoulian e. , 2010 , mnras , 401 , 371 d. , valageas p. , van waerbeke l. , heavens a. , 2008 , phys , 462 , 67 j. f. , frenk c. s. , white s. d. m. , 1996 , apj , 462 , 563 y. , umetsu k. , futamase t. , 2007 , apj , 660 , 995 , 2008 , apj , 680 , 1 l. c. , hoekstra h. , hudson m. j. , van waerbeke l. , mellier y. , 2007 , apj , 669 , 21 c. y. , ho l. c. , impey c. d. , rix h. , 2002 , aj , 124 , 266 a. , 2003 , mnras , 338 , 35 a. , bacon d. , 2003 , mnras , 338 , 48 j. d. , massey r. j. , albert j. , collins n. , ellis r. s. , heymans c. , gardner j. p. , kneib j. , koekemoer a. , leauthaud a. , mellier y. , refregier a. , taylor j. e. , van waerbeke l. , 2007 , apjs , 172 , 203 b. p. t. , 2008 , phd thesis , the university of edinburgh p. , rix h. , 1997 , apj , 474 , 25 t. , erben t. , simon p. , miralles j. , schneider p. , heymans c. , eifler t. , fosbury r. a. e. , freudling w. , hetterscheidt m. , hildebrandt h. , pirzkal n. , 2007 , a&a , 468 , 823 t. , hartlap j. , joachimi b. , kilbinger m. , simon p. , benabed k. , brada m. , eifler t. , erben t. , fassnacht c. d. , high f. w. , hilbert s. , hildebrandt h. , hoekstra h. , kuijken k. , marshall p. j. , mellier y. , morganson e. , schneider p. , semboloni e. , van waerbeke l. , velander m. , 2010 , a&a , 516 , a63 + n. , abraham r. g. , aussel h. , barnes j. e. , benson a. , blain a. w. , calzetti d. , comastri a. , capak p. , carilli c. , carlstrom j. e. , carollo c. m. , colbert j. , daddi e. , ellis r. s. , elvis m. , ewald s. p. , fall m. , franceschini a. , giavalisco m. , green w. , griffiths r. e. , guzzo l. , hasinger g. , impey c. , kneib j. , koda j. , koekemoer a. , lefevre o. , lilly s. , liu c. t. , mccracken h. j. , massey r. , mellier y. , miyazaki s. , mobasher b. , mould j. , norman c. , refregier a. , renzini a. , rhodes j. , rich m. , sanders d. b. , schiminovich d. , schinnerer e. , scodeggio m. , sheth k. , shopbell p. l. , taniguchi y. , tyson n. d. , urry c. m. , van waerbeke l. , vettolani p. , white s. d. m. , yan l. , 2007 , apjs , 172 , 38 j. l. , 1968 , atlas de galaxias australes l. , mellier y. , 2003 , preprint ( arxiv : astro - ph/0305089 ) c. o. , brainerd t. g. , 2000 , apj , 534 , 34 the figures shown in this appendix complement figure [ fig : flashes_results ] in the main paper and provide additional detail on the results from running the mv pipeline on flashes , with @xmath98 . the parameters @xmath65 and @xmath77 are defined through @xmath94 and similarly for the flexions , where @xmath95 is the component . we use two different techniques to estimate the average distortion on each we show the multiplicative bias of the second component for each of shear , f flexion and g flexion as a function of the different simulation branches . for these results we use a shapelets order of @xmath98 . this bias behaves as the multiplicative bias of the first component ( figure [ fig : flashes_results ] ) , as expected . the biases of all distortion measurements , and in particular f flexion , are severely dependent on s / n and brightness profile . the additive bias @xmath77 is minimal for shear and f flexion ( see figures [ app : fig : flashes_results_c1 ] and [ app : fig : flashes_results_c2 ] ) indicating that the psf is either well corrected for or not significantly affecting these two measurements . for the g flexion the offset is larger . to maximise the number of lens - source pairs we use all objects with assigned photometric redshifts as sources , but imposing a redshift cut of for lenses . additionally we use sources without individual redshifts ( s10 redshift bin 6 ) , assigning mean angular diameter distance ratios ( @xmath107 ) to these lens - source pairs according to the estimated redshift distribution employed by s10 . we then weight all pairs with their individual lensing efficiency , similar to the weighting scheme in e.g. @xcite ( see appendix [ app : cosmos_signal ] ) . this downweights pairs that are close in redshift and naturally removes pairs where the `` source '' is in front of the `` lens '' . to the source catalogues we apply the following cuts : * @xmath142 . this cut is important as the f flexion measurement in particular gets heavily biased towards low s / n ( see section [ sec : tests_flashes_results ] ) . * if the centroid can not be determined accurately the shapelets decomposition will be inferior . therefore objects where the code is forced to move the centroid compared to the one estimated by sextractor by more than half a pixel are excluded . * the summed power in constant @xmath65 of the polar shapelets provides an indicator of the shapelet fit being affected by a neighbouring object . if the fractional power is particularly high at high orders the object is excluded . * if the fwhm or scale radius of the object is too small compared to the scale radius of the psf the object is excluded . * if @xmath143 , @xmath144 or @xmath145 then the object is excluded . these numbers are based on the measured distributions and the cuts are applied to remove outliers with very noisy shape measurements . * finally , we remove faint objects with an assigned photometric redshift of @xmath105 that have a prominent secondary peak at @xmath146 , as discussed in . the acs psf fluctuates both spatially and temporally ( e.g. * ? ? ? * ; * ? ? ? * ) , a variation mostly driven by changes in telescope focus caused for example by the breathing of the telescope . we can map the psf using stars , but in high - galactic latitude acs fields typically only @xmath112 stars are present . this number is too low for the standard approach of a polynomial interpolation . instead , we closely follow the analysis of , who conducted a principal component analysis ( pca ) of the acs psf variation as measured in dense stellar fields . they found that @xmath147 of the psf variation can be described with a single parameter ( the first principal component ) . this parameter is related to the hst focus position , and we therefore dub it ` focus ' here we make use of the measurement of the hst focus in all cosmos exposures and the investigated stellar field exposures . we also obtain shapelets versions of the focus - dependent psf models , by decomposing the dense stellar field stars into shapelets and interpolating between them with polynomials which are varied both spatially and with different powers of the focus principal component coefficient . from these models and from the cosmos focus estimates we then compute a shapelets psf model for each cosmos exposure , which we then combine to obtain a model for the stacked psf at all galaxy positions . ( left ) and @xmath148 ( right ) in an example star field . the black stars show the mean of the coefficients as a function of stellar flux after subtraction of a spatial third - order polynomial model derived from bright stars to separate psf and cti effects . each coefficient has been scaled to a reference number of @xmath149 parallel readout transfers . the purple curves show the parametric cti model , jointly determined from 700 stellar field exposures . the horizontal dashed line indicates an offset corresponding to the mean cti model for the bright stars used for the polynomial interpolation . the green triangles indicate the corrected coefficients after subtraction of the cti model.,title="fig:",width=162 ] ( left ) and @xmath148 ( right ) in an example star field . the black stars show the mean of the coefficients as a function of stellar flux after subtraction of a spatial third - order polynomial model derived from bright stars to separate psf and cti effects . each coefficient has been scaled to a reference number of @xmath149 parallel readout transfers . the purple curves show the parametric cti model , jointly determined from 700 stellar field exposures . the horizontal dashed line indicates an offset corresponding to the mean cti model for the bright stars used for the polynomial interpolation . the green triangles indicate the corrected coefficients after subtraction of the cti model.,title="fig:",width=162 ] our correction for cti again closely follows , who derive parametric corrections for the change in polarization for both galaxies and stars . the correction for stars is important in order to measure the actual psf , independent of the non - linear cti effects . in the stellar field analysis we therefore correct the psf cartesian shapelet coefficients for cti before generating the pca psf model . in order to estimate the influence of cti on the different shapelet coefficients , we follow and spatially fit each coefficient within one exposure with polynomials . due to the limited depth of the charge traps , cti is non - linear , and has a larger relative impact on faint sources than on bright ones . the cti effect can thus be estimated from the flux - dependent residuals , after the polynomial model has been used to subtract both the flux - independent psf and the flux - averaged cti signal . figure [ fig : cte ] shows these residuals as a function of stellar flux for the stellar shapelets coefficients @xmath150 and @xmath148 in one example stellar field . here the residuals were scaled to the same number of readout transfers ( 2048 ) . the cti effect on the coefficients is clearly visible ( black stars ) , but with our power law model ( curve ) it can be well corrected for ( green triangles ) . the model is fit simultaneously from all 700 stellar fields as a function of stellar flux , sky background , time and number of readout transfers ( see s10 ) . cti affects object shapes in the readout direction , which also after drizzling roughly matches the @xmath151-direction . thus cti residuals are expected to be roughly symmetric about the @xmath151-axis and hence vanish for coefficients @xmath15 with odd @xmath100 . in the drizzled images the readout direction is up for the upper and down for the lower chip and the cti trails occur in the opposite directions . this leads to a sign switch for coefficients with basis functions that are not symmetric about the @xmath152-axis ( odd @xmath101 ) , and we have taken this into account for @xmath148 in figure [ fig : cte ] . we have detected ( and modeled ) a significant signature of cti on the following stellar shapelets coefficients : @xmath153 , @xmath154 , @xmath148 , @xmath155 , @xmath156 , @xmath150 , @xmath157 , @xmath158 , @xmath159 , and @xmath160 . the correction of galaxy shapes for cti again closely follows s10 . here we fit power - law corrections to the shear and ( now in addition ) flexion estimates as a function of galaxy flux , flux radius , sky background , time , and number of readout - transfers . note that @xcite introduced a more advanced cti correction scheme operating directly on the pixel level . this is expected to yield higher precision , enabling for example the correction of the @xmath161 component , which can not be estimated with our method due to its degeneracy with a simple shift in object position . however , we are confident that our correction scheme is sufficiently accurate for the analysis presented here , in particular as potential residuals cancel to first order for the azimuthally averaged galaxy - galaxy lensing signal . for the navarro - frenk - white ( nfw ) profile @xcite , the strength of the shear signal scales as @xmath113 where @xmath114 , @xmath115 and @xmath116 are the angular diameter distances to the lens , to the source , and between lens and source respectively @xcite . the flexion signals scale as @xmath117 . we therefore weight the signals accordingly , scale them to a reference lens and source redshift and compute the weighted average in 25 logarithmic distance bins as follows : @xmath162[eq : weighted_average ] and similar for the flexions , with the shear estimator and weight @xmath163[eq : shear_estimator ] where @xmath164[eq : lenseff ] is the geometric lensing efficiency and @xmath165 is the error on the shape measurement with @xmath166 the intrinsic shear noise . by contrast we use the following f flexion estimator and weight : @xmath167[eq : fflex_estimator ] and similarly for the g flexion . as specified in the main paper , the lens catalogue we use has a redshift cut of @xmath105 . this is to avoid having to go too close to the lens on the sky in order to see a flexion signal . within an angular radius of @xmath168 we have low confidence in the results ; we are simply too close to the lensing galaxies and it becomes difficult to account for effects induced by the lens light . bor corrects for light leakage at larger radii , but the correction is most likely incomplete very close to the lens due to deviations from a smooth srsic profile . for objects beyond our lens sample , the median redshift is close to 1.0 . at this redshift the angular distance limit of @xmath168 on the sky corresponds to a physical distance of about @xmath169 . the f flexion falls off to low values already at about @xmath170 for a typical galaxy , so we are left with a very low signal within a narrow ring around the lens . imposing the redshift cut of @xmath105 on lenses gives us a median lens redshift of @xmath171 at which the inner limit corresponds to @xmath172 , leaving a wider distance interval in which we can investigate the f flexion signal . in figure [ app : fig : cosmos_highz ] we show the galaxy - galaxy signal for the high redshift sample , i.e. for lenses with @xmath173 . the bins that are within of the average lens in this sample , and which are most likely contaminated by lens light , are marked with dotted lines . the f flexion signal outside of this limit does agree well with the profile predicted by the shear , but falls off quickly . we compare our galaxy - galaxy shear signal to the one we get using the shears from , using all the cuts normally applied in each analysis so that only common objects are used . the bias correction described in their paper is incorporated in their shears whilst our measurements have no correction applied . however , due to our s / n cut ( see appendix [ app : cosmos_catalogues ] ) their correction is always less than 4.2% . as shown in figure [ app : fig : tim ] the difference between the results from the two pipelines , ksb and shapelets , is negligible . this provides an independent confirmation that the mv pipeline produces shears of as high a quality as the state - of - the - art weak lensing analysis presented in .

we construct one - zone steady - state models of cosmic ray ( cr ) injection , cooling , and escape over the entire dynamic range of the fir - radio correlation ( frc ) , from normal galaxies to starbursts , over the redshift interval . normal galaxies with low star - formation rates become radio - faint at high , because inverse compton ( ic ) losses off the cmb cool cr electrons and positrons rapidly , suppressing their nonthermal radio emission . however , we find that this effect occurs at higher redshifts than previously expected , because escape , bremsstrahlung , ionization , and starlight ic losses act to counter this effect and preserve the radio luminosity of galaxies . the radio dimming of star - forming galaxies at high is not just a simple competition between magnetic field energy density and the cmb energy density ; the cmb must also compete with every other loss process . we predict relations for the critical redshift when radio emission is significantly suppressed compared to the frc as a function of star - formation rate per unit area . for example , a milky way - like spiral becomes radio - faint at , while an m82-like starburst does not become radio - faint until . we show that the `` buffering '' effect of non - synchrotron losses improves the detectability of star - forming galaxies in synchrotron radio emission with evla and ska . additionally , we provide a quantitative explanation for the relative radio brightness of some high- submillimeter galaxies . we show that at fixed star formation rate surface density , galaxies with larger cr scale heights are radio bright with respect to the frc , because of weaker bremsstrahlung and ionization losses compared to compact starbursts . we predict that these `` puffy starbursts '' should have steeper radio spectra than compact galaxies with the same star - formation rate surface density . we find that radio bright submillimeter galaxies alone can not explain the excess radio emission reported by arcade2 , but they may significantly enhance the diffuse radio background with respect to a naive application of the frc .

0910.0478

the radio and fir luminosities of star - forming galaxies are linearly correlated over three decades in luminosity @xcite . this fir - radio correlation ( frc ) is driven by star - formation . starlight emitted by young , massive stars in the optical and uv is reprocessed by dust into far - infrared radiation . at the same time , the supernova remnants of massive stars accelerate cosmic rays ( crs ) , which produce synchrotron radio emission in the galaxies magnetic fields . the relation at @xmath2 is linear except in the very lowest luminosity galaxies @xcite . since the frc has less than a factor of 2 scatter in the local universe , radio luminosity is often used as a tracer of star - formation at both low and high redshift ( e.g. , * ? ? ? * ; * ? ? ? although the frc has been mainly studied in the low-@xmath1 universe , there has been recent interest in understanding how it applies to star - forming galaxies at high @xmath1 . there have been several conflicting results observationally . a number of studies have found that the frc holds unchanged or with little evolution at high redshifts ( e.g. , @xcite ; @xcite suggest radio - dim ulirgs at high @xmath1 ) . submillimeter galaxies , however , seem to be radio - bright , by a factor of @xmath5 ( @xcite ) . in particular , @xcite and @xcite argue that their samples of submillimeter galaxies are intrinsically radio bright with respect to the local frc , with no significant contamination from radio bright agn . in addition to the work on individual star - forming galaxies , there have also been new investigations into both the recently - resolved fir and the unresolved ghz radio backgrounds . the fir background was detected by cobe and has long been attributed to star - formation ( see the reviews by * ? ? ? * ; * ? ? ? the star - formation origin of the fir background was recently confirmed by blast @xcite . a detection of the extragalactic ghz radio background has also been recently reported by arcade2 @xcite . predictions for the strength of the diffuse radio background from star formation rely on the frc holding out to @xmath6 ( e.g. , * ? ? ? * ; * ? ? ? * ) , where most star - formation occurs . in this paper , we describe how and why the linearity and normalization of the frc are broken ( or preserved ) for different types of galaxies at high @xmath1 . we describe the basic physics of our model and the elements of the high-@xmath1 universe that physically allow the frc to break down in [ sec : theory ] . in [ sec : procedure ] , we briefly describe our underlying procedure . in [ sec : results ] , we discuss our results , providing predictions for when star - forming galaxies should be more or less radio luminous than the @xmath2 frc suggests . we also summarize the implications of our predictions for radio emission as a star formation tracer in [ sec : sftracer ] and estimate the effects of radio - bright smgs on the diffuse radio background in [ sec : radiobackground ] . throughout this paper , primed quantities denote the source s rest - frame and unprimed quantities are for the observer - frame . @xcite ( ltq ) recently presented a theory of the frc over its entire span at low redshift , from normal galaxies like the milky way to the densest ulirgs like arp 220 . the linearity and normalization of the frc provide strong constraints on the magnetic energy density in galaxies , which governs the strength of synchrotron losses . indeed , the energy density of starlight in dense galaxies is so large that inverse compton losses alone shorten the cr electron and positron lifetime to @xmath7 in galaxies like arp 220 @xcite . the magnetic field in arp 220-like starbursts must exceed mg for the cosmic ray electrons and positrons to radiate a large enough fraction of their energy in synchrotron emission to account for the observed ghz radio emission . such galaxies are likewise sufficiently dense that ionization and bremsstrahlung losses are also rapid . these considerations imply that the magnetic energy density must increase dramatically from normal spirals to dense bright starbursts . the observed schmidt law allows us to quantify these dependences since it connects the star - formation rate surface density ( @xmath8 ) , which is proportional to the energy density in starlight ( @xmath9 ) , to the gas surface density : @xmath10 according to @xcite , or @xmath11 according to @xcite . using the @xcite schmidt law , ltq found that the magnetic field strength @xmath12 must increase as @xmath13 or @xmath14 , where @xmath15 , implying almost a thousandfold increase in @xmath12 from milky way - like galaxies to arp 220-like ulirgs . this strong dependence of @xmath12 on the global properties of the galaxy is largely responsible for the radio emission from starbursts and is required by the existence of the frc . it is also crucial to the physics of how the frc is maintained or broken at high z. ltq was unable to distinguish between the possibilities of @xmath12 scaling with @xmath16 and @xmath12 scaling with @xmath17 . ltq found that over most of the range of the frc , normal galaxies and starbursts are calorimetric . they are electron calorimeters , meaning cr electrons lose most of their energy before escaping . they are also uv calorimeters , meaning uv light is efficiently absorbed by dust and converted to fir light . if cr electrons cool mainly by synchrotron losses , then the ratio of fir emission to radio emission from primary cr electrons should be constant , since each is some fraction of the power released by star - formation . this is the essence of calorimeter theory , first proposed by @xcite . however , ltq found that even in this strict calorimeter limit , @xmath18 should decrease systematically by @xmath19 from low surface brightness galaxies to dense ulirgs . the critical synchrotron frequency causes this `` @xmath20 effect '' : at fixed frequency , as the magnetic field strength increases with increasing surface density , lower energy electrons dominate the synchrotron emission because the synchrotron frequency of electrons , @xmath20 , is proportional to @xmath21 , where @xmath22 is the electron energy . since the injected cr spectrum is likely steeper than @xmath23 , there is more power radiated at low energies . thus , radio emission increases with surface density of gas and star formation , even including only synchrotron losses in the pure electron and uv calorimeter limit . on top of calorimetry , ltq proposed that two different conspiracies combine to produce the linear frc , each working in a different surface density regime . the `` low-@xmath16 conspiracy '' arises as both uv calorimetry and electron calorimetry gradually fail at relatively low surface densities . ] ( @xmath24 ; @xmath25 ) . cosmic ray electrons escape diffusively without radiating all of their energy , decreasing the radio emission from the calorimetric expectation . however , some uv photons also escape without being reprocessed into fir by absorption . quantitatively , each effect alters the luminosity by a factor of @xmath26 in the very lowest surface density galaxies ( @xmath27 ; @xmath28 ) . therefore , both the fir and radio emission decrease roughly equally , to maintain a constant @xmath18 ( see also * ? ? ? * ) . at the other extreme , high surface density compact starbursts have two completely different opposing processes that affect the radio emission . compact starbursts are expected to be proton calorimeters : cr protons lose almost all of their energy through pion interactions with the dense ism of starbursts , emitting @xmath29-rays , neutrinos , and secondary electrons and positrons ( * ? ? ? * ; * ? ? ? * ; * ? ? ? the secondary electrons and positrons themselves emit synchrotron radio emission , and outnumber the primary cr electrons . proton calorimetry therefore greatly enhances the radio emission , all else being equal . although escape losses are negligible in starbursts , synchrotron does compete with the other important cooling processes : inverse compton , bremsstrahlung , and ionization ( * ? ? ? these are particularly important in bright , dense starbursts , because they have very high gas densities and photon energy densities . these other losses reduce the share of cr electron and positron power going into radio by a factor of @xmath30 . between the secondaries and the @xmath20 effect increasing @xmath31 , and the non - synchrotron losses suppressing @xmath31 , a linear frc correlation is produced . this is the `` high-@xmath16 conspiracy '' of ltq , and it operates to produce a linear frc even though these starbursts are completely calorimetric ( for protons , electrons , and positrons ) and escape losses are negligible . galaxies at high redshift can be used to test the theory of the frc presented in ltq . at high @xmath1 , we expect that both the low- and high-@xmath16 conspiracies are unbalanced for some galaxies . at low @xmath16 , the energy density of the cmb ( @xmath32 ) competes with the energy density in both starlight and magnetic fields . crs in galaxies face increasing losses from inverse compton since the cmb is stronger . normal galaxies therefore should be radio dim , as inverse compton losses divert power from synchrotron . starbursts , however , where crs already face intense ic losses from starlight , should not be affected except at the highest redshifts ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in this paper , we quantify how much the radio emission is dimmed by increased ic losses from the cmb as a function of @xmath1 . on the other hand , high-@xmath1 starbursts often have different morphologies than those observed of low-@xmath1 ulirgs . local starbursts including ulirgs are typically compact , with radii of a few hundred parsecs and vertical scales heights of less than a hundred parsecs @xcite . these properties do not hold for submillimeter galaxies ( smgs ) , very bright ( @xmath33 ) starbursts at @xmath3 mostly powered by star - formation and not agns ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? smgs are usually several kpc in diameter ( several times larger than low-@xmath1 ulirgs ) ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , and have much lower surface densities than their low-@xmath1 counterparts of the same luminosity ( @xcite , but see @xcite ) . submillimeter galaxies have large random velocities compared to their rotation speeds , implying scale heights of @xmath34 @xcite . at high redshift , we therefore must consider a class of _ _ puffy starbursts _ _ , with @xmath35 ( @xmath36 ) but scale heights of @xmath37 , in addition to the @xmath38 `` compact starbursts '' typical of local ulirgs considered in ltq . importantly , puffy starbursts will have a smaller volume density for a given @xmath16 and @xmath8 . since bremsstrahlung , ionization , pion , and possibly synchrotron losses all depend on the volume density instead of surface density , these loss processes may all be weaker in puffy starbursts like smgs . with these losses suppressed , the high-@xmath16 conspiracy will become unbalanced , and the frc can be broken . these starbursts can either be radio - bright or radio - dim , depending on their magnetic field strengths . since the scale height is so large and the volume density is relatively low for smgs , we can determine whether @xmath12 scales with @xmath16 or @xmath17 with these galaxies , breaking the degeneracy in ltq . because @xmath15 , if @xmath12 increases with @xmath17 , then the magnetic field in puffy starbursts will be weak and synchrotron radio emission will be dim compared to compact starbursts with the same @xmath16 . however , if @xmath12 increases with @xmath16 , the magnetic field will be strong and synchrotron radio emission will be bright in puffy starbursts compared to compact starbursts with the same @xmath16 . we model galaxies and starbursts as uniform disks of gas , with scale height @xmath39 , star formation rate surface density is directly proportional to @xmath8 . the conversion factor is @xmath40 , where @xmath41 is the star formation rate surface density in units of @xmath42 . ] @xmath8 , and gas surface density @xmath16 . we solve the diffusion - loss equation to find the steady - state equilibrium cr spectra in galaxies and starbursts . injection , escape , and cooling losses all compete at each energy to determine the final cr spectrum ( see ltq for details ) . the relevant scale height is the height of the volume in which crs are confined and produce synchrotron radiation . normal galaxies ( @xmath43 or @xmath44 ) have large radio halos with @xmath34 , which we adopt as the cr scale height , even though the gas disk is much thinner . compact starbursts are much smaller , with @xmath45 ( e.g. , * ? ? ? * ; * ? ? ? * ) ; we use this as their scale height because they are probably good enough calorimeters to prevent most of the cr electrons / positrons from escaping out into the galactic haloes at high enough surface densities ( @xmath46 or @xmath36 ) . the cr disk scale height for our prototypical puffy starbursts , the smgs , is not yet directly measured , but it must be at least the gas scale height , if the star formation is distributed throughout the gas . the gas scale height for smgs is kinematically inferred to be @xmath34 ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , which we adopt as the cr scale height of puffy starbursts . our models span the entire observed range in @xmath16 for galaxies and starbursts , from @xmath47 to @xmath48 . combined with the scale height , @xmath16 determines the rate of injection , escape , and cooling at all energies . the total rate of injection per unit volume is proportional to the volumetric star - formation rate , @xmath49 . the observed schmidt law directly connects @xmath16 and @xmath8 @xcite . we consider both @xmath10 ( * ? ? ? * hereafter k98 ) and @xmath11 ( * ? ? ? * hereafter b07 ) , using the normalizations given by k98 and b07 respectively . the b07 relation was explicitly derived for high-@xmath1 galaxies , including smgs . protons and electrons are injected into our model galaxies with an energy spectrum @xmath50 ; in this paper , we use @xmath51 in almost all cases .. the slightly different @xmath52 makes small quantitative differences , but does not change our conclusions . ] escape , synchrotron , inverse compton , bremsstrahlung , and ionization losses are included for electrons and positrons , along with escape , ionization , and pion losses for protons . diffusive escape times are normalized to the milky way value , with a loss time of @xmath53 . we also consider variants with advective escape , as in starburst super - winds . we assume that cosmic rays travel through gas with density @xmath54 , where @xmath55 is the mean ism number density and @xmath56 . the magnetic field is parametrized as @xmath57 or @xmath58 , normalized by the milky way magnetic field strength . we searched for models that satisfy the local @xmath2 frc for normal galaxies and compact starbursts . the cr proton spectrum is then normalized by milky way cr proton constraints . the parameters chosen for each variant are used to predict the frc for puffy starbursts and at high @xmath1 . in this paper , we consider several variants to get a sense of how the frc varies with redshift : * the ltq standard model , with no winds , @xmath57 , and the k98 star - formation relation . * the ltq model with winds of @xmath59 in starbursts ( @xmath35 ) , @xmath58 , and the k98 star - formation relation . * a variant with no winds , @xmath57 , and the b07 star - formation relation . * a variant with winds of @xmath59 in starbursts ( @xmath35 ) , @xmath57 , and the b07 star - formation relation . * a variant with winds of @xmath59 in starbursts ( @xmath35 ) , @xmath58 , and the b07 star - formation relation . we show in table [ table : models ] that we can reproduce an acceptably linear local frc for each variant . we are able to adopt parameters consistent with milky way - derived cr proton constraints such that @xmath60 varies by a factor of 1.7 - 2.2 from normal galaxies to compact starbursts . this variation is consistent with the factor of @xmath19 scatter in the frc and winds , in which the winds invariably make electron escape too efficient when @xmath61 ( @xmath62 with b07 ) , suppressing its radio luminosity ( see the detailed discussion in ltq , appendix a.2 ) . we are unable to find any parameters where the variation in @xmath63 is less than 2.0 , but we consider a variation of 2.2 adequate for this paper s purposes . [ ftnt : windvariant ] ] ( e.g. , * ? ? ? * ) . as in ltq , the need for a linear local frc strongly constrains the magnetic field in galaxies to scale as either @xmath64 or @xmath65 ( see [ sec : theory ] ) . the dependence of the fir emission at a given wavelength is beyond the scope of this paper , since it depends on the exact sed of the fir emission in the galaxy . instead , we simply use the total infrared ( tir ) emission , assuming it is all of the uv light reprocessed by the dust , and use @xmath66 as a proxy for @xmath67 . @xcite estimate that @xmath68 , a correction we apply to the local value of @xmath69 from @xcite to find the normalization of @xmath63 at @xmath2 ( see ltq ) . observable defined in @xcite , we then divide our calculated @xmath67 by @xmath70 to get the true fir luminosity . however , we do not account for different @xmath70 values for normal galaxies or smgs . ] we assume that the total tir emissivity ( defined here as luminosity per volume ) has been measured and corrected for redshift to its rest - frame value , @xmath71 . we can also calculate the observable quantity @xmath72 as @xmath73 @xcite . for the radio emission , we calculate both the rest - frame radio emissivity @xmath74 where @xmath75 , and the rest - frame radio emissivity as estimated by an observer using a k - correction . note that we are _ not _ actually trying to calculate the specific flux in the observer frame , but instead what an observer will infer for the _ rest - frame _ flux after using a k - correction galaxy at 2 ghz and trying to infer the 1 ghz flux by using the spectral slope . the observer has an observed radio flux at 1.4 ghz , and can easily calculate the rest - frame flux at 1.4 ( 1 + z ) ghz flux , but wants the rest - frame flux at 1.4 ghz . ] : the rest - frame flux is what matters when we are considering the true , intrinsic evolution of the frc . hence , although the specific flux in the observer - frame will have a bandwidth compression factor , we assume the observer will take it back out to get the inferred rest - frame specific flux . to calculate the observer - inferred rest - frame emissivity at rest - frame frequency @xmath75 , we start with @xmath76 , where @xmath77 and @xmath78 . since the synchrotron spectra are typically expected to fall off as @xmath79 from observations of local star - forming galaxies , the radio luminosity can be k - corrected by multiplying by @xmath80 . we therefore calculate @xmath81 , which we will refer to as the `` inferred '' radio emissivity . the ratio of the tir and radio luminosities @xmath82 is then @xmath83 in the rest frame , and would be inferred to be @xmath84 in the rest - frame by observers . [ sec : discussion ] we show the rest - frame frc for our model with @xmath85 , the b07 star - formation law , and no winds , in figure [ fig : lfirradiorest ] ( _ left panel _ ) as an example . the solid dark red line is the @xmath86 frc . all solid lines assume that starbursts are compact , with @xmath38 . it is clear from figure [ fig : lfirradiorest ] that compact starbursts show little evolution in the frc , while low surface density galaxies have lower radio luminosities at high redshift . this behavior is robust in all of the variants . the cause of evolution in the frc is inverse compton losses off the cmb for cr electrons and positrons . figure [ fig : frcevolution ] ( _ left panel _ ) shows that the frc should display relatively little evolution out to @xmath87 , except for the lowest surface brightness galaxies . however , normal galaxies have suppressed synchrotron radio emission at @xmath3 , a factor of @xmath19 for @xmath88 ( @xmath89 ) and of order 10 for @xmath90 ( @xmath91 ) . the radio luminosities continue to fall with redshift . at @xmath92 , ic off the cmb starts to matter even for the weaker starbursts ( @xmath61 ; @xmath93 ) . dense starbursts ( @xmath94 ; @xmath95 for the k98 law or @xmath96 for the b07 law ) remain on a linear frc even at @xmath97 . the strong cooling from bremsstrahlung , ionization , and ic off starlight implied by the high-@xmath16 conspiracy ( [ sec : theory ] ) acts as a buffer against ic losses off the cmb . similarly , diffusive escape provides a similar buffering effect in low-@xmath16 galaxies , so that the increased ic losses off the cmb does not suppress the radio emission quickly . in other words , most of the power from ic scattering of cmb photons does not come at the expense of synchrotron , but of other losses . the radio dimming can not be described by a simple competition between magnetic field energy density and the cmb energy density ; the cmb energy density must also compete with _ every other loss process_. the buffering actually serves as an important test for both conspiracies described in [ sec : theory ] . in high-@xmath16 galaxies and starbursts , we predict that bremsstrahlung , ionization , and inverse compton of starlight already take a large portion of a electron s energy budget ; hence , the high-@xmath16 conspiracy works to hold @xmath63 constant out to quite high redshift . the buffering essentially doubles the redshift that compact starbursts remain on the frc ; the weakest starbursts ( @xmath61 ; @xmath93 ) are radio - dim by a factor of @xmath98 at @xmath99 with buffering , instead of @xmath100 without the buffering . in low-@xmath16 galaxies , electrons and positrons can easily escape before they lose energy to inverse compton off the cmb at low enough @xmath1 ; thus , the low-@xmath16 conspiracy also works to hold @xmath63 constant . the suppression due to ic losses off the cmb seen in figure [ fig : lfirradiorest ] can be estimated by examining the ratios of the synchrotron cooling time to the total loss time , including both escape and cooling losses . in appendix [ sec : cmbradiodim ] , we derive the ratios of loss times and we show that for a given schmidt law , a critical redshift @xmath101 can be defined for each @xmath16 and @xmath8 at which radio emission is suppressed . we define @xmath101 to be the redshift when the rest - frame radio luminosity at @xmath75 is suppressed by a factor of 3 compared to @xmath86 . for our model with the b07 star - formation law , no winds , and @xmath85 , we find the critical redshifts are approximated as @xmath102 where @xmath8 is in units of @xmath42 . we refer the reader to appendix [ sec : cmbradiodim ] , where we present similar relations for our other models of the frc . we note that the radio suppression could , conceivably , be used to measure the temperature of the cmb at high redshift . in principle , this method could apply to any galaxy with a radio and fir detection . however , the conspiracies would have to be accounted for , and they are affected significantly by both the gas surface density @xmath16 and scale height @xmath39 ( see [ sec : puffy ] ) . any measurement of the cmb temperature would depend on assumptions of the galaxy properties and would be model - dependent . as figure [ fig : frcevolution ] ( _ right panel _ ) shows , the inferred rest - frame values of @xmath103 show additional apparent evolution , simply because the spectral slopes of the galaxies are not exactly @xmath104 . compact starbursts have flatter spectra and @xmath105 are instantaneous spectral slopes , not the measured spectral slopes between two observed frequencies , unless otherwise noted . [ ftnt : alphanotation ] ] with @xmath106 at 1.4 ghz , so by adopting @xmath107 they _ appear _ to become slightly radio brighter until @xmath87 , after which at higher frequencies their spectra steepen , and their radio emission appears to dim again at higher @xmath1 . normal galaxies have steeper spectra with @xmath108 at 1.4 ghz , so their apparent radio luminosity sinks below their true radio luminosity at high redshift . at higher redshifts , our models predict that galaxies have intrinsically steeper radio spectra , because of increased ic losses from the cmb . the rest - frame spectral slope @xmath109 of normal galaxies at 1.4 ghz and asymptotes at @xmath110 by @xmath111 , when losses are dominated by ic for our injection spectrum of crs with @xmath51 . there is much less intrinsic rest - frame evolution of starburst radio spectra ; even at @xmath92 , @xmath109 increases only by 0.1 for the weakest compact starbursts . the observable @xmath112 shows much more pronounced evolution with redshift for starbursts , increasing by @xmath113 to @xmath6 , and @xmath114 to @xmath115 . this effect arises simply because we predict the cr electron and positron spectra steepen with rest - frame frequency as synchrotron and ic losses become stronger ( e.g. , * ? ? ? * ltq ) : at higher redshift and fixed observing frequency , we are seeing higher energy electrons and positrons . an important effect that we do not consider is the thermal free - free radio emission . this will set a minimum total observed radio emission that is directly proportional to the star - formation luminosity . thus , the true total radio deficit at ghz will not be as big as the synchrotron - only deficits plotted in figures [ fig : lfirradiorest ] and [ fig : frcevolution ] . thermal free - free emission will also flatten the spectrum , especially at high frequencies . however , free - free emission is much fainter than the synchrotron luminosity at ghz frequencies , except in the faintest star - forming galaxies @xcite . [ sec : puffy ] as shown in figure [ fig : lfirradiorest ] , puffy starbursts fall on a linear fir - radio correlation of their own ( dotted lines ) , in line with observations of smgs @xcite . the radio luminosity of these galaxies is nonetheless the result of a conspiracy , between ic losses on starlight , which decrease the radio luminosity , and the enhanced radio emission from secondary electrons and positrons , and the @xmath20 effect ( [ sec : theory ] ) . the variation in @xmath63 for puffy starbursts alone is usually less than a factor of @xmath116 over the range @xmath117 ( see table [ table : models ] ; in most variants the variation is @xmath118 ) . like compact starbursts , escape plays essentially no role in most of the models , except that winds can slightly decrease the radio emission in relatively tenuous @xmath61 ( @xmath93 ) starbursts ( footnote [ ftnt : windvariant ] ) . it is plain from figure [ fig : lfirradiorest ] ( _ left panel _ ) that the normalization of the puffy starburst frc ( dotted lines ) is different than the frc of the compact starbursts and normal galaxies ( solid lines ) when @xmath85 . we show in table [ table : models ] that models with @xmath119 have radio - bright puffy starbursts compared to the observed local frc , by a factor of @xmath120 at @xmath86 . like the compact starbursts , puffy starbursts show little rest - frame evolution in @xmath63 , except at relatively low surface densities ( @xmath61 ; @xmath93 ) , where they become radio dim at high redshifts because of ic losses on cmb photons . therefore , we predict that puffy starbursts , which are mainly observed at high @xmath1 , have _ intrinsically _ different radio properties not caused by their redshift . we propose a natural explanation of this radio excess in the framework of ltq ( [ sec : theory ] ) . in dense starbursts , protons are efficiently converted into secondary electrons and positrons through inelastic proton - proton scattering , which contribute to the synchrotron emission . furthermore , the @xmath20 effect increases @xmath63 for starbursts which have larger @xmath12 . compact starbursts lying on the @xmath86 frc balance these effects with increased bremsstrahlung , ionization , and ic losses , which compete with the synchrotron losses and suppress the excess radio luminosity ( [ sec : theory ] ) . in puffy starbursts with relatively low volume densities compared to their compact cousins at fixed @xmath16 , however , bremsstrahlung and ionization are not strong enough to compensate for these effects . only synchrotron and ic losses remain , upsetting the conspiracy . the radio excess predicted by this picture is systematically greater when using the k98 star - formation relation , because the ic loss rate on starlight is smaller at fixed surface density by a factor of 2.5 to 11 from weak starbursts to the densest starbursts . this freedom to vary the radio - excess with @xmath12 does not exist in the standard calorimeter model , where all cr electron / positron energy goes into synchrotron emission , and the radio emission saturates . the balance between synchrotron and the other forms of cooling can be changed in starbursts to alter the normalization of the frc , even though escape is negligible . there is a reason to expect our allowed @xmath119 specifically : radiation pressure may drive turbulence and enhance the magnetic field until its energy density is comparable to radiation ( e.g. , * ? ? ? if the k98 relation holds , then since the magnetic energy density scales as @xmath121 , @xmath85 , and if the b08 relation holds , then @xmath122 . this explanation is more problematic for the b07 relation and @xmath85 . to the _ total _ milky way magnetic field strength of @xmath123 near the solar circle . if most of the magnetic field strength in starbursts is driven by turbulence , perhaps scaling to the disordered galactic magnetic field strength ( @xmath124 ; * ? ? ? * ) near the solar circle would make more sense , because the ordered magnetic field arises from a different process . the lower normalization for @xmath12 then would partly compensate for the steeper dependence on @xmath16 . ] however , while we assume that there is a parametrization for @xmath12 that applied to both compact starbursts and puffy starbursts , the radio excess should arise more generally . the radio excess arises simply because synchrotron cooling time is shorter than the bremsstrahlung and ionization cooling times in puffy starbursts , but longer in compact starbursts at fixed @xmath16 : at fixed @xmath125 , @xmath126 and @xmath127 . we therefore expect there to be a radio excess with respect to the frc in any puffy starburst with a strong enough magnetic field , with the exact enhancement depending on magnetic field strength because of the remaining competition , after ionization and bremsstrahlung are sub - dominant , from the ic losses on starlight . , however , there will be more scatter in @xmath103 for puffy starbursts at fixed @xmath39 and @xmath1 . smgs would not form their own frc if @xmath12 was the same for all smgs regardless of @xmath16 , or if @xmath12 increased _ very _ steeply with @xmath16 , because there still must be a conspiracy with ic losses off starlight . since smgs do appear to form their own frc ( e.g. , * ? ? ? * ; * ? ? ? * ) , this could be evidence specifically that their magnetic field strengths increase with @xmath16 ( or @xmath17 ) . ] our results imply that a moderate radio excess at the factor of @xmath5 level alone is not a safe indicator of the presence of a radio - loud agn , especially at high redshifts where smgs are observed . while radio excess with respect to the local frc has been suggested as a selection criterion for radio - loud agns ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , our models with @xmath128 imply that @xmath129 for puffy starbursts powered by star - formation alone . a radio excess is inexplicable in our models only when the source is an order of magnitude brighter ( @xmath130 ) in the radio than predicted from the fir emission . while smgs are relatively rare and may not be a problem in small samples , we recommend that other means be used to be sure that the radio - excess is caused by an agn , such as a flat radio spectrum , radio morphology , mid - ir colors , or the presence of strong x - ray emission ( see also * ? ? ? a magnetic field dependence of @xmath131 appears to hold for galactic molecular clouds ( e.g. , * ? ? ? * ) , and ltq found that the fir - radio correlation was consistent with this magnetic field dependence . the existence of galaxies with different scale heights allows us to distinguish the two possibilities for magnetic field scaling . if @xmath57 , then the magnetic field strength will be the same for all galaxies with the same @xmath16 , regardless of scale height . in models with @xmath14 , by contrast , the magnetic field strength is weaker in puffy starbursts than in compact starbursts with the same @xmath16 . as seen in figure [ fig : lfirradiorest ] ( _ right panel _ , dotted lines ) , puffy starbursts again form their own frc . in models with @xmath14 , they are radio dim compared to the @xmath2 frc . we show in table [ table : models ] that the normalization of the frc is radio - dim by a factor of @xmath132 . we can explain this in the ltq theory of the frc as well . the magnetic field strength must increase more slowly with @xmath17 than with @xmath16 to reproduce the @xmath2 frc , because @xmath133 is 10 times smaller in compact starbursts than normal galaxies and puffy starbursts . compact starbursts are highly compressed with respect to normal galaxies , so they have strong magnetic fields and synchrotron radio emission is strong enough to compete with the other losses . puffy starbursts are not compressed , so that their magnetic fields are weak and synchrotron losses can not keep up with ic losses , nor with bremsstrahlung and ionization as 1.4 ghz emission traces ever lower electron energies at higher magnetic field strengths . puffy starbursts therefore turn out to be radio dim compared to compact starbursts on the @xmath2 frc , if @xmath14 . as before , the b07 star - formation relation predicts greater ic losses and therefore weaker radio emission . because of the claims in the literature that smgs are radio bright , we do not favor these models . the suggested relative radio brightness of high-@xmath1 smgs therefore provides some evidence that , in fact , @xmath134 rather than @xmath14 . however , the matter of whether high-@xmath1 smgs are in fact radio bright is not yet settled . although ltq concluded that @xmath12 must increase dramatically from normal galaxies to dense starbursts ( [ sec : theory ] ) , they were unable to distinguish between these two possibilities with the @xmath2 frc alone . for this reason , high-@xmath1 starbursts and their qualitatively different morphologies compared to those at @xmath2 can distinguish theories of the frc . a prediction of all of our variants is that puffy starbursts like submillimeter galaxies should have steep non - thermal radio spectra , with @xmath135 ( see table [ table : models ] ) . the steep spectra are caused by strong synchrotron cooling in the @xmath119 case and the relatively stronger ic cooling off starlight in the @xmath14 case . in general , puffy starbursts should have roughly the same @xmath136 as normal galaxies in the local universe , which tends to be somewhat higher ( @xmath137 ) than in compact starbursts ( @xmath138 ) . the slope should hold even out to extremely high @xmath16 , as long as starbursts are puffy . in contrast , we find that @xmath139 in compact starbursts , because of efficient ionization and bremsstrahlung losses , which flatten the equilibrium cr spectrum because of their energy dependence . as we note in ltq , our predicted spectral index for normal galaxies is somewhat too high , and this difference in @xmath136 may carry over to the puffy starbursts . however , the significant difference in @xmath136 between compact and puffy starbursts should remain as a general prediction of our model : compact starbursts should have flatter spectra than puffy starbursts . the high spectral slopes can be observed either with direct measurements of multifrequency data of individual submillimeter galaxies , or with single frequency observations at a variety of redshifts . there are relatively few measurements of @xmath136 for submillimeter galaxies specifically ; faint radio sources have @xmath140 @xcite , though that sample includes both compact starbursts and agns . @xcite do find that @xmath141 for smgs , comparable to our predictions . they also find that submillimeter galaxies have a radio - excess , in agreement with figure [ fig : lfirradiorest ] . more recently , @xcite found an average @xmath142 , which is somewhat flatter than our models . these spectral slopes are not different from normal star - forming galaxies , but are noticeably steeper than local ulirgs @xcite . however , we do not account for free - free absorption , which probably flattens the spectra of local ulirgs like arp 220 at low frequency @xcite , and is not well understood in smgs . since puffy starbursts have steeper spectra than compact starbursts , we expect their inferred @xmath63 will increase with redshift : if the true radio spectral slopes of smgs are greater than the assumed @xmath136 by @xmath143 , they will appear to become radio dimmer by a factor @xmath144 , or up to @xmath145 at @xmath146 . in figure [ fig : starburstspectra ] , we show the expected radio synchrotron spectra of starburst galaxies , without correcting for thermal absorption or thermal emission . at a rest - frame frequency of 1 ghz , puffy starbursts ( dashed ) have steeper radio spectra than compact starbursts ( solid ) . note that at high frequencies ( @xmath147 ) , the ratio of the radio luminosities per unit star formation of the compact and puffy starbursts asymptotes to a value set by the ratio of @xmath148 and @xmath149 in these starbursts . at these high frequencies , only synchrotron and ic cooling are effective , and ic cooling would be the same for puffy and compact starbursts because of the schmidt law ( [ sec : theory ] ) . for @xmath57 , @xmath148 is the same for puffy and compact starbursts , but for @xmath58 , puffy starbursts have much smaller @xmath148 . thus , measurements of the synchrotron radio emission of smgs at high @xmath125 could determine the magnetic field strength of smgs and determine which scenario applies . of course , there are unlikely to be two perfectly distinct populations of compact starbursts and puffy starbursts . instead , there may be a continuum variation in scale heights from tens to thousands of parsecs . we would then expect to see a larger scatter , both in @xmath63 and @xmath136 , in a full sample of both the most compact and the most puffy starbursts . @xcite find that submillimeter galaxies do have a larger scatter in @xmath150 than other galaxies . however , @xcite find a relatively small scatter of @xmath151 in smg radio spectral index . importantly , for larger @xmath39 , both @xmath63 and @xmath109 asymptote as cr electron and positron losses are entirely determined by synchrotron and ic ; @xmath34 starbursts are already near this limit . in other words , for arbitrarily large @xmath39 , the radio excess with respect to the @xmath2 frc asymptotes to a value of @xmath152 , depending on the assumed schmidt law , and the radio spectral slope asymptotes to @xmath110 for @xmath51 . relying on the frc , a number of studies have used the ghz radio emission as a tracer of star - formation ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? radio emission has the advantage that it is unaffected by dust obscuration , making it potentially very useful in starbursts , and therefore for most of the star - formation at @xmath153 ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? our models let us evaluate the theoretical basis for radio as a sfr indicator at all redshifts . we show the predicted radio emissivity as a function of star formation in figure [ fig : sfrradio ] . at low surface densities ( @xmath154 ; @xmath25 ) , synchrotron radio has a non - linear dependence on star - formation at all redshifts . the weak radio emission is caused by electrons escaping their host galaxies , as inferred by @xcite and discussed in ltq . that is , normal galaxies are not perfect electron calorimeters , so the radio emission is not a reliable star - formation tracer at low @xmath8 . at higher redshift , synchrotron emission is diminished by inverse compton off the cmb . for a milky way - like galaxy , we find that radio is a good star - formation tracer at @xmath155 , but underestimates it significantly by @xmath156 ( eq . [ eqn : b07zcrit ] ) . already galaxies with star formation rates similar to galactic levels are beginning to be observed in the radio at high redshift @xcite , so that the ic suppression may soon be observed . however , ic losses off the cmb should not be important even in the weakest starbursts until @xmath157 , as is also visible by the redshift evolution of the frc in figure [ fig : frcevolution ] . radio emission does grow linearly with star - formation rate between normal galaxies with @xmath88 ( at @xmath158 ) and compact starbursts . therefore , it serves as an acceptable star - formation indicator for these galaxies . however , if @xmath119 , puffy starbursts like smgs have about @xmath120 times the radio emission at any given star - formation rate than compact starbursts . therefore , we expect that if @xmath119 , the usual radio emission estimate based on the @xmath2 frc will _ overestimate _ their star formation rates by a factor of @xmath159 . the excess is greatest at @xmath160 corresponding to @xmath161 , typical of observed smgs . among the puffy starbursts themselves , the radio emission grows linearly with star - formation rate . if instead @xmath14 , the radio emission will underestimate the star - formation rate . however , because the smgs lie on their own frc , there is little real redshift evolution in the radio emissivity of puffy starbursts , because of the buffering provided by ic losses off starlight ( see [ sec : frcevolution ] , appendix [ sec : cmbradiodim ] ) . assuming a spectral slope @xmath107 will also underestimate the radio emissivity , since these galaxies can have steep radio spectra . this explains the apparent evolution with @xmath1 of puffy starbursts in figure [ fig : sfrradio ] : we have applied a k - correction using a typically employed @xmath107 when , in fact , the true synchrotron spectra are steeper . if the radio star - formation tracer could be calibrated to the special conditions in puffy starbursts like smgs , taking into account their different scale height , radio emissivity , and spectral slopes , we predict that radio would be a more accurate star - formation tracer for them . we can also directly calculate the _ observed _ ghz radio flux density @xmath162 from synchrotron emission of star - forming galaxies . we show the predicted flux density per unit star - formation at observer - frame 1.4 ghz in figure [ fig : ghzradiofluxes ] . surveys with the expanded very large array ( evla ) will have a continuum sensitivity of @xmath163 at frequencies of 1 - 50 ghz , and it should be able to directly detect galaxies with @xmath164 and star - formation rates of @xmath165 out past @xmath166 . as stated in @xcite , a starburst like m82 with sfr @xmath167 will become undetectable past @xmath87 . however , the buffering effect we emphasize in this paper preserves the radio emission of dense starbursts at high @xmath1 , so that bright starbursts will be detectable further : starbursts with sfr @xmath168 will be detectable with evla at 1.4 ghz out to @xmath169 and the most intense starbursts ( sfr @xmath170 and @xmath171 ) will be detectable out past @xmath97 in synchrotron emission . @xcite predicts the square kilometer array ( ska ) will be sensitive to star - forming galaxies with flux densities of @xmath172 . if this sensitivity is attained , then milky way - like galaxies ( @xmath173 ; sfr @xmath174 ) will be directly detectable at 1.4 ghz in synchrotron emission out to @xmath3 , and starbursts with sfr @xmath174 will be detectable at 1 ghz beyond @xmath175 . even at @xmath97 , the synchrotron emission of dense , compact starbursts ( @xmath176 ) with sfr greater than @xmath177 should be detectable at 1.4 ghz with ska ( models with @xmath178 have radio - dim puffy starbursts , and these are detectable at @xmath99 with the ska only for sfr greater than @xmath179 ) . by contrast , @xcite found a sensitivity of @xmath180 , based on the free - free emission ; this limit will apply to normal galaxies and weak starbursts where the synchrotron emission is suppressed . these sensitivities assume that natural confusion , in which radio sources overlap , will not hamper the ska ; estimates for the natural confusion limit vary from njy to @xmath181 levels ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? ska and evla will also have good spectral coverage , which may help measurements of the spectral index . if a galaxy is detected at @xmath182 ( @xmath183 for evla and @xmath184 for ska ; @xcite ) at two different frequencies @xmath185 and @xmath186 with @xmath187 , then @xmath188 can be constrained to @xmath189 at the @xmath190 level . evla will be better at high observer - frame frequencies ( 1 - 50 ghz ) . a problem for the evla will be the increasing fraction of thermal emission , which is expected to dominate the emission of starbursts at @xmath191 . evla will therefore not easily measure the nonthermal spectral indices of galaxies at high z. on the other hand , the ska will face free - free absorption when observing low - z starbursts ; for example , arp 220 may be optically thick even at 1 ghz @xcite . at high redshift , however , the rest - frame frequencies ska will observe will be less affected by free - free absorption . star - formation in galaxies over cosmic time produces a diffuse radio synchrotron background . our models indicate that submillimeter galaxies and other puffy starbursts ought to have enhanced radio emission . in principle , this means that the synchrotron radio background could be up to @xmath159 times higher than usually predicted from the cosmic infrared background ( cib ) and a naive application of the @xmath2 frc . the magnitude of this implied radio excess is interesting , because arcade2 recently reported an excess radio background at 3 , about five times higher than expected from star formation @xcite . @xcite found that constraints on inverse compton emission require the excess to come from regions with galactic - level ( @xmath192 ) magnetic fields , and suggest an evolution of the frc as the source of the reported excess . however , while smgs are individually very bright and contribute much to the cosmic star - formation rate at @xmath193 , they are not typical starbursts @xcite . instead , they seem to represent a transient phase that can survive about 100 before their gas is depleted @xcite . we can estimate the total radio background enhancement by scaling to the contribution of smgs to the cib , which is largely reprocessed starlight from galaxies at @xmath153 ( e.g. , * ? ? ? * ; * ? ? ? * ) , and adjusting by the smg radio excess . submillimeter galaxies do provide the majority of light at @xmath194 , but this is only a small fraction of the total ir background . at the peak of the cib ( @xmath195 ) , submillimeter galaxies provide @xmath196 of the total power , and possibly only @xmath197 @xcite . optimistically , the radio excess from smgs would be @xmath198 . this is significant , but not enough to explain the very large arcade2 excess . more conservatively , the excess is more likely @xmath199 , and could be as little as a few percent . since the number of radio sources down to several @xmath181 is well known , we can estimate the fraction of the expected radio background comes from bright smgs . @xcite find a tir background of about @xmath200 ; from the normalization of the fir - radio correlation we expect a 1.4 ghz background of @xmath201 . @xcite found an average radio flux density of @xmath202 for bright smgs ( @xmath203 ) . the number counts of @xmath204 sources imply that they have a density of @xmath205 , contributing roughly @xmath206 to the 1.4 ghz radio background ( e.g. , * ? ? ? bright smgs have an approximate density of @xmath207 ( e.g. , * ? ? ? * ; * ? ? ? * ) , so they constitute about @xmath145 of the background from @xmath204 sources , or @xmath208 of the expected 1.4 ghz background from star - formation . this is roughly in line with our estimate of @xmath209 , although the uncertainties are large enough that it could be consistent with smgs lying on the frc . in any case , it is fairly clear that bright smgs are not the source of the arcade excess . the total excess could be greater if most starbursts at @xmath153 are puffy , and not just smgs . we do not expect this simply because most current studies show that the local frc _ does _ hold out to high redshift for most observed galaxies and starbursts ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? the spectral slope would also be a problem : arcade2 inferred @xmath210 , while we predict a spectral slope @xmath211 , steeper than local compact starbursts . resolved radio sources in the range @xmath212 , which are expected to be star - forming galaxies @xcite , contribute about @xmath213 to the 1.4 ghz radio background @xcite , which is consistent with the total energetics expected from the frc . @xcite also find consistency between the frc and the observed radio source counts at @xmath214 . stacking studies of fainter radio sources have given somewhat ambiguous results on whether the frc applies @xcite , but @xcite find no large evolution in the frc down to @xmath215 . any large radio excess would have to come either from a new population of low luminosity galaxies or very high-@xmath1 galaxies ( see figure [ fig : ghzradiofluxes ] ) ; extrapolations of the higher flux source populations do not predict a large radio excess ( e.g. * ? ? ? * ) . nonetheless , our work indicates that a radio excess from smgs can be significant . conversely , if the luminosity function of galaxies was very steep , most galaxies could be intrinsically radio - dim with respect to the frc , because of ic losses off the cmb ( see [ sec : frcevolution ] and figures [ fig : lfirradiorest ] and [ fig : frcevolution ] ) . then the frc would overestimate the strength of the radio background . however , most star - formation at high @xmath1 is believed to have occurred in starbursts @xcite , so this possibility is unlikely . we have applied the theory of ltq to predict the frc for redshifts @xmath0 . we use one - zone models of galaxies and starbursts with cr injection , cooling , and escape to predict the equilibrium , steady - state radio spectra of galaxies and starbursts over the entire range of the frc . our goals were to determine how and why the low- and high-@xmath16 conspiracies crucial to the @xmath2 frc ( [ sec : theory ] ) affect the frc at high redshift , and to provide a quantitative model for predicting the critical redshifts at which galaxies deviate from the @xmath2 frc . we find the following : 1 . for compact starbursts ( @xmath45 ) , we find relatively little evolution in the fir - radio correlation out to @xmath216 ( figure [ fig : lfirradiorest ] ) . this is partly because the magnetic energy density in galaxies is strong enough to dominate the cmb even at high redshifts . however , the high-@xmath16 conspiracy ( [ sec : theory ] ) also acts as a buffer against ic losses off the cmb ; the increased ic losses must compete with the already present bremsstrahlung , ionization , and ic off starlight in addition to synchrotron losses . the rest - frame radio spectral slope @xmath109 at fixed @xmath125 does not change with @xmath1 , but the observed @xmath136 at fixed @xmath217 increases because the non - thermal synchrotron radio spectrum steepens at higher rest - frame frequency . 2 . we derive in appendix [ sec : cmbradiodim ] the critical redshifts when inverse compton losses off the cmb suppress the radio luminosity of galaxies compared to the @xmath2 frc . these relations are given for our standard model in equation [ eqn : b07zcrit ] . the non - thermal radio luminosity is suppressed severely in milky way - like galaxies ( @xmath89 ) at @xmath3 and the weakest compact starbursts at @xmath218 . the spectrum at ghz steepens to @xmath219 because of these enhanced ic losses . nonetheless , the low-@xmath16 conspiracy ( [ sec : theory ] ) also acts to prevent the radio emission from steeply falling with redshift , since inverse compton losses off the cmb must be more efficient than diffusive escape , not just synchrotron losses ( see [ sec : frcevolution ] ) . ltq found that the @xmath86 frc demands that @xmath12 scales with @xmath17 or @xmath16 in galaxies lying on the schmidt law . in models with @xmath119 , we find that puffy starbursts with @xmath37 such as smgs are radio bright compared to the @xmath86 frc by a factor of @xmath159 . this follows from a breakdown of the high-@xmath16 conspiracy ( [ sec : theory ] ) : bremsstrahlung and ionization cooling are weak in puffy starbursts relative to the compact starbursts that predominate in the @xmath86 universe . in contrast , in models with @xmath14 , we find that puffy starbursts are radio dim compared to the observed frc , because of weak synchrotron cooling relative to the ic losses . since several studies have reported radio excesses for smgs , we favor the @xmath119 scaling ; however the issue of whether smgs are radio - bright is still not fully resolved . in either case , puffy starbursts show little true evolution with @xmath1 , though they may appear to have fainter rest - frame radio luminosities at high @xmath1 because of their steep spectra . puffy starbursts inevitably have high @xmath136 ( @xmath220 ) , since bremsstrahlung and ionization losses are weak with respect to synchrotron and ic . a key prediction of our scenario with @xmath119 is that the variations in @xmath63 will be correlated with scale height at fixed @xmath8 , since the radio - excess in our models is a direct consequence of the large cr scale height and the small bremsstrahlung and ionization losses it causes . radio - excess ( low @xmath221 ) puffy starbursts will have steeper radio spectral slopes ( bigger @xmath136 ) , larger velocity dispersions @xmath222 compared to their rotation speeds @xmath223 , and possibly moderately cooler dust temperatures ( smaller @xmath224 ) . 4 . as previously expected , radio emission can be a poor tracer of star formation in low surface density galaxies , because of electron escape and ic losses off the cmb . for our preferred @xmath119 scaling , radio emission overestimates star - formation rate by a factor of @xmath120 in puffy starbursts . star - formation rate is underestimated by synchrotron radio emission with the @xmath14 scaling . while smgs may be individually radio bright compared to the local frc , they contribute a relatively small fraction of the cosmic infrared background and the total star - formation luminosity of the universe . this means that they enhance the star - formation radio background by @xmath225 , and possibly around @xmath209 , with respect to a naive application of the @xmath86 frc . as in ltq , we did not exactly match the observed radio spectral slopes of galaxies , with @xmath108 in normal galaxies and @xmath106 in compact starbursts . this will have a slight effect on the k - correction . an error of @xmath226 in @xmath136 should only affect inferred radio luminosity by 30% at @xmath146 and 60% at @xmath100 . nevertheless , the prediction of steeper radio spectra in puffy starbursts with respect to compact starbursts at all relevant @xmath1 should be robust . our explanation for the small @xmath63 ratio in submillimeter galaxies as a breakdown of the high-@xmath16 conspiracy is based purely on the steady - state spatially - averaged synchrotron emission , but the details of the fir emission may also matter . throughout this paper , we have simply assumed that the bolometric fir luminosity could be correctly inferred from observations , and have assumed the same uv opacity for all galaxies at all redshifts . the total fir emission is also likely to depend on the metallicity , and may be lower at high @xmath1 for the lowest surface density galaxies . the exact far - infrared sed is important in determining the fir emission when observations have only been made at only a few wavelengths . the presence of agns , a different imf at high @xmath1 , and selection biases may also affect the inferred @xmath221 of smgs . we did not include the effects of galaxy evolution on the cr spectrum in our models . we argued in ltq that it should not matter for quiescent spirals or for extreme starbursts , because the cr lifetime is much shorter than the time dependence of stellar populations . however , galaxy evolution may play a role in weaker starbursts @xcite and in post - starburst galaxies @xcite . studies of merging normal galaxies and galaxies in clusters have indeed found that they are radio bright with respect to the frc , possibly because of compression of magnetic fields or shock acceleration @xcite . we also assumed that the magnetic field strength at a given density does not depend on redshift . it is not entirely clear how long normal galaxies take to build up their magnetic fields , or even what process is at work ( see the reviews in * ? ? ? * ; * ? ? ? * ) , though there are theoretical mechanisms that can rapidly generate strong magnetic fields . studies at @xmath3 indicate that normal milky way - like galaxies had magnetic fields with similar strengths to the present @xcite . at the very highest redshifts , magnetic field strengths might be weaker , because the seed fields were essentially zero compared to the present strengths . starbursts also may build their magnetic fields up in much shorter times than normal galaxies , and through a different process than normal galaxies @xcite . finally , we have used one - zone models , which are appropriate if the crs sample all of the gas phases in each galaxy s ism . however , the ism is known to be clumpy in the milky way , in compact starbursts like arp 220 ( e.g. , * ? ? ? * ) , and even in the smgs themselves @xcite . a full understanding of the frc will probably require models that take into account the inhomogeneity of star - forming galaxies . we thank eric murphy and michal michaowski for critical readings of the text , eliot quataert for many stimulating conversations , and the galprop team for making their code and its subroutines freely available . galprop is available at http://galprop.stanford.edu . we would like to thank igor moskalenko for sharing his group s estimates of the milky way s total @xmath29-ray luminosity for normalization of our models of the frc . t. a. t. is supported in part by an alfred p. sloan fellowship . at high redshift , the nonthermal synchrotron radio luminosity of galaxies is suppressed by the cmb . this is because the inverse compton process shortens the lifetime that crs have to radiate synchrotron ; equivalently , their energy goes into inverse compton photons instead of synchrotron radio . because of the shorter times the cr electrons and positrons have to radiate synchrotron , the redshift @xmath101 when the radio luminosity of a galaxy is quenched by a factor @xmath227 is given by @xmath228 the loss time includes all cooling and escape losses , so that @xmath229 in some regimes , some losses may be neglected . for example , diffusive losses are unimportant in starbursts , as are ic losses off the cmb at @xmath86 . advective losses may be unimportant in normal galaxies and also for the densest starbursts . we can immediately see that the presence of losses besides synchrotron and ic off the cmb prevent the suppression of radio . in the case when no losses other than synchrotron and ic off the cmb exist , equation [ eqn : radiosuppress ] reduces to @xmath230 the presence of other loss processes , whether escape or cooling , requires a higher @xmath101 for a fixed quench factor @xmath227 : @xmath231 essentially the ic losses off the cmb must compete with not only synchrotron losses , but every other cooling and escape process as well . with the high-@xmath16 conspiracy , @xmath232 can be of order @xmath233 , so the energy density of the cmb must be @xmath30 times greater to suppress the radio emission than would be naively expected . we give the loss times for each process in ltq . from those lifetimes , we can find the ratios of the synchrotron cooling timescale to the other loss timescales : @xmath234 where @xmath235 , @xmath236 , @xmath237 is the wind speed in units of 300 @xmath238 , and @xmath239 is the rest - frame frequency divided by 1.4 ghz . our models parameterize @xmath12 in terms of the gas surface density @xmath16 : @xmath240 finally , the schmidt law allows us to convert between gas surface density @xmath16 and surface density of star - formation rate @xmath8 : @xmath241 we can approximately solve for the redshift @xmath101 when @xmath242 for galaxies and starbursts in each of the scenarios in table [ table : models ] . for the scenario with the b07 star - formation law , @xmath85 , and no winds , we find : @xmath243 where @xmath8 is in units of @xmath42 . for the scenario with the b07 star - formation law , @xmath244 , and winds in starbursts , we find : @xmath245 for the scenario with the b07 star - formation law , @xmath246 , and winds in starbursts , we find : @xmath247 for the standard model of ltq , with the k98 star - formation law , @xmath85 , and no winds , we find : @xmath248 finally , for the model of ltq with the k98 star - formation law , @xmath131 , and winds in starbursts , we find : @xmath249 condon , j. j. 2009 , sensitive continuum survey with the ska : goals and challenges ( manchester : ska program development office ) , ska memo # 114 ( http://www.skatelescope.org/pdf/memos/114_memo_condon.pdf ) lllccccc k98 & n & @xmath250 & 1.7 ( 1.5 ) & 3.9 ( -0.59 ) & 0.92 - 0.96 & 0.44 - 0.64 & 0.69 - 0.92 + k98 & y & @xmath251 & 1.7 ( 1.7 ) & 0.85 ( 0.07 ) & 0.88 - 0.94 & 0.43 - 0.61 & 0.78 - 0.85 + b07 & n & @xmath250 & 1.7 ( 1.5 ) & 2.2 ( -0.35 ) & 0.87 - 0.96 & 0.57 - 0.67 & 0.86 - 0.98 + b07 & y & @xmath252 & 2.2 ( 1.7 ) & 3.4 ( -0.53 ) & 0.85 - 0.91 & 0.45 - 0.59 & 0.75 - 0.87 + b07 & y & @xmath253 & 2.0 ( 2.6 ) & 0.49 ( 0.31 ) & 0.87 - 0.91 & 0.49 - 0.64 & 0.88 - 0.91 + [ table : models ]

we investigate lattices that can be represented as sublattices of the lattice of all convex subsets of a linearly ordered set and as lattices of convex subsets of .

math0609114

a _ lattice of sets _ is a family @xmath1 of sets which is a lattice with respect the inclusion relation @xmath2 . a _ ring of sets _ is a family of sets closed under finite unions and intersections . it is well known that a lattice is isomorphic to a ring of sets and only if it is distributive ( @xcite * vii * 8 , or @xcite theorem 10.3 ) . on the other hand any lattice is isomorphic to a lattice of sets closed under finite intersections ( see section 3 , theorem [ threp ] ) . a _ linear interval _ , or simply an _ interval _ , is a convex subset of linearly ordered space @xmath0 . the set of all intervals of @xmath0 is denoted by @xmath3 or simply by @xmath4 . the lattice @xmath5 is a simple example of a non modular lattice of sets . if the set @xmath6 is finite then @xmath7 is a planar lattice and its diagram was described by a. r. schweitzer ( @xcite ) . the lattice @xmath7 , where @xmath8 is partial ordered set , was studied by g. birkhoff and m. k. bennet ( @xcite ) . the class * sub * of all lattices that can be embedded into some lattice of the form @xmath7 was studied by m. semenova and f. wehrung in @xcite which proved that * sub * forms a variety , defined by three indentites denoted by ( s ) , ( u ) and ( b ) . in @xcite the same authors extended this result for sublattices of products of lattices of convex subsets o we say that a lattice @xmath9 with 0 is _ representable as a lattice of intervals _ if there exists a linearly ordered set @xmath8 and an order - embedding @xmath10 such that @xmath11 and @xmath12 . if moreover @xmath13 , which means that @xmath14 is isomorphic to a sublattice of the lattice of all intervals of @xmath0 , we shall say that @xmath14 is _ faithfully representable as a lattice of intervals_. in this note we investigate lattices with 0 ( finite or infinite ) which can be represented or faithfully represented as lattices of intervals . we give three lattice theoretical conditions which are described in the following definition : [ def11 ] a lattice @xmath9 is said to be a * loc - lattice * if it has the following properties : 1 . for every @xmath15 we have that @xmath16 or @xmath17 or @xmath18 . 2 . if @xmath19 are pairwise incomparable elements of @xmath20 and @xmath21 then @xmath22 . 3 . if @xmath19 are elements of @xmath20 such that @xmath21 , @xmath23 and @xmath24 then @xmath25 . a _ semi - prime filter _ of a lattice @xmath9 is a proper filter @xmath26 of @xmath14 such that for every @xmath27 with @xmath28 either @xmath29 or @xmath30 or there exists an @xmath31 such that @xmath32 and @xmath33 . the lattice @xmath9 is said to be _ well separated _ if for every semi - prime filter @xmath26 of @xmath20 and every @xmath34 there exists an @xmath35 such that @xmath36 . in figure fig1 ( a ) we have an example of a loc - lattice with five semi - prime filters , @xmath37 , where @xmath38 . this lattice is not well separated ( @xmath39 and for every @xmath40 , @xmath41 ) and it is representable ( but not faithfully representable ) as lattice of intervals . the lattice of fig1 ( b ) is a loc - lattice , it also has five semi - prime filters , @xmath37 , it is not well separated and it is faithfully representable as a lattice of intervals . the main result of this note is the following : [ ath]@xmath42 1 . every loc - lattice @xmath9 is representable as a lattice of intervals . 2 . a well - separated lattice is faithfully representable as a lattice of intervals if and only if it is a loc - lattice . a _ lattice of sets _ on a set @xmath6 is a family @xmath43 of subsets of @xmath6 such that @xmath44 is a lattice . we say that the family @xmath43 _ separates _ the set @xmath6 if for every @xmath45 with @xmath46 either there exists an @xmath47 with @xmath48 and @xmath49 or there exists an @xmath50 with @xmath51 and @xmath52 . we also say that the family @xmath43 _ well - separates _ @xmath6 if it separates @xmath6 and for every @xmath47 and every @xmath53 with @xmath54 there exists an @xmath55 with @xmath56 and @xmath57 . in section 2 we show ( theorem [ thlocset ] ) that if @xmath43 is family of subsets of a set @xmath6 with some special properties - called a _ loc - lattice of sets _ ( definition [ def21])- which separates @xmath6 , then we may construct a linear ordering @xmath58 of the set @xmath6 such that every @xmath47 is a convex subset of @xmath0 . moreover if @xmath43 well - separates @xmath6 then @xmath43 is a sublattice of the lattice of all convex sets of @xmath0 . in section 3 we introduce the notion of the _ semi - prime filter _ of a lattice and we show that every lattice is isomorphic to a lattice @xmath43 of subsets of a set @xmath6 ( the set of all semi - prime filters the lattice ) which is closed under finite intersections and separates the points of @xmath6 . moreover , if @xmath14 is loc - lattice then @xmath14 is isomorphic to a loc - lattice @xmath44 of subsets of a set @xmath6 which separates @xmath6 and theorem [ ath ] follows from the results of section 2 . finally in section 4 we present some applications of theorems [ ath ] and [ thlocset ] in general topology . a family of sets @xmath43 is said to be a _ lattice of sets _ if @xmath44 is a lattice . the lattice sum of two elements @xmath59 of @xmath43 is denoted by @xmath60 and their lattice product by @xmath61 . [ def21 ] a family @xmath43 of subsets of a set @xmath6 is called a * loc - lattice of sets * if it is a lattice of subsets of @xmath6 satisfying the following conditions : 1 . for every @xmath62 , @xmath63 . 2 . if @xmath64 and @xmath65 then @xmath66 . 3 . for every @xmath67 we have that either @xmath68 or @xmath69 or @xmath70 . 4 . if @xmath71 are pairwise incomparable elements of @xmath44 and @xmath69 then @xmath72 . given a family @xmath43 of subsets of a set @xmath6 we define a binary relation @xmath73 as follows . we write @xmath74 if for every @xmath47 with @xmath48 we have that @xmath75 . we say that a family @xmath43 _ separates _ the points of @xmath6 ( or that @xmath43 is a _ separating family _ for @xmath6 ) if for every @xmath45 with @xmath46 we have that @xmath76 or @xmath77 , where @xmath78 is the negation of @xmath74 . this notion of a separating family is introduced by a. renyi @xcite which has showed that the minimal size of a separating family of a finite set @xmath6 is exactly @xmath79 . let @xmath45 . we say that @xmath80 are _ completely separated _ if there exist @xmath62 with @xmath81 , @xmath82 , @xmath83 and @xmath84 and that @xmath85 are _ totally separated _ if there exist @xmath62 with @xmath86 , @xmath87 and @xmath88 . we also say that the family @xmath43 _ completely separates _ ( resp . _ totally separates _ ) the set @xmath6 if for every two distinct points of @xmath6 are completely separated ( resp . totally separated ) . finally , we say that the family @xmath43 _ well separates _ the set @xmath6 if for every @xmath47 and @xmath54 there exists an @xmath50 such that @xmath89 and @xmath90 . note that if @xmath43 completely separates @xmath6 then it well separates @xmath6 . the main result of this section is the following : [ thlocset ] let @xmath6 be a set and @xmath43 be a loc - lattice of subsets of a set @xmath6 which separates the points of @xmath6 . 1 . there exists a linear ordering @xmath58 of @xmath6 such that every @xmath47 is a convex subset of @xmath6 . if @xmath43 well separates @xmath6 then for every @xmath62 , @xmath91 and so @xmath44 is a sublattice of @xmath92 . the following lemma summarizes the basic properties of a loc - lattice of sets : [ lemp1 ] let @xmath43 be a loc - lattice of sets and @xmath93 . then 1 . if the sets @xmath71 are pairwise incomparable , @xmath68 and @xmath69 then @xmath94 . therefore , if the sets @xmath71 are pairwise incomparable then only one of the relations @xmath68 , @xmath69 , @xmath70 occurs . 2 . if @xmath71 are pairwise disjoint and @xmath95 then @xmath69 . 3 . if @xmath96 and @xmath97 then @xmath98 . 4 . if @xmath99 and @xmath100 then @xmath101 . \(1 ) if @xmath68 and @xmath69 then @xmath102 , by condition ( 4 ) of definition [ def21 ] . + ( 2 ) if @xmath68 then by condition ( 4 ) of definition [ def21 ] we have that @xmath103 . but then @xmath104 , a contradiction . if @xmath70 then @xmath105 and so @xmath106 , a contradiction . therefore by condition ( 3 ) of definition [ def21 ] ) we must have @xmath69 . + ( 3 ) since @xmath96 and @xmath97 we have that @xmath107 . so , @xmath108 . + ( 4 ) we may suppose that @xmath65 . the case @xmath109 is impossible since then @xmath110 which contradicts the assumption @xmath111 . the case @xmath112 is also impossible . therefore , @xmath113 . a _ ternary relation _ in a set @xmath6 is a subset @xmath114 of @xmath115 . we shall use the notation @xmath116 instead of @xmath117 and that @xmath118 instead of @xmath119 . we define a ternary relation in @xmath6 setting @xmath120 if for every @xmath47 with @xmath121 we have also that @xmath122 . let @xmath123 . given a linear ordering @xmath58 of @xmath6 we set @xmath124 let @xmath43 be a family of subsets of a set @xmath6 and @xmath125 . we say that a linear ordering @xmath58 of @xmath126 of @xmath6 is @xmath43- * consistent * if @xmath127 . clearly in order to prove theorem [ thlocset ] we must find an @xmath43-consistent linear ordering @xmath128 of @xmath6 [ lem04 ] let @xmath43 be a loc - lattice of subsets of a set @xmath6 . then for every @xmath129 either @xmath120 or @xmath130 or @xmath131 . if @xmath132 and @xmath133 then there exist @xmath134 such that @xmath135 , @xmath136 , @xmath137 and @xmath138 . let @xmath139 such that @xmath140 . then @xmath141 and @xmath142 . by lemma [ lemp1 ] ( 4 ) we have that @xmath143 and so @xmath144 . if @xmath59 are subsets of a linear ordered set @xmath0 we write @xmath145 if for every @xmath86 and @xmath87 we have that @xmath146 . similarly if @xmath147 and @xmath148 we write @xmath149 if @xmath146 for every @xmath87 . [ lemlocsep ] let @xmath43 be a loc - lattice @xmath43 of subsets of @xmath6which well - separates @xmath6 and @xmath58 an @xmath43-consistent linear ordering of @xmath6 . then @xmath43 is a sublattice of the lattice @xmath150 of intervals of @xmath0 . clearly @xmath43 is a subset of @xmath150 . it remains to show that if @xmath62 then @xmath91 where @xmath151 denotes the convex hull of a subset @xmath152 of @xmath6 . since @xmath60 is convex we have that @xmath153 . suppose that @xmath154 . then @xmath88 and therefore either @xmath145 or @xmath155 . suppose that @xmath145 . let @xmath156 . then we must have that either @xmath157 or @xmath158 . suppose that @xmath158 and let @xmath159 with @xmath160 and @xmath161 . the sets @xmath71 are incomparable , @xmath69 and @xmath70 which contradicts lemma [ lemp1 ] ( 1 ) . the following theorem , due to m. altwegg @xcite ( see also @xcite ) , is usefull since it characterizes linear orderings of sets in terms of ternary relations . note that by lemma [ lem04 ] @xmath162 always satisfies condition ( 3 ) of the theorem of altwegg . [ thalt ] let @xmath114 be a ternary relation in a set @xmath6 which satisfies the following postulates : 1 . @xmath163 if and only if @xmath164 . 2 . if @xmath116 and @xmath165 then either @xmath166 or @xmath167 . 3 . for every @xmath129 either @xmath116 or @xmath168 or @xmath169 . then there exists a linear ordering @xmath58 of @xmath6 such that for every @xmath170 we have that @xmath171 if and only if @xmath172 . moreover , every other linear ordering @xmath173 of @xmath6 satisfying the preceding condition is equal to @xmath58 or to the inverse order @xmath174 of @xmath58 . the ternary relation @xmath162 does not satisfy in general the conditions of theorem [ thalt ] since for a loc - lattice @xmath43 of subsets of @xmath6 there are probably many and even non - isomorphic @xmath43-consistent linear orderings . indeed , the simplest example of a loc - lattice of subsets of a set @xmath6 is a chain of subsets of @xmath6 . suppose that @xmath175 and that @xmath176 where @xmath177 then the usual ordering @xmath178 of @xmath179 and the ordering @xmath173 given by @xmath180 are non - isomorphic @xmath43-consistent linear orderings of @xmath179 . let @xmath6 be a set and @xmath43 a family of subsets of @xmath6 . if @xmath181 is a sequence of pairwise distinct points of @xmath6 we say that @xmath182 is a * representative family * for @xmath183 if for every @xmath184 with @xmath185 we have that @xmath186 , @xmath187 and @xmath188 . it is clear that if the family @xmath43 is closed under finite intersections then for every finite sequence @xmath183 of pairwise completely separated points of @xmath6 there exists a _ representative family _ for @xmath183 . [ lemp3 ] let @xmath43 be a loc lattice of subsets of a set @xmath6 , @xmath189 three distinct points of @xmath6 which are pairwise completely separated . let @xmath190 a representative family for the triple @xmath191 . then 1 . @xmath120 holds if and only @xmath69 . 2 . if @xmath192 then @xmath193 , @xmath194 and @xmath195 . ( @xmath196)we shall show that if @xmath120 holds then for every representative family @xmath197 of @xmath191 we have that @xmath69 . if @xmath68 then @xmath198 and so @xmath199 . since @xmath120 we have that @xmath200 and so @xmath201 , a contradiction . in the same manner the case @xmath70 is excluded . therefore , we must have @xmath69 . suppose now that there exists a representative family @xmath197 for @xmath191 such that @xmath69 and @xmath120 does not hold . then by lemma [ lem04 ] either @xmath131 or @xmath202 and by ( a ) , either @xmath68 or @xmath70 , but this contradicts lemma [ lemp1 ] ( 1 ) . ( @xmath203 ) suppose that @xmath96 then @xmath204 and since @xmath69 we have that either @xmath201 or @xmath122 , which contradicts the assumption that @xmath197 is a representative family for @xmath191 . if @xmath205 since @xmath206 we shall have @xmath207 , a contradiction . similarly we see that @xmath208 [ lem02 ] let @xmath43 be a loc - lattice of subsets of a set @xmath6 which completely separates @xmath6 . then @xmath43 totally separates @xmath6 . let @xmath85 be distinct points of @xmath6 and @xmath209 a representative family for @xmath210 . suppose that @xmath211 and let @xmath212 . then we may choose a representative family @xmath197 for @xmath191 with @xmath213 , @xmath214 , @xmath215 . since @xmath216 we can not have @xmath217 . similarly , the case @xmath218 is impossible . therefore by property ( 1 ) of definition[def11 ] we must have that @xmath70 . by lemma [ lemp3 ] ( 2 ) we conclude that @xmath88 . [ lemp4 ] let @xmath43 be a loc lattice of subsets of a set @xmath6 , @xmath219 five distinct points of @xmath6 such that every two of them are completely separated and let @xmath220 a representative family for @xmath221 . then 1 . @xmath69 and @xmath222 if and only if @xmath223 and @xmath224 . 2 . @xmath69 and @xmath225 if and only if @xmath223 and @xmath226 . 3 . if @xmath69 and @xmath227 then either @xmath228 or @xmath229 . \(1 ) suppose that @xmath69 and @xmath222 . then either @xmath230 or @xmath223 or @xmath231 . since @xmath69 and @xmath71 are pairwise incomparable we have that @xmath232 and so @xmath195 . if @xmath231 then @xmath233 and since @xmath222 we have that @xmath70 , a contradiction . in the same manner the case @xmath230 is excluded . so , @xmath69 . + ( 2)similar to ( 1 ) . + ( 3)it follows readily from ( 1 ) and ( 2 ) . [ prop ] let @xmath43 be a loc - lattice of subsets of a set @xmath6 which completely separates the points of a subset @xmath126 of @xmath6 . then there exists a consistent linear ordering @xmath58 of @xmath126 . moreover , every other consistent linear ordering @xmath173 of @xmath126 is equal to @xmath58 or equal to the inverse order of @xmath58 . let @xmath234 . by lemmas [ lem04 ] , [ lemp3 ] and [ lemp4 ] we see that the ternary relation @xmath235 of @xmath126 satisfies the conditions of altwegg s theorem [ thalt ] . so we may find a @xmath236-consistent linear ordering of @xmath126 which will be @xmath43-consistent too . by lemma [ lemlocsep ] we have as an immediate corollary the following . [ thsep ] let @xmath6 be a set and @xmath43 be a loc - lattice of subsets of a set @xmath6 which completely separates @xmath6 . then there exists a linear ordering @xmath58 of @xmath6 such that @xmath43 is a sublattice of the lattice @xmath237 of all intervals of @xmath0 . moreover , every other linear ordering @xmath173 of @xmath6 satisfying the preceding condition is equal to @xmath58 or to the inverse order @xmath174 of @xmath58 . the general case of theorem [ thlocset ] , when the family @xmath43 does not completely separate the set @xmath6 is much more complicated . + * proof of theorem [ thlocset ] * before we proceed to the proof of theorem [ thlocset ] we shall introduce some further notation . let @xmath0 be a linearly ordered set . if @xmath238 we set @xmath239 a _ section _ of a linearly ordered set @xmath240 a pair @xmath241 of subsets of @xmath6 such that @xmath242 and @xmath145 ( for the theory of sections see also @xcite ) . the the pairs @xmath243 and @xmath244 are considered as sections too . the set of all sections of a linearly ordered set @xmath0 is denoted by @xmath245 . the section @xmath241 is said to be 1 . a _ gap _ if @xmath152 has no greatest element and @xmath246 has no least element , 2 . a _ jump _ if @xmath152 has a greatest element and @xmath246 has a least element , 3 . a _ left cut _ if @xmath152 has a greatest element and @xmath246 has no least element , 4 . right cut _ if @xmath152 has no greatest element and @xmath246 has a least element . given a section @xmath241 we also denote by @xmath247 the first member @xmath152 of the section and by @xmath248 the second member @xmath246 of the section . + we define a binary relation @xmath73 on @xmath6 by the rule that @xmath74 if every element @xmath249 of @xmath43 containing @xmath250 also contains @xmath251 . it is trivial that @xmath252 is reflexive and transitive , and since @xmath43 separates the set @xmath6 it is also antisymmetric ; that is @xmath252 is a partial order on @xmath6 . we write @xmath78 if @xmath74 does not hold . two points @xmath80 of @xmath6 is said to be _ independent _ if @xmath253 and @xmath254 . a subset @xmath126 of @xmath6 is said to be _ independent _ if every two distinct elements of @xmath126 are independent . + let @xmath126 be a subset , @xmath58 a linear ordering of @xmath126 , @xmath255 the set of all sections and @xmath256 any mapping from the remainder set @xmath257 to the set @xmath255 . suppose further that for any @xmath258 we have defined a linear ordered set @xmath259 with @xmath260 . then we may extend the linear ordering @xmath58 of @xmath126 to a linear ordering @xmath173 of @xmath261 by the following rules : 1 . if @xmath80 in @xmath126 we set @xmath262 if and only if @xmath263 . + 2 . if @xmath264 with @xmath241 and @xmath265 we set @xmath262 if @xmath82 and @xmath266 if @xmath267 . + [ 0.6ex ] 3 . if @xmath264 with @xmath241 , @xmath268 with @xmath269 and @xmath270 we set @xmath271 if and only if @xmath272 . + [ 0.6ex ] 4 . if @xmath264 with @xmath241 , @xmath268 with @xmath269 and @xmath273 we set @xmath271 if and only if @xmath274 . we shall call the linear ordering @xmath173 the * extension * of @xmath58 to the set @xmath261 . let @xmath43 be a family of subsets of a set @xmath6 . a triple @xmath275 is said to be an @xmath43-*consistent triple * if the following conditions are fulfilled : 1 . @xmath126 is a subset of @xmath6 and @xmath58 is an @xmath43-consistent linear ordering of @xmath126 . 2 . @xmath276 is a map from the set @xmath257 to the set @xmath277 of all sections of @xmath278 so that for every family @xmath279 with @xmath260 and @xmath280 to be an @xmath43-consistent linear ordering of @xmath281 , the extension of @xmath58 to the set @xmath282 is also @xmath43-consistent . let @xmath283 the class of all ordinal numbers . given a loc - family @xmath43 of subsets of @xmath6 which separates the points of @xmath6 we shall construct a family @xmath284 with the following properties : 1 . for every @xmath285 , @xmath286 is an @xmath43-consistent triple . 2 . if @xmath287 then @xmath288 is an extension of @xmath289 . 3 . if @xmath290 then @xmath291 . clearly if a such family @xmath292 exists , then for some @xmath293 we shall have that @xmath294 and theorem [ thlocset ] has been proved . * step 0 . * in this step of the proof we shall construct an @xmath43-consistent triple @xmath295 . by zorn s lemma the set @xmath6 contains a maximal independent subset @xmath126 . if @xmath126 is an independent subset of @xmath6 then by proposition [ prop ] there exists an @xmath43-consistent linear ordering of the set @xmath126 . let @xmath126 be a maximal independent subset of @xmath6 and @xmath58 an @xmath43-consistent linear ordering of @xmath126 . if @xmath296 then it is easy to see that the relation @xmath252 is an @xmath43-consistent linear ordering of @xmath6 . so we may assume that @xmath297 . + for every @xmath298 we set @xmath299 note that for every @xmath298 we have that @xmath300 if and only if @xmath301 and that @xmath302 is a convex subset of @xmath278 . by lemma [ lemp3 ] ( ii ) we have that @xmath303 and if @xmath304 with @xmath305 then @xmath306 is the immediate successor of @xmath251 . we classify the points of @xmath257 in eight classes or types by the following rules : a point @xmath298 is said to be a point of + @xmath307 if @xmath308 , @xmath309 and @xmath310 , + @xmath311 if @xmath308 , @xmath309 and @xmath312 , + @xmath313 if @xmath308 , @xmath314 and @xmath310 , + @xmath315 if @xmath316 , + @xmath317 if @xmath318 , + @xmath319 if @xmath320 and @xmath251 is neither the first nor the last element of @xmath126 , + @xmath321 if @xmath320 and @xmath251 is the first element of @xmath126 , + @xmath322 if @xmath320 and @xmath251 is the last element of @xmath126 . + for every @xmath298 with @xmath308 . we set @xmath323 l_2(x)=\{y\in l(x ) : \text{there exists a } \ ; y'\in l^+(x ) \ ; \text{with}\ ; \neg(yxy')_{{\mathcal{m}}}\}.\\ \end{array}\ ] ] let @xmath298 such that @xmath308 and @xmath324 . then the set @xmath325 is an initial segment of @xmath302 ( if it is not empty ) and @xmath326 is a final segment of @xmath302 ( if it is not empty ) . moreover , @xmath327 let @xmath328 and @xmath329 with @xmath330 . then there exists a @xmath331 and @xmath47 with @xmath332 and @xmath54 . since @xmath333 and @xmath58 is consistent we have that @xmath334 and so @xmath335 . this shows that @xmath325 is an initial segment of @xmath302 . by the same reasoning we show that @xmath326 is a final segment of @xmath302 . suppose that there exists @xmath336 and let @xmath337 with @xmath338 . we may find @xmath339 with @xmath340 , @xmath341 , such that @xmath342 is a representative triple for @xmath343 . clearly @xmath344 and @xmath345 . but then @xmath346 and @xmath347 which contradicts lemma [ lemp3 ] . finally suppose that there exists a @xmath348 . let @xmath337 with @xmath338 and let @xmath342 be a representative triple for @xmath349 with @xmath54 . since @xmath350 we have that @xmath351 and since @xmath352 we have that @xmath353 . but then @xmath354 , a contradiction . for every @xmath298 we associate a section @xmath355 of @xmath126 as follows : + * if @xmath250 is a point of type @xmath356 we set @xmath357 b_x = l_1^+(x)=l_2(x)\cup l_2^+(x ) . \end{array}\ ] ] * if @xmath250 is a point of type @xmath358 we set @xmath359 b_x = l_2(x)\cup l_2^+(x ) . \end{array}\ ] ] * if @xmath250 is a point of type @xmath360 we set @xmath361 b_x = l_1^+(x ) . \end{array}\ ] ] let @xmath362 be the set of all points of @xmath6 of type @xmath363 , @xmath364 the set of all points of @xmath257 of type @xmath358 with @xmath365 and @xmath366 the set of all points of @xmath257 of type @xmath360 with @xmath367 . for every @xmath368 we set @xmath369 [ lemp5 ] let @xmath370 is a subset of @xmath257 with @xmath371 then @xmath372 . moreover @xmath373 and @xmath374 . suppose that there exist three independent points @xmath375 then we can find a representative triple @xmath376 for @xmath377 . but then @xmath378 which contradicts lemma [ lemp3 ] . suppose that there exist two independent points @xmath379 . we may suppose that @xmath380 . let @xmath381 , @xmath382 . then we may find a representative triple @xmath383 for @xmath384 with @xmath75 . but then @xmath385 which again contradicts lemma [ lemp3 ] . [ lemp6 ] every point of @xmath364 is independent from any point of @xmath366 . suppose that there exist @xmath386 and @xmath387 such that @xmath388 . then we see that @xmath389 , a contradiction . * there exist @xmath390 and @xmath386 such that @xmath391 . + in that case clearly @xmath392 ( otherwise @xmath389 ) and so by lemmas [ lemp5 ] and [ lemp6 ] we have that @xmath393 for every @xmath387 . we set @xmath394 let any @xmath395 . if @xmath396 is independent from @xmath397 then @xmath398 from every @xmath386 and we set @xmath399 if @xmath396 is not independent from @xmath397 we set @xmath400 * there exist @xmath390 and @xmath387 such that @xmath401 . + in that case clearly @xmath402 ( otherwise @xmath403 ) and so by lemmas [ lemp5 ] and [ lemp6 ] we have that @xmath398 for every @xmath404 . we set @xmath405 let any @xmath395 . if @xmath396 is independent from @xmath397 then @xmath393 from every @xmath387 and we set @xmath406 if @xmath396 is not independent from @xmath397 we set @xmath400 * @xmath407 and for every @xmath408 , @xmath404 and @xmath387 we have that @xmath409 and @xmath410 . + we select two independent points @xmath411 of @xmath362 and we set @xmath412 for any other @xmath395 we set @xmath413 if the points @xmath414 are not completely separated and @xmath415 if the points @xmath416 are not completely separated . + * @xmath417 is point of type @xmath418 and @xmath419 . in that case we set @xmath420 [ lemp11 ] if @xmath250 is a point of type @xmath421 with @xmath422 then one and only one of the following cases occurs : 1 . for every @xmath423 we have @xmath424 2 . for every @xmath425 we have @xmath424 . suppose that for some @xmath423 we have that @xmath426 . let @xmath427 and let @xmath428 a representative triple for @xmath429 with @xmath430 . then @xmath431 since @xmath430 and @xmath432 we have that @xmath433 and so @xmath434 . the same argument shows that there exist no @xmath435 with @xmath436 such that @xmath437 and @xmath438 . * @xmath417 is a point of type @xmath421 and for every @xmath423 we have @xmath424 . + in that case we set @xmath439 * @xmath417 is a point of type @xmath421 and for every @xmath425 we have @xmath424 . + in that case we set @xmath440 * @xmath417 is a point of type @xmath441 , @xmath442 where @xmath443 is the first element of @xmath126 and for some @xmath444 we have that @xmath445 then we set @xmath446 * @xmath417 is a point of type @xmath441 , @xmath442 where @xmath443 is the first element of @xmath126 and for every @xmath444 we have that @xmath447 then we set @xmath448 * @xmath417 is a point of type @xmath449 , @xmath450 where @xmath451 is the last element of @xmath126 and for some @xmath452 we have that @xmath453 then we set @xmath454 * @xmath417 is a point of type @xmath449 , @xmath450 where @xmath451 is the last element of @xmath126 and for every @xmath452 we have that @xmath455 then we set @xmath456 [ lemp7 ] let @xmath298 with @xmath355 . then for every @xmath457 we have that @xmath458 . suppose that @xmath250 is a point of type @xmath356 . then we may find @xmath459 and @xmath460 with @xmath461 . let @xmath339 with @xmath462 , @xmath463 , @xmath48 @xmath464 , @xmath465 and @xmath466 . since @xmath467 there exists a @xmath75 with @xmath468 . the triple @xmath469 is a representative family for @xmath470 and so @xmath346 . if @xmath55 with @xmath471 then @xmath472 and @xmath473 by lemma [ lemp1 ] ( 3 ) we have that @xmath474 and so @xmath56 . this shows that @xmath458 . the other cases can also be checked easily . [ lemp8 ] if @xmath475 with @xmath476 and @xmath477 then @xmath478 . in particular , for every section @xmath241 of @xmath278 there exists at most one point @xmath250 such that @xmath479 , @xmath480 and @xmath481 . clearly @xmath482 . if @xmath47 with @xmath48 then @xmath483 and @xmath484 and by lemma [ lemp7 ] we shall have that @xmath485 . so @xmath486 . similarly , @xmath487 and so @xmath478 , since the family @xmath43 separates the set @xmath6 . let us call the points of type @xmath418 , the points @xmath250 of type @xmath441 such that @xmath488 ( i.e. the points of case ( viii ) ) and the points @xmath250 of type @xmath449 with @xmath489 ( i.e. the points of case ( x ) ) _ exceptional points _ and the other points of @xmath257 _ ordinary points_. we set @xmath490 our next step is to extend the linear ordering @xmath58 to an @xmath43-consistent linear ordering @xmath491 of @xmath492 . before doing this for every section @xmath241 of @xmath278 we shall define a consistent linear ordering @xmath493 to the set @xmath494 of all ordinary points @xmath250 of @xmath257 with @xmath479 . 1 . @xmath495 . in that case an ordinary point @xmath496 is a point of type @xmath358 or of the type @xmath363 . by lemma [ lemp5 ] and the previous construction we have that for any @xmath497 either @xmath486 or @xmath487 . we set @xmath498 if @xmath74 . @xmath499 . in that case for @xmath500 we set @xmath263 if and only if @xmath501 . let @xmath241 to be a section of @xmath278 such that @xmath502 and @xmath503 . we set @xmath504 \widetilde{\mathbf{s}}_1=\{x\in \widetilde{\mathbf{s } } : l(x)\cap a\neq \emptyset\ } , \;\ ; \widetilde{\mathbf{s}}_2=\{x\in \widetilde{\mathbf{s } } : l(x)\cap b\neq \emptyset\}\\[0.7ex ] \widetilde{\mathbf{s}}_3=\{x\in \widetilde{\mathbf{s } } : m(x ) \cap a\neq \emptyset\ } , \;\ ; \widetilde{\mathbf{s}}_4=\{x\in \widetilde{\mathbf{s } } : m(x)\cap b\neq \emptyset\}.\\ [ 0.7ex ] \end{array}\ ] ] [ lemp9 ] for every @xmath505 , @xmath506 , @xmath507 we have that @xmath508 , @xmath509 , @xmath510 and @xmath511 . the fact that @xmath508 follows from lemma [ lemp7 ] . let @xmath512 , @xmath82 and @xmath443 to be the last element of @xmath152 , which exists since @xmath513 . suppose that there exists a @xmath47 with @xmath332 and @xmath54 . let @xmath55 with @xmath514 and @xmath515 . then @xmath516 and so @xmath517 . but then @xmath518 and @xmath519 which contradicts lemma [ lemp7 ] . the other cases are treated similarly . [ lemp10 ] @xmath520 . suppose that there exist two independent points @xmath521 of @xmath522 . since @xmath523 then the ( non - empty ) sets @xmath524 are not initial segments of @xmath126 . also we may suppose that @xmath525 . so there exist @xmath328 , @xmath526 and @xmath527 such that @xmath528 and @xmath529 . the set @xmath530 is independent . therefore there exists a representative triple @xmath531 for @xmath532 with @xmath75 . but then @xmath533 which contradicts lemma [ lemp3 ] . so @xmath534 . in the same manner we show that @xmath535 . the ordinary points of @xmath536 are those of @xmath537 , @xmath538 and @xmath539 . given @xmath540 whenever we say that we set @xmath541 , we shall mean that we define that @xmath542 for any @xmath144 and any @xmath543 . we set @xmath544 if @xmath545 we set @xmath274 if for every @xmath82 we have that @xmath546 ( see lemma [ lemp11 ] ) . if @xmath547 we set @xmath274 if for every @xmath267 we have that @xmath546 ( see lemma [ lemp11 ] ) . if @xmath548 we set @xmath274 if @xmath487 . if @xmath549 we set @xmath274 if @xmath486 . by lemmas [ lemp7 ] and [ lemp10 ] we see that @xmath493 is an @xmath43-consistent linear ordering of all the ordinary points of @xmath550 . let @xmath551 to be the set of all ordinary points of @xmath257 and let @xmath552 we extend the linear ordering @xmath58 of @xmath126 to an @xmath43-consistent linear ordering @xmath58 of @xmath492 as follows : 1 . if @xmath553 then we set @xmath554 if and only if @xmath263 . + 2 . if @xmath555 and @xmath265 we set @xmath554 if @xmath556 . + 3 . if @xmath557 and @xmath265 we set @xmath558 if @xmath559 . + 4 . if @xmath560 and @xmath561 then we set @xmath562 if @xmath563 . if @xmath560 and @xmath564 then we set @xmath562 if @xmath565 . using lemma [ lemp9 ] we may check that @xmath58 is an @xmath43-consistent linear ordering of @xmath566 . now we define the mapping @xmath567 from the set of all exceptional points of @xmath6 to the set @xmath568 of all sections of @xmath569 as follows : + let @xmath570 . + 1 . if @xmath571 then we set @xmath572 where @xmath573 2 . if @xmath574 then @xmath575 where @xmath576 3 if @xmath577 with @xmath502 , @xmath578 then @xmath575 where @xmath579 * step @xmath580 * let @xmath581 with @xmath582 and suppose that for any ordinal @xmath583 an @xmath43-consistent triple @xmath584 is constructed such that if @xmath287 then @xmath585 is an extension of @xmath586 . we consider two cases : + * case 1 . * the ordinal @xmath580 is a limit ordinal . + we set @xmath587 the ordering @xmath588 is defined by @xmath589 finally for every @xmath590 we set @xmath591 it is straightforward to check that the triple @xmath592 is an @xmath43-consistent triple . + * case 2 . * there exists an ordinal @xmath593 such that @xmath594 . + for every @xmath595 such that @xmath596 we consider the family @xmath597 which is the restriction of @xmath43 in the set @xmath598 . clearly , @xmath599 is a loc - lattice of subsets of @xmath598 . + we repeat the construction of * step 0 * for every pair @xmath600 , @xmath595 , and we obtain for every section @xmath536 of @xmath601 an @xmath599-consistent triple @xmath602 . let @xmath603 let @xmath588 to be the extension of @xmath604 to the set @xmath605 . the mapping @xmath606 is defined as follows : 1 . if @xmath607 then we set @xmath608 . 2 . if @xmath609 and @xmath590 then @xmath610 . let @xmath611 and @xmath612 be the corresponding @xmath599 -consistent triple . if @xmath613 we set @xmath614 with @xmath615 it is straightforward to check that @xmath592 is an @xmath43-consistent triple . clearly if @xmath616 we have that @xmath617 and the proof of part ( a ) of theorem [ thlocset ] is complete . the part ( b ) of theorem [ thlocset ] follows from part ( a ) and the lemma [ lemlocsep ] . in this final section we show that every lattice is isomorphic to a lattice @xmath43 of subsets of a set @xmath6 which is closed under finite intersections and separates the points of @xmath6 ( theorem [ threp ] ) . we this representation in order to prove theorem [ ath ] . let @xmath9 be a lattice . a _ filter _ of @xmath14 is a proper subset @xmath26 of @xmath20 such that : if @xmath29 and @xmath618 then @xmath30 , and if @xmath619 then @xmath620 . an _ ideal _ of @xmath14 is a proper subset @xmath621 of @xmath20 such that : if @xmath29 and @xmath622 then @xmath30 , and if @xmath619 then @xmath623 . a _ prime filter _ is a filter @xmath26 such that if @xmath27 and @xmath623 then @xmath29 or @xmath30 . let @xmath9 be a lattice and @xmath27 . an element @xmath624 is said to be an * internal element * of @xmath85 if @xmath21 and @xmath625 . by @xmath626 we shall denote the set of all internal elements of @xmath85 and by @xmath627 the set @xmath628 . [ def32 ] a filter @xmath26 of a lattice @xmath9 is called a * semi prime filter * if for every @xmath27 with @xmath623 we have that @xmath629 . a lattice @xmath9 is said to be 1 . * well separated * if for every semi prime filter @xmath26 of @xmath14 and every @xmath630 there exists @xmath30 such that @xmath631 ; 2 . * completely separated * if for any @xmath632 semi prime filters of @xmath14 with @xmath633 we have that @xmath634 and @xmath635 . a completely separated lattice is well separated . indeed , if @xmath26 is a semi prime filter and @xmath630 we select a semi prime filter @xmath636 with @xmath637 . since @xmath634 there exists an element @xmath638 . it is plain that @xmath639 and @xmath640 . the lattice @xmath44 with @xmath641 is an example of a well separated loc - lattice which is not completely separated . note also that if @xmath14 is completely separated loc - lattice then @xmath14 is isomorphic to a completely separated loc - lattice of sets , which by lemma [ lem02 ] is totally separated . we conclude that a loc - lattice @xmath14 is completely separated if and only if for any @xmath632 distinct semi prime filters of @xmath14 there exist @xmath29 and @xmath642 with @xmath643 . [ lem31 ] given a lattice @xmath644 , an ideal @xmath621 of @xmath14 and a filter @xmath26 of @xmath14 such that @xmath645 there exists a semi prime filter @xmath646 such that @xmath647 and @xmath648 . let @xmath649 the set of all filters of @xmath14 and @xmath650 let @xmath646 be a maximal element of @xmath651 . we shall show that @xmath646 is a semi prime filter . for every filter @xmath26 and @xmath624 we set @xmath652 clearly , @xmath653 is a filter , @xmath654 and @xmath655 . let @xmath27 such that @xmath656 and @xmath657 . since @xmath658 and by the maximality of @xmath646 we have that @xmath659 and @xmath660 . let @xmath661 such that @xmath662 . let @xmath663 . clearly @xmath31 , @xmath32 and @xmath664 . so , @xmath665 and therefore @xmath666 . let @xmath9 be a lattice and let @xmath667 be the set of all semi prime filters of @xmath14 . we call the mapping @xmath668 defined by @xmath669 the * stone map * of @xmath14 . the topology @xmath670 of @xmath667 which has as a basis the set @xmath671 is called the * stone topology * of @xmath667 . the space @xmath672 is a @xmath673 topological space called the * stone space * of @xmath14 . [ threp ] let @xmath674 be a lattice and @xmath675 the stone map of @xmath14 . then @xmath676 is an isomorphism onto @xmath677 . moreover @xmath678 is a set lattice closed under finite intersections , @xmath671 separates @xmath667 and for every @xmath27 we have that @xmath679 & f(a)\vee f(b)= f(a\vee b)=\bigcup\left\{f(x ) : x\in \bar{\mathbf{i}}(a , b)\right\}.&\;\;\;\;\;\;(2 ) \end{array}\ ] ] clearly , if @xmath27 and @xmath618 then @xmath680 . suppose that @xmath681 and @xmath631 . by lemma [ lem31 ] there exists an @xmath682 such that @xmath683 and @xmath684 , in particular @xmath29 and @xmath685 . but then @xmath686 , a contradiction . so , @xmath618 if and only if @xmath681 and @xmath676 is an isomorphism onto @xmath671 . let @xmath687 with @xmath633 . then either @xmath634 or @xmath688 . suppose that @xmath634 . clearly , the family @xmath671 separates @xmath667 . finally the equations ( 1 ) , ( 2 ) are obvious from that facts that for every semi - prime filter @xmath26 and every points @xmath85 of @xmath20 we have that @xmath689 if and only if @xmath29 and @xmath30 and that @xmath690 if and only if there exists @xmath691 such that @xmath692 . by duality , it is clear that every lattice @xmath9 can also be represented as lattice of of subsets of a set @xmath693 , closed under finite unions . the set @xmath693 will be the set of all semi prime filters of the dual lattice @xmath694 , where @xmath695 is the inverse order of @xmath696 . the elements of @xmath693 are called the * semi prime ideals * of @xmath14 . [ lemfinal ] let @xmath674 be a lattice and let @xmath675 be the stone map of @xmath14 . then @xmath697 is a loc - lattice of sets . moreover , if @xmath14 is well separated then the family @xmath671 well separates the set @xmath667 . by theorem [ threp ] the stone mapping @xmath676 is an isomorphism so @xmath697 is lattice of subsets of @xmath667 and satisfies the properties ( 1 ) , ( 3 ) and ( 4 ) of definition[def21 ] . we show that @xmath698 satisfies property ( 2 ) of definition[def21 ] . we note that @xmath699 if and only if @xmath700 . suppose that for @xmath701 , @xmath702 we have that @xmath65 . then @xmath703 which implies that @xmath704 . indeed , if @xmath705 there exists an internal element @xmath706 of @xmath85 . then @xmath189 must be incomparable and so by the property ( 2 ) of definition [ def11 ] we have that @xmath707 which implies that @xmath708 . but then @xmath709 and @xmath710 and by the property ( 3 ) of definition [ def11 ] we shall have that @xmath711 , a contradiction . so , @xmath712 suppose that the lattice @xmath14 is well separated and let @xmath713 and @xmath714 . let @xmath543 such that @xmath640 and @xmath30 . clearly @xmath715 and @xmath716 . therefore the family @xmath671 well separates @xmath667 . it is plain that theorem [ ath ] follows from theorem [ thlocset ] and lemma [ lemfinal ] . an _ orderable topological space _ is a topological space @xmath717 such that there exists a linear ordering @xmath58 of @xmath6 with the property that the open intervals of @xmath0 is a base for @xmath670 . weakly orderable topological space _ is a @xmath673 topological space @xmath717 such that there exists a linear ordering @xmath58 of @xmath6 with the property @xmath670 has a base consisting of convex sets . a _ generalized orderable space _ or a _ suborderable space _ ( @xcite ) is a hausdorff topological space @xmath717 such that there exists a linear ordering @xmath58 of @xmath6 with the property that @xmath670 has a basis of convex subsets of @xmath6 . it is known ( @xcite ) that the class of generalized ordered spaces coincides with the class of subspaces of linearly ordered topological spaces . a linear ordering of @xmath58 of @xmath6 is said a _ ( dedekind ) complete linear ordering _ if every nonempty subset of @xmath6 with an upper bound has a least upper bound ( supremum ) . complete orderable topological space _ is a topological space @xmath717 such that there exists a complete linear ordering @xmath58 of @xmath6 with the property that the open intervals of @xmath0 is a base for @xmath670 . an immediate consequence of theorems [ ath ] and [ thlocset ] is the following : 1 . every loc - lattice is isomorphic to a basis of a weakly orderable space . 2 . a @xmath673 topological space @xmath717 is weakly orderable if and only if has a basis which is a loc- lattice of subsets of @xmath6 . the problem of characterization of orderable spaces is considered by many authors . r. l. moore ( @xcite ) and a. wallace ( @xcite ) characterized the orderable continua , s. eilenberg ( @xcite ) characterized the connected orderable spaces , h. herrlich ( @xcite ) the countable and the totally disconnected metric orderable spaces . a characterization of general orderable and suborderable spaces is given by j. van dalen and e. wattel ( @xcite ) and e. dek ( @xcite ) ( see also @xcite , for another approach ) . we can obtain further characterizations of orderable and suborderable spaces using theorems [ ath ] and [ thlocset ] . before this we shall investigate which elements of a lattice can be representated as open intervals . an element @xmath718 of a lattice @xmath719 is said to be _ accessible from below _ if there exists a subset @xmath152 of @xmath20 such that @xmath720 and @xmath721 . we say that @xmath718 is _ inaccessible from below _ if it is not accessible from below . respectively we say that @xmath718 is _ accessible from above _ if there exists a subset @xmath152 of @xmath20 such that @xmath722 and @xmath721 and that @xmath718 is _ inaccessible from above _ if it is not accessible from above . [ lem41 ] let @xmath9 be a totally separated loc - lattice and let @xmath727 be an inaccessible from above element of @xmath20 . if @xmath668 is the stone map of @xmath14 and @xmath58 the @xmath671-consistent linear ordering of @xmath667 then @xmath728 is an open interval suppose that @xmath729 and that the set @xmath730 is not empty . we shall show that @xmath728 has a first element @xmath731 . suppose that @xmath728 has no a first element . . then @xmath733 . indeed , if there exists a @xmath734 then we select a @xmath735 such that @xmath736 and @xmath737 such that @xmath738 , @xmath739 and @xmath740 . then @xmath741 which contradicts the assumption that @xmath742 . this implies that @xmath743 . since @xmath676 is an embedding from @xmath719 onto @xmath744 we conclude that @xmath745 , which contradicts the assumption that @xmath718 is inaccessible from above . let @xmath9 be a totally separated loc - lattice such that every @xmath727 is inaccessible from above . if @xmath668 is the stone map of @xmath14 and @xmath58 the @xmath671-consistent linear ordering of @xmath667 then @xmath672 is a complete linear ordered topological space . by lemma [ lem41 ] for every @xmath727 we have that @xmath728 is an open interval . since @xmath9 is a totally separated loc - lattice , then the family @xmath671 totally separates the set @xmath667 and therefore @xmath671 is a basis for the topology @xmath749 generated by the open intervals of @xmath750 . so @xmath751 and so @xmath672 is an ordered space . it remains to show that @xmath0 has no gaps . suppose that @xmath752 is a gap of @xmath750 . it is easy to see that the set @xmath753 is a semi prime filter of @xmath14 and so an element of @xmath667 . we set @xmath754 . then either @xmath755 and so @xmath756 is the last element of @xmath757 or @xmath758 and so @xmath756 is the first element of @xmath759 . every case contradicts the assumption that @xmath536 is a gap . 1 . every totally separated loc - lattice such that every @xmath727 is inaccessible from above is isomorphic to a basis of a complete orderable space . 2 . a @xmath760 topological space @xmath717 is orderable if and only if has a basis @xmath43 such that @xmath44 is a loc- lattice of subsets of @xmath6 such that every @xmath47 is inaccessible from above . a family @xmath43 of sets is called * interlocking * ( see @xcite ) provided that every set @xmath47 which is an intersection of strictly larger members of @xmath43 has a representation as a union of strictly smaller members of @xmath43 . so , we may call a lattice @xmath9 to be an * interlocking lattice * if every element of @xmath20 either is inaccessible from above or it is accessible from above and below . in such lattices the sets @xmath728 , @xmath727 are either open intervals or unions of open intervals . therefore an interlocking and totally separated lattice has a representation as a basis of an ordered space . g. birkhoff and m. k. bennet _ the convexity lattice of a poset _ , order * 2 * ( 1985 ) , 223 - 242 . e. ech , _ topological spaces _ , wiley , new york ( 1966 ) . j. van dalen , e. wattel _ a topological characterization of ordered spaces _ , general topology and applications , * 3 * ( 1973 ) , 347 - 354 . e. dek _ theory and applications of directional structures _ , topics in topology ( proc . , keszthely , 1972 ) . janos bolyai , vol . 8 , north - holland , amsterdam , 1974 . , general topology and applications 187211 . b. a. davey and a. priestley , _ introduction to lattices and order _ , cambridge university press ( 1990 ) . s. eilenberg , _ ordered topological spaces _ , am . j. math . * 63 * ( 1986 ) 39 - 45 . v. felouzis _ on the characterization of totally orderable spaces and their subspaces _ , elefteria * 4b * ( 1986 ) 13 - 20 . h. herrlich _ ordnungsfhigkeit topologischer rume _ , inaugural dissertation , berlin ( 1962 ) . d. j. lutzer _ on generalized ordered spaces _ , dissertationes math . , rozprawy mat . * 89 * ( 1971 ) . r. l. moore , _ concerning simple continuous curves _ , trans . of ams . * 21 * ( 1920 ) 333 - 347 . a renyi , _ on random generating elements of a finite boolean algebra , _ acta sci ( szeged ) * 22 * ( 1961 ) 75 - 81 . a. r. schweitzer , _ a theory of geometrical relations , _ am . j. math . * 31 * ( 1909 ) , 365 - 410 . m. semenova and f. wehrung , _ sublattices of lattices of order - convex sets , i. the main representation theorem , _ j. algebra * 277 * ( 2004 ) , no . 2 , 825 - 860 . r. vaidyanathaswamy , _ set topology , _ chelsea ( 1960 ) . unabridged republication by dover publications , inc mineola , new york ( 1999 ) a. wallace , _ monotone transformations , _ duke math . j. * 9 * ( 1942 ) 487 - 506 .

we study the periodic orbits and the escapes in two different dynamical systems , namely ( 1 ) a classical system of two coupled oscillators , and ( 2 ) the manko - novikov metric ( 1992 ) which is a perturbation of the kerr metric ( a general relativistic system ) . we find their simple periodic orbits , their characteristics and their stability . then we find their ordered and chaotic domains . as the energy goes beyond the escape energy , most chaotic orbits escape . in the first case we consider escapes to infinity , while in the second case we emphasize escapes to the central `` bumpy '' black hole . when the energy reaches its escape value a particular family of periodic orbits reaches an infinite period and then the family disappears ( the orbit escapes ) . as this family approaches termination it undergoes an infinity of equal period and double period bifurcations at transitions from stability to instability and vice versa . the bifurcating families continue to exist beyond the escape energy . we study the forms of the phase space for various energies , and the statistics of the chaotic and escaping orbits . the proportion of these orbits increases abruptly as the energy goes beyond the escape energy .

1203.1010

the escapes from a dynamical system refer to a problem of basic interest for dynamics : the problem of chaotic scattering . this problem has attracted the attention of many authors in the previous decades ( some typical references are by @xcite ) . in particular we found @xcite the forms of the escape regions and the escape rates in the simple dynamical system @xmath0 for various values of the energy @xmath1 above the escape energy @xmath2 . we found also regions of orbits that never escape , or escape after a very long time . in particular the stable periodic orbits are surrounded by islands of stability that never escape . this problem is typical of escapes to infinity " . in fact , similar results were found in general polynomial hamiltonians representing two perturbed harmonic oscillators . in the present paper we explore the escapes in the classical system ( [ func : ham ] ) and a very different problem of escapes , using a system from the domain of general relativity . in particular the system we use belongs to the so - called manko - novikov ( mn ) metric family . @xcite found an exact vacuum solution of einstein s equations which describes a stationary , axisymmetric , and asymptotically flat spacetime with arbitrary mass - multipole moments . the mn metric subclass that we use can be considered as a perturbation of the kerr metric @xcite and it was introduced by @xcite . the kerr metric represents a rotating black hole of mass @xmath3 and spin @xmath4 . the mn perturbation is expressed by a parameter @xmath5 , which measures how much the mn quadrupole moment @xmath6 departs from the kerr quadrupole moment @xmath7 ( i.e. @xmath8 ) . while the kerr metric describes an integrable system , the mn metric describes a non - integrable system that allows chaos . the line element of the mn metric expressed in the weyl - papapetrou cylindrical coordinates @xmath9 is of the form @xmath10,\ ] ] where @xmath11 are considered as functions of the prolate spheroidal coordinates @xmath12 , while the coordinates @xmath13 can be expressed as functions of @xmath12 as well . thus @xmath14 and @xmath15 ^ 2+[(1+w)a+(1-w)b]^2,\label{fb } \\ c & = & ( v^2 - 1)(1+a~b)[(b - a)-w(a+b ) ] \nonumber \\ & & + ( 1-w^2)(b - a)[(1+a~b)+v(1-a~b ) ] , \\ \psi & = & \beta \frac{p_2}{r^3 } , \label{fc}\\ \gamma^\prime & = & \ln{\sqrt{\frac{v^2 - 1}{v^2-w^2}}}+\frac{3\beta^2}{2 r^6 } ( p_3 ^ 2-p_2 ^ 2 ) \nonumber \\ & + & \beta \left(-2+\displaystyle{\sum_{\ell=0}^2 } \frac{v - w+(-1)^{2-\ell}(v+w)}{r^{\ell+1}}p_\ell\right ) , \label{fgampr}\\ a & = & -\alpha \exp { \left[-2\beta\left(-1+\displaystyle{\sum_{\ell=0}^2 } \frac{(v - w)p_\ell}{r^{\ell+1}}\right)\right ] } , \label{fa}\\ b & = & \alpha \exp { \left[2\beta\left(1+\displaystyle{\sum_{\ell=0}^2 } \frac{(-1)^{3-\ell}(v+w)p_\ell}{r^{\ell+1}}\right)\right ] } , \label{fb}\\ r & = & \sqrt{v^2+w^2 - 1 } , \label{fr}\\ p_\ell & = & p_\ell ( \frac{v~w}{r } ) . \label{flega}\end{aligned}\ ] ] here @xmath16 is the legendre polynomial of order @xmath17 @xmath18 while the parameters @xmath19 are related to the mass @xmath3 , the spin @xmath4 , and the quadrupole deviation @xmath5 through the expressions @xmath20 while @xmath21 is the dimensionless spin parameter @xmath22 . these formulae give the kerr metric when @xmath23 . contrary to the system ( [ func : ham ] ) , the escapes in the mn system refer not only to infinity , but also to orbits that fall into the central compact object called `` bumpy black hole '' by @xcite . the mn central compact object is not really a black hole , because its horizon is broken along the equator and regions of closed timelike curves exist outside the horizon @xcite . the mn problem is very different from the simple hamiltonian ( [ func : ham ] ) ; however , the two problems have several common properties concerning the role of the periodic orbits and their relations to escapes . in the present paper we study in detail the periodic orbits in the systems ( [ func : ham ] ) and ( [ func : mnmetric ] ) and their connection with the escape phenomena . we found that the periodic orbits in general do not lead to escapes , except in very special cases . most of the families of periodic orbits generate an infinity of unstable , but non - escaping , periodic orbits . on the other hand , the asymptotic curves of the various unstable periodic orbits play an important role in generating chaos and most chaotic domains lead to escapes when the energy goes beyond the escape energy . the paper is organized as follows . in section [ sec : perorbham ] we calculate the periodic orbits of the system ( [ func : ham ] ) . then we find the asymptotic curves of the unstable periodic orbits and the corresponding chaos . in section [ sec : escham ] we find the escapes from the system ( [ func : ham ] ) and their statistics . in section [ sec : perorbmn ] we find the periodic orbits in the system ( [ func : mnmetric ] ) and the chaotic domains . in section [ sec : escmn ] we find the corresponding escapes and their statistics . finally , in section [ sec : concl ] we compare the two systems and draw our main conclusions . a ) simple periodic orbits ( of period 1 ) for energy @xmath24 . ( [ fig:01]b ) a surface of section @xmath25 for @xmath24 containing ordered and chaotic orbits . on the @xmath26 axis there are two stable periodic orbits ( @xmath27 ) and an unstable orbit ( @xmath28 ) ( at the intersection of the red and black curves ) , together with one stable ( red ) and one unstable ( thick black ) asymptotic manifolds . there are also two stable orbits ( a ) on the @xmath29 axis and triple and double period stable orbits that have bifurcated from the orbits ( @xmath30 ) and ( @xmath31 ) on the @xmath32=0 axis . , title="fig:",scaledwidth=51.0% ] a ) simple periodic orbits ( of period 1 ) for energy @xmath24 . ( [ fig:01]b ) a surface of section @xmath25 for @xmath24 containing ordered and chaotic orbits . on the @xmath26 axis there are two stable periodic orbits ( @xmath27 ) and an unstable orbit ( @xmath28 ) ( at the intersection of the red and black curves ) , together with one stable ( red ) and one unstable ( thick black ) asymptotic manifolds . there are also two stable orbits ( a ) on the @xmath29 axis and triple and double period stable orbits that have bifurcated from the orbits ( @xmath30 ) and ( @xmath31 ) on the @xmath32=0 axis . , title="fig:",scaledwidth=49.0% ] for small values of the energy the system ( [ func : ham ] ) has 4 periodic orbits of period 1 . there are two straight line periodic orbits @xmath33 that are stable ( orbits ( @xmath30 ) and ( @xmath31 ) in fig . [ fig:01]a ) . there is also a stable orbit ( @xmath27 ) , like an ellipse , described clockwise or counterclockwise , a simple unstable orbit crossing the @xmath34 perpendicularly ( orbit ( @xmath28 ) ) , and the orbit @xmath35 ( orbit ( d ) ) . in fig . [ fig:01]b we see also @xmath36 stable triple orbits @xmath37 and @xmath36 stable double orbits @xmath38 that have bifurcated from the orbits ( @xmath30 ) and ( @xmath31 ) . the orbits @xmath39 undergo an infinity of transitions to instability and stability as _ h _ tends to @xmath40 while the period of these orbits tends to infinity @xcite . the periodic orbits @xmath30 and @xmath31 terminate at @xmath41 when their period is infinite , in accordance with the strmgren termination principle @xcite . for @xmath42 the curve of zero velocity ( czv ) @xmath43 opens and the orbits @xmath30 and @xmath31 escape to infinity . the families ( @xmath30 ) and ( @xmath31 ) at every transition to stability or instability generate by bifurcation new families of equal or double period . families of higher order are also bifurcated from the families ( @xmath30 ) and ( @xmath31 ) . on a surface of section @xmath44 with @xmath45 the stable orbits are surrounded by islands of stability , while the unstable orbit is surrounded by a set of chaotic orbits ( fig . [ fig:01]b ) . in particular the orbits ( @xmath30 ) and ( @xmath31 ) appear on the axis @xmath29 of fig . [ fig:01]b . near the unstable orbit ( @xmath28 ) there is a chaotic region ( fig . [ fig:01]b ) . as @xmath1 increases the chaotic region around the unstable orbit ( c ) increases , while various sets of higher order periodic orbits bifurcate from the straight line orbits ( ( @xmath30 ) , ( @xmath31 ) ) ( fig . [ fig:01]b ) . , scaledwidth=70.0% ] a ) the hnon stability parameter _ a _ of the families of fig . [ fig:02 ] as a function of the energy _ h _ ; ( [ fig:03]b ) details of fig . [ fig:03]a . , title="fig:",scaledwidth=49.5% ] a ) the hnon stability parameter _ a _ of the families of fig . [ fig:02 ] as a function of the energy _ h _ ; ( [ fig:03]b ) details of fig . [ fig:03]a . , title="fig:",scaledwidth=50.5% ] the characteristics of the various families of periodic orbits are shown in fig . [ fig:02 ] . these characteristics give the value of _ x _ when @xmath35 and @xmath46 . we mark in particular the orbits @xmath37 , @xmath38 and @xmath47 bifurcating from the orbits ( @xmath30 ) and ( @xmath31 ) as the energy @xmath1 increases . the stability of these families is given in fig . [ fig:03 ] . the orbits are stable if their hnon parameter @xmath30 is between @xmath48 and @xmath49 . the family @xmath50 ( a ) is stable for @xmath51 , then unstable in the interval @xmath52 , stable again for @xmath53 , and has infinite more transitions to instability and stability until @xmath54 . at the first transition to instability @xmath55 the value of _ a _ goes below @xmath56 and one stable family of double period bifurcates there . its stability is represented by the red curve in fig . [ fig:03 ] , starting at @xmath57 . at the first transition from instability to stability along the family ( a ) @xmath58 _ a _ goes above @xmath56 and there is a bifurcation of an unstable family @xmath38 ( blue curve above @xmath57 ) . at the second transition to instability the value of _ a _ goes above @xmath57 @xmath59 and two different stable families of equal period bifurcate there ( black 1:1 ) . the orbits of these families deviate from the straight line @xmath60 . at the second transition to stability there are bifurcations of two unstable families ( magenta 1:1 ) , and so on . symmetric bifurcations are formed also around the family ( @xmath31 ) . there are also higher order bifurcations from the families ( @xmath30 ) and ( @xmath31 ) , e.g. the family @xmath37 ( figs . [ fig:02 ] and [ fig:03]a ) . a ) some periodic orbits for @xmath61 . four periodic orbits of period 1 , ( @xmath27 ) , ( @xmath28 ) , and the lyapunov orbits @xmath62 , @xmath63 , and a periodic orbit of period 2 , belonging to the family bifurcating from the family ( @xmath30 ) . ( [ fig:04]b ) an ( unstable ) periodic orbit ( @xmath27 ) for @xmath64 . , title="fig:",scaledwidth=51.0% ] a ) some periodic orbits for @xmath61 . four periodic orbits of period 1 , ( @xmath27 ) , ( @xmath28 ) , and the lyapunov orbits @xmath62 , @xmath63 , and a periodic orbit of period 2 , belonging to the family bifurcating from the family ( @xmath30 ) . ( [ fig:04]b ) an ( unstable ) periodic orbit ( @xmath27 ) for @xmath64 . , title="fig:",scaledwidth=49.0% ] the families bifurcating from the orbit ( a ) consist of orbits that reach the czv at two points . these families do not terminate for @xmath1 larger than @xmath65 , but extend all the way to @xmath66 . an example of an orbit of such a family @xmath67 is shown in fig . [ fig:04]a for @xmath61 . this family is bifurcated from the family ( a ) for @xmath68 and it is stable in the interval @xmath69 . at @xmath70 the family @xmath38 becomes unstable and generates a family with double period ( i.e. 4 times the original period ) . for @xmath71 the family @xmath38 crosses again the . @xmath56 axis ( fig . [ fig:03]a ) and becomes again stable generating an unstable double period family ( i.e. 4 times the original period ) . finally , at @xmath72 the family @xmath38 crosses the @xmath57 axis and becomes unstable for larger _ h_. at this crossing point it generates an equal period stable family ( 2 times the original period ) . this also becomes unstable for a slightly larger _ h _ , followed by a cascade of period doubling bifurcations , that generate an infinity of unstable families , that exist for arbitrarily large _ h_. when @xmath73 the czvs open above and below the axis @xmath74 ( fig . [ fig:04]a , b ) and several orbits escape to infinity . however , the periodic orbits of figs . [ fig:04]a , b do not escape . e.g. the orbit @xmath38 in fig . [ fig:04]a that bifurcated from the orbit ( a ) at @xmath68 does not escape , although the orbit ( a ) has escaped . in fig . [ fig:04]b we show a non - escaping orbit ( b ) for @xmath64 when the openings are very large . . the orbits in the red regions escape directly to infinity without any intersection with the axis @xmath35 . blue dot corresponds to the orbit ( c ) . , scaledwidth=70.0% ] at every opening of the czv there is an unstable orbit ( called lyapunov orbit ) crossing it ( fig . [ fig:04]a ) . any orbit , crossing a lyapunov orbit outwards , escapes to infinity @xcite . the regions of direct escape are shown in red in fig . [ fig:05 ] for @xmath75 . orbits starting in these regions escape directly , without any intersections with the axis @xmath35 . the boundaries of the direct escape regions are defined by the asymptotic curves of the lyapunov orbits . these asymptotic curves make infinite rotations around the escape regions @xcite . in that paper we described in detail the asymptotic curves from the lyapunov orbits . we found that these asymptotic curves include an infinity of branches that start and end with infinite rotations around the escape regions . the stable asymptotic curves of any other unstable periodic orbit leading to escapes approach the same limiting asymptotic curves , i.e. the boundaries of the red regions as in fig . [ fig:05 ] . for relatively small values of the energy @xmath1 , however , most of the chaotic orbits escape after one or more intersections with the axis @xmath35 ( fig . [ fig:05 ] ) . the asymptotic curves of the unstable orbit ( @xmath28 ) surround the red escape regions . examples of such asymptotic curves are shown in figs . [ fig:01]b and [ fig:05 ] . the unstable asymptotic curves start in opposite directions and the stable asymptotic curves start also in opposite directions . in figs . [ fig:01]b and [ fig:05 ] we give in black the unstable asymptotic curve that starts towards the right downwards and surrounds clockwise the lower red region . as it comes again close to the point ( @xmath28 ) it makes several oscillations , up and down close to the point ( @xmath28 ) . the stable asymptotic curve ( red ) starts to the left and downwards , surrounding counterclockwise the red escape region and intersects the unstable asymptotic curve at several homoclinic points . the oscillations of the stable and unstable asymptotic curves are much larger in fig . [ fig:05 ] @xmath76 than in fig . [ fig:01]b @xmath77 . but the main difference between figs . [ fig:01]b and [ fig:05 ] is that in the first case there are no escapes , while in the second case most of the chaotic orbits escape to infinity . the orbits that escape directly without any intersection with the axis @xmath35 , are marked in red . the stable asymptotic curve ( red curve ) of the unstable periodic orbit ( c ) ( blue dot in fig . [ fig:05 ] ) makes an infinity of rotations around the lower red region . this asymptotic curve never intersects the red regions . this happens because orbits starting on the stable asymptotic curve have further intersections along the same asymptotic curve , until they reach asymptotically the periodic orbit . therefore they can not escape immediately , hence they can not be found inside a red region . on the other hand , the unstable asymptotic curve ( black curve ) intersects the red region after a number of oscillations below and above the stable asymptotic curve ( fig . [ fig:05 ] ) . the stable and unstable asymptotic curves intersect at an infinity of homoclinic points . a few points @xmath78 are marked in fig . [ fig:05 ] . the corresponding homoclinic orbits approach asymptotically the periodic orbit ( c ) as @xmath79 and @xmath80 . the unstable asymptotic curve ( black ) makes outer and inner loops above and below the stable asymptotic curve . the areas of these lobes are equal . the homoclinic intersections approach closer and closer the periodic orbit ( c ) , their distances decreasing proportionally to @xmath81 , where @xmath82 is the larger ( positive ) eigenvalue of the orbit ( c ) . thus the lengths of the lobes become longer proportionally to @xmath82 . the value of @xmath82 increases as the energy increases . the inner lower lobes make some rotations , back and forth , around the lower red escape region . the inner loop starting downwards from the homoclinic point 5 makes @xmath83 rotation around the red region and returns closer than the point 5 to the periodic orbit ( c ) . the next inner loop makes about @xmath84 rotations around the red region and the subsequent loop intersects the red region ( fig . [ fig:05 ] ) . then the orbits passing through points of the black curve inside the red region , escape to infinity without any further intersections with the surface of section . as a point along the black curve approaches the red region its next image on the surface of section makes an infinity of rotations around the upper red region of fig . [ fig:05 ] approaching asymptotically the boundary of the red region . this phenomenon is described in detail in the paper of @xcite . for values of @xmath1 up to about @xmath85 there are some stable families , generated from the stable family ( @xmath27 ) , of orbits starting on the @xmath35 axis . this family consists of almost elliptical orbits that are described either clockwise or counterclockwise ( figs . [ fig:01]a , b ) . the family ( @xmath27 ) is stable up to about @xmath86 and then it becomes unstable at a period doubling bifurcation ( @xmath56 ) . the double period families bifurcating from ( @xmath27 ) start at @xmath87 and become unstable at another period doubling bifurcation ( 4 times the original period ) for about @xmath88 . thus the interval between the first and the second period doubling bifurcations is @xmath89 . this is followed by a cascade of period doubling bifurcations that produce an infinity of unstable families , existing for arbitrarily large @xmath1 . it is known that the intervals between successive period doubling bifurcations in conservative systems decrease by a universal ( asymptotic ) factor @xmath90 @xcite . therefore , the total interval @xmath91 that contains higher order bifurcation of the family ( b ) is approximately @xmath92 , and beyond @xmath93 there is no stable periodic orbit generated by bifurcations from the family ( b ) . ) on the surface of section for @xmath94 , when the periodic orbit has just become unstable , and two stable orbits that bifurcated from it . the asymptotic curves from the orbit ( @xmath27 ) surround both islands of stability , but they do not extend to larger distances.,scaledwidth=60.0% ] when the orbit ( @xmath27 ) has just become unstable there are two islands of stability around it . for @xmath94 ( fig . [ fig:06 ] ) the asymptotic curves of the orbit ( @xmath27 ) surround the two islands , but they do not extend to large distances . in fact , for this value of _ h _ there are invariant curves surrounding the orbit ( @xmath27 ) and the islands around it . however , for still larger _ h _ the asymptotic curves of ( @xmath27 ) extend very far ( fig . [ fig:07]a ) surrounding the whole available area on the surface of section ( fig . [ fig:07]b ) . the invariant curves surrounding the point ( b ) for @xmath94 , have been destroyed for @xmath95 and the asymptotic curves come close to the limiting curve @xmath96 . extend well beyond the areas around the two islands ( fig . [ fig:07]a ) and go all around the available area on the surface of section ( fig . [ fig:07]b ) . , title="fig:",scaledwidth=51.0% ] extend well beyond the areas around the two islands ( fig . [ fig:07]a ) and go all around the available area on the surface of section ( fig . [ fig:07]b ) . , title="fig:",scaledwidth=49.0% ] the same pattern of transition to instability is followed by most stable periodic orbits . namely most stable orbits become unstable at a period doubling bifurcation followed by a cascade of infinite period doublings that lead to an infinity of unstable periodic orbits . an exception are the families ( @xmath30 ) ( @xmath31 ) , which have a non - universal bifurcation ratio @xmath97 and lead to escapes @xcite . a theoretical explanation of this particular bifurcation ratio was provided by @xcite . a review of previous theoretical and numerical work on bifurcations is provided by @xcite . as the energy @xmath1 increases the proportion of orbits , that escape directly through the openings of the the czv increases . in fig . [ fig:05 ] @xmath76 the upper red region represents orbits that escape through the upper opening of the czv , and the lower red region represents orbits that escape through the lower opening of the czv , without ever intersecting the axis @xmath35 . however , most of the chaotic orbits , represented by scattered dots in fig . [ fig:05 ] , escape from the system after a small or large number of intersections with the @xmath35 axis . . , scaledwidth=70.0% ] as @xmath1 increases to @xmath98 the red regions increase in relative size ( proportion of the total area ) . in fig . [ fig:08 ] are shown the direct escape regions for @xmath98 and the asymptotic curves ( black , unstable and red , stable ) , from the periodic orbit ( @xmath28 ) , in the lower part of the figure . the regions in fig . [ fig:08 ] are colored according to the number of intersections an orbit has with the axis @xmath35 before the orbit escapes . the stable asymptotic curve ( red ) again does not intersect the red region , but the ( black ) unstable asymptotic curve intersects the red region on the left after half a rotation around it . then it intersects it again ( after an inner and an outer loop , below and above the stable asymptotic curve ) on the right side of the red region and so on . the orbits starting at points of the black curve inside the red region escape again without any further intersections with the axis @xmath35 . a ) escape regions for @xmath99 either directly ( red ) , or after a number of intersections of the @xmath35 axis upwards @xmath100 , as indicated in the color scale . the white region represents an island of stability . we mark also an unstable asymptotic curve ( @xmath101 ) from the periodic orbit ( @xmath28 ) . ( [ fig:09]b ) the same , as in fig . [ fig:09]a , for @xmath102 , together with an initial arc of the stable asymptotic curve of the orbit ( @xmath27 ) on the left ( red ) . , title="fig:",scaledwidth=49.5% ] a ) escape regions for @xmath99 either directly ( red ) , or after a number of intersections of the @xmath35 axis upwards @xmath100 , as indicated in the color scale . the white region represents an island of stability . we mark also an unstable asymptotic curve ( @xmath101 ) from the periodic orbit ( @xmath28 ) . ( [ fig:09]b ) the same , as in fig . [ fig:09]a , for @xmath102 , together with an initial arc of the stable asymptotic curve of the orbit ( @xmath27 ) on the left ( red ) . , title="fig:",scaledwidth=50.5% ] for larger _ h _ ( @xmath103 ) ( fig . [ fig:09]a ) and @xmath102 ( fig . [ fig:09]b ) the two regions of direct escapes increase in proportion to the total available space . in fig . [ fig:09]b we have marked one unstable asymptotic curve of the unstable periodic orbit ( c ) ( black ) , and one stable asymptotic curve of the unstable periodic orbit ( b ) on the left ( red ) . the two curves are almost tangent at a point . we have checked that for @xmath104 the two asymptotic curves do not intersect and for @xmath105 they clearly intersect close to this point . thus , for a value of @xmath1 a little smaller than @xmath102 the two curves are tangent . then according to the newhouse theorem @xcite close to the tangency point there is a stable periodic orbit , generated at a tangent bifurcation . in figs . [ fig:09]a , b we see that the periodic orbits ( @xmath27 ) and ( @xmath28 ) are in a region where escapes occur after several intersections of the axis @xmath35 by the orbits . in fact , the periodic orbits never escape , and orbits close to them escape after a time that increases to infinity as the initial conditions approach the periodic orbits . in particular the stable asymptotic curves do not intersect the red regions of direct escapes . in fact , all the orbits starting on the stable asymptotic curves never escape . on the other hand , the unstable asymptotic curve @xmath101 of ( @xmath28 ) in figs . [ fig:09]a , b enters the red region at a rather small distance from the orbit ( @xmath28 ) , and the orbits starting inside the red region escape directly . however , this unstable asymptotic curve comes out of the red region several times , hence the corresponding orbits escape in general after some intersections with the axis @xmath106=0 . the fact that there are chaotic orbits that escape after long times is related to the phenomenon of stickiness along the unstable asymptotic curves @xcite . as @xmath1 increases further new tangencies between the unstable and the stable asymptotic curves of the unstable orbits are formed , thus new stable periodic orbits appear . however , the islands around these periodic orbits are extremely small , since almost @xmath107 of the orbits escape for large @xmath1 . this situation is similar to the case where no escapes exist at all , but tangencies between the asymptotic curves generate new stable periodic orbits @xcite . a ) the proportion of chaotic ( including escaping ) orbits @xmath108 as a function of the energy @xmath1 . ( [ fig:10]b ) the proportion of the directly escaping orbits @xmath109 ( without any intersection with the axis @xmath35 ) as a function of the energy h.,title="fig:",scaledwidth=49.5% ] a ) the proportion of chaotic ( including escaping ) orbits @xmath108 as a function of the energy @xmath1 . ( [ fig:10]b ) the proportion of the directly escaping orbits @xmath109 ( without any intersection with the axis @xmath35 ) as a function of the energy h.,title="fig:",scaledwidth=50.5% ] a ) the logarithmic proportions @xmath110 of escapes for @xmath111 and @xmath112 as functions of the number of intersections @xmath113 . ( [ fig:11]b ) the probability @xmath114 of escapes as a function of @xmath113 for various values of the energy @xmath1.,title="fig:",scaledwidth=49.5% ] a ) the logarithmic proportions @xmath110 of escapes for @xmath111 and @xmath112 as functions of the number of intersections @xmath113 . ( [ fig:11]b ) the probability @xmath114 of escapes as a function of @xmath113 for various values of the energy @xmath1.,title="fig:",scaledwidth=50.5% ] we proceed now to a statistical analysis of the escapes . this is done in four ways : 1 . first we separate the orbits into ordered and chaotic @xmath115 escaping . in fig . [ fig:10]a we give the proportion of the chaotic orbits @xmath116 ( including the escaping orbits ) as a function of the energy @xmath1 . namely we populate the whole circle inside the limiting curve @xmath117 by initial conditions of @xmath118 orbits ( where @xmath119 ) and we find the total numbers of chaotic @xmath115 escaping orbits @xmath120 . then the number of ordered orbits is @xmath121 . when @xmath1 is very small the proportion of ordered orbits is almost @xmath107 . as @xmath1 increases the proportion of chaotic orbits increases and for @xmath122 this proportion is about @xmath123 ( fig . [ fig:10]a ) . when @xmath1 goes beyond the escape perturbation @xmath122 the proportion of chaotic orbits increases , but most chaotic orbits escape to infinity . only a very small proportion of chaotic orbits near higher order unstable periodic orbits inside islands of stability do not escape ( like the orbits near the unstable periodic orbit ( b ) of fig . [ fig:06 ] ) . + the proportion of chaotic orbits ( including the escaping orbits ) can be given by the approximate formula ( fig . [ fig:10]a ) @xmath124}.\ ] ] + for small @xmath1 this proportion tends to zero , while for large @xmath1 it tends to 1 . we calculate the proportion of direct escapes ( fig . [ fig:10]b ) , @xmath125 ( where @xmath126 is the number of orbits that escape without any intersection with the axis @xmath35 ) as a function of the energy @xmath1 . the direct escape regions are red in figs . [ fig:05 ] , [ fig:08 ] , [ fig:09]a , b . this proportion is small for @xmath1 slightly larger than @xmath127 and increases fast with @xmath1 , reaching more than @xmath128 for @xmath129 . 3 . then we study the escape time . the time is represented by the number of intersections @xmath113 before escape . most escapes take a long time for @xmath1 slightly above @xmath40 and take place faster as @xmath1 increases . in fig . [ fig:11]a we give the logarithm of the proportion of escaping orbits @xmath130 where @xmath131 is the number of escapes after the @xmath113th intersection ( upwards , @xmath46 ) and before the @xmath132 intersection ( i.e. there is no @xmath132 intersection ) . we see that the escape rates are decreasing with @xmath113 . the escape rates @xmath130 are large for small @xmath113 and decrease considerably for larger @xmath113 . for @xmath133 the escapes take place quite fast . finally , we calculate the probability of escape as a function of time represented by the number of intersections @xmath113 for various values of the energy . namely the probability @xmath134 is the ratio @xmath135 where @xmath136 is the number of orbits that have not yet escaped before the @xmath113th intersection @xmath137 this probability was calculated for the system ( [ func : ham ] ) and other similar systems in previous papers @xcite . it was found that @xmath134 tends to a constant value @xmath138 , independent of the initial conditions of non - ordered orbits . for large times ( large @xmath113 ) the probability of escape tends to zero if the energy is smaller than a critical value @xmath139 , larger than the escape energy @xmath40 . however , if @xmath1 is larger than @xmath139 , the probability @xmath134 for large @xmath113 , tends to a constant value @xmath138 larger than zero . this quantity depends only on the energy @xmath1 . in our present notation the limiting probability @xmath138 is proportional to a power of the quantity @xmath140 , i.e. @xmath141 where @xmath139 is the critical value of @xmath1 , and the exponent @xmath142 is approximately @xmath143 . this value of @xmath142 was found to be the same for different dynamical systems @xcite . [ fig:11]b we give the value of @xmath134 as a function of @xmath113 for various values of @xmath1 . when @xmath144 the probability @xmath134 tends to @xmath145 . on the other hand when @xmath146 the probabilities tend to @xmath147 . these values can be approximated by the formula @xmath148 where @xmath149 . this formula is consistent with the universal formula found by @xcite . + the statistics of escapes in various dynamical systems has been a subject of considerable interest in recent years . the main question is whether the various chaotic phenomena ( like poincar recurrences , correlations and escapes ) decay exponentially in time or according to power laws ( e.g. * ? ? ? * ; * ? ? ? the power laws are supposed to be connected to the existence of stickiness in mixed dynamical systems that contain islands of stability . however a detailed study of the effects of stickiness on the escapes should be the study of a future research . the manko - novikov ( mn ) spacetime depends on a parameter @xmath5 that measures the quadrupole moment deviation of this metric from the kerr metric with the same mass @xmath3 and spin @xmath4 . there have been proposals and attempts to constrain such deviations from observational data ( see e.g. @xcite and references therein ) . the final stages before the plunge ( escape ) in a non - kerr spacetime like the mn have certain astronomical interest ( see @xcite ) . from this point of view our work aims to exploit the non - integrability of the mn metric in order to give a detailed example of a such a final stage . the geodesic orbits of a test particle of mass @xmath150 are described as equations of motion of the following lagrangian @xmath151 the mn metric has two integrals of motion @xcite , namely the energy ( per unit mass ) @xmath152 and the z - component of the angular momentum ( per unit mass ) @xmath153 where the dots mean derivatives with respect to the proper time . the kerr metric has one more integral of motion , the so - called carter constant @xcite , thus it is an integrable system . however , the mn model is non - integrable and it allows the appearance of chaos . the motion on a meridian axis @xmath154 in the mn system satisfies the relation @xmath155 where the effective potential @xmath156 depends on @xmath5 , @xmath157 and @xmath158 . thus the motion takes place inside the czv @xmath159 = 0.\ ] ] studies of the orbits in the mn metric were made by @xcite . in the present paper we study on detail the periodic orbits , chaos and escapes in this system . in the following we use the values @xmath160 , @xmath161 and @xmath162 for the spin . we find the orbits for a fixed value of the z - angular momentum @xmath163 , and various values of @xmath157 . , @xmath164 and @xmath165 and two periodic orbits inside the inner czv @xmath166 ( stable ) and @xmath165 ( unstable ) . there are also 5 czvs that reach the axis @xmath167 . the line segment @xmath168 on the axis @xmath167 represents the horizon of the bumpy black hole . the horizon is broken at the point @xmath169 . , scaledwidth=50.0% ] the form of the czv in the case @xmath170 is given in fig . [ figorpeinoutczv ] . it consists of two main closed curves and 5 more curves of small extent reaching the axis @xmath171 . the line segment @xmath172 ( @xmath173 for @xmath174 ) , along the axis @xmath175 , represents the horizon and any orbit reaching this axis escapes into the bumpy black hole . all orbits inside the 5 curves that are close to the horizon escape into the bumpy black hole . on the other hand , the orbits starting inside the main closed czvs remain inside these curves for ever . in particular there are two simple periodic orbits inside the outer czv , and two more simple periodic orbits inside the inner czv . the periodic orbits @xmath176 ( outer ) and @xmath166 ( inner ) intersect perpendicularly the @xmath177 - axis while the other two periodic orbits are straight lines along the @xmath165 axis . furthermore , we mark a stable resonant orbit @xmath164 . the orbits @xmath178 and the outer @xmath165 orbit are stable , while the inner @xmath165 is unstable . @xmath179 for @xmath170 . ( [ figpsinout]a ) islands and chaos in the inner region . ( [ figpsinout]b ) invariant curves and islands in the outer region . the unstable periodic orbits are marked with big dots . ] if we take a surface of section @xmath180 the intersections @xmath181 of most orbits inside the outer czv are along closed invariant curves around the point @xmath176 that represents the periodic orbit @xmath176 ( fig . [ figpsinout]b ) . however , there are also islands of stability around the stable ( resonant ) periodic orbits , and some chaos between these islands , around the unstable periodic orbits ( fig . [ figpsinout]b ) . for @xmath182 . the boundary represents the orbit @xmath165 , which is stable in this case . , scaledwidth=50.0% ] on the other hand , most orbits in the inner region ( inside the inner czv ) are chaotic ( fig . [ figpsinout]a ) although there is a large island of stability and three small islands of type @xmath164 . the chaotic character of these orbits is related to the instability of the orbit @xmath165 , which is represented by the boundary of the inner region on the surface of section . however , for a large range of values of @xmath157 smaller than @xmath183 the orbit @xmath165 is stable and most orbits around @xmath166 are ordered ( e.g. for @xmath184 , fig . [ figpse05 ] ) . a ) the two main czv are joined into one for @xmath185 . the point ` x ` represents an unstable orbit that is generated at the saddle point when the two curves join . ( [ fige096]b ) the corresponding surface of section @xmath180 . in this case the boundary , which represents the orbit @xmath165 , is unstable . chaos is dominant in the inner region and close to the boundary . ] if @xmath157 increases beyond @xmath186 the inner and outer czvs are joined into a common czv ( fig . [ fige096]a for @xmath185 ) . when the two curves join , an unstable periodic orbit is formed , which intersects perpendicularly the @xmath165 axis ( point x in fig . [ fige096]a ) and exists for larger values of e. at the same time the two periodic orbits @xmath165 join into one unstable orbit . the inner and outer regions are also joined on the surface of section @xmath187 ( fig . [ fige096]b ) . the periodic orbit @xmath165 is represented now by the common boundary of the two regions and close to it there is some chaos extending all the way around the orbit @xmath176 . chaos is also dominant inside ( i.e. on the left of ) the orbit x ( fig . [ fige096]b ) . and the boundaries of the main permissible regions ( @xmath177 is given as a function of the energy e ) . , scaledwidth=50.0% ] figure [ figepmfle ] gives the characteristics of the periodic orbits @xmath176 and @xmath32 and the boundaries of the permitted motions for @xmath188 . the inner boundary is at an almost constant @xmath189 . the outer boundary increases considerably with increasing @xmath157 and tends to infinity as @xmath190 . when @xmath191 the orbits escape to infinity . when @xmath192 the permissible region splits into two and we have a middle outer boundary ( mob ) ( fig . [ figepmfle ] ) which is the inner boundary of the outer region , and a middle inner boundary ( mib ) which is the outer boundary of the inner region . as @xmath157 decreases the outer region shrinks , and disappears for @xmath193 . then the orbit @xmath176 also disappears after becoming just one point . then the orbit in 3 dimensions is a circle on the equatorial axis around the center @xmath175 . for smaller @xmath157 only the inner closed czv persists . the position of the orbit @xmath166 is close to the inner boundary , and can not be distinguished from the inner boundary in the scale of fig . [ figepmfle ] . in fact , the distance of this orbit from the inner boundary decreases , as @xmath157 decreases , and goes to zero at about @xmath194 ( fig . [ figinrech]b ) . for the same value of @xmath157 the outer boundary of the inner region ( mib ) reaches the inner boundary . then the orbit @xmath166 in 3 dimensions is circular on the equatorial plane . a ) periodic orbits @xmath166 and @xmath195 for @xmath196 . the family @xmath195 was generated at the transition of the orbit @xmath165 from stability to instability as @xmath157 increases . ( [ figinrech]b ) the characteristics of the families @xmath197 and the boundaries of the ( inner ) permissible region . ] as we have mentioned above the orbit @xmath165 of the inner region becomes stable when @xmath157 decreases below @xmath198 . as @xmath157 increases above this critical value , a stable family @xmath195 that crosses perpendicularly the @xmath165 axis bifurcates from @xmath165 ( fig . [ figinrech]a , b ) . as @xmath157 increases further this family becomes unstable , generating by bifurcation higher order periodic orbits . e.g. for @xmath170 one can see 3 small islands around this unstable family ( on the right hand side of fig . [ figpsinout]a ) . and of some resonant bifurcations from the family @xmath199 for relatively small values of the energy e. the dotted region indicates chaos . the dashed line gives the position of the first homoclinic point of the orbit @xmath200 , and the gray region indicates escaping orbits , while the red region gives the initial conditions ( for @xmath165 ) of orbits that escape directly . , scaledwidth=70.0% ] when e decreases below @xmath194 the inner czv increases again ( left side of fig . [ figinrech]b ) , and the orbit @xmath166 deviates from the inner boundary . but , for @xmath201 the distance of @xmath166 from the inner boundary decreases again ( fig . [ figepmfs ] ) , although it does not tend to zero . the most strange evolution occurs when e becomes smaller than @xmath202 . then a couple of simple periodic orbits crossing perpendicularly the @xmath165 axis is formed out of nothing . one orbit @xmath203 is stable , and the other @xmath204 unstable ( fig . [ figepmfs ] ) . this is called a `` tangent bifurcation '' @xcite . a ) a surface of section @xmath180 for @xmath205 ( just below the value of @xmath202 when the orbits @xmath206 are formed at a tangent bifurcation ) . invariant curves surround the orbits @xmath207 and are also near the boundary . a little chaos appear near the unstable orbit @xmath200 and its asymptotic curves . ( [ figsstb]b ) a surface of section for @xmath208 . it contains several islands around periodic orbits that have bifurcated from @xmath209 , like @xmath210 and @xmath211 . several chaotic regions appear between the islands , but they are separated from each other by invariant curves around @xmath209 . for this value of e the orbit @xmath209 is unstable , while the boundary ( representing the orbit @xmath165 ) is stable . ] from the unstable point @xmath200 start two stable and two unstable asymptotic manifolds ( fig . [ figsstb]a ) . the inner stable and unstable asymptotic manifolds surround the orbit @xmath209 and intersect at an infinite number of homoclinic points . the first homoclinic point is on the axis @xmath212 , on the left of the orbit @xmath209 . the other couple of stable and unstable asymptotic curves approach the boundary and surround the orbit @xmath166 on the left of the figure . near the unstable orbit @xmath200 and its homoclinic points there is some chaos ( figs . [ figepmfs ] and [ figsstb]a ) . as @xmath157 decreases several periodic orbits bifurcate from @xmath209 and near all the unstable orbits there are chaotic regions . these chaotic regions increase as @xmath157 decreases . two examples of bifurcating families , namely @xmath164 and @xmath213 are shown in fig . [ figepmfs ] . the orbit @xmath209 remains stable from its generation at @xmath202 down to @xmath214 . at @xmath214 the orbit @xmath209 becomes unstable and a double period @xmath83 stable orbit bifurcates there towards smaller values of @xmath157 . in fig . [ figsstb]b this orbit has receded from @xmath209 and two islands of stability are formed , one around each point . the orbits of the islands go alternatively from one island to the other . the orbit @xmath209 remains unstable for a small interval @xmath215 and then it becomes again stable , generating an unstable periodic orbit @xmath83 . as e comes close to @xmath216 the family @xmath209 becomes again unstable for a small interval @xmath215 . at the transition to instability it generates two different stable periodic orbits of equal period @xmath211 . in fig . [ figsstb]b the orbit @xmath209 is unstable and the two islands around it refer to different orbits . for a little smaller @xmath157 the orbit @xmath209 becomes again stable and generates two unstable periodic orbits @xmath211 . as @xmath157 decreases further there is an infinity of transitions to instability and stability of orbits of double period @xmath83 and of equal period @xmath211 . at the limit of this infinite sequence of transitions the orbit @xmath209 reaches the escape region ( fig . [ figepmfs ] ) and does not exist for smaller values of @xmath157 . this phenomenon of infinite transitions to instability and stability along the same family was found also in other dynamical systems , e.g. the system given by eq . ( [ func : ham ] ) of the present paper @xcite . a ) a surface of section @xmath180 for @xmath217 . in this case most of the chaotic regions have joined . however , there are also islands , like @xmath218 . ( [ figsses]b ) a surface of section for @xmath219 below the escape value of @xmath220 where escapes start to appear . the chaotic region of fig . [ figsses]a is now a region of escapes , but we see also the same islands . ] the chaotic regions of fig . [ figsstb]b are separated from each other by invariant curves closing around @xmath209 . for a little smaller @xmath157 most of these invariant curves are destroyed , as @xmath157 decreases , and a large chaotic region is formed ( figs . [ figepmfs ] and [ figsses]a ) for @xmath217 . this large chaotic region contains several islands of stability , like @xmath221 etc ( fig . [ figsses]a ) , that recede from @xmath209 as @xmath157 decreases ( see the characteristics of the corresponding periodic orbits in fig . [ figepmfs ] ) . in the mn system ( including the kerr system ) there are two types of escapes of orbits : ( a ) escapes to infinity when @xmath222 , and ( b ) escapes to the central bumpy black hole for small values of e. the first type of escapes ( escapes to infinity ) is similar to the escapes in the keplerian two - body problem ( escapes along hyperbolic orbits ) . the permissible region increases in size as the energy increases . this size tends to infinity as the energy approaches the energy of the parabolic motions . on the other hand , the escapes to the bumpy black hole occur when the czvs open inwards as the energy decreases and allow motions towards the bumpy black hole . a ) downwards @xmath223 and fig . ( [ figescor]b ) upwards @xmath224 . ] in the mn case , that we study here , when @xmath219 the czv of the inner region ( the outer region does not exist ) joins two czvs , above and below it and orbits from the inner region are allowed to escape to the bumpy black hole @xmath225 . the aforementioned czvs above and below the central region are the second and fourth leaf - like czvs respectively ( counting clockwise ) in fig . [ figorpeinoutczv ] from the set of five czvs touching the broken horizon of the bumpy black hole @xmath226 . these two czvs expand outwards on both sides of the axis @xmath227 , while @xmath157 decreases , until ( for @xmath228 ) they reach the expanding central inner region ( fig . [ figepmfs ] ) , which till then contains only bounded orbits . after the connection , orbits of the inner region may escape downwards ( fig . [ figescor]a ) or upwards ( fig . [ figescor]b ) to the central bumpy black hole . in the mn case the escaping orbits are chaotic orbits that undergo chaotic scattering . in fact , if we compare fig . [ figsses]a @xmath229 where no escapes are permitted , and fig . [ figsses]b where we have many escapes we see a great similarity . that is the large chaotic domain around @xmath209 of fig . [ figsses]a is transformed into an escape domain in fig . [ figsses]b . we see several islands in fig . [ figsses]b that are slightly changed from the islands of fig . [ figsses]a ( the main change is that the islands @xmath211 in fig . [ figsses]a are surrounded by 3 secondary islands in fig . [ figsses]b ) . this similarity explains why the chaotic domain of fig . [ figepmfs ] ( around @xmath209 ) is transformed into an escape domain for @xmath230 . in particular the periodic orbit @xmath209 of fig . [ figsses]a reaches the point where the main czv of the inner region joins the upper and lower czvs and for smaller @xmath157 this orbit escapes to the bumpy black hole , i.e. the periodic orbit does not exist any more . as a function of the energy @xmath157 . , scaledwidth=50.0% ] this case is very similar to the cases @xmath231 ( orbits @xmath232 and @xmath233 ) of the system ( [ func : ham ] ) . there are two more similarities between the two cases . first is the sequence of infinite transitions to instability and stability of the orbits @xmath234 and the orbits @xmath232 and @xmath233 . the second similarity is in the periods of the orbit @xmath209 and of the orbits @xmath232 and @xmath233 , which both tend to infinity as the orbits approach the termination point . the period of the orbits @xmath209 is given in fig . [ figenpein ] , which increases considerably as @xmath157 tends to @xmath219 and presumably it tends to infinity . the period of the orbits @xmath231 was also found to tend to infinity @xcite . this behavior is consistent with the strmgren termination principle of the families of periodic orbits @xcite . according to this principle the families either join other families , or terminate when their size , or their energy , or their period becomes infinite . a ) the proportion of the chaotic orbits ( including the escape orbits ) , with initial conditions along the axis @xmath32 , as a function of the energy @xmath157 . ( [ figcharep]b ) the proportion @xmath235 of direct escapes as a function of @xmath157 . , title="fig:",scaledwidth=50.0%]a ) the proportion of the chaotic orbits ( including the escape orbits ) , with initial conditions along the axis @xmath32 , as a function of the energy @xmath157 . ( [ figcharep]b ) the proportion @xmath235 of direct escapes as a function of @xmath157 . , title="fig:",scaledwidth=50.0% ] the proportion of chaotic @xmath115 escaping orbits @xmath108 increases as the energy decreases ( fig . [ figcharep]a ) . we calculate the proportion @xmath108 by taking initial conditions of orbits along the @xmath32-axis . this proportion tends to zero for relatively large @xmath157 but it increases abruptly as @xmath157 goes beyond @xmath236 ( to smaller @xmath157 ) . an approximate formula for this proportion is @xmath237\ ] ] and can be applied in the interval @xmath238 ( fig . the curve ( [ func : approx1 ] ) is very similar to the corresponding curve of the hamiltonian ( [ func : ham ] ) ( fig . [ fig:10]a ) . however , for @xmath239 the proportion of escaping orbits increases further and tends to @xmath107 . in fig . [ figcharep]a two more abrupt increases can be seen . the first of these two increases takes place near @xmath240 , and around it one can use an approximate formula @xmath241 . the second increase takes place near @xmath242 and the corresponding approximate formula is @xmath243 . this formula gives values near @xmath244 for @xmath2450.180 . when @xmath246 most chaotic orbits escape to the bumpy black hole . however , only a proportion @xmath109 escape directly from the system without any further intersection with the axis @xmath165 . this proportion increases as the energy @xmath157 decreases ( fig . [ figcharep]b ) . the curve @xmath109 of fig . [ figcharep]b is very similar to the corresponding curve of fig . [ fig:10]b . in fig . [ figepmfs ] we mark in red the initial conditions of the directly escaping orbits . most of the other chaotic orbits escape after one or more intersections beyond the original point , with the @xmath165 axis . but , as we see in fig . [ figepmfs ] , for @xmath246 there are also orbits that do not escape . in particular the periodic orbit @xmath166 and orbits close to it do not escape for much smaller values of @xmath157 . similarly the orbits that bifurcated from @xmath209 , like the orbits @xmath247 ( fig . [ figepmfs ] ) exist also for @xmath246 . however , all these families become unstable for small @xmath157 . in any case the existence of an infinity of families of unstable periodic orbits is a similar phenomenon as the one observed in the system ( [ func : ham ] ) . we considered the periodic orbits and escapes in two quite different dynamical systems in order to find their common features . the first system is a system of two coupled oscillators , while the second system is a perturbation of the kerr metric in general relativity . in the system ( [ func : ham ] ) escapes occur when the curves of zero velocity open and many orbits escape to infinity . in the second case we have escapes to infinity , but we emphasized the escapes to the central bumpy black hole . in both cases when the energy changes and reaches the escape value , there is a periodic orbit that approaches the point where the czv opens and its period tends to infinity . thus , the family of periodic orbits terminates at the escape energy . before the escape energy this family undergoes an infinity of transitions from stability to instability and vice - versa , and at every transition there is a bifurcation of an equal or double period family of periodic orbits , that does not escape as the energy varies . another common feature is that most chaotic orbits become escape orbits beyond the escape energy ( beyond means larger energy in the first system , but smaller energy in the second system ) . however , for energies slightly beyond the escape energy most orbits take a rather long time to escape . as the energy increases ( decreases ) the proportion of fast escapes increases considerably the proportion of chaotic ( @xmath115 escaping ) orbits increases abruptly as the energy goes beyond the escape energy . we studied the structure of the phase space and we distinguished the ordered and chaotic domains . the ordered domains surround the positions of the stable periodic orbits . thus we found these periodic orbits and their bifurcations as the energy varies . on the other hand , the chaotic domains surround the asymptotic curves from the unstable periodic orbits . in the first case ( of two coupled oscillations ) we found the characteristics of the periodic orbits and their stability . then we found the structure of the phase space on a surface of section for various values of the energy . as the energy increases the chaotic domains increase , while all stable periodic orbits become unstable . ( however , new stable periodic orbits appear for larger energies at tangent bifurcations . ) the chaotic domains are covered by the asymptotic curves of a main unstable periodic orbit . as the energy goes beyond the escape energy most chaotic orbits escape , either directly , or after a small or a large number of intersections with the @xmath106=0 axis . some asymptotic curves are split into successive pieces , each piece making infinite rotations around an escape domain . the statistics of chaos and escape are given by different graphs : ( a ) ( b ) the proportion of chaotic and escaping orbits and the proportion of the directly escaping orbits as functions of the energy , ( c ) ( d ) the proportion of the escaping orbits at the @xmath113th intersection with respect to the initial total number of orbits and with respect to the remaining orbits , as functions of the number of iterations ( which represents time ) . in the system ( [ func : mnmetric ] ) ( manko - novikov metric ) we found various forms of the czvs for various values of the energy , and the corresponding simple periodic orbits . we found also surfaces of section with mostly ordered orbits , for relatively large energies @xmath157 ( but @xmath2451 ) , or a mixture of ordered and chaotic orbits for smaller energies . we found the characteristics of the main families of periodic orbits and their bifurcations . we emphasized the appearance of a couple of periodic orbits ( one stable , and one unstable ) at a tangent bifurcation , and their evolution . as the energy decreases the stable family undergoes an infinity of bifurcations ( at transitions from stability to instability and vice versa ) . as the energy goes below a critical value the period of this orbit goes to infinity and then this orbit escapes to the bumpy black hole and this family of periodic orbits terminates . on the other hand , the bifurcating families exist and do not escape , but become unstable . we found the proportion of the chaotic regions as the energy decreases . beyond the critical energy most of the chaotic region contains escape orbits . the proportion of the chaotic@xmath115escaping orbits increases abruptly below the escape energy and tends to @xmath107 for smaller energies . it is remarkable that these quite different systems have very similar forms of the proportion of chaotic orbits ( figs . [ fig:10]a and [ figcharep]a ) and of the proportion of directly escaping orbits ( figs . [ fig:10]b and [ figcharep]b ) . the relations of the periodic orbits with the chaotic and escaping orbits are also quite similar . these similarities indicate that the properties of periodic , ordered , chaotic and escaping orbits that we studied in the present paper are very similar in generic dynamical systems . bambi , c. : constraint on the quadrupole moment of super - massive black hole candidates from the estimate of the mean radiative efficiency of agn . d 83 , 103003 ( 2011 ) . doi:10.1103/physrevd.83.103003 contopoulos , g. , papadaki , h. , polymilis , c. : the structure of chaos in a potential without escapes . 60 , 249 ( 1994 ) . doi:10.1007/bf0069332 contopoulos , g. , lukes - gerakopoulos , g. , apostolatos , t. a. : orbits in a non - kerr dynamical system . j. bifurc . chaos 21 , 2261 ( 2011 ) . doi:10.1142/s0218127411029768 manko , v. s. , novikov , i. d. : generalizations of the kerr and kerr - newman metrics possessing an arbitrary set of mass - multipole moments . class . quantum gravity 9 , 2477 - 2487 ( 1992 ) . doi:10.1088/0264 - 9381/9/11/013 newhouse , s. e. : the creation of non - trivial recurrence in the dynamics of diffeomorphisms . in : iooss , g. , helleman , r. h. g. , stora , r. ( eds . ) chaotic behaviour of deterministic systems , pp . north - holland , amsterdam ( 1983 )

we study the phase diagram of the hubbard model in the weak - coupling limit for coexisting spin - density - wave order and spin - fluctuation - mediated superconductivity . both longitudinal and transverse spin fluctuations contribute significantly to the effective interaction potential , which creates cooper pairs of the quasi - particles of the antiferromagnetic metallic state . we find a dominant-wave solution in both electron- and hole - doped cases . in the quasi - spin triplet channel , the longitudinal fluctuations give rise to an effective attraction supporting a-wave gap , but are overcome by repulsive contributions from the transverse fluctuations which disfavor-wave pairing compared to . the sub - leading pair instability is found to be in the-wave channel , but complex admixtures of and are not energetically favored since their nodal structures coincide . inclusion of interband pairing , in which each fermion in the cooper pair belongs to a different spin - density - wave band , is considered for a range of electron dopings in the regime of well - developed magnetic order . we demonstrate that these interband pairing gaps , which are non - zero in the magnetic state , must have the same parity under inversion as the normal intraband gaps . the self - consistent solution to the full system of five coupled gap equations give intraband and interband pairing gaps of structure and similar gap magnitude . in conclusion , the gap dominates for both hole and electron doping inside the spin - density - wave phase .

1505.03003

pairing of electrons by exchange of spin fluctuations is a popular paradigm proposed for unconventional superconductivity including fe - based superconductors , cuprates , and heavy fermion systems . since many of these systems exhibit an ordered magnetic phase coexisting with , or in close proximity to , the superconducting phase , a small number of studies have addressed the subsidiary problem of pairing of quasi - particles in the symmetry broken spin - density - wave ( sdw ) phase . while this problem has a long history , recent experimental and theoretical developments have led a number of authors to revisit it.@xcite for the simplest case of a doped one - band hubbard model with standard commensurate @xmath4 ordering , the existence of pairing and its consequences for the symmetry of the superconducting order parameter was initially investigated in a seminal paper by schrieffer _ et al . _ , @xcite where the effective pairing interaction was obtained within the random phase approximation ( rpa ) arising from longitudinal spin fluctuations in the magnetically ordered phase . these authors neglected the contribution to pairing from the transverse spin fluctuations corresponding to the goldstone mode of the spin symmetry broken state , arguing that while such modes lead to a divergent contribution to the spin susceptibility at the ordering vector @xmath5 , the coherence factors of the sdw phase screen the bare electron - electron interaction vertex , which therefore vanishes at @xmath5 . soon after , frenkel and hanke @xcite showed that the transverse fluctuations do contribute to the pairing interaction in the same order as the longitudinal fluctuations ; the divergence of the transverse spin susceptibility is eliminated by the coherence factors , but a residual constant interaction remains . for electron - doped cuprates , the one - band hubbard model seems to provide a reasonable minimal model since the doped electrons reside primarily on the copper sites . furthermore , calculations of band parameters for the electron - doped cuprates point to the fact that the coulomb interaction is smaller than the bandwidth , @xcite in contrast to their hole - doped counterparts . in consequence , the mean field treatment of the sdw order works quite well for the normal state properties of the electron - doped systems . in particular , arpes reports show a fermi surface evolution upon increased electron doping which agrees well with the band reconstruction of the one - band hubbard model due to commensurate @xmath4 order . upon electron doping , the fermi surface consists of electron pockets around @xmath6 , @xmath7 . close to critical electron doping , the emergence of hole pockets at ( or close to ) the fermi level around @xmath8 , @xmath9 may occur . @xcite theoretically , the study of spin - fluctuation pairing for the electron - doped cuprates has been addressed mostly in the paramagnetic phase , @xcite where the mechanism was found to give rise to a gap with @xmath0 symmetry with strong non - monotonic features as a function of momentum as a result of fermi surface intersections with the magnetic zone boundary , the so - called hot spots . at these positions the pairing becomes particularly pronounced . this behavior was found to agree qualitatively with arpes @xcite and raman @xcite observations . in ref . the possibility for a cross - over to @xmath10 symmetry was found at large dopings . in a later work , krotkov and chubukov @xcite studied the spin - mediated pairing gap close to the quantum critical point of antiferromagnetic ( af ) order and found an anisotropic @xmath0 gap symmetry , but the anisotropic behavior was not related to the hot spots , as opposed to previous work . a few studies also addressed the coexistence of superconductivity and long - range af order but treated the pairing phenomenologically , and also found a @xmath0-wave solution . @xcite for hole - doped cuprates , stronger interactions imply that application of a weak - coupling approach to pairing is somewhat less justified . in addition , calculations of the spin wave spectrum suggest that the commensurate @xmath4 order in this case is not the ground state solution , @xcite complicating the theoretical modelling . spin - fluctuation - mediated pairing has , however , been extensively applied to study also hole - doped cuprates within the one - band hubbard model . first , for the paramagnetic phase scalapino _ et al . _ @xcite generalised the approach of berk and schrieffer @xcite and found a dominant @xmath0-wave pairing instability . later , the study was extended by various methods and numerical approaches @xcite including discussions of the possibility of other superconducting pairing symmetries arising from spin fluctuations . from a strong coupling approach , as derived from the @xmath11 model , analysis of the effective pairing at low hole doping @xcite suggests a @xmath3-wave superconducting ground state in any coexistence phase . this is consistent with rigorous perturbative weak - coupling calculations in the presence of weak density - wave order @xcite as well as a study of hole - doped cuprates in the coexistence phase . @xcite in addition , a recent study of spin - fluctuation mediated superconductivity in the sdw ordered metal attacked the problem by analytical rpa calculations in the large magnetization ( small pocket ) limit @xcite and found that interactions were dominated by longitudinal fluctuations on the electron - doped side , supporting a nodeless @xmath0 gap , whereas on the hole - doped side both longitudinal and transverse fluctuations support a nodal @xmath0 gap . recently , lu _ et al . _ @xcite studied the case of underdoped cuprates in a @xmath11 like model and came to a rather different conclusion . in this paper , the coexistence of commensurate af and superconductivity was investigated in a phenomenological model where the pairing interaction arises based on nearest - neighbor magnetic exchange neglecting the double occupancy constraint . in the coexistence phase , this study found the leading superconducting instability to be triplet @xmath1-wave in the case of hole doping , providing a potential explanation for recent photoemission measurements indicating a nodal gap " , a state with a full gap near the usual positions of the @xmath0 gap nodes whose existence is well - established at higher dopings.@xcite it is worth noting , however , that in the original strong - coupling study of unconventional superconductivity driven by the spin waves , studied within the @xmath11 model , the @xmath0-wave symmetry of the superconducting gap was found to be the only stable solution . @xcite we also note that there exist other potential explanations of the existence of the nodeless gap in the literature . @xcite this controversy suggests the need for a better understanding of the phase diagram of the single - band hubbard model within a single , reliable approximation scheme which can encompass paramagnetic , superconducting and coexistence phases , and which is capable of identifying the strength of pairing by both longitudinal and transverse spin fluctuations and charge fluctuations , and their relative importance for pairing across a large doping range , for different electronic structures , and for both weak and strong magnetism . therefore , we extend the work of ref . by a more complete , fully numerical solution of the problem to both confirm the analytical calculations and extend them to the larger phase diagram . we address the question of how the pairing interactions and resulting gap symmetry in the one - band hubbard model , treated within the full spin - fluctuation approach with self - consistently determined sdw order , evolve as a function of doping throughout the phase diagram , including the coexistence dome of sdw order and superconductivity . we show that the @xmath0 solution is in fact the leading superconducting instability for all electron doping levels . in the hole - doped case , we limit ourselves to small hole doping only and force the @xmath4 order to be stable by a suppression of additional intrapocket nesting contributions to the transverse spin susceptibility . this approach also yields a coexistence phase with @xmath0 order . thus , the weak - coupling approach where pairing is mediated by spin fluctuations gives qualitatively different results than the `` strong - coupling '' approach of ref . in the case of hole doping . we find that the sub - leading instability is a singlet @xmath2-wave solution , which shares common nodes with the @xmath0 solution along the zone diagonals , and has additional nodes along the momentum axes . in the hole - doped case where the fermi pockets are located far away from the zone axis , @xmath2 and @xmath0 become nearly degenerate . however , since the two solutions share common nodes , there is no gain in condensation energy by e.g. a time - reversal - symmetry broken solution of the form @xmath12 , and the solution of the full gap equation in fact favors @xmath0 over @xmath2 . we find further that extended @xmath13-wave is always suppressed in the coexistence phase , in agreement with the findings in ref . for small doping levels , and on the electron - doped side we do not encounter a leading triplet gap at any moderate doping away from half - filling . the spin - fluctuation - pairing mechanism becomes strongly suppressed above the critical doping for which long - range magnetic order vanishes , since the nesting conditions are rapidly weakening as the fermi surface segments of the paramagnetic phase move apart . in a recent study of the iron - based superconductors by hinojosa _ et al . _ @xcite it was pointed out that additional interband gaps may develop in the coexistence phase due to pairing of two fermions residing on different bands . such pairs are naively expected to be negligible because they involve fermionic states far from the fermi energy and do not alone manifest a cooper instability . in this regard , these interband pairs can be considered as `` anomalous '' . however , they are non - zero in the sdw state due to a combined effect of the sdw gap and coupling to the intraband gaps ( dubbed in the following as normal gaps ) . in the iron - based system , this effect is predicted to create a different superconducting phase which explicitly breaks time - reversal symmetry by development of chiral gaps . we find from a self - consistent treatment of the coupled gap equations that the one - band hubbard model also supports substantial anomalous interband pairings . the structural factor of the anomalous gaps is @xmath0 as for the intraband gaps , but the phase structure of the intraband and interband gaps can be different . however , unlike the case of iron - based system , @xcite the chiral solution @xmath14 does not appear to be favored in our numerical calculations . the model is the one - band hubbard hamiltonian @xmath15 where @xmath16 creates an electron on site @xmath17 with spin @xmath18 . the interaction term @xmath19 denotes the energy cost associated with having two electrons on the same site . in reciprocal space the hamiltonian reads @xmath20 with @xmath21 - 4t'\cos(k_x)\cos(k_y)-\mu.\ ] ] the parameter @xmath22 is the energy gain corresponding to hopping between neighboring sites , and @xmath23 denotes the energy gain by hopping to next - nearest neighbor sites . the doping level of the system is controlled by changing the chemical potential @xmath24 . the interaction between two electrons is first treated in the hartree - fock approximation giving rise to af ordering of the spins . we therefore consider the mean field hamiltonian @xmath25 where @xmath26 $ ] is the af order parameter , and @xmath27 refers to summation over the reduced brillouin zone only . diagonalization of the mean field hamiltonian leads to the energy spectrum @xmath28 the magnetic gap equation is solved self - consistently given the hopping integrals @xmath29 , @xmath30 , the coulomb repulsion @xmath19 and the doping . a gapless goldstone mode at @xmath31 in the transverse spin channel is obtained automatically , since the magnetic gap equation is equivalent to the condition @xmath32 . @xcite higher order interactions in @xmath19 generate superconductivity on top of the af order through longitudinal and transverse spin fluctuations , following the original proposals of refs . . since @xmath19 scatters the bare electrons , the diagrammatics are performed in terms of the bare electron green s functions , but the cooper pairing takes place between the quasi - particles of the af state . the quasi - particle operators are related to the bare electron operators by the transformation : @xmath33.\end{aligned}\ ] ] transverse and longitudinal spin fluctuations give rise to fundamentally different interactions . inspection of the interaction vertex formulated in real space @xcite shows that the charge and longitudinal interaction vertices give rise to no spin flips , whereas the transverse interaction does . in the latter channel , we have the gapless goldstone mode of the af phase which gives rise to a divergent interaction potential between the bare electrons . however , when we consider pairing between the quasi - particles of the af state , this divergence is removed by the coherence factors as noted in earlier works . @xcite in the case of pairing between opposite spin electrons , spin flip processes are possible and the effective interaction is mediated both by longitudinal and transverse spin fluctuations . if pairing occurs between same spin electrons , only longitudinal spin fluctuations contribute and the pairing potential does not include the bare coulomb repulsion @xmath19 since this acts only between opposite spin electrons . the interaction hamiltonian is formulated in terms of the sdw quasi - particles , and in line with earlier work @xcite we show the interactions in the longitudinal and transverse channel individually , with the transverse part of the interaction stated as a spin flip vertex explicitly , @xmath34 with @xmath35l^2({{\bf k}},{{\bf k}}')-[2u - v_c({{\bf k}}-{{\bf k}}'+{{\bf q}})]m^2({{\bf k}},{{\bf k } } ' ) + v_z({{\bf k}}-{{\bf k}}')l^2({{\bf k}},{{\bf k}}')-v_z({{\bf k}}-{{\bf k}}'+{{\bf q}})m^2({{\bf k}},{{\bf k } } ' ) \nonumber , \\ & & \label{eq : gammaz}\\ \gamma^{+-}_{{{\bf k}},{{\bf k}}'}&=&v_{+-}({{\bf k}}-{{\bf k}}')n^2({{\bf k}},{{\bf k}}')-v_{+-}({{\bf k}}-{{\bf k}}'+{{\bf q}})p^2({{\bf k}},{{\bf k}}')\label{eq : gammapm } , \\ & & \nonumber \\ \gamma^{\rm ss}_{{{\bf k}},{{\bf k}}'}&=&-v_{c}({{\bf k}}-{{\bf k}}')l^2({{\bf k}},{{\bf k}}')-v_{c}({{\bf k}}-{{\bf k}}'+{{\bf q}})m^2({{\bf k}},{{\bf k}}')-v_{z}({{\bf k}}-{{\bf k}}')l^2({{\bf k}},{{\bf k}}')-v_{z}({{\bf k}}-{{\bf k}}'+{{\bf q}})m^2({{\bf k}},{{\bf k}}').\label{eq : gammass}\end{aligned}\ ] ] the main ingredients in the pairing interactions are the spin and charge susceptibilities within the rpa approximation @xmath36 @xmath37 @xmath38 where the spin susceptibilities are defined by @xmath39 and @xmath40 . as seen from eqs . ( [ eq : gammaz])-([eq : gammass ] ) the bare interaction vertices are modified by coherence factors of the sdw phase given by @xmath41 with @xmath42 and @xmath43 . the interaction hamiltonians stated in eqs . ( [ eq : hc])-([eq : hss ] ) are restricted to cooper pairing between quasi - particles residing in the same band , i.e. @xmath44 and @xmath45 . in section [ sec : anomalouspairing ] we introduce an extended model which includes cooper pairing between fermions residing on different pockets , i.e. mean fields of the form @xmath46 . for now we restrict ourselves to normal intraband cooper pairs and include pair scattering interactions within each band as well as between the bands . the expression for the interband couplings between pairs of @xmath47 with @xmath48 residing on the @xmath49 band and @xmath50 on the @xmath51 band , which are dubbed @xmath52 , @xmath53 and @xmath54 are obtained by interchanging the coherence factors @xmath55 and @xmath56 in the eqs . ( [ eq : gammaz])-([eq : gammass ] ) in agreement with refs . . the presence of magnetic order breaks spin - rotational symmetry , but inversion symmetry is preserved . this allows us to express the superconducting gap in an even and odd parity form which corresponds to a quasi - spin singlet and a quasi - spin triplet gap , respectively . we label the gaps by the superscript `` s '' for even parity ( singlet ) and superscript `` t '' for the odd parity ( triplet ) gap , i.e. @xmath57 . as we show later when discussing the interband ( anomalous ) pairing terms , it is more useful to classify the superconducting gaps in the sdw background by parity rather than by spin quantum numbers . in this regard calling cooper pairing spin singlet or spin triplet in the sdw background actually refers to the even or odd parity of the wave function , respectively . in the derivation of the superconducting gap equations the splitting of these two channels leads to the symmetrization / antisymmetrization of the pairing potential in the quasi - spin singlet / triplet gap equation @xmath58\nonumber,\\ \label{eq : scgapequation}\end{aligned}\ ] ] and similarly for @xmath59 by interchanging @xmath60 . in the case of opposite spin interaction the effective pairing interactions for the singlet and triplet channel are given by @xmath61 for the intraband contributions and equivalent expressions for the interband , @xmath62 , contributions . similarly , for same spin electron interactions , the singlet and triplet potential is obtained directly by a symmetrization / antisymmetrization of the potential stated in eq . ( [ eq : gammass ] ) . the strength of the pairings is calculated by evaluating the real part of the rpa susceptibilities at zero energy . note that due to broken spin rotation symmetry there is a difference between @xmath63 and @xmath64 . the pairing vertex @xmath65 diverges at @xmath66 due to the goldstone mode in the transverse channel . this divergence is removed , however , by the coherence factor @xmath67 of the sdw phase , as discussed above . in the paramagnetic phase where @xmath68 , pairing takes place between the bare electrons and the gap equation then reduces to @xmath69\frac{\delta^{\rm ( s / t)}_{{{\bf k}}'}}{e^{\rm ( s / t)}_{{{\bf k}}'}}\tanh\big(\frac{e^{\rm ( s / t)}_{{{\bf k}}'}}{2k_bt}\big),\nonumber\\ \label{eq : scgapequationns}\end{aligned}\ ] ] with @xmath70 . the effective pairings are given by @xmath71\pm v_{+-}({{\bf k}}-{{\bf k}}').\nonumber\\ \label{eq : pmpot}\end{aligned}\ ] ] in the sdw phase , the effective pairings contain umklapp terms , which are the terms in eqs . ( [ eq : gammaz])-([eq : gammass ] ) containing the argument @xmath72 . due to these terms , there are attractive pair scatterings @xmath47 on the reconstructed fermi surface . this contrasts with the situation in the paramagnetic phase , where the spin - fluctuation part of singlet potential is always purely repulsive for all pairs of @xmath47 , as seen from eq . ( [ eq : pmpot ] ) . however , despite the presence of a partially attractive potential , a conventional @xmath13-wave superconducting gap is not possible . this is due to a symmetry constraint on the pairing potentials , which must obey @xmath73 , a property that is fulfilled for both the longitudinal and transverse pairing interactions . this symmetry also carries over to the superconducting gap , which must satisfy @xmath74 for the two dimensional square lattice , the gap solutions can be classified according to the five irreducible representations of the @xmath75 group that are even under reflection through the horizontal plane , i.e. extended @xmath13-wave ( @xmath76 ) , @xmath0 , @xmath10 , @xmath2 and the triplet solution , @xmath77 , which is doubly degenerate . first we consider whether these solutions all comply with the additional symmetry constraint , eq . ( [ eq : dsym ] ) present in the sdw phase . this disqualifies the @xmath10 solution and as a result we consider the leading gap symmetries from among the set @xmath78\sin(k_x)\sin(k_y ) , \label{eq : gxy}\\ e_{u } : & ~ & p_{x}=\sin(k_x ) , \quad p_{y}=\sin(k_y ) \label{eq : p}.\end{aligned}\ ] ] we emphasize that these basis functions are only the lowest order functions corresponding to the given irreducible representations . in the construction of higher order solutions in the sdw phase , it is important to note that an odd number of the above basis functions must be multiplied in order for the resulting higher order gap function to comply with eq . ( [ eq : dsym ] ) . further allowed higher order solutions can also be achieved by a multiplication of the @xmath10 basis function [ @xmath79 with any @xmath80 , @xmath81 or @xmath82 basis functions . a relevant triplet solution is constructed by a multiplication of the triplet @xmath1-wave with the lowest order a@xmath83 and b@xmath83 basis functions , @xmath84[\cos(k_x)+\cos(k_y)]\sin(k_x ) , \label{eq : extf } \nonumber \\\end{aligned}\ ] ] which for tetragonal symmetry belongs to the same @xmath85 symmetry representation as the original @xmath1-wave solution . therefore , we dub this state @xmath86-wave . note that in the literature on sr@xmath87ruo@xmath88 this function is sometimes called @xmath89-wave . @xcite we find that this is the leading solution among the triplet solutions for any doping . however , as we shall see below , it does not dominate over the singlet solutions . and ( b ) small electron doping @xmath90 for @xmath91 and @xmath92 . ] the consequences of a @xmath0 solution are manifested very differently depending on the fermi surface geometry . in the presence of only electron pockets , as in fig . [ fig : fermisurf](b ) , it is nodeless at the fermi surface , as opposed to the other solutions , which all display nodes . if both electron and hole pockets are present at the fermi surface as in fig . [ fig : solutionslge](e ) , the @xmath0 solution has nodes only at the hole pockets , whereas the extended @xmath13-wave as well as the @xmath2-wave solutions exhibit nodes at both types of pockets . navely , we might expect the preferred solution to have the minimum number of nodes at the fermi surface . in this respect , the @xmath0 solution clearly wins , but ultimately the leading solution relies on a detailed investigation of the structure of the pairing interaction in the sdw phase and the solution to the full gap equation . finally , when only hole pockets are present at the fermi surface as shown in fig . [ fig : fermisurf](a ) , the @xmath93 solution becomes nodeless , whereas none of the singlet gap symmetries will provide nodeless superconductivity . as we show below , the @xmath93 solution is supported from the structure of the longitudinal fluctuations , but becomes overall less favorable compared to @xmath0 due to an effective repulsion arising from the transverse spin fluctuations . in order to determine the leading and sub - leading instabilities , we linearize the full gap equation stated in eq . ( [ eq : scgapequation ] ) to obtain the eigenvalue problem @xmath94 the eigenvector of this equation @xmath95.\end{aligned}\ ] ] is a function of momentum @xmath48 . the momentum is located either on the electron pocket around @xmath6 , which we denote by @xmath49 , or on the two hole pockets , denoted by @xmath96 , around @xmath8/@xmath9 , respectively , see fig . [ fig : fermisurf ] . we solve the eigenvalue problem eq . ( [ eq : lge ] ) for @xmath48 and @xmath50 on these three closed pockets , which is equivalent to solving it in the whole magnetic brillouin zone . all intraband and interband interactions are included in @xmath97 : @xmath98 . \nonumber\ ] ] here , @xmath99 denotes the length of the fermi surface line segment represented by the point @xmath50 and @xmath100 is the fermi velocity at @xmath50 . the intraband ( @xmath101 ) and interband ( @xmath102 ) pairings must be expressed in the singlet and triplet symmetrized versions in accordance with eqs . ( [ eq : sympots])-([eq : sympott ] ) . the largest eigenvalue , @xmath103 , gives the leading instability since it corresponds to the largest superconducting critical temperature and the symmetry of the gap is given by the corresponding eigenvector . the linearized gap equation does not allow for a determination of complex gap solutions , which are time - reversal - symmetry - broken ( trsb ) solutions , since higher order interactions in @xmath104 are removed in the linearization process . therefore , we have additionally solved the full nonlinear gap equation as given in eq . ( [ eq : scgapequation ] ) . this allows us to compare the real solutions and trsb solutions energetically . the latter type of solutions arise naturally in situations where the pairing potential allows for two degenerate eigenfunctions of @xmath97 . therefore , they are likely to appear in the triplet channel , where all solutions to the linearized gap equation are two - fold degenerate . since the linearized equation in fact also exhibits accidentally degenerate solutions in the singlet channel upon hole doping , this opens the possibility of trsb states in this case . the prospect of observation of transitions to trsb states as a function of doping or other control parameters , potentially the first observation of this kind , has recently been the subject of considerable attention in fe - based superconductors . @xcite so far , we have restricted the discussion to cooper pairing between quasi - particles residing on the same band , which are cooper pairs of the form @xmath44 and @xmath45 . now we introduce an extension of the model to also include anomalous pairings between fermions belonging to different pockets , i.e. @xmath46 . the complete interaction hamiltonian includes pair scattering processes between normal and anomalous gaps , as well as scatterings between anomalous gaps . anomalous pairs of the form @xmath46 involve fermions far from the fermi level , since the two bands are gapped by @xmath105 . nevertheless , the anomalous gaps become sizeable due to the coupling to the normal intraband pairs and exist only in the sdw phase since the coupling between normal and anomalous gaps are proportional to @xmath106 . a detailed analysis of the interaction hamiltonian reveals that _ even _ parity intraband gaps , i.e. @xmath107 couple to _ even _ parity anomalous interband gaps . in the iron - pnictide study of ref . the reported anomalous gap was dubbed quasi - spin triplet and coupled to an even parity singlet intraband gap . however , in that case the triplet gap is actually of even parity , which is allowed due to the band index . as a matter of fact , in the sdw background it is more useful to classify the superconducting gaps by parity rather than by spin quantum numbers , as spin - rotational symmetry is explicitly broken . thus , in our case as well as in ref . one has an even - parity normal intraband gap coupling to even - parity anomalous pairing contributions . as we shall see below , calculations of the normal intraband gaps reveal an even - parity @xmath0-wave solution . because of this , and also as a result of the structure of the interaction hamiltonian , we restrict ourselves to the even - parity channel which we label by a quasi - spin singlet index , @xmath13 . in the even - parity channel , the mean - field interaction hamiltonian takes the form @xmath108 \nonumber\\ & & \nonumber \\ & & \hspace{.8cm}+\delta_{\alpha\beta{\uparrow}{\downarrow}}^s({{\bf k}})\alpha_{-{{\bf k}}{\downarrow}}\beta_{{{\bf k}}{\uparrow } } -\delta_{\alpha\beta{\downarrow}{\uparrow}}^s({{\bf k}})\beta_{-{{\bf k}}{\downarrow}}\alpha_{{{\bf k}}{\uparrow } } + h.c . \big],\nonumber\\ \label{eq : hmffull}\end{aligned}\ ] ] with the mean fields @xmath109 + i\sigma_{\gamma\delta}^y\gamma^{\alpha\beta , s}_{{{\bf k}},{{\bf k } } ' } \langle \alpha_{{{\bf k } } ' \delta } ^\dagger \beta_{-{{\bf k } } ' \gamma}^\dagger + \beta_{{{\bf k } } ' \delta } ^\dagger \alpha_{-{{\bf k } } ' \gamma}^\dagger \rangle , \label{eq : daa}\\ & & \nonumber\\ \delta_{\beta\beta}^s({{\bf k}})&=&-\frac{1}{8n}\sideset{}{'}\sum_{{{\bf k } } ' } i\sigma_{\gamma\delta}^y[\gamma^{\rm s}_{{{\bf k}},{{\bf k}}'}\langle \beta_{{{\bf k } } ' \gamma}^\dagger \beta_{-{{\bf k } } ' \delta}^\dagger \rangle + \tilde\gamma^{\rm s}_{{{\bf k}},{{\bf k}}'}\langle \alpha_{{{\bf k } } ' \gamma } ^\dagger \alpha_{-{{\bf k } } ' \delta } ^\dagger \rangle ] - i\sigma_{\gamma\delta}^y \gamma^{\alpha \beta , s}_{{{\bf k}},{{\bf k } } ' } \langle \alpha_{{{\bf k } } ' \delta } ^\dagger \beta_{-{{\bf k } } ' \gamma}^\dagger + \beta_{{{\bf k } } ' \delta } ^\dagger \alpha_{-{{\bf k } } ' \gamma}^\dagger \rangle , \label{eq : dbb}\\ & & \nonumber \\ \delta_{\alpha\beta}^s({{\bf k}})&=&-\frac{1}{8n}\sideset{}{'}\sum_{{{\bf k } } ' } i\sigma_{\gamma\delta}^y \gamma^{\alpha \beta , s}_{{{\bf k}},{{\bf k}}'}\langle \alpha_{{{\bf k } } ' \gamma } ^\dagger \alpha_{-{{\bf k } } ' \delta } ^\dagger - \beta_{{{\bf k } } ' \gamma } ^\dagger \beta_{-{{\bf k } } ' \delta } ^\dagger \rangle , \label{eq : gapano } \\ \delta_{\alpha\beta { \uparrow}{\downarrow}}^s({{\bf k}})&=&-\frac{1}{4n}\sideset{}{'}\sum_{{{\bf k } } ' } \gamma^{\alpha\beta\alpha\beta,1,s}_{{{\bf k}},{{\bf k } } ' } \langle \alpha_{{{\bf k } } ' { \uparrow}}^\dagger \beta_{-{{\bf k } } ' { \downarrow}}^\dagger \rangle + \gamma^{\alpha\beta\alpha\beta,2,s}_{{{\bf k}},{{\bf k } } ' } \langle \beta_{{{\bf k } } ' { \uparrow}}^\dagger \alpha_{-{{\bf k } } ' { \downarrow}}^\dagger \rangle , \\ \delta_{\alpha\beta { \downarrow}{\uparrow}}^s({{\bf k}})&=&-\frac{1}{4n}\sideset{}{'}\sum_{{{\bf k } } ' } \gamma^{\alpha\beta\alpha\beta,1,s}_{{{\bf k}},{{\bf k } } ' } \langle \alpha_{{{\bf k } } ' { \downarrow}}^\dagger \beta_{-{{\bf k } } ' { \uparrow}}^\dagger \rangle + \gamma^{\alpha\beta\alpha\beta,2,s}_{{{\bf k}},{{\bf k } } ' } \langle \beta_{{{\bf k } } ' { \downarrow}}^\dagger \alpha_{-{{\bf k } } ' { \uparrow}}^\dagger \rangle . \label{eq : gapanodownup}\end{aligned}\ ] ] in these gap equations we have introduced three new spin - fluctuation - mediated pairing interactions @xmath110 - [ 2v_{+-}({{\bf k}}-{{\bf k}}')\mu({{\bf k}},{{\bf k } } ' ) + 2v_{+-}({{\bf k}}-{{\bf k}}'+{{\bf q}})\nu({{\bf k}},{{\bf k } } ' ) ] , \nonumber \\ & & \label{eq : pairingaaab } \\ \gamma^{\alpha\beta\alpha\beta,1}_{{{\bf k}},{{\bf k}}'}&= & -[v_{lo}({{\bf k}}-{{\bf k}}')p^2({{\bf k}},{{\bf k}}')+v_{lo}({{\bf k}}-{{\bf k}}'+{{\bf q}})n^2({{\bf k}},{{\bf k}}')+2v_{+-}({{\bf k}}+{{\bf k}}')n^2({{\bf k}},{{\bf k}}')+2v_{+-}({{\bf k}}+{{\bf k}}'+{{\bf q}})p^2({{\bf k}},{{\bf k}}')],\nonumber \\ & & \\ \gamma^{\alpha\beta\alpha\beta,2}_{{{\bf k}},{{\bf k}}'}&= & v_{lo}({{\bf k}}-{{\bf k}}')l^2({{\bf k}},{{\bf k}}')+v_{lo}({{\bf k}}-{{\bf k}}'+{{\bf q}})m^2({{\bf k}},{{\bf k } } ' ) + 2v_{+-}({{\bf k}}+{{\bf k}}')m^2({{\bf k}},{{\bf k}}')+2v_{+-}({{\bf k}}+{{\bf k}}'+{{\bf q}})l^2({{\bf k}},{{\bf k}}')],\nonumber\\ \label{eq : gammaabab2 } \end{aligned}\ ] ] where @xmath111 , and @xmath112 in eq . ( [ eq : pairingaaab ] ) refers to the quasi - spin singlet and triplet , respectively . in eqs . ( [ eq : daa])-([eq : gapanodownup ] ) the superscript @xmath13 refers to the symmetrized potentials , e.g. @xmath113 . this ensures that all mean fields are of even parity . we have introduced two new coherence factors , which are proportional to the magnetic order parameter @xmath106 @xmath114 note that the coupling between the normal intraband gaps and the anomalous pairings occurs due to the pairing stated in eq . ( [ eq : pairingaaab ] ) which is proportional to the sdw order parameter . , in the sdw phase as a function of doping , for the parameters @xmath92 and @xmath91 . ] in the next sections , we discuss the results of spin - fluctuation - mediated pairing in the sdw phase . first , we restrict the model to normal intraband cooper pairs only . in section [ sec : pairingstrength ] we study how the effective interactions mediated by transverse and longitudinal spin fluctuations evolve as a function of doping . thereafter , we turn to the implications for the evolution of different gap symmetry solutions in section [ sec : lgeresults ] . we focus on electron doping , since this is the most relevant regime for a coexistence phase of commensurate af order and superconductivity . in the case of hole doping , we restrict the investigation to small hole doping levels below 10 % . lastly , we provide a discussion of the extended model , where anomalous pairings are included , in section [ sec : anomalouspairingresults ] . first we focus on the strength and structure of the pairing potentials between opposite spin electrons arising from longitudinal and transverse spin fluctuations for hole and electron doping . in fig . [ fig : potphasediagram ] we plot the value of the pair scattering @xmath115 for which the pairing strength is maximal for any pair of momenta @xmath47 on the fermi surface . in the low doping limit , we observe that the interaction through transverse fluctuations is much stronger for hole doping than for electron doping . for interactions in the longitudinal channel , the situation is opposite , since in this case the pairings are strongest for small electron doping compared to small hole doping . generally , for the hole - doped system , transverse fluctuations are quantitatively stronger than the longitudinal fluctuations . for electron dopings , the longitudinal fluctuations dominate close to half filling , whereas both types of pairings contribute more equally at larger doping levels . in fact , the strength of the pairing interactions in both channels builds up towards the critical electron doping where the af order disappears . this navely suggests an increase in superconductivity close to critical electron doping . however , a closer inspection of the @xmath47 dependence of the pairing reveals that it develops a sign change for a part of the intrapocket potential and becomes strongly repulsive for nearby points at the fermi surface . this leads to an overall gap suppression despite the increased interaction strength . . the black cross denotes the position of @xmath50 and the value of @xmath115 is shown as a function of @xmath48 around the hole pocket . negative potential contributions correspond to effective attraction . the zone diagonal is indicated by a dashed line . the filling is @xmath116 and @xmath117 , @xmath92 . note the different color scale for pairing potentials in the longitudinal and transverse channels . ] . the black cross denotes the position of @xmath50 and the value of @xmath115 is shown as a function of @xmath48 around the electron pocket . the filling is @xmath90 and @xmath117 , @xmath92 . note the different color scale for singlet and triplet pairing potentials . ] in general , knowledge of the @xmath47 structure of the pairing potential is crucial in order to decide what kind of superconducting instabilities are favored in the af phase . to address this question , in figs . [ fig : potbeta ] and [ fig : potalpha ] we map out both the singlet and triplet pair scattering potentials due to longitudinal and transverse spin fluctuations for hole- and electron doping , respectively . in the case of small hole doping ( fig . [ fig : potbeta ] ) , the singlet pairing interactions from longitudinal and transverse spin fluctuations are similar in structure ; there is an effective attractive ( negative ) interaction for @xmath48 and @xmath50 on the same side of the zone diagonal , whereas the interaction becomes repulsive ( positive ) when @xmath48 and @xmath50 are on opposite sides of the zone diagonal , as deduced from fig . [ fig : potbeta](a - b ) . this structure supports @xmath3-wave as well as @xmath2-wave states . in this case , the pairing interaction due to transverse spin fluctuations is much stronger than the pairing arising from longitudinal spin fluctuations , as was also deduced from fig . [ fig : potphasediagram ] , and expected from refs . , and . in the triplet channel , the effective interaction arising from the two different types of spin fluctuations has very different character . the longitudinal fluctuations in fact give rise to a locally attractive pairing on the entire pocket , as shown in fig . [ fig : potbeta](c ) . this type of pairing supports a nodeless gap of @xmath1-wave character , i.e. @xmath118 . however , due to the strong intrapocket repulsive interaction mediated by transverse spin fluctuations shown in fig . [ fig : potbeta](d ) , the total pairing interaction does not allow for a @xmath1-wave solution . @xcite therefore the triplet solution becomes higher order and will display nodes at the fermi level . ) , ( c , d ) only electron pockets ( 5 % electron doping and @xmath91 ) or ( e , f ) both types of pockets ( 5% electron doping and @xmath119 ) . the last situation is obtained very close to the af quantum critical point , i.e. for @xmath120 . the leading solutions are shown in the first column and the sub - leading solutions are shown in the second column . in all three cases the leading solution is @xmath0 , and the sub - leading solution is @xmath2-wave . the latter differs from @xmath13-wave by being odd under @xmath121 and @xmath122 . note that in the case of hole doping the two solutions @xmath0 and @xmath2 are nearly degenerate with @xmath123 and @xmath124 . in the case of small and intermediate electron doping , the @xmath0 solution becomes strongly dominant . upon increased electron doping approaching the quantum critical point , the @xmath2-wave solution becomes increasingly important , although always subdominant . ] for electron doping , the pairing interaction in the singlet channel is locally attractive for @xmath48 and @xmath50 residing on the same electron pocket , as shown in fig . [ fig : potalpha](a - b ) . attractive pairing occurs due to both longitudinal and transverse spin fluctuations . as opposed to the hole - doped case , it is the longitudinal fluctuations that mediate the strongest effective pairing . in this regime , it is clearly the @xmath0 solution which will be favored due to the symmetry @xmath125 . upon increasing the electron doping the purely attractive intrapocket interaction is replaced by partly repulsive regions and this will diminish the resulting superconducting gap value . in the triplet channel the pairings from both types of fluctuations will be quite weak , see fig . [ fig : potalpha](c - d ) . as in the case of hole doping , the longitudinal fluctuations support a @xmath1-wave gap since it is attractive for @xmath48 and @xmath50 located at the same side of the fermi pocket and repulsive for @xmath48 and @xmath50 on opposite sides , as shown in fig . [ fig : potalpha](c ) . the transverse fluctuations display the reverse structure as evident from in fig . [ fig : potalpha](d ) , which is the reason why the @xmath1-wave solution becomes suppressed also on the electron - doped side . keeping in mind the structure of the pairing interactions described above , we now turn to the solutions to the linearized gap equation eq . ( [ eq : lge ] ) in order to determine the subleading solutions and the doping evolution of the three leading solutions . we consider three qualitatively different types of fermi surfaces . first , we study a fermi surface consisting of hole pockets at @xmath126 which occurs on the hole - doped side , see fig . [ fig : solutionslge](a ) . second , we turn to electron pockets at @xmath127 and @xmath128 which occur for small and intermediate electron doping levels , see fig . [ fig : solutionslge](c ) , and third , we study the occurrence of both electron and hole pockets very close to critical doping as in fig . [ fig : solutionslge](e ) , i.e. the doping level for which af order disappears . as discussed above , the presence of hole pockets destabilizes the commensurate af by additional intrapocket contributions to the bare spin susceptibilities . in this case , we force stability of the commensurate af order by turning off the intrapocket contributions by hand , which is justified from the work by chubukov and frenkel@xcite showing that vacuum renormalization leads to a stability of the commensurate af order at small hole dopings . by inspection of the intrapocket potential structure on the hole pocket shown in fig . [ fig : potbeta](a , b ) , we expect either of the two solutions @xmath0 or @xmath2 for the weakly hole - doped system . from the intrapocket potential structure it is not possible to qualitatively distinguish between these two solutions . the only other way that the system might choose one solution over the other would be from the structure of the pairing potential for @xmath48 on one hole pocket and @xmath50 on the neighboring pocket . however , from the numerical potential evaluation it turns out that the interpocket pairing contribution is an order of magnitude smaller than the intrapocket pairing and its symmetry does not favor one of the solutions over the other . this is the reason we obtain two nearly degenerate singlet solutions , namely @xmath0 and the @xmath2 solutions , in fig . [ fig : solutionslge](a , b ) . turning to the case of 5 % electron doping shown in fig . [ fig : solutionslge](c , d ) , we see that the @xmath0 solution is clearly favored over the sub - leading solution @xmath2 , which becomes strongly suppressed . this is a direct consequence of the intrapocket attraction between @xmath48 and @xmath50 residing on the same electron pocket . upon increasing doping , the electron pockets grow in size , but throughout the sdw region , the @xmath0 solution continues to be the leading solution although the sub - leading @xmath2-wave gets closer . also close to the crossover to the paramagnetic phase , where both electron and hole pockets are present at the fermi surface , the leading solution remains @xmath0 as shown in fig . [ fig : solutionslge](e , f ) . ) , are shown . the @xmath0 solution ( red line ) dominates at all fillings . the next - nearest neighbor hopping is @xmath92 and the bare coulomb interaction is @xmath91 . ] the complete doping evolution of the three leading superconducting order parameters , @xmath0 , @xmath2 , and @xmath129-wave , is shown in the phase diagram in fig . [ fig : phasediagramu3nt0p35 ] . on the hole - doped side , we limit the study to the underdoped region . very close to half filling we observe a near - degeneracy of the @xmath0 and @xmath2-wave solutions . the existence of two degenerate solutions in principle allows for their mixture as a trsb solution , or for orthorhombic distortions as a direct mixture . at the mean field level , the preference of a low temperature trsb solution might be expected when it is possible to remove gap nodes at the fermi surface , since removal of gaps from @xmath130 leads to a gain in the condensation energy . in the present case of the degenerate solutions @xmath0 and @xmath2 , however , we note that these share the same nodes along the zone diagonal . thus , there will be no apparent gain of energy by constructing a trsb solution of the form @xmath131 . in fact , solving the full gap equation as stated in eq . ( [ eq : scgapequation ] ) reveals that the @xmath0 solution is energetically favored . upon larger hole doping , the near - degeneracy of the @xmath0 and @xmath2-wave solutions is split and the @xmath0 solution becomes clearly dominant in this regime . in the case of electron doping , the @xmath3-wave solution is strongest very close to half filling even though this is not where the longitudinal and transverse pairing potentials achieve their maximum strengths , see fig . [ fig : potphasediagram ] . the reason is that in the limit where electron pockets are small , the structure of the intrapocket pairing potentials is purely attractive , as shown in fig . [ fig : potalpha](a , b ) , and this strongly supports a @xmath3-wave solution . we note that this feature of a well - developed gap in the limit of small electron dopings is an inherent result of the weak - coupling approach to the coexistence phase . at critical electron doping for which @xmath132 , the fermi arcs just touch the magnetic zone boundary and as a consequence , nesting by @xmath133 on the paramagnetic side is rapidly weakened upon increased electron doping . the ordering of the leading solutions remains the same as in the sdw phase , with the @xmath129 solution , which in the paramagnetic phase takes the simpler form @xmath134\sin(k_x)$ ] , the least favorable . the paramagnetic fermi surface , which is a hole pocket centered at @xmath4 is roughly circular thereby preventing nesting not only at @xmath133 , but at any @xmath135-vector . as a result spin - fluctuation - mediated superconductivity rapidly dies off . the evolution of the @xmath0 solution was previously discussed in ref . , where the possibility of a different pairing mechanism close to the crossover between sdw and paramagnetism was speculated . in ref . the pairing was treated at a phenomenological level and here we note that also the full treatment of the spin - mediated pairing gives rise to a rapid decrease in the superconductivity upon entering the paramagnetic region . finally , we show estimations for the critical temperatures of the commensurate af order , @xmath136 , as well as the superconducting ordering temperature , @xmath137 in fig . [ fig : tempphasediagramu3nt0p35 ] . the @xmath137 is estimated from the eigenvalue @xmath103 obtained for the leading @xmath3-wave solution as a function of doping . whereas @xmath137 drops off upon increasing electron doping it shows the opposite evolution as a function of hole doping . at @xmath138 the superconducting instability is absent due to a full gapping of the fermi surface by the magnetic order . similar behavior was found by variational cluster perturbation theory in the work by snchal and coworkers . @xcite , with the energy cut off set to @xmath139 and @xmath140 mev . the superconducting instability is @xmath0 at all fillings . the sdw region which is determined from the mean field gap equation is shown by the green area . the next - nearest neighbor hopping is @xmath92 and the bare coulomb interaction is @xmath91 . note that within the weak - coupling approach the sdw ground state is always metallic away from @xmath138 and therefore , does not cover the mott insulating behavior of in the underdoped region of hole - doped cuprates . ] as discussed in the introduction , the results on the hole - doped side contradicts the recent findings of ref . , where a @xmath1-wave solution appears as the leading instability . in fact , the cooper - pairing in ref . is mediated by an exchange interaction which involves only nearest neighbor sites , while in our approach the interaction potential in the cooper channel invokes effectively also sites which are farther apart . the similarity of the treatment of the coexistence phase in ref . and our approach allows for a direct comparison of the interaction hamiltonians . in the simplified case of only hole pockets at the fermi surface , the interaction hamiltonian reads generally @xmath141 in the _ t - j_-like model without double occupancy constraint employed in ref . , the effective interaction entering eq . takes the form @xmath142 \nonumber \\ & & -j({{\bf k}}+{{\bf k}}')[n^2({{\bf k}},{{\bf k}}')+p^2({{\bf k}},{{\bf k } } ' ) ] , \nonumber \\ \label{eq : lugamma}\end{aligned}\ ] ] with @xmath143 $ ] and @xmath144 . by contrast , in the hubbard model the effective interaction entering eq . is given by @xmath145 with the longitudinal and transverse effective interactions stated in eqs . ( [ eq : gammaz ] ) and ( [ eq : gammapm ] ) . the lower sign of eq . ( [ eq : betahub ] ) belongs to the triplet channel and this is the source of an effective intrapocket repulsion on the hole - doped side as shown in fig . [ fig : potbeta](d ) . on the contrary , the effective interaction as stated in eq . ( [ eq : lugamma ] ) gives rise to a purely attractive triplet potential for @xmath48 and @xmath50 residing on the same hole pocket , and therefore the @xmath1-wave solution would indeed appear to be the dominating instability . note , however , that the model of ref . can not be considered strictly as a strong - coupling limit of the single band hubbard model as it does not include a constraint for no double occupancies of the fermions explicitly . this could play an important role . for example , as mentioned in the introduction , the original strong - coupling study of unconventional superconductivity driven by the spin waves , studied within the _ t - j _ model with the constraint of no double occupancies , does find the d@xmath146-wave symmetry of the superconducting gap to be the only stable solution . @xcite we introduced the possibility of additional pairing gaps in the coexistence phase in sec . [ sec : anomalouspairing ] . such anomalous pairings were neglected in the previous discussion , see figs . [ fig : potphasediagram]-[fig : tempphasediagramu3nt0p35 ] , where we demonstrated that the dominating symmetry of the normal intraband pairing gaps is @xmath0 at all doping values . now we address the question of whether the inclusion of anomalous interband gaps introduces essential modifications to this result . the anomalous gaps develop under the constraint that they must have the same parity as the normal intraband gaps . in the current case , this means that anomalous pairing gaps must necessarily be of even parity . ) -([eq : gapanodownup ] ) for the electron - doped system with @xmath51-bands outside the energy cut off . the superconducting gap symmetry is @xmath0 for both normal and anomalous gaps . the next - nearest neighbor hopping is @xmath92 and the bare coulomb interaction is @xmath91 . the energy cut off set to @xmath147 . we show the absolute value of the gap averaged over @xmath48-states within the energy range @xmath148 , e.g. @xmath149 . within this energy range no normal superconducting gap develops on the @xmath51-band , i.e. @xmath150 . the dashed blue line shows @xmath151 in the case where all anomalous gaps are excluded . ] we solve the system of five coupled gap equations as stated in eqs . ( [ eq : daa])-([eq : gapanodownup ] ) numerically to obtain the fully self - consistent solutions in the coexistence phase . since we have not included any energy dependence of the effective pairing vertices , we restrict the momentum sums of eqs . ( [ eq : daa])-([eq : gapanodownup ] ) to states in the vicinity of the fermi surface . the effective potential for pair scattering between anomalous gaps , @xmath152 stated in eq . ( [ eq : gammaabab2 ] ) , formally includes a divergent contribution for @xmath153 . to avoid such contributions we cut off all pairings @xmath152 to a maximum value of @xmath154 . by this procedure , we find that also the anomalous gaps acquire a @xmath0 structure . in fig . [ fig : electronselfconsistent ] we present the normal and anomalous mean field gaps given in eqs . ( [ eq : daa])-([eq : gapanodownup ] ) in the case of an electron - doped system in the regime of well - developed sdw order . we show the gap value averaged over @xmath48-states close to the fermi surface , e.g. @xmath149 , where @xmath155 is the number of @xmath48-states with @xmath156 and @xmath157 is the cut off energy . upon approaching half filling where @xmath158 for this band , the self - consistent solutions gradually decrease . this differs from the results obtained for the linearized gap equation in fig . [ fig : phasediagramu3nt0p35 ] where the decrease is very abrupt . the gradual decrease in fig . [ fig : electronselfconsistent ] follows the decrease in the number of states within the energy range around the fermi surface towards half filling . and @xmath159 and ( b ) the anomalous gap , @xmath160 . all gaps have @xmath0 structure . whereas the normal gaps show the same overall phase on electron and hole pockets , the anomalous gaps display an internal @xmath161-shift between the electron and hole pockets . ] from the results presented in fig . [ fig : electronselfconsistent ] , it is evident that the mean field @xmath162 is the most important contributor to anomalous pairing whereas @xmath163 and @xmath164 are negligible . this illustrates the point that it is the linear coupling to the normal gaps via the pairing interaction @xmath165 of eq . ( [ eq : pairingaaab ] ) which generates the anomalous pairings . without this coupling term also the subdominant mean fields @xmath163 and @xmath164 would vanish . furthermore , the inclusion of anomalous pairings slightly enhances the size of the intraband gap . this is seen from fig . [ fig : electronselfconsistent ] by compairing @xmath151 as deduced from the self - consistent determination of all five gaps shown by blue circles with a full line with the self - consistent calculation of the normal intraband gaps alone , which is shown by the blue dashed line . in the latter case the gap magnitudes are slightly smaller . if hole pockets are included within the energy cut off , i.e. @xmath166 , the self - consistent calculation finds a @xmath161-shift of the anomalous superconducting phase at the hole pockets compared to the electron pockets , see fig . [ fig : phase ] ( b ) . this is a robust feature arising from the momentum structure of the normal - anomalous pair scattering potential and as a result of the internal sign between @xmath49- and @xmath51-operators in eq . ( [ eq : gapano ] ) . we have tested numerically that this result is insensitive to the relative ratio of normal and anomalous pairing potentials . lastly , we mention that the self - consistent gap solutions are all real - valued . therefore unlike the iron - based case studied in ref . , we do not find a trsb superconductor in the coexistence phase . in conclusion , the incorporation of anomalous pairings in the effective interaction hamiltonian gives only small quantitative changes to the intraband gap results presented in the previous sections . this study represents a very detailed investigation of spin - fluctuation - mediated superconductivity in a system with well - developed itinerant antiferromagnetic order and includes a calculation of the complete coexistence phase diagram of the hubbard model . the spin fluctuations bear important fingerprints of the spin order with the transverse spin fluctuations corresponding to the goldstone mode of the spin - symmetry - broken state . we find that longitudinal and transverse spin fluctuations are equally important for the development of an effective pairing glue between the fermionic quasi - particles of the spin - density - ordered metal . both types of spin fluctuations act in concert in the even parity quasi - spin singlet channel and this gives rise to a robust gap solution of @xmath0 structure at all doping levels within the coexistence doping region . the situation is quite different in the case of odd parity quasi - spin triplet superconductivity . here longitudinal spin fluctuations promote a nodeless @xmath1-wave solution in the hole - doped system due to an effective intrapocket attraction . however , a strong intrapocket repulsive contribution from transverse fluctuations destroys this effective attraction and destabilizes the @xmath1-wave solution . thus , the coexistence phase treated within the hubbard model does not support a nodeless gap on the hole doped side , in contrast to a recent study of the coexistence phase within the @xmath11 model without double occupancy constraint . @xcite a modified version of the hubbard model in which the relative contribution from the transverse fluctuations is suppressed , could also stabilize a nodeless @xmath1-wave state . finally , we have investigated the additional interband pairing amplitudes that appear in the sdw phase as a consequence of the linear coupling to the normal intraband gaps . such anomalous pairing gaps are required to have the same parity under inversion as the normal intraband gaps . specifically , the anomalous pairing amplitudes also manifest a @xmath0 structure and the gap magnitude is similar to the normal pairing gaps . we thank a.v . chubukov , a. kreisel , and s. mukherjee for useful discussions , and w. rowe for her contributions to the early stages of this project . a. t. r. and b. m. a acknowledge support from a lundbeckfond fellowship ( grant a9318 ) . j. h. acknowledges support from nsf - dmr-1005625 . the work of i. e. was supported by the focus program 1458 eisen - pniktide of the dfg , and by the german academic exchange service ( ppp usa no . 57051534 ) . i. e. acknowledges the financial support of the ministry of education and science of the russian federation in the framework of increase competitiveness program of nust misis ( no . 2 - 2014 - 015 ) y. peng , j. meng , d. mou , j. he , l. zhao , y. wu , g. liu , x. dong , s. he , j. zhang , x. wang , q. peng , z. wang , s. zhang , f. yang , c. chen , z. xu , t. k. lee , and x. j. zhou , nat . comms . * 4 * , 2459 ( 2013 ) . m. yu . kuchiev and o.p . sushkov , physica c * 218 * , 197 ( 1993 ) ; v.v . flambaum , m.yu . kuchiev , and o.p . sushkov , physica c * 227 * , 267 ( 1994 ) ; v. i. belinicher , a. l. chernyshev , a. v. dotsenko , and o. p. sushkov , phys . b * 51 * , 6076 ( 1995 ) .

a semiconductor nanowire based superconductor - normal - superconductor ( ) junction is modeled theoretically . a magnetic field is applied along the nanowire axis , parallel to the current . the bogoliubov - de gennes equations for andreev bound states are solved while considering the electronic subbands due to radial confinement in the-section . the energy - versus - phase curves of the andreev bound states shift in phase as the-section quasiparticles with orbital angular momentum couple to the axial field . a similar phase shift is observed in the continuum current of the junction . the quantum mechanical result is shown to reduce to an intuitive , semi - classical model when the andreev approximation holds . numerical calculations of the critical current versus axial field reveal flux - aperiodic oscillations that we identify as a novel form of josephson interference due to this orbital subband effect . this behavior is studied as a function of junction length and chemical potential . finally , we discuss extensions to the model that may be useful for describing realistic devices . [ qcalc_standalone ] the josephson effect is characterized by a current - phase relationship ( cpr ) linking macroscopic current flow to the phase gradient of the superconducting order parameter . the precise form of the cpr for a superconducting weak link depends on intrinsic factors such as junction geometry , material properties , coherence lengths , etc . , in addition to extrinsic variables like temperature and magnetic field . in superconductor - normal - superconductor ( ) junctions in which the-section is long enough to suppress direct tunnelling of cooper pairs , but shorter than the-section phase coherence length , a supercurrent may be carried by quasiparticles undergoing andreev reflection at the- interfaces . planar junctions of width large compared to the-section superconducting coherence length have been studied in great detail ( width refers to the dimension perpendicular to the current ) . these have revealed , for example , fraunhofer oscillations of the critical current with respect to an externally applied out - of - plane magnetic field . for junction widths comparable to the-section coherence length , i.e. the narrow junction limit , this becomes a quasi - gaussian , monotonic decay of the critical current . recently , attention has been given to nanoscale , quasi one - dimensional ( 1d ) junctions , such as those readily engineered by contacting semiconductor nanowires with superconducting leads . gating the semiconducting-section allows for modulating the supercurrent by controlling the chemical potential . the oscillations of the magnetoresistance of a nanowire junction in the voltage - biased state ( i.e. no dc supercurrent ) versus an axial magnetic field have been studied . efforts to realize majorana fermion quasiparticles in 1d semiconductors with strong spin - orbit interaction and proximity coupling to a superconductor have further raised interest in this type of junction . theoretical results have indicated that the behaviour of the critical current in such a junction versus magnetic field and chemical potential can be used to identify topological phases . + previous theoretical descriptions of quasi-1d junctions have not fully considered the effects of nanoscale confinement on the cpr , in particular the implications of orbital angular momentum coupling to an external magnetic flux . here we provide a quantum mechanical description of an idealized junction with a flux applied along the nanowire axis ( parallel to the current ) . for a planar junction , no significant modification of the cpr with an axial flux is expected , as azimuthal motion of the carriers is absent . however , for a cylindrical geometry , azimuthal motion leads to a non - trivial effect which we identify as a previously unstudied form of josephson interference . this is due to the coupling between andreev quasiparticles ( bound states and continuum states ) with orbital angular momentum and the axial flux , which results in phase shifts of the energy - versus - phase for these current carrying states . the total current summed over all channels ( occupied orbitals ) can display interference . in contrast to fraunhofer interference in wide planar junctions , the flux is aligned _ with _ the current and the oscillations are _ not _ periodic in the flux quantum . this effect is only present in nanoscale junctions with lateral dimensions ( i.e. diameter ) smaller than the london penetration depth . this is a regime in which the general theorem of byers and yang does not apply . it is shown that the supercurrent from continuum states also contributes to this interference . for certain junction parameters , the interference effect can dominate the vs characteristics . semiclassically , the effect is intuitively understood by the pickup of a magnetic phase by andreev pairs with an azimuthal velocity component as they cross the junction ballistically . the aim of this paper is to theoretically describe this type of josephson interference in a fully quantum mechanical way . in particular , we are interested in understanding the effect in isolation from the additional complications of real devices , such as non - cylindrical contact geometry , interfacial potential barriers , etc . we consider in the discussion section how to modify the present model to better describe realistic devices . here , we consider the case where the diameter is smaller than the superconducting coherence length in the-section , so that the phase of the order parameter is uniform around the-section circumference in any magnetic field up to the critical field of the leads , . spin - orbit and zeeman effects in the-section ( e.g. relevant to iii - v semiconductor nanowires ) are neglected , and we assume no barriers at the- interfaces . furthermore , we neglect magnetic depairing effects .

1411.3054

[ fig : fig1_schematic ] junction of length @xmath6 . it is modelled as a cylinder with a nanoscale diameter @xmath7 smaller than both the london penetration length and the @xmath2-section phase coherence length . an axial magnetic field @xmath8 penetrates the cylinder . b ) the superconducting order parameter has the magnitude @xmath9 in the @xmath2-section , and is zero in the normal section , with a jumplike variation at the boundaries.,width=326 ] consider an @xmath0 junction created by a semiconducting nanowire contacted by superconducting leads . a cylindrical coordinate system @xmath10 is used , with the nanowire axis along @xmath11 . the junction is modelled as a cylinder of radius @xmath12 . the diameter @xmath13 of the cylinder is assumed smaller than the @xmath2-section london penetration depth @xmath14 , and the @xmath2-section superconducting coherence length @xmath15 . in figure [ fig : fig1_schematic]a , we divide the cylinder into three regions , with region 1 the superconducting section corresponding to the left lead ( @xmath16 ) , region 2 the normal section corresponding to the nanowire ( @xmath17 ) , and region 3 the @xmath2-section corresponding to the right lead ( @xmath18 ) . the leads are connected to bulk superconductors at temperature @xmath19 . figure [ fig : fig1_schematic]b shows the corresponding variation of the superconducting order parameter inside the nanowire . + we assume uniform electrostatic potentials in each section ( i.e. no scattering potential is included and transport is ballistic ) , and no potential barrier at the @xmath2-@xmath1 interfaces . the effective mass of the electron is assumed to have the same value @xmath20 in both @xmath2- and @xmath1-sections so that there is no fermi wavevector mismatch ( fwvm ) at the @xmath2-@xmath1 interfaces . this assumption allows us to use kulik s method @xcite to calculate the supercurrent of the junction . the advantage of this method is that it gives an analytical expression for the bound state energies , which provides us an intuitive way to understand the interference effect and connects our model to an approximate semiclassical picture . we note that inclusion of fwvm or interfacial barriers would not alter the basic mechanism of orbital josephson interference that is demonstrated by this simpler model , but would require use of a more complicated transmission matrix formalism to calculate supercurrents . our expectations for the qualitative effects of barriers are discussed in section [ sec:4_disc ] . + an axial magnetic field @xmath21 penetrates the cylinder . any screening of the magnetic field in the @xmath2-sections is neglected , as we have @xmath22 . in the coulomb gauge , the vector potential is @xmath23 . using the superscript @xmath24 to refer to the three sections of the junction ( with the @xmath1-section corresponding to @xmath25 ) , the single - electron hamiltonian ( excluding the superconducting pairing potential ) in the presence of the magnetic field can be written as : [ eq : h0 ] @xmath26 here , @xmath27 describes the kinetic energy of motion along the axis of the cylinder , @xmath28 the kinetic and magnetic energies of the azimuthal motion around the cylinder , and @xmath29 the radial confining potential of the cylinder in section @xmath30 . + radial confinement results in charge carriers occupying transverse subbands denoted by a pair of quantum numbers . we use the pair @xmath31 in the @xmath1-section , and @xmath32 in the @xmath2-sections , where @xmath33 and @xmath34 are the radial quantum numbers , and @xmath35 the orbital angular momentum quantum numbers . the chemical potential in the cylinder is defined as the energy difference between the bottom edge of the lowest subband and the fermi energy , and is denoted by @xmath36 ( figure [ fig : fig2_subbands ] ) . the numerical calculations were performed using the electron effective mass @xmath37 corresponding to inas . zeeman and spin - orbit effects on the critical current of the junction ( studied in ref . @xcite ) are not considered here , in order to focus on the effects of orbital angular momentum . in section [ sec:4_disc ] , we discuss the conditions under which either the orbital effect or the zeeman + spin orbit effects could be more dominant . + in this paper , we do not write out an explicit form for @xmath29 , and do not solve for the radial wavefunctions corresponding to the subbands @xmath31 in any section of the cylinder . instead , we use a shell conduction model for the @xmath1-section . this is appropriate for certain iii - v nanowires ( such as inas or inn ) , where the charge carriers are typically confined near the surface due to a positive surface potential , forming a surface accumulation layer @xcite . assuming a strong downward surface band bending ( @xmath38 mev ) @xcite , the radial position of the carriers in all subbands @xmath31 is taken to be @xmath39 . this greatly simplifies the calculation of the eigenvalues of @xmath28 ( eq . [ eq : h0]c ) . however , we emphasize that the qualitative results obtained here should not be limited to this shell conduction model , particularly since we find a weak dependence of the interference effect on @xmath40 . + superconductivity in the leads is described by the order parameter ( pairing potential ) @xmath41 , which in the general case , must be calculated self - consistently . we use a simplified model in which @xmath42 is constant , so that proximity effects such as the reduction of @xmath42 near the @xmath2-@xmath1 interfaces due to ` reverse ' proximity are neglected . for @xmath43 , there is a jumplike variation at the boundary of each section ( figure [ fig : fig1_schematic]b ) : @xmath44 outside the radius of the cylinder , the order parameter is zero : @xmath45 when @xmath46 . here , @xmath9 is the superconducting energy gap value in the leads , and @xmath47 is the phase of the superconducting condensate in the left ( right ) lead . the order parameter is zero in the @xmath1-section because of the lack of attractive electron - electron interactions ( repulsive interactions are present in general , but neglected here . ) [ eq : delta ] , a spatially uniform @xmath42 is assumed in the @xmath2-sections in all magnetic fields up to @xmath5 . this is justified if the following two conditions hold : ( i ) the diameter of the cylinder is smaller than the superconducting coherence length in the @xmath2-section , @xmath48 . the change in the phase of the order parameter around the circumference of the cylinder , @xmath49 , is constrained to integer multiples of @xmath50 , because @xmath42 has to be single valued ( i.e. due to fluxoid quantization @xcite ) . when @xmath51 , one can assume @xmath52 , since @xmath15 sets the length - scale for the spatial variations of the order parameter @xcite , so @xmath42 must be uniform . the validity of the assumption @xmath51 in experimental devices is discussed in section [ sec:4_disc ] . ( ii ) the injected current in the @xmath0 junction is much smaller than the critical current of the superconducting leads . otherwise , the superfluid flow in the @xmath2-sections can not be neglected , and a self - consistent determination of @xmath42 is required , as performed in ref . @xcite . here , we assume the critical current of the junction is bottle - necked in the @xmath1-section , which is reasonable given that the critical currents of nanowire junctions are typically small compared to those of the @xmath2 leads . -section ( i.e. the eigenenergies of @xmath53 in eq . [ eq : h0 ] at @xmath54 ) versus the normalized magnetic flux @xmath55 . the @xmath56 and @xmath57 subbands with @xmath58 are shown . the energies are parabolic because of the @xmath59 contribution to the energy by @xmath28 ( see eqs . [ eq : h0]c , [ eq : single - particle - energy]a ) . the upper dashed line is the fermi energy . the chemical potential @xmath36 is the fermi energy measured from the bottom of the lowest subband , @xmath60 ( lower dashed line ) . we have assumed @xmath61 , so the bottoms of the @xmath58 subbands have the same energy . the effective chemical potential for electrons at flux @xmath4 , @xmath62 , is the difference between the subband energy and the fermi energy at that flux . this is shown for the subband @xmath63 at @xmath64 . for the hole - like states ( not shown ) , the subband energies ( eq . [ eq : single - particle - energy]b ) are inverted ( mirrored ) with respect to the @xmath65 line , and the @xmath66 quantum number negated ( @xmath67 ) . the effective chemical potential for holes is given by @xmath68.,width=297 ] we wish to calculate the current - phase relationship ( cpr ) of the junction in the presence of an axial magnetic field . first , the spectrum of discrete levels ( andreev bound states ) in the junction is obtained . next , the current from the discrete levels , as well the continuum " levels with energy @xmath69 is calculated . it is shown that energy eigenstates corresponding to these energy levels ( both the bound states and the continuum states ) follow the single - electron subband structure imposed by the hamiltonian @xmath53 ( eq . [ eq : h0 ] ) . in particular , we show that the cpr is modified by the axial magnetic flux in a way that depends on the orbital angular momentum of the subbands . this leads to a form of josephson interference when one or more subbands with orbital angular momentum are occupied . the wavefunctions of the elementary excitations of the @xmath0 junction are identified as the solutions to the bogoliubov - de gennes @xcite ( bdg ) equations : @xmath70 where @xmath53 is given by eq . [ eq : h0 ] , and @xmath71 and @xmath72 are particle- and hole - like wavefunctions . the asterisk ( @xmath73 ) denotes complex conjugation . + the solution strategy for eq . [ eq : bdg ] starts with finding the solutions to @xmath53 . let us first consider the @xmath1-section . given the simple forms of @xmath27 and @xmath28 in eq . [ eq : h0 ] , it is clear that the single - particle eigenfunctions are plane - waves in the @xmath74 directions of the form @xmath75 . the linear momentum along the axis of the cylinder is given by @xmath76 , and the radial eigenfunction in the subband @xmath31 is given by @xmath77 . as discussed above , we use a shell conduction model for the @xmath1-section , so @xmath78 is not written out explicitly , but assumed to result in a radial position @xmath39 for the carriers . the most general solutions @xmath79 to eq . [ eq : bdg ] are expansions over these single - particle solutions @xcite . however , since @xmath80 in the @xmath1-section , ( @xmath81 ) are good quantum numbers and the single particle energies are given by : [ eq : single - particle - energy ] @xmath82 ( l^2 + \phi ^2 ) \right . \nonumber \\ \ ; & \left . - \varepsilon_{l } + \zeta_{n , l } - \mu \right\rbrace u({\boldsymbol}r),\\ -h_0^ * v({\boldsymbol}r ) = & \left\lbrace- \hbar^2 k^ 2 / ( 2m^ * ) - \left [ \hbar^2/(2m^ * r^2 ) \right ] ( l^2 + \phi ^2 ) \right . \nonumber \\ \ ; & \left . -\varepsilon_{l } - \zeta_{n , l } + \mu \right\rbrace v({\boldsymbol}r ) . \end{aligned}\ ] ] here , @xmath55 is the normalized magnetic flux enclosed by the charge carriers , @xmath83(2l \phi)$ ] , and @xmath84 is the radial confinement energy associated with @xmath77 . the electron subband energies ( i.e. the eigenvalues of @xmath85 at @xmath54 ) are plotted in figure [ fig : fig2_subbands ] , and are parabolic in shape versus the magnetic flux . for the corresponding eigenvalues for @xmath86 , the parabolas are inverted ( mirrored with respect to the fermi energy , @xmath87 ) . this gives @xmath72 its hole - like character : its group velocity @xmath88 is opposite to its wave vector @xmath89 ( i.e. it is retroreflected see ref . @xcite ) . note , however , that for a given @xmath90 , charge is transported in the same direction by the two wavefunctions , as the retroreflected hole has opposite charge to the electron . the eigenvalues associated with @xmath79 are not equal in magnitude , as the term @xmath91 has the same sign in both lines of eq . [ eq : single - particle - energy ] . this follows because of the complex conjugation of the diagonal term on the second row of eq . [ eq : bdg ] , and is a manifestation of the breaking of time - reversal symmetry in the presence of a magnetic field : the retroreflected particle sees the same magnetic field as the incident particle , rather than a time - reversed field @xmath92 . + following the original work of kulik @xcite we calculate the spectrum of the bound states of the nanowire @xmath0 junction , however , here we allow the solutions to carry finite orbital angular momentum . + suppose there is a solution @xmath93 to eq . [ eq : bdg ] , with energy @xmath94 within the gap , @xmath95 . since we assume no fwvm or barriers at the @xmath2-@xmath1 interface , the right- and left - moving solutions @xmath96 can be separated @xcite . we disallow superpositions of @xmath31 subbands in the @xmath1-section . this is justified because : ( i ) we have assumed the ballistic regime , so no scattering - induced subband - mixing occurs , and ( ii ) the pairing potential , eq . [ eq : delta ] , is zero in the @xmath1-section , and so it does not mix the @xmath31 states ( see ref . + in the @xmath2-sections , cooper pairing generally mixes subbands with different radial quantum numbers @xmath34 , but the orbital angular momentum number @xmath66 remains a good quantum number @xcite . the latter follows from the cylindrical symmetry of eq . [ eq : bdg ] , which in turn follows from a cylindrically symmetric @xmath53 and a spatially uniform @xmath42 . for a given quantum number @xmath66 and energy @xmath94 , the most generic single - particle wavefunction in the leads is given by @xmath97 . in each term of the sum , @xmath98 adjusts itself such that the energy of that term is @xmath94 . for @xmath99 considered here , @xmath98 also has an imaginary component , resulting in an exponential decay of the wavefunction inside the @xmath2-section @xcite . in general , there is significant freedom in the choice of the expansion coefficients @xmath100 which will allow matching of the radial wavefunctions in the @xmath1 and @xmath2 sections . + the wavefunctions @xmath96 can be written as : @xmath101 the wavenumbers @xmath102 and @xmath103 represent the momenta of the electron - like and hole - like components in the @xmath1-section , respectively . @xmath104 is the bcs coherence factor in the leads , and @xmath105 is its complex conjugate . @xmath106 are , in general , superpositions of the @xmath107 functions with different @xmath94 , due to the inter - subband mixing induced by @xmath42 . since there is no fwvm or barrier at the @xmath2-@xmath1 interfaces , @xmath96 has to be continuous at @xmath108 . in order for this to be possible , we must have @xmath109 . a solution can always be achieved by a correct choice of the expansion coefficients @xmath100 , so the form given in eq . [ eq : psi_plus ] for @xmath110 is valid . + we now concentrate on the @xmath1-section , and derive the quantization rules for the energies of the bound states . asserting that each term of @xmath110 in the @xmath1-section has energy @xmath94 , the wavenumbers @xmath111 are obtained as a function of energy : [ eq : wavenumbers ] @xmath112 where we have defined an effective chemical potential for an electron - like ( hole - like ) particle in the subband @xmath31 in the @xmath1-section @xmath113 . the minus ( plus ) sign in the parentheses refers to the electron - like ( hole - like ) particle , and @xmath84 is the radial confinement energy due to @xmath78 . the effective chemical potential is the difference between the energy of the subband @xmath31 and the fermi energy at a given magnetic field ( see figure [ fig : fig2_subbands ] ) , and is a positive quantity for any subband that is occupied . + the energy quantization rules can be obtained @xcite by finding the set of coefficients @xmath114 that make eq . [ eq : psi_plus ] continuous at @xmath108 . this is only possible if the following relation holds : @xmath115 where the junction phase @xmath116 enters with a minus sign when for @xmath117 , and a plus sign for @xmath118 . note that the superscript @xmath119 denotes the right- and left - moving solutions , respectively . the complex phase of the right - hand side of eq . [ eq : quantization - rule ] must equal @xmath120 , with @xmath121 etc . since @xmath111 depend explicitly on the bound state energy @xmath94 ( eq . [ eq : wavenumbers ] ) , this results in a quantization rule for @xmath94 , and yields the bound state spectrum . this procedure is carried out in section [ sec:3_results ] to numerically solve for @xmath94 as a function of @xmath122 . + quite generally , if @xmath123 is an eigensolution of the bdg equation ( eq . [ eq : bdg ] ) with energy @xmath94 , then @xmath124 is also an eigensolution@xcite with energy @xmath125 . if @xmath126 is a right - moving solution , then @xmath127 is left - moving , and vice versa . let @xmath128 denote the conjugate of @xmath129 . the wavefunctions @xmath127 and @xmath130 are degenerate at zero field , for all @xmath122 ( figure [ fig : fig3_bs_phase_shift]a ) . this degeneracy is lifted in the presence of the magnetic field , which induces a finite phase shift , as will be discussed below . the pair of solutions ( @xmath131 ) are phase shifted _ together _ in one direction , while the opposite pair ( @xmath132 ) are phase shifted in the _ opposite _ direction ( figure [ fig : fig3_bs_phase_shift]b ) . + deriving an analytical expression for the bound state spectrum by inserting eq . [ eq : wavenumbers ] into eq . [ eq : quantization - rule ] becomes intractable , because of the complicated dependence of @xmath133 on @xmath94 . in order to gain insight into the behaviour of the bound states , we invoke below the well - known andreev approximation @xcite , in which @xmath134 is considered a small quantity compared to @xmath135 and @xmath136 . this approximation is widely used in the literature for a variety of situations @xcite , but can be violated in some regimes of our @xmath0 junction . in particular , when the subband energy is close to the fermi energy , @xmath102 and @xmath103 become small and the assumption @xmath137 is not justified . keeping these restrictions in mind , we now look at how the cpr of the junction is modified in the presence of the axial magnetic field . + the effective chemical potential for electron - like ( hole - like ) particles in the subband @xmath31 in the @xmath1-section can be written @xmath138 , where @xmath139 ( 2l \phi)$ ] enters with a plus sign for electron - like particles . it reflects the coupling of the orbital motion and the axial field . the andreev approximation translates to the following condition : @xmath140 , i.e. the quasi - particle energy and the coupling to the field are small perturbations on the single - particle energies . [ eq : wavenumbers]a , [ eq : wavenumbers]b can be expanded in a taylor series in the powers of @xmath141 . we calculate @xmath133 to first order : @xmath142 where @xmath143 is the velocity of a particle in the subband @xmath31 travelling along the cylinder axis in the @xmath1-section . + by inserting @xmath144 into eq . [ eq : quantization - rule ] and equating the complex phase of the left hand side of eq . [ eq : quantization - rule ] with @xmath120 , where @xmath145 etc . , we obtain the following expression for the spectrum of bound states : @xmath146 where @xmath147 is the healing length @xcite for the subband ( @xmath148 ) , and the energy of the bound state depends on three quantum numbers @xmath149 , and the junction phase difference @xmath122 . @xmath150 refers to the eigenenergy of the right - moving ( left - moving ) solution . note that the energy corresponding to @xmath151 is @xmath152 , for example . + in a short junction , @xmath153 , eq . [ eq : spectrum ] allows only one @xmath154 value per solution , and there are four bound states ( @xmath155 ) per subband @xmath31 . at zero field , there are two positive , and two negative solutions at any given @xmath122 ( figure [ fig : fig3_bs_phase_shift]a ) . for long junctions there are more than four bound state energies per subband , with different @xmath154 numbers @xcite . + the bound state spectrum eq . [ eq : spectrum ] gives , for the case of no magnetic field ( @xmath91 = 0 ) , a result similar to the well known andreev levels of a ballistic @xmath0 junction @xcite , but with a different value of the healing length for each subband . an example is shown in figure [ fig : fig3_bs_phase_shift]a for subband @xmath156 in a short junction with @xmath157 . figure [ fig : fig3_bs_phase_shift]b shows the energy - versus - phase curves of the bound levels at a finite flux , @xmath158 . the curves are now phase shifted by an amount @xmath159 , where @xmath83(2l \phi)$ ] . that is , @xmath160 . the phase shift @xmath161 can be understood semiclassically as the phase picked up by azimuthal travel around the circumference of the cylinder in the presence of the magnetic field . in this picture , for a subband @xmath31 with @xmath162 , the particles ( both electron- and hole - like ) travel in a spiral path as they traverse the junction length @xmath6 . in the shell - conduction model the spiral has radius @xmath40 . the velocity along the axis is @xmath163 , while the azimuthal velocity is @xmath164 . the semiclassical phase @xmath165 is is due to the coupling of @xmath166 and the vector potential @xmath167 , and is calculated from the ginzburg - landau formula for the phase : @xmath168 where the differential element @xmath169 is along the spiral path , @xmath170 is the velocity , @xmath171 is time , and the second integral is taken from @xmath172 corresponding to the particle leaving one @xmath2-section , to @xmath173 , when it arrives at the other @xmath2-section . the result is @xmath174 , which equals @xmath161 . this shows that when the andreev approximation holds and there is shell conduction , the semiclassical result coincides with the quantum mechanical one . note that in the expression for @xmath165 there is no explicit dependence on @xmath40 . the dependence of the phase shift on @xmath40 comes only through @xmath163 , and is a weak dependence within the andreev approximation . more generally , we numerically calculate the energy spectra using eq . [ eq : wavenumbers ] without the andreev approximation and find similar phase shifts that are always proportional to the junction length @xmath6 and to the angular momentum quantum number @xmath66 . + ) , cylindrical @xmath0 junction with no barriers at the @xmath2-@xmath1 interfaces , vs the superconducting phase difference @xmath122 of the leads . two values of normalized magnetic flux , @xmath175 , are shown . we concentrate on one subband @xmath31 with @xmath58 and @xmath176 . a ) zero magnetic field , @xmath177 . the energies correspond to the four allowed wavefunctions ( defined in the main text ) , and the states are pairwise degenerate for all @xmath122 . b ) @xmath158 , where the degeneracies are lifted , and there is a phase shift with opposite directions for each pair of states . the phase shift is small because of the short length of the junction . the following parameters were used for both panels : @xmath178.,width=336 ] the current due to the andreev bound states in the subband ( @xmath148 ) at temperature t is calculated @xcite from the formula @xmath179 where @xmath180 is the fermi - dirac occupation probability of a given energy level ( @xmath181 is the boltzmann constant ) . energies corresponding to both types of wavefunctions @xmath182 and @xmath183 must be inserted into eq . [ eq : current - subband ] . the total bound state current is the sum of supercurrent amplitudes from all occupied subbands @xcite ( ` open channels ' in the language of ref . @xcite ) : @xmath184 the continuous spectrum of states with energies @xmath69 also contributes to the junction current . a continuum level can be viewed as a leaky " solution @xcite to the andreev bound state problem described above , with a complex - valued eigenenergy @xmath185 , with @xmath186 . the leaky level follows the same subband structure as the andreev bound states . the imaginary component of energy results in a finite lifetime for the continuum level , reducing its contribution to the junction current , but for a long junction @xmath187 , this contribution is significant , and can not be ignored @xcite . + the continuum current due to the subband @xmath31 , @xmath188 , is calculated using the transmission formalism @xcite . we calculate the transmission coefficients for the electrical currents carried by electron - like and hole - like excitations incident on the @xmath2-@xmath1 interfaces , resulting in leaky solutions in the @xmath1-section . the details of the calculation are given in appendix [ sec : appendix_cont ] . for the rest of this section the subscripts @xmath148 are dropped for the sake of simplicity ; it is implicitly assumed that all quantities pertain to the subband @xmath31 . the result for the continuum current is : @xmath189 where @xmath94 is the real part of the energy of the continuum level and @xmath190 is the fermi - dirac distribution at temperature @xmath19 , and @xmath191 are real - valued bcs coherence factors : @xmath192 [ eq : coherence - factors ] the functions @xmath193 depend on energy , through the wavenumbers @xmath194 and @xmath195 ( eq . [ eq : wavenumbers ] ) as well as the coherence factors @xmath191 : @xmath196 $ ] . equation [ eq : continuum - current ] can be intuitively understood in terms of the leaky solutions to the bdg equation : the terms containing @xmath197 pertain to the contribution of leaky states of type @xmath198 , while those containing @xmath199 pertain to @xmath200 . the junction phase @xmath122 enters with a plus ( minus ) sign for left ( right ) moving solutions . at zero magnetic field , @xmath201 . this is analogous to the degeneracy of @xmath202 in figure [ fig : fig3_bs_phase_shift]a . therefore , eq . [ eq : continuum - current ] reduces to eq . 17 in ref . @xcite , up to an application of the andreev approximation . + in the presence of the magnetic field , the terms @xmath203 shift the functions @xmath193 in phase relative to the zero - field case , in the same manner as the phase shifts found previously for the bound states . employing the andreev approximation ( eq . [ eq : wavenum - approx ] ) , we obtain @xmath204 , \label{eq : d_cont_approx}\ ] ] which is shifted in phase with respect to the zero - field case . explicitly for the subband @xmath31 we have @xmath205 , with the phase shift @xmath161 defined previously . + the total continuum current of the junction is @xmath206 the critical current @xmath3 of the junction is defined as the maximum of total bound state + continuum currents with respect to @xmath122 : @xmath207 . \label{eq : current - critical}\ ] ] we numerically solve the continuum and bound state currents of an @xmath0 junction at finite magnetic fields , using the shell conduction approximation with the shell at a radius @xmath208 . temperature is set to @xmath209 in all calculations . from this point on , only @xmath31 subbands with @xmath58 are assumed to be occupied in the @xmath1-section . the andreev approximation is not used in calculating the cpr . the critical current of the junction is calculated from eq . [ eq : current - critical ] , and its behaviour versus axial magnetic flux @xmath210 is studied . note that our assumption of no barriers at the @xmath2-@xmath1 interfaces also implies full andreev reflection . , continuum current @xmath211 , and the sum @xmath212 for the subband @xmath156 vs the superconducting phase @xmath122 at zero magnetic field , @xmath177 . since the junction is long , @xmath213 , the cpr is triangular . the kinks in @xmath214 at @xmath215 are due to andreev bound states crossing the gap edge into the continuum levels , but the total subband current is a smooth function of @xmath122 . b ) total subband current @xmath216 vs the superconducting phase as a function of the normalized axial magnetic flux @xmath4 . at zero flux the maximal value occurs near @xmath217 . at finite flux , the bound state and continuum currents are phase shifted . for @xmath218 , two discontinuities can be seen in @xmath216 because of the phase shifts , and the maximal value no longer occurs near @xmath217 . at @xmath219 , the phase shifts amount to @xmath220 and the zero - field curve is recovered . the maximal current at this flux is slightly smaller than the zero field case , due to a decrease in the average momentum of the andreev quasiparticles with increasing flux . the following parameters were used in both panels : @xmath221.,width=336 ] as an illuminating example we study a 500 nm long junction with a chemical potential of @xmath222 . this value for @xmath36 is chosen because it allows @xmath223 subbands to be occupied at @xmath177 ; at @xmath64 the @xmath224 subbands depopulate . , see figure [ fig : fig2_subbands ] . in this example , as @xmath4 approaches 1 , @xmath225 go to zero , at which point andreev quasiparticles can no longer be supported by the @xmath226 subbands . ] in this section we concentrate on the cpr obtained for one subband , namely @xmath227 , and discuss how the coupling of the finite angular momentum with the axial field modifies the subband cpr . since the junction length is greater than the healing length of the populated subbands ( @xmath228 ) , the long junction limit applies . many bound states are present in the junction ( @xmath229 ) . in figure [ fig : fig4_currents]a we show the bound state current @xmath230 , continuum current @xmath211 , and their sum , for the subband @xmath156 at zero magnetic field . as expected of a long junction @xcite , @xmath230 and @xmath211 are of the same order of magnitude , and the cpr is triangular in shape . an additional group of 4 bound states ( @xmath231 ) appear in the junction at @xmath232 , and exit at @xmath233 , giving rise to discontinuities in @xmath230 , @xmath211 . however , the total subband current @xmath216 is always a smooth function of @xmath122 . it is maximal near @xmath217 , ( exactly at @xmath217 at zero temperature ) regardless of the junction length @xcite . note that the continuum current is zero at @xmath217 . + in figure [ fig : fig4_currents]b we show the subband current @xmath216 as a function of the magnetic flux . as the flux is increased from zero , two discontinuities develop in the subband current ( shown for @xmath234 ) . the bound state current is modified as the eigenenergies corresponding to states @xmath235 are shifted in phase in the opposite direction to those of states @xmath236 , similarly to figure [ fig : fig3_bs_phase_shift]b . an equivalent process happens for the continuum current , as explained in eq . [ eq : d_cont_approx ] . as a result , the subband current @xmath216 also shows two discontinuities , and is no longer necessarily maximal near @xmath217 . + the amounts of the phase shifts in @xmath230 and @xmath211 depend on the quantity @xmath237 . the wavenumbers @xmath111 are subband parameters defined in eq . [ eq : wavenumbers ] , and the length @xmath6 is device dependent . therefore , the fluxes at which phase shifts equal integer multiples of @xmath220 need not occur at integer multiples of @xmath238 or @xmath239 ; they can occur at any value of @xmath4 . an example is shown in figure [ fig : fig4_currents]b for @xmath219 , where the phase shifts equal @xmath220 and the cpr recovers its shape at @xmath177 . notice , however , that the maximal value of the subband current at @xmath219 is smaller than at @xmath177 . this can be intuitively understood as follows : as the flux increases , the effective wavenumbers of the electrons and holes change according to eq . [ eq : wavenumbers ] . it can be seen that the average momentum of the electron - hole pair and therefore the healing length @xmath240 are always smaller for higher fluxes . since the magnitude of josephson current in a long junction scales approximately linearly @xcite with @xmath241 , it is suppressed at higher fluxes . this suppression is stronger near the depopulation point of a given subband , where the average momentum decreases significantly . + in order to elucidate the mechanism of the josephson interference between subbands , we show in figure [ fig : fig5_slices_8p5]a the critical current versus axial flux of the junction studied in section [ sec:31_one_subband ] . the length @xmath242 is chosen because it allows for a relatively large amount of phase pickup , since the phase pickup is proportional to the length of the junction ( @xmath237 in eq . [ eq : quantization - rule ] ) . hence , several oscillations of the critical current occur prior to the depopulation of the @xmath226 subbands at @xmath64 . the supercurrent of each subband is shown versus the phase difference @xmath122 in figure [ fig : fig5_slices_8p5 ] panels b - f , at particular values of the magnetic flux . the @xmath176 subband current equals that of @xmath243 at all fluxes . at zero flux the current of each subband is maximal near @xmath217 ( exactly @xmath217 for zero temperature ) , as discussed in section [ sec:31_one_subband ] . this can be clearly seen in panel b. the total current of the junction is the sum of the contributions from the @xmath223 subbands , and is therefore roughly three times the contribution of each . the dotted vertical lines show the phase at which the critical current occurs . + as the flux is increased , the crp of the @xmath224 subbands are modified , similarly to figure [ fig : fig4_currents]a . the critical current of the junction decreases , since the @xmath244 subbands no longer interfere constructively ( panel c ) . at @xmath245 the maximal current switches from a phase @xmath246 to @xmath247 , as shown in panel d. this is a feature of the triangular cpr . the critical current increases until @xmath248 ( panel e ) , at which point the junction current is maximal near @xmath249 . we call this a peak a secondary peak as it occurs roughly in the middle of the main period of oscillations ( see discussion below on periodicity ) , when the magnetic phase pickup of the @xmath250 subbands equals roughly @xmath251 . the magnitude of this peak is roughly two thirds the total current as zero field , as the @xmath252 subbands contribute maximally , and the @xmath56 subband current is close to zero . the process reverses itself for @xmath253 , until at @xmath219 the phase pickup of the @xmath250 subbands equals @xmath220 and all subbands interfere constructively again ( panel f ) . we refer to the peak at @xmath219 a primary peak . other primary peaks occur at @xmath254 . as in figure [ fig : fig4_currents ] , the contribution of the @xmath250 subbands decrease as @xmath4 is increased , because the decrease in the average quasiparticle momentum . this is the mechanism behind the slow decay of the magnitude of the primary peaks as flux increases . + _ aperiodicity _ the @xmath255 subbands interfere constructively when the magnetic phase pickup of the @xmath250 subbands equals an integer multiple of @xmath220 . this corresponds to the main period of the critical current oscillations with @xmath4 , which we estimate below . in other words , we want to find @xmath4 such that @xmath256 for integer @xmath257 . the wavenumbers @xmath111 are defined in eq . [ eq : wavenumbers ] . for the general case this is not easy to do analytically , as @xmath111 themselves depend on the flux through the effective chemical potentials @xmath258 . however , we can get an estimate of the expected period by invoking the andreev approximation ( eq . [ eq : wavenum - approx ] ) , and assuming the effective fermi velocities do not depend on the flux , so can be evaluated at some fixed @xmath4 , e.g. @xmath177 . these are reasonable assumptions when the flux is much smaller than the depopulation point of a given subband ( e.g. , @xmath64 for @xmath224 subbands in the example of figure [ fig : fig5_slices_8p5 ] ) . the result for the position of the first primary peak @xmath259 is : @xmath260 where @xmath163 is defined below eq . [ eq : wavenum - approx ] . for the example of figure [ fig : fig5_slices_8p5 ] , this evaluates to @xmath261 , deviating from the numerically calculated value by only @xmath262 . however , we see that the numerical positions of the next primary peaks at @xmath263 , @xmath264 can not be accurately approximated as integer multiples of @xmath259 : the period becomes shorter as @xmath4 is increased . this is because the field dependence of @xmath163 can not be ignored at higher values of @xmath4 , and the andreev approximation breaks down . intuitively , the effective fermi velocity is noticeably lower at higher fields , resulting in more time spent in the junction by the andreev pair and more phase pickup , therefore a smaller period for @xmath3 oscillations . consequently , even in the simple case of a few subbands , the oscillations of @xmath3 versus @xmath4 are not strictly periodic . + a higher chemical potential results in the occupation of a greater number of angular momentum subbands . the rich interplay between the different @xmath66-subbands results in a complex pattern of the oscillation of @xmath3 with @xmath4 . assume subbands with @xmath265 up to @xmath266 are occupied . each @xmath265 subband s supercurrent oscillates with a flux dependent ` period ' approximated by eq . [ eq : period_andreev ] , which depends on the subband velocity @xmath163 , and is therefore generally anharmonic with other subbands . typically , a peak in @xmath3 as a function of @xmath4 occurs under one of two circumstances : ( i ) when some ( at least two ) of the subbands with different @xmath265 values interfere constructively , or ( ii ) when the critical current occurs near @xmath267 , a secondary peak is formed as described in figure [ fig : fig5_slices_8p5 ] . there are @xmath268 choices for pairs of constructively interfering subbands ( parentheses indicate the binomial coefficient ) . as each @xmath265 subband s oscillations can be anharmonic with those of all other subbands , it follows that there are @xmath268 different flux - dependant ` periods ' in the @xmath3 oscillations due to condition ( i ) , with another @xmath266 due to condition ( ii ) . the @xmath3 curves therefore can display complex , aperiodic structures . + in figure [ fig : fig6_slices_20]a we plot an example of an @xmath3 versus @xmath4 curve for a junction with the same parameters as that of figure [ fig : fig5_slices_8p5 ] , except the chemical potential is raised from 8.5 mev to 20 mev . at zero flux , subbands up to @xmath269 are occupied . the @xmath270 states depopulate at @xmath271 , and the @xmath272 states at @xmath273 . as an example of configurations that can give rise to a peak in @xmath3 , subband cprs are shown in figure [ fig : fig6_slices_20 ] panels b - f , for 5 peaks indicated in panel a with vertical dotted lines . the peak at @xmath274 ( panel b ) satisfies condition ( ii ) , while the other examples are due to constructive interference of two subbands , i.e. condition ( i ) : subbands with @xmath244 at @xmath275 ( panel c ) , @xmath276 at @xmath277 ( panel e ) , and @xmath278 at @xmath279 and @xmath280 ( panels d , f ) . , for @xmath222 ( panels a - d ) and @xmath281 ( panels e - h ) , and several values of the junction length @xmath6 . the following parameters were used : @xmath282 . in panels a - d , the vertical line at @xmath283 indicates the depopulation of the @xmath224 subbands . similarly in panels e - h the vertical lines at @xmath284 indicate the depopulation of the @xmath285 subbands , respectively.,width=336 ] we now discuss the effect of the junction length @xmath6 on the pattern of @xmath3 oscillations . the left column in figure [ fig : fig7_res ] ( panels a - d ) shows the numerically obtained @xmath3 versus @xmath4 for a junction with @xmath222 , as @xmath6 is varied . all other junction parameters are the same as in figure [ fig : fig5_slices_8p5 ] . the critical current is normalized to its value at zero magnetic flux . two processes affect the behaviour of @xmath3 : ( i ) the depopulation of the @xmath224 subbands at @xmath64 , shown by vertical dotted lines in panels a - d , and ( ii ) the josephson interference effect described above . the first of these processes results in step - like discontinuities in @xmath3 at @xmath283 , a drop to roughly one - third of the zero - field @xmath3 value as the @xmath286 subbands depopulate . since the @xmath287 subband does not couple to the axial flux , @xmath3 is almost constant above @xmath64 . the slow decay of the primary peak heights below @xmath64 , and of @xmath3 above @xmath64 , are both due to the decreasing average momentum at higher fluxes , as discussed in section [ sec:31_one_subband ] . for a short junction with @xmath288 ( panel a ) and for @xmath289 , the phase shifts in the cpr of the @xmath224 subband are small ( the phase shifts are dependent on @xmath237 ) , but become more significant close to @xmath64 , where the axial velocities of the quasiparticles are smaller and the time of flight across the junction longer . this results in the observed decrease in @xmath3 before the step - like discontinuity . as @xmath6 is increased to 50 nm , the value of this phase shift increases and modulation due to interference starts to emerge . for @xmath290 ( figure [ fig : fig7_res ] panel c ) , the first primary peak occurs at @xmath291 , and for @xmath242 ( panel d ) at @xmath219 . the decrease by a factor of @xmath292 in the position of the first primary peak is approximately equal ( within @xmath293 ) to the reciprocal ratio of lengths @xmath6 , as expected from the estimate in eq . [ eq : period_andreev ] . + the right column of figure [ fig : fig7_res ] ( panels e - h ) shows the @xmath3 versus @xmath4 curves for a junction with @xmath294 . the subbands with @xmath285 depopulate at @xmath295 respectively ( vertical dotted lines in panels e - h ) , resulting in step - like discontinuities in @xmath3 at those flux values . similarly to above , the behaviour of @xmath3 is dominated by this effect for a @xmath296 junction ( panel e ) , but as @xmath6 is increased the josephson interference becomes visible ( panels f - h ) . note that between the flux values @xmath273 and @xmath297 the @xmath3 curves in the right column of figure [ fig : fig7_res ] are qualitatively similar to their counterparts in the left column , because similarly to the @xmath222 case , subbands with @xmath298 are occupied within this flux window . note that for @xmath281 , @xmath299 is much larger than for the @xmath222 case ; the oscillation amplitude appears to be smaller in panels ( e - h ) only because the relative contribution of each subband to the total current is smaller when there are more subbands occupied . + in summary , for a short junction , the depopulation of subbands is more visible than the josephson interference effect . however , as the junction length is increased , the interference effect becomes apparent . the periods of oscillation decrease slightly as flux is increased . for a junction with a low chemical potential ( only a few transverse subbands occupied , e.g. an @xmath0 point - contact @xcite ) , the pattern of @xmath3 modulation is simpler and the period longer than the case of a high chemical potential ( many transverse subbands occupied ) . a long , low-@xmath36 junction is optimal for experimental observation of this josephson interference effect . we have described the theory of a previously unstudied form of the josephson interference effect that can occur in nanoscale @xmath0 junctions due to the coupling of the orbital angular momentum of transverse electronic subbands with an axial magnetic flux . we found in section [ sec : effect_of_length ] the regimes in which this interference effect dominates the @xmath3 vs @xmath4 characteristics of the junction . an idealized model of an @xmath0 junction was used , with several simplifying assumptions , in order to elucidate the mechanism of the effect . we discuss generalizations of the model below , in particular those modifications that may be necessary to directly model experimental devices . + _ fwvm and barriers at the interfaces _ no barriers were assumed at the @xmath2-@xmath1 interfaces , and fwvm was neglected . the effective mass for electrons @xmath20 was assumed to be uniform throughout the junction . these assumptions allowed kulik s method of matching the wavefunctions at the interfaces to be used to calculate the bound state and the continuum currents . we stress that the basic mechanism of the orbital interference effect , i.e. the modification of the @xmath1-section wavenumbers in the presence of the axial field , is independent of fwvm and interfacial barriers . therefore , the main features of the @xmath3 oscillations ( periodicity , amplitude ) should only be modified by fwvm and barriers as higher order corrections . the exact shape of the junction cpr and the @xmath3 vs @xmath4 curves , however , depends on the details of the interfaces . accurate modeling of experiments must take this into account , based on the material and interfacial properties specific to a particular experimental implementation . + as was discovered in studies of andreev reflection at nb - inas interfaces @xcite , fwvm modifies the cpr of superconductor / semiconductor / superconductor junctions . in the general case , where fwvm and barriers are included at the @xmath2-@xmath1 interfaces , the bound state energies and the continuum current must be calculated from the transmission matrix formalism @xcite , in which the transmission matrix includes a normal ( specular ) reflection coefficient as well as an andreev ( retro ) reflection coefficient . the values of these coefficients depend on the material details of the junction . the cprs of junctions with fwvm @xcite and barriers @xcite have been previously calculated ( numerically ) at zero magnetic field . the effect of both mechanisms is to make the cpr more closely resemble a sinusoidal curve . since the @xmath1-section wavenumbers have the same coupling to the axial field shown in eq . [ eq : wavenumbers ] regardless of fwvm or barriers , we expect the phase - shift mechanism leading to interference to remain . however , the shape of the oscillations in @xmath3 should appear more sinusoidal , following the shape of the cpr . + normal reflection of the quasiparticle wavefunctions from the @xmath2-@xmath1 barriers will result in a higher order correction to the periodicity of the interference effect . this is due to an increased average phase pickup as the quasiparticle spends more time in the junction . on the other hand , this should also lead to some randomization of the phase . the former would lead to shorter period oscillations ( i.e the effect occurs at lower fields ) , while the latter will partially smear out the interference effect , reducing the amplitude of oscillations . similar effects are expected from elastic backscattering occurring in the @xmath1-section in non - ballistic junctions . are beyond the scope of this paper , and are left to future work . + _ general form of @xmath42 _ a spatially uniform pairing potential was assumed in the @xmath2-sections at all magnetic fields ( eq . [ eq : delta ] ) . this was justified by assuming cylindrical @xmath2-sections , and restricting the cylinder diameter to be smaller than the superconducting coherence length in the @xmath2-sections . however , experimental fabrication of nanoscale @xmath0 junctions is usually done by evaporating or sputtering metallic ( e.g. al or nb ) thin film contacts onto a semiconducting nanowire . in this case , the geometry of the @xmath2-section is not cylindrical but @xmath300 shaped . the lack of cylindrical symmetry necessitates , in principle , a 3-dimensional numerical calculation of @xmath301 using self - consistent methods . however , as long as the @xmath1-section can be assumed to be cylindrically symmetric , our model should closely approximate the experimental situation . this is because the interference effect depends mainly on the eigensolutions in the @xmath1-section , particularly the orbital angular momentum states and their coupling to the flux . the details of the eigensolutions in the @xmath2-section do not play a direct role , other than asserting the form of the wavefunction ansatz ( eq . [ eq : psi_plus ] ) is valid . similarly , the axial field can induce a non - uniformity in @xmath42 . if the thickness of the metallic film becomes larger than the @xmath2-section coherence length , fluxoid quantization can result in a @xmath302-dependent phase for @xmath42 in the presence of the field . inserting a @xmath302-dependent @xmath42 in the bdg equations ( eq . [ eq : bdg ] ) will affect the bound state solutions , likely requiring a 3-dimensional numerical solution . however , we still expect the interference effect to depend mainly on the states in the @xmath1-section . + _ general radial wavefunctions _ the shell - conduction model was used in order to simplify computations , and to help gain intuitive insight into the problem ; it is not strictly necessary for the main arguments of the paper . indeed , we found in section [ sec : semiclassical ] that the semiclassical phase shift @xmath165 is only weakly dependent on the radius @xmath40 , so the interference effect should be present for general radial wavefunctions . in future work , the radial wavefunctions in the @xmath1-section , @xmath78 , and the corresponding single particle energies in the presence of the field will be numerically calculated , yielding the appropriate wavenumbers @xmath111 for electron- and hole - like solutions . we expect the term @xmath237 appearing in eqs . [ eq : quantization - rule ] , [ eq : continuum - current ] will continue to result phase shifts similar to those seen in the present model . + _ zeeman and spin - orbit effects _ in order to study the the orbital josephson interference effect in isolation , zeeman and spin - orbit effects were neglected in our analysis . it is useful to ask under what circumstances should the orbital effects or the zeeman + spin - orbit effects dominate ? the critical current of a short , insb @xmath0 junction , including spin - orbit and zeeman effects , was studied in ref.@xcite . the bound state energies were solved , and a phase - shift was observed in the energy - versus - phase curves due to the zeeman effect . similar to the mechanism described in our analysis , the zeeman effect modifies the @xmath1-section wavenumbers , but based on the spin state rather than the orbital state . this results in an oscillation of @xmath3 , with the first minimum occurring at @xmath303 , where @xmath304 is the fermi velocity , @xmath305 is the effective land g - factor , and @xmath306 is the bohr magneton . for a 200 nm insb junction with @xmath307 , this evaluates to @xmath308 t for @xmath304 corresponding to @xmath309 mev , but for other materials @xmath310 is typically much larger . considering inas with a moderately large @xmath311 gives @xmath312 t ( again for @xmath313 nm and @xmath309 mev ) . since we used the effective mass for inas in our calculations , we can directly compare with the results of figure [ fig : fig7_res]c for a 200 nm long junction . the orbital effect should dominate in this case , as the first minimum of @xmath3 is at a flux corresponding to @xmath314 t. the consequence of the inclusion of the spin - orbit coupling is a smaller correction : the so called anomalous josephson effect , in which the current is no longer an even function of the superconducting phase : @xmath315 . + we conclude that for insb devices , the zeeman effect could easily dominate . this is especially true if either of the following conditions hold : ( i ) if only @xmath316 subbands are occupied ( i.e. small chemical potential ) , or ( ii ) in a perpendicular field , where the zeeman effect is present but the orbital effect is suppressed . on the other hand , for most low spin - orbit semiconductor materials , @xmath317 and we would expect the orbital subband effect to dominate in an axial field experiment , unless only @xmath287 subbands are occupied . + _ magnetic depairing _ field - induced depairing suppresses both superconductivity in the @xmath2-sections and proximity superconductivity in the @xmath1-section @xcite . for a type - ii superconductor with a relatively large gap such as nb , we can assume depairing in the @xmath1-section should dominate . for diffusive junctions in the narrow junction limit , the usadel equations predict a monotonic gaussian decay @xmath318 for a perpendicular magnetic flux @xmath4 through the @xmath1-section @xcite , where @xmath319 is a numerical constant . a similar effect should apply to the axial field case , except with a slower magnetic field decay due to a smaller cross - sectional area ( smaller flux ) . this is expected to produce a monotonic decay envelope superimposed on the critical current oscillations , and should be taken into account when modeling experimental data . + _ summary _ the idealized model studied here serves to demonstrate a novel form of josephson interference due to orbital angular momentum states , with the unusual property of flux - aperiodic oscillations . extensions to the model discussed above , most importantly fwvm and barriers at the @xmath2-@xmath1 interfaces , will be useful for describing experimental implementations . we thank a. j. leggett and s. frolov for helpful discussions . this work was supported by nserc and the ontario ministry for research and innovation . the electrical current transmission amplitudes for the continuum states are obtained by matching the solutions of the bdg equation ( eq . [ eq : bdg ] ) in the three regions of the cylinder ( @xmath320 ) while assuming an incident source term " on the @xmath2-@xmath1 interface at @xmath321 with energy @xmath69 . these transmission amplitudes are then used to calculate the continuum current . this section follows appendices a , b in ref . @xcite , but is generalized to account for finite magnetic field and orbital angular momenta . + hereafter all quantities are presumed to pertain to one subband , @xmath31 , unless explicitly stated otherwise . we drop the subscripts @xmath148 for simplicity . consider the quasiparticle excitation spectrum of the bdg hamiltonian . in the left @xmath2-section ( @xmath322 ) , the electron - like solutions with energy @xmath323 are given by @xmath324 and the hole - like solutions by @xmath325 here , @xmath191 are given in eq . [ eq : coherence - factors ] , and @xmath326 have the form given above eq . [ eq : psi_plus ] , with different expansion coefficients @xmath327 for the electron- and hole - like wavefunctions . we do not reproduce these coefficients here as they do not enter the calculation . let the source term @xmath328 be incident from the left on the @xmath2-@xmath1 interface at @xmath321 . the wavefunctions generated due to this source term are grouped into two categories ( depending on which @xmath2-section they belong to ) , with coefficients @xmath329 , following bagwell s notation @xcite . the first group pertains to the left @xmath2-section . the electron - like source term incident on the @xmath2-@xmath1 interface is @xmath330 the andreev reflected hole - like wavefunction is @xmath331 for the electrons in the normal region , the wavefunction is @xmath332 and for the holes @xmath333 here , @xmath334 is radial part of the wavefunction in the @xmath1-section , defined below eq . [ eq : bdg ] . the explicit energy dependence of the wavenumbers @xmath335 is given in eq . [ eq : wavenumbers ] . + the transmitted wavefunction into the right contact is @xmath336 this is supported by electrons in the @xmath1-section with the wavefunction @xmath337 and holes @xmath338 the normal reflection processes have not been considered , as no fwvm or barriers are assumed at the @xmath2-@xmath1 interfaces . + the coefficients @xmath329 can be found by connecting together eqs . [ eq : app_elec_left ] , [ eq : app_elec_right ] , and eqs . [ eq : app_hole_left ] , [ eq : app_hole_right ] at any point @xmath339 within the @xmath1-section or at the @xmath2-@xmath1 interfaces . we use @xmath340 . the result is @xmath341 where @xmath116 . by definition , @xmath342 is the transmission amplitude from the left contact to the right contact due to an electron - like incident source term with energy @xmath343 . the corresponding transmission coefficient @xmath344 is @xmath345 with the function @xmath197 given below eq . [ eq : coherence - factors ] . + the coefficient for transmission from right to left ( due to a left - moving source ) can be found by making the transformation @xmath346 in the above formula . that is , @xmath347 repeating these calculations for an electron - like source term with energy @xmath348 shows that eqs . [ eq : app_t_lr ] , [ eq : app_t_rl ] give the correct transmission coefficients for the negative energy case as well . + similar to the case of the bound states , if @xmath349 is a solution at energy @xmath94 , then @xmath350 gives a solution at energy @xmath125 . both types of solutions must be taken into account when calculating the total transmission coefficients . consider the ( left - moving ) source term @xmath351 obtained by applying the above transformation on eq . [ eq : app_psi_e ] : @xmath352 all relevant wavefunctions due to this source term ( i.e. the andreev reflected , @xmath1-section electron- and hole - like , and transmitted wavefunctions ) can be constructed by applying the transformation @xmath353 to eqs . [ eq : app_andreev_wf][eq : app_hole_right ] . crucially , the resulting wavefunctions contain the wavenumbers @xmath354 , while having energy @xmath125 . repeating the above calculation , the transmission coefficient @xmath355 due to the source term @xmath351 is found : @xmath356 or equivalently , @xmath357 with @xmath358 . + the current @xmath359 due to electron - like excitation source terms is calculated @xcite using the formula @xmath360 f(e ) \mathrm{d } e. \label{eq : app_jnl}\end{aligned}\ ] ] here , @xmath190 is the fermi - dirac distribution at temperature @xmath19 . it can be seen that eq . [ eq : app_jnl ] is equal to eq . [ eq : continuum - current ] . in deriving eq . [ eq : app_jnl ] we used only the electrical current transmitted due to electron - like source terms . repeating the above calculations for hole - like source terms results in a current @xmath361 which is equal to @xmath359 . naively , one might then think the total continuum current obtained is too large by a factor of two . however , note that the density of excitations ( electron- plus hole - like ) in the @xmath2-section is twice as large as the density of states in the @xmath1-section . consequently , the subband s continuum current is @xmath362 and we recover eq . [ eq : continuum - current ] . this disparity in the density of states in the @xmath2- and @xmath1-sections was noted in ref . @xcite , see eq . b8 in that paper . + equation [ eq : app_jnl ] is used in section [ sec:3_results ] to numerically calculate the continuum current @xmath363 due to each subband . for a junction much shorter than the subband s healing length , @xmath364 , the bound state current @xmath365 is much larger than @xmath363 . when @xmath366 , @xmath230 and @xmath211 are of the same order of magnitude , and we obtain a triangular cpr . at finite magnetic fields , phase shifts appear in both @xmath365 and @xmath363 as described in section [ sec:31_one_subband ] , but the cpr retains its triangular shape . + notice that at zero magnetic field , we have @xmath367 , so @xmath363 simplifies and can be written as @xmath368 f(e ) \mathrm{d } e.\end{aligned}\ ] ] the extra factor of 2 here is usually attributed to the spin degree of freedom of the electrons / holes . by including both types of coefficients @xmath369 in eq . [ eq : app_jnl ] we are taking into account this degree of freedom . this can be further elucidated by noting that in the spinful version of this problem , the particle - hole symmetry is manifested as two ( nambu - spinor type ) solutions @xmath370 and @xmath371 with opposite energies , spins , and coherence factors ( the pauli matrices @xmath372 act on the spin and particle - hole manifolds , respectively ) . that is , @xmath373 and @xmath350 generalize to states of opposite spin , and the spin degree of freedom is correctly accounted for by considering both types of solutions . the effective chemical potentials follow @xmath374 , see figure [ fig : fig2_subbands ] . in this example , as @xmath375 approaches 1 , @xmath376 go to zero , at which point andreev quasiparticles can no longer be supported by the @xmath226 subbands .

we present the results obtained from an objective search for stellar clusters , both in the currently active nuclear starburst region , and in the post - starburst disk of m82 . images obtained with the advanced camera for surveys ( acs ) aboard the _ hubble space telescope _ ( hst ) in f435w(b ) , f555w(v ) , and f814w(i ) filters were used in the search for the clusters . we detected 653 clusters of which 393 are located outside the central 450 pc in the post - starburst disk of m82 . the luminosity function of the detected clusters show an apparent turnover at b=22 mag ( ) , which we interpret from monte carlo simulations as due to incompleteness in the detection of faint clusters , rather than an intrinsic log - normal distribution . we derived a photometric mass of every detected cluster from models of simple stellar populations assuming a mean age of either an 8 ( nuclear clusters ) or 100 ( disk clusters ) million years old . the mass functions of the disk ( older ) and the nuclear ( younger ) clusters follow power - laws , the former being marginally flatter ( ) than the latter ( ) . the distribution of sizes ( full width at half maximum ) of clusters brighter than the apparent turn - over magnitude ( mass ) can be described by a log - normal function . this function peaks at 10 pc for clusters more massive than , whereas for lower masses , the peak is marginally shifted to larger values for the younger , and smaller values for the older clusters . the observed trend towards flattening of the mass function with age , together with an over - abundance of older compact clusters , imply that cluster disruption in m82 is both dependent on the mass and size of the clusters .

0802.1922

the high spatial resolution of the hubble space telescope ( hst ) observations have allowed the detection of compact star clusters in starburst regions of several galaxies @xcite . the most massive of them , often referred to as super star clusters ( sscs ) , have masses and sizes comparable to those of galactic globular clusters ( gcs ) . the sscs span a wide range of ages : those associated with currently active star forming regions are as young as a few million years old , while the others are as old as 500 myr @xcite . ever since their discovery by the hst - based observations , they have been thought to be progenitors of gcs . there is little doubt about the survival of the most massive of the sscs over a hubble time , but an interesting question is what fraction of them will survive ? some star clusters are relatively weakly bound objects , and vulnerable to disruption by a variety of processes that operate on three different timescales @xcite . on short timescales ( @xmath6 yr ) , when the clusters and protoclusters are partly gaseous , the exploding supernovae and the resulting superwinds remove gaseous mass from clusters leading to cluster expansion and disruption , a process popularly dubbed as infant mortality . on intermediate timescales ( @xmath7 yr ) , the mass - loss from evolving stars leads to the disruption of the clusters . on even longer timescales ( @xmath8 yr ) , stellar dynamical processes , especially evaporation due to two - body scattering , and tidal effects on a cluster as it orbits around the galaxy , known as gravitational shocks , come into play in the removal of stellar mass from clusters . the first two processes are directly related to the evolution of massive and intermediate - age stars , and hence their timescales are less uncertain . on the other hand , the exact timescale for the gravitational shocks is difficult to calculate given that it depends on factors external to the cluster , such as a detailed modeling of the local gravitational potential . another external process that can play a role in disrupting clusters is their encounters with giant molecular clouds ( gmcs ) @xcite . the timescale for this process depends on the molecular gas content and is typically less than that for the gravitational shocks in gas - rich environments . the gcs have already gone through these processes , whereas the sscs are at the right age to see the different disruption processes at work . how does the various cluster disruption processes affect the cluster mass distribution function ? @xcite have addressed this issue in detail and found that as long as there is no dependence of the fractional gas mass or the stellar initial mass function ( imf ) on the cluster mass , the first two disruption processes discussed above are not expected to affect the cluster mass function ( cmf ) . on the other hand , disruption due to gravitational shocks and encounters of clusters with gmcs is capable of changing the cluster mass function . luminosity and mass functions of gcs follow log - normal distributions peaking at @xmath9 mag and @xmath10 @xmath4 , respectively @xcite . meanwhile , young sscs are found to obey a power - law distribution of luminosities ( @xmath11 with @xmath12 ) over a wide range of cluster luminosities @xcite . @xcite found that a log - normal or quasi log - normal distribution function can be obtained from an initial power - law type of distribution , through selective removal of low - mass clusters from the initial distribution as the cluster population evolves . they propose that the sscs can still be progenitors of gcs in spite of them having different functional forms of mass distribution . the clue to understand the evolution of sscs towards gcs lies in establishing the cmfs of intermediate age clusters . such studies have been carried out for the clusters in the antennae galaxy @xcite , m51 @xcite , and region b of m82 @xcite . these studies are consistent with a power - law cmf that is truncated at cluster masses of @xmath13 @xmath4 , giving the appearance of a quasi - log - normal distribution as predicted by the @xcite scheme . however , difficulties in determining reliable ages of slightly evolved clusters , combined with the incompleteness affecting the mass function for low cluster masses , have hindered the interpretation of the observed mass functions . recent observations of m82 with the hst / acs instrument , covering the entire optical extent of the galaxy , offers an excellent opportunity to investigate the role of disruption processes on mass function . being the nearest galaxy with a large population of sscs , the luminosity and size distribution functions can be studied better in this galaxy than in any other galaxy . in addition , younger and older populations are spatially segregated : present starburst activity ( age@xmath14 myr ) is rather exclusive to the central zone @xcite , whereas the disk lacks any recent star formation . almost all the star formation in the disk took place in a violent disk - wide burst about 100 - 500 myr ago , following the interaction of m82 with the members of m81 group @xcite . cluster formation is known to be efficient during the burst phase of star formation @xcite , and hence we expect large number of clusters of age @xmath15 myr in the disk . recent determinations of age of the clusters in m82b region by @xcite confirm that the peak epoch of cluster formation occurred @xmath16 myr ago . the presence of two distinct epochs of cluster formation , well separated spatially from each other , makes m82 an ideal candidate for a study of evolutionary effects on the cluster mass function . in 2 , we give a brief summary of the observational material used in this study , strategies used for cluster selection , and the analysis carried out for obtaining their size and luminosities . there , the observed luminosity and size distribution functions are also presented . results of the cluster simulation are presented in 3 . methods for deriving the mass of individual clusters , and the construction of the mass distribution function , is described in 4 . in 5 , we compare the observed mass and size distribution functions in m82 with the functions obtained for other galaxies , and discuss the most important disruption process active during the first few @xmath17 yr in m82 . the conclusions from this study are summarized in the observational data used in this work were part of the hst s 16th anniversary , which was celebrated with the release of the color composite image of m82 . the data were obtained by the hubble heritage team @xcite using the acs wide field channel in 2006 march , and released in fits format . observations consisted of 96 individual exposures in f435w , f555w , f814w , and f658n filters covering a field of view of @xmath18 centered on the galaxy nucleus , and cover the entire optical disk of the galaxy . bias , dark , and flat - field corrections were carried out using the standard pipeline process ( calacs ) by the heritage team . the iraf / stsdas multidrizzle task was used to combine images for each filter and to produce weight maps , which indicate the background and instrumental noise . also this task was used to identify bad pixels , to perform sky subtraction , cosmic ray rejection and to eliminate artifacts ( see more details in * ? ? ? the final data set of fits files of science quality images have a spatial sampling of 0.05 arcsecpixel@xmath19 , which corresponds to 0.88 pcpixel@xmath19 at the m82 s distance of 3.63 mpc @xcite . in this work , we used the images in f435w , f555w , f814w bands , which we refer to as @xmath20 , @xmath21 and @xmath22 images , hereinafter . the exposure times , estimated detection limits for point and extended sources in each filter are given in table 1 . a circle of 500 pixels ( 450 pc ) radius centered on the starburst nucleus is used to separate the nuclear region from the disk . the clusters inside this radius are associated with strong h@xmath23 emitting complexes , and hence are younger than 10 myr @xcite . on the other hand , the disk outside the 450 pc radius shows characteristic signatures of post - starburst conditions , with hardly any h@xmath23 emission . the named regions a , c , d , e and h of m82 fall into the nuclear cluster class , whereas the b , f , g and l are disk clusters ( see * ? ? ? * ; * ? ? ? * for identification charts ) . at the resolution provided by the hst , only regions h , f and l retain their identity as bright knots , the rest being resolved into complexes of compact knots . selection of an unbiased sample of cluster candidates requires the use of an automatic object detection code . we used sextractor @xcite for this purpose . an _ area _ consisting of at least 5 adjacent background subtracted pixels with intensity above @xmath24 was defined as a source ( @xmath25 being the local rms value ) . the number of detected sources depends critically on the value of @xmath26 . we found @xmath275 in the @xmath20 and @xmath21-bands , and @xmath28 for the @xmath22-band as optimum values in the search for clusters . this condition establishes a signal - to - noise ratio greater than 50 for a typical source of 100 pixel area . in relatively isolated regions , the _ area _ is determined by the intensity profile of the source , and a pre - defined value of @xmath24 , whereas in crowded regions , it is determined by a deblending parameter . another critical parameter controlling the source detection is the value of the local background , which is measured using a _ boxsize _ of @xmath29 pixels . sextractor determines full width at half maximum ( fwhm ) for each source by fitting a gaussian profile . we carried out independent searches of candidate sources on each of the @xmath30 and @xmath22 images . this resulted in 44274 , 82515 and 151565 sources in the @xmath30 and @xmath22-band images , respectively . for sources identified in a given band , we have carried out multiple aperture photometry in the other two bands . the list of identified sources in each band contains both resolved ( extended ) and unresolved ( stellar - like ) objects . the point sources have a size distribution that peaks at a fwhm of 2.1 pixels , with the tail of the distribution extending to 3.0 pixels , which at the distance of m82 corresponds to a physical size of 2.6 parsec . hence , clusters with a gaussian fwhm larger than 2.6 parsec are resolved objects , and can be easily separated from the stars . we found that 17% ( @xmath20 ) , 23% ( @xmath21 ) , and 40% ( @xmath22 ) of the sources in the initial sample are star - like ( fwhm@xmath31 pixels ) . a visual inspection of the extended sources reveals that all the bright ones are symmetrical as expected for bound systems . however , a considerable fraction of the fainter sources lack such symmetry , and the likelihood of them being clusters is rather low . we found three kinds of contaminating sources 1 ) sources formed by improper subtraction of a local background , 2 ) elongated sources formed by chance superposition of several point sources , and 3 ) groups of stars without a well - defined peak . the first kind of source is due to the presence of intersecting dust filaments of different sizes . contamination by this kind of sources , as expected , is maximal for the @xmath20-band , and is less important at longer wavelengths . on the other hand , the second and third types of contaminations are most severe in the @xmath22-band image due to a large number of red stars present . we adopted the following selection criteria to choose real clusters from the sextractor source list : \1 . all sources should have an @xmath32 pixels , which ensures that pseudo - clusters ( faint superposed stars ) are rejected . however , this criterion rejects all clusters fainter than @xmath33 mag . the most compact clusters ( fwhm=3 pix ) should be brighter than @xmath34 mag in order to satisfy this condition ( see table 1 ) . the sources should obey the condition @xmath35 , which automatically rejects all diffuse artificial sources created by a residual local background . it also removes faint groups of stars on all scales . if the _ area _ is less than 100 pixels , additionally we demand that the sources should be nearly circular ( ellipticity @xmath36 ) , a condition that rejects linearly superposed double or multiple stars . the second selection criterion works exceptionally well in discarding non - cluster sources at the fainter magnitudes . however , it has the disadvantage that it rejects genuine clusters in the crowded nuclear region . this is because the sextractor @xmath37 in these regions is delimited by the presence of a close neighbor , before the intensity profile reaches the @xmath24 limit . clusters rejected due to crowding have typical areas less than 100 pixels . we found that these clusters distinguish themselves from the artificial sources by showing a relatively higher peak surface brightness . we used the aperture magnitude , @xmath38 , within a diameter of 2 pixel as a proxy of the source s peak surface brightness . in order to recover these small - area nuclear clusters , we have established the following additional criterion : \4 . @xmath38 should be brighter than @xmath39 for small - area clusters ( @xmath40 pixels ) , irrespective of conditions 2 and 3 . the values of @xmath39 were chosen to be 25 , 24 and 22 magnitude for the @xmath41 , and @xmath22 images , respectively . it is important to note that the number of nuclear clusters selected depends critically on the value of @xmath39 . the chosen values represent a compromise between overpopulating the sample by un - physical clusters , and rejection of genuine clusters . every condition adopted in the present work was set after an elaborate interactive process of visual inspection of several selected and rejected sources . in table 2 , we present the statistics of the sources detected in every band . the total number of sources identified by the sextractor in each of the three bands is given in the first row of the table . the number of extended sources in this list is given in the second row . the third row gives the number of candidate clusters , which is a subset of the extended sources that occupy an area of at least 50 pixels . the last row gives the number of clusters detected after applying all the selection criteria listed above . it can be seen that only around 6% of the candidate clusters survive our selection criteria . from a visual inspection , we have confirmed that none of the bright sources are rejected , and the large rejection fraction is mainly because of over - dominance of the faint sources in the list . as discussed earlier in this section , the contaminating sources at the faint end are formed by multiple stars , either physical systems or sources formed by chance superposition of individual stars , which is expected given that the disk of m82 is almost oriented along the line of sight . a list of cluster candidates is obtained from the sum of the sources in the three filters , the common sources being counted once . background extragalactic sources seen in the halo of m82 were removed from this list . our final list contains 653 clusters , 260 of them belonging to the nuclear region . majority of these clusters ( 65% ) was identified in the @xmath20-band image . the remaining 35% of the sources were identified in the @xmath21 and @xmath22-band images . though these latter sources failed to satisfy the @xmath24 criterion in the @xmath20-band , they have sufficiently good quality photometry for a reliable estimation of their masses . the fwhm of a source is taken from the list in the shortest of the three bands where it is detected above @xmath24 limit . @xcite , using the hst / wfpc2 images of m82 , searched for nuclear sscs with associated h@xmath23 emission , and found 197 sscs . about 50% of these clusters overlap with at least one of our nuclear sources . we found that the remaining sources are bright in h@xmath23 with only a diffuse emission in the @xmath20-band , which are not considered as genuine clusters in our selection criteria . in total , about 150 nuclear clusters of our sample are cataloged for the first time . though our final sample contains a tiny fraction of the potential cluster candidates , the cluster size distribution function ( csf ) obtained from clusters brighter than @xmath42 mag is a true representation of the intrinsic function , as we demonstrate using monte carlo simulations in 3 . the csf obtained from this bright cluster sample may not necessarily hold for the entire sample . future methods of disentangling multiple stars from clusters ( e.g. analysis of 2-dimensional profiles along with colors ) would be required to investigate this issue . nevertheless , given that the bright clusters in the nucleus and the disk of m82 are also the most massive members in the respective zones , a comparison of the derived csf in the two zones would be invaluable in understanding the formation mechanisms and the subsequent disruption of the clusters . with this in mind , we carry out the analysis of the csf for clusters brighter than @xmath42 mag in this work . with the help of sextractor , we carried out aperture photometry in a number of concentric apertures for each cluster . as clusters span a wide range of sizes , and the background is highly non - uniform , a varying aperture diameter should be adopted in order to minimize the error on the extracted photometry . in the crowded nuclear region , special care should be taken in order to ensure that a given aperture does not include a neighboring source . we identified some isolated clusters and studied their growth profiles in order to estimate the flux fraction that would lie outside an aperture of 5 pixel radius , as a function of the fwhm . in general , this correction factor lies between 0.51.0 magnitude with a monotonic dependence on the fwhm , and showing a dispersion as large as 0.5 mag . the corrected magnitudes agree well with the @xmath43 magnitudes obtained by sextractor by summing all pixels above the @xmath24 contour limit . hence , we adopted the @xmath43 magnitudes as our photometric value . the colors were obtained from the photometry of the brightest part of the cluster , corresponding to the magnitude of the fixed aperture of 5 pixel radius , as we do not expect color variations across the face of a cluster . yet , the enigmatic cluster f and its neighbor l , do show color variations of such kind , but those cases are exceptions rather than the rule @xcite . the procedure adopted in this work ensures that the error on a color is significantly smaller than that on a magnitude . in table 3 , we list the results of the aperture photometry for the detected clusters . column 1 contains our identification numbers , which run from 1n to 265n for the nuclear clusters , followed by 1d to 393d for the disk ones ( 5 nuclear clusters 84n , 139n , 147n , 149n and 258n were rejected as the error on at least one of the colors was unreasonably high ) . the clusters are arranged in increasing order of @xmath20 magnitude , separately for the nuclear and disk sources . columns 25 contains the @xmath20 magnitude , its error , @xmath44 , and @xmath45 colors , respectively . the error column lists an estimation of the formal error , as given by the sextractor , errors on the colors are expected to be of this order . the systematic errors due to an improper background subtraction , which affects a magnitude measurement , and not a color measurement , could be as large as 0.2 mag for sources fainter than @xmath46 mag . the last column of the table contains the cross identifications with @xcite ( numbers following the letter m ) , and other named regions mentioned in the beginning of this section . in figure [ fig_stamps ] , we present the stamp - size images of @xmath47 pixel format centered on all the detected disk clusters . it can be seen that even the relatively faint sources have the morphology resembling a cluster . identity of a few of the relatively fainter sources as genuine clusters is debatable . however , we find that the contamination from non - cluster sources is less than 5% . in figure [ fig_mag_hist_ssc ] , we show the luminosity function of sources satisfying our selection criteria of clusters for each of the @xmath48 and @xmath22 images , independently . the luminosity functions in every band show an apparent turn - over with the distributions being consistent with a power - law form of index @xmath49 on the brighter side , and falling steeply on the fainter side . selection criterion 1 is responsible for the steep fall as discussed in 2.1 . the turn - over magnitudes are equal to @xmath5021.5 , 20.5 , and 18.5 in the @xmath48 and @xmath22-bands , respectively . at the very bright end of the luminosity function , when the numbers expected from the extrapolation of the power - law are below 10 , the observed numbers are systematically lower in all the bands . it seems that we are missing around 5 bright clusters . this may suggest the existence of an upper limit to the cluster masses as advocated by @xcite . on the other hand , it is also possible that the under - abundance of bright clusters is due to the combined effects of high extinction , and small number statistics . the nuclear and disk cluster samples present similar luminosity functions , with the apparent turn - over magnitude @xmath51 mag fainter for the latter sample . sextractor provides the fwhm of a fitted gaussian profile for each source . we have adopted this figure of merit as a measure of the size of the clusters , instead of the often used half - light radius ( @xmath52 ) . this is because the gaussian fwhm is uniquely defined for a source , and can be measured with the same accuracy for bright and faint regions , whereas the value of @xmath52 is very sensitive on the parameters of the fitted analytical function @xcite . sextractor calculates the half - light radius , that we found is related to the fwhm ( both in pixel units ) by the equation : @xmath53 . the values of @xmath52 calculated from this relation may differ slightly from those values determined based on an analytical function . we use this equation only for the purpose of comparing the mean fwhm obtained in this work with @xmath52 quoted in the literature for other galaxies . the histograms of gaussian fwhms in the @xmath54 photometric bands , for the total , nuclear and disk samples , are presented in figure [ fig_fwhm_hist_all ] . there , it is evident that the observed csf peaks at a characteristic value of about 10 pixels fwhm or @xmath55 pc . @xcite found a mean radius value @xmath56 pc for the m82 nuclear clusters . however , their values can not be directly compared with ours as they had defined the radius as the inflexion point of the photometric growth curve in @xmath20 and h@xmath23 bands . on the other hand , our values are in good agreement with those measured in a sample of 18 nearby spiral galaxies , where @xcite reported a mean value of @xmath57 pc . the presence of a characteristic cluster size is quite different to what was found in m51 , where the size distribution follows a power - law function @xcite . the observed cluster luminosity function follows a power - law at the bright end , turning over sharply at faint magnitudes . similarly , the csf peaks at a characteristic value of @xmath5010 pixels fwhm . in order to investigate whether observational biases are responsible for these turn - overs , we carried out detailed monte carlo simulations . we have used simulations in order to generate a @xmath20-band acs image of m82 containing n@xmath58 clusters . the luminosity function of the simulated clusters is defined by a power - law distribution in the 1825 magnitude range : @xmath59 . a power - law index @xmath49 was adopted , that corresponds to a slope of 0.4 in the @xmath60 versus magnitude plot . in the simulations , each cluster is assumed to be round and to have a gaussian intensity profile of a given fwhm . we used two functional forms for the size distribution of the clusters : the first one adopting a power - law function ( @xmath61 ) with an index @xmath62 . in the second simulation , we adopted a log - normal function peaking at 10 pixels with a @xmath25 value of 2 pixels . we generated n@xmath63 clusters , the number of extended objects in the observed @xmath20-band source list with an area of at least 35 pixels . the location of the observed sources were used as their reference positions of the simulated clusters . this procedure simulates properly the crowding of sources observed in m82 . we also generated point sources as gaussian profiles of fwhm fixed at 2.1 pixels , with their positions and magnitudes corresponding to the observed values . the precise position of every source ( stars and clusters ) is generated by randomly placing them within an rms of 10 pixels around that in the source list . in order to emulate as closely as possible the observed conditions , we degraded the noise - free simulated image by adding the observed rms noise ( @xmath64 count / pixel / sec ) , and the sextractor - generated background image of m82 . the sextractor was run on the simulated image with the same set of parameters that were adopted for the observed @xmath20-band image . the resulting catalog was passed through the same selection filters described in 2.1 in order to obtain a catalog of simulated clusters . we compared the extracted @xmath43 magnitude and fwhm size of every cluster with the simulated values . results are summarized in figure [ fig_fwhm_mag_dif ] . the histograms of the differences have a maximum at @xmath65 , indicating that sextractor accurately recovers magnitudes and sizes of the simulated sources . for 90% of the recovered sources , magnitude difference is less than 0.1 mag , while fwhm size difference lie within 0.75 pixel of the simulated values . for the remaining 10% of the sources , the extracted magnitudes and sizes turn out to be systematically brighter and larger . both effects are consequence of the superpositions of more than one simulated source due to crowding that extractor code identifies as a single source . photometric recovery is better than 0.1 magnitude for both extended and compact sources , which illustrates that there is no systematic underestimation of flux , as measured by @xmath43 , from extended sources . the luminosity function of the simulated clusters is plotted in figure [ fig_hist_sim_mag ] , separate histograms are shown for the nuclear and disk samples . at the bright end , luminosity functions of the recovered sources follow the same power - law that was used to generate them . the distribution function for the extracted disk clusters begins to depart from the simulated one at @xmath66 magnitude , remains nearly flat for another 2 magnitudes before falling sharply at fainter luminosities . the fraction of the simulated sources recovered is larger than 90% for sources brighter than 21.0 mag , falling to @xmath67% at 22 magnitude , which we refer to as the apparent turn - over magnitude . the luminosity distribution function for the nuclear clusters is very similar to that for the disk population except that the apparent turn - over magnitude is @xmath51 mag brighter than that for the disk . the higher nuclear background , and greater crowding as compared to those of the disk are responsible for the difference . the simulated luminosity function resembles very much the observed one , implying that the observed apparent turn - over of the luminosity function is due to incompleteness at the faint end and not intrinsic to the cluster population . hence , the intrinsic turn - over in the luminosity function , if any , would correspond to a magnitude fainter than @xmath42 mag . we find that the power - law index ( @xmath12 ) can be recovered to within an error of 0.1 using only the data brighter than the apparent turn - over magnitude , independently for the nuclear and disk clusters . this ensures that reliable value of @xmath23 can be determined from the observed data . having established that the turn - over in the luminosity function is caused by the incompleteness in the detection of faint clusters , we now investigate whether the csf is also affected by our selection criteria . from simple analytical calculation of the type presented in table 1 , we find that the clusters that survive our first selection criterion all the way to @xmath33 mag are those with intermediate sizes ( fwhm@xmath50511 pixels ) , with the rest of the clusters having cut - off magnitudes between 2224 . the rejection of clusters is mainly because they fail to satisfy the requirement of at least 50 pixels area . thus , the csf derived from a sample that includes clusters fainter than the apparent turn - over ( @xmath42 mag ) is not expected to represent the true distribution . we used the results of monte carlo simulations to check whether the csf obtained using clusters on the brighter side of the apparent turn - over represents the true distribution . we plot the size distribution function for the simulated data in figure [ fig_hist_sim_fwhm ] . only clusters brighter than @xmath42 mag in the disk , and @xmath68 mag in the nucleus , were used in creating these histograms . separate histograms are shown for the nuclear and disk clusters . the top and bottom panels show the histograms for a log - normal and power - law types of distribution functions , respectively . the recovery of the simulated size distribution functions is very good for both types of functional forms for fwhm@xmath69 pixels . in particular , a power - law distribution retains its form and index value both for the disk and nuclear samples . thus , the observed log - normal distribution of sizes could nt have been the result of selection effects transforming an intrinsically power - law into a log - normal distribution . clusters with fwhm@xmath70 pixels are not detected even at magnitudes brighter than the apparent turn - over . this explains the absence of such clusters in the observational data ( figure [ fig_fwhm_hist_all ] ) . from the analysis of the simulated luminosity function we have inferred that around 10% of clusters brighter than @xmath46 mag , and 40% of clusters brighter than @xmath42 mag are not detected . analysis of simulated data presented in figure [ fig_hist_sim_fwhm ] offers us an opportunity to understand the reasons for the non - detection . one of the reasons is that they are too extended ( fwhm@xmath70 pixels ) . among the rest , the cluster detection fraction does nt dependent on the value of the fwhm . rather , the reason for the non - detection is that either the cluster is situated in a region with higher than average background value or that there is a bright cluster in its vicinity . thus , our simulations illustrate that there is no size - dependent bias against the cluster detection as long as they are brighter than @xmath42 mag and have fwhm@xmath71 pixels . hence , the size function derived using data of clusters brighter than the apparent turn - over is a true representation of the intrinsic function . m82 is a spiral galaxy seen almost edge - on @xcite , consequently the effect of obscuration is severe . hence , the physical parameters derived from observational data critically depend on the treatment of the interstellar extinction . the issue of extinction toward the nuclear starburst of this galaxy has long been the subject of study . @xcite , using near infrared recombination lines integrated over the central starburst region , estimated visual extinction values between 12 and 27 magnitudes , the exact value depending on the adopted reddening model . @xcite found @xmath72 mag in selected knots using a uniform foreground screen model , and @xmath73 mag for models where dust and stars are mixed . in this section , we analyze the observed colors and brightness of the clusters with the aim of deriving extinction corrected photometry , and compare them with a population synthesis model . star clusters can be considered as simple stellar populations ( ssps ) , and some of their physical parameters such as mass and age can be obtained from the analysis of their colors and luminosity with the help of population synthesis models . in this work , we have used the solar metallicity ssp models of @xcite . these authors had carried out the evolution of colors and magnitudes for the instrumental hst / acs filters , a fact that enables us a direct comparison with the observed data . the @xcite initial mass function ( imf ) in its corrected version has been used . it has nearly a salpeter slope ( 2.30 instead of 2.35 ) for all masses higher than 1 @xmath4 . the derived masses depend on the assumption of the lower cut - off mass of the imf . in the case of standard kroupa s imf , the derived masses would be around 2.5 times higher . the observed cluster colors are plotted in the @xmath45 vs @xmath44 plane in figure [ fig_col_vs_col ] . the filled circles correspond to the nuclear clusters , whereas open ones correspond to the disk clusters . the solid line represents the locus of an ssp evolutionary track . time tags represented by vertical lines correspond to ssp ages 10 , 50 , 100 , 500 myr and 5 gyr . there , a reddening vector corresponding to @xmath74 mag is also shown . the histograms corresponding to the distribution of the colors are shown in the top and right side panels , separately for the disk ( solid line ) and the nuclear clusters . it can be seen that the distribution of colors of disk clusters peaks at lower values and has a smaller spread as compared to those of the nuclear clusters . the color distributions have a long red tail for both classes of clusters . the red tail of the distribution reaches colors even greater than those of a 510 gyr old population . thus , the large observed color range is produced by a large spread in the reddening rather than a spread in the age . hence , the observed colors are useful in deriving the amount of reddening , given a value for the cluster age . the stellar population in the nuclear region of m82 has been the subject of innumerable studies . @xcite found that the multi - wavelength data for the central 400 pc region of m82 can be explained by a few starburst events over the last 30 myr . more recent study by @xcite favors the presence of two short - term bursts over the last 10 myr . @xcite studied 12 nuclear star clusters of m82 , and found their ages to be between 410 myr . @xcite found that a great majority of the nuclear star clusters are associated with h@xmath23 emission , suggesting that they are younger than the lifetime of ionizing o stars . based on these results , we adopt an age of 8 myr for the nuclear clusters . the h@xmath23 emitting regions are exclusively located in the central starburst region , and in the cone along the minor axis of m82 . the absence of h@xmath23 emission in the disk implies that currently there is no significant amount of star formation . the disk is known to be rich in the balmer absorption lines , suggesting an intense star formation in the past @xcite . @xcite made use of these spectral features to infer a rather young age for the disk in m82 . the disk stars were formed in a burst as a consequence of a tidal interaction with members of m81 group during the last 500 myr . were the clusters formed simultaneously with the disk - wide star formation ? to answer this question , spectroscopically derived ages of clusters would be required . such data are available only for four disk clusters regions f and l , and two clusters in the region b @xcite . derived ages range between 5065 myr for f and l , and @xmath75 myr for the two optically bright knots in region @xmath20 . we have used the acs photometric data to check whether the ages of these bright clusters are representative of that of the bulk of the population . as discussed previously , the colors in m82 are strongly affected by reddening , and hence any age inference based on colors of individual clusters is not expected to give a unique solution . we investigated whether the color differences between the clusters and the surrounding disk could be used to derive a statistical age for the cluster population in the disk . if the clusters have a similar age as that of the stellar disk , and if the reddening is caused by the dust clouds external to the clusters , then the color distribution of the clusters would be similar to those of the local disk . we carried out experiments to test this idea . we measured the colors of the disk surrounding every cluster in a 10-pixel wide annulus with an inner radii of 20 , 30 and 40 pixels , that correspond to 1 , 1.5 , and 2@xmath76 , respectively . we found that the colors in these three annuli are very similar . in figure [ fig_col_disk ] , we compare the colors of the inner most annulus with that of the corresponding cluster . in the bottom panel , the disk @xmath44 color is plotted against that of the cluster . there , we can see that the disk colors span a much shorter range as compared with that of the clusters . in fact , the color difference ( middle panel ) of the cluster and the local disk reflects the color of the cluster . the peak of the cluster color distribution is at @xmath77 mag bluer than the surrounding disk . systematically blue colors of the clusters suggest that the clusters are younger than the disk . interestingly enough , the disk surrounding the red clusters is not as red , suggesting that the dust causing the reddening of the clusters is local to the cluster , perhaps associated with the interstellar medium left over from the star formation episode . very similar results are found from the analysis of the @xmath45 colors . the bluer colors of the clusters could be due to metallicity effects , if the clusters have systematically lower metallicity content as compared to the local disk . this case is unlikely , yet it could happen if the clusters were formed from the accreted metal - poor gas , after the formation of the disk . in this scenario , clusters would be at the most as old as the disk . on the other hand , if the cluster stars are of higher metal abundance , they could be as young as 50 myr . this rather young age compares very well with the 5065 myr determined by @xcite for the bright clusters @xmath78 and @xmath79 from spectroscopic indices . consequently , the ages of the disk clusters are most likely lie in the 50500 myr range . for the determination of reddening and mass of individual clusters , we fix their age at 100 myr , and discuss the error on these quantities if they were as young as 50 myr or as old as 500 myr . once the reddening is derived using the colors , we obtain masses of individual clusters with the help of the ssp for the assumed age ( 8 myr for the nuclear clusters , and 100 myr for the disk clusters ) . the method we have followed is illustrated in figure [ fig_cmds ] . for a given position in the color magnitude diagram ( cmd ) , we derived the reddening by comparing the observed colors with those of the ssp . @xcite extinction curve with rv=3.1 is used to convert reddening to visual extinction @xmath80 . the mass is then calculated using the extinction - corrected @xmath20-band luminosity and the mass - to - light ratio of the ssp for the assumed age . in table 3 , we list the derived quantities ( visual extinction , extinction corrected b - band absolute magnitude and mass ) along with their estimated errors for every cluster . we have independently derived extinction values from @xmath44 and @xmath45 colors , and the tabulated values are the mean of the two measurements . the dispersion from the mean value is tabulated as the error on @xmath80 . the absolute @xmath20 magnitude listed is extinction - corrected . the tabulated mass is photometric and is obtained by the method described above , using an age of 8 myr for the nuclear clusters , and 100 myr for the disk clusters . the derived @xmath80 would be systematically higher by 0.12 mag if the disk clusters are as young as 50 myr , and lower by 0.58 mag if they are as old as 500 myr for the ssps used in this work @xcite . the derived disk masses would also be different if we had used a different age , as both the extinction and mass - light - ratio are sensitive to the assumed age of the population . fortunately , these two corrections act in opposite directions , with the net result that the masses would always be within a factor of 1.5 of the tabulated values for the range of ages between 50500 myr . the errors on @xmath20 magnitude and masses are calculated by quadratically summing the errors on @xmath80 and the @xmath20 magnitude , the latter taken as 0.2 mag for all clusters . the histogram showing the distribution of visual extinction for an assumed age of 8 myr for the nuclear , and 100 myr for the disk clusters , is plotted in figure [ fig_physpar ] . the distribution for the nuclear clusters is nearly flat between @xmath811.04.0 mag , whereas it is peaked at @xmath82 mag for the disk clusters . there are a few clusters in both the disk and the nucleus with inferred @xmath836 mag . the derived @xmath80 is less than 2.0 mag for 85% of the disk clusters , whereas only 20% of the nuclear clusters are below this value . the results remain practically the same if all the disk clusters are as young as 50 myr . on the other hand , if all the disk clusters are as old as 500 myr , the distribution of their @xmath80 would peak at a value lower by 0.58 mag ( @xmath84 bin width ) with respect to that in figure [ fig_physpar ] . this would further increase the mean difference in the extinction values between the disk and nuclear clusters . the derivation of the cluster masses for our entire sample enables us to derive the present - day cmf . in figure [ fig_mass_fun ] , we plot the cmf separately for the nuclear and disk clusters . the nuclear cmf is scaled up to match the disk cmf at @xmath85 @xmath4 . poissonian error bars are indicated . the distribution for both samples follows a power - law over almost two orders of magnitude in mass for cluster masses above @xmath86 @xmath4 . however , the power - law index for the disk and nuclear cluster populations shows a marginal difference , @xmath2 for the nuclear clusters , and @xmath87 for the disk population . a different choice of age ( within the range 410 myr for nuclear clusters and 50500 myr for the disk clusters ) in the derivation of the photometric mass would have only changed the masses by less than the size of the bin width used in figure [ fig_mass_fun ] . hence , the slopes of the mass functions are not affected by our assumption that all the clusters in each zone have the same age . studies of young star clusters in nearby galaxies yield a value for @xmath23 close to 2.0 @xcite . our derived value of @xmath2 for the m82 nuclear clusters is marginally flatter than this . on the other hand , the mass function of the disk clusters is clearly flatter than that for young clusters , implying that the low - mass clusters are selectively lost as the cluster population evolves . in order to shed more light on the missing population , we analyze the cluster size distribution below . in figure [ fig_size_fun ] , we compare the csf for the nuclear and disk samples . following the discussion in 3.2 , only clusters brighter than the apparent turn - over luminosity ( @xmath42 mag for the disk and @xmath68 mag for the nucleus ) were considered . the mass that corresponds to the apparent turn - over is @xmath86@xmath4 , and hence the plotted functions are representative of the cluster population more massive than this limit . the csf for the nuclear region can be fitted very well with a log - normal function centered at fwhm of 11 pixels ( @xmath88 pc ) . on the other hand , the csf for the disk clusters is quasi log - normal : a log - normal fit to the distribution of clusters with sizes larger than 11 pixels underestimates the actual number of observed compact clusters , while a fit to the distribution of compact clusters shows a deficiency of extended or loose clusters . in order to investigate the possible origin of the skewness in the distribution , we have analyzed the size distribution function in three separate mass bins , i.e. low - mass ( mass@xmath89@xmath4 ) , high - mass ( mass@xmath90@xmath4 ) and intermediate mass ( all in between ) . the csf for the nuclear clusters follows the log - normal distribution in each of the three mass bins ( taking into account the statistical errors caused by the small number of clusters in the highest mass bin ) . the disk clusters in the highest mass bin , also follow a log - normal distribution . the csf of the lower mass clusters of the disk , however , depart from that of the nuclear clusters : the largest clusters , as well as the mean cluster size , are systematically smaller for the lower mass bins . this tendency is illustrated in figure [ fig_stats ] , where the mean cluster size for each mass bin has been plotted against the mean mass of clusters in that bin , for the young and old ones , separately . for the highest mass bin , the mean sizes of the young and old clusters are similar . the mean size decreases systematically with decreasing cluster mass for the old clusters , whereas the inverse is true for the young clusters . systems in virial equilibrium are expected to follow a power - law radius - mass relation ( @xmath91 ) with a @xmath92 , the most common examples being the giant molecular clouds ( gmcs ) and elliptical galaxies . stars of a cluster are well mixed and reach virial equilibrium after just a few crossing times @xcite . m82 clusters are around ten crossing times old and hence are expected to be in virial equilibrium @xcite . the relationship seen in figure [ fig_stats ] corresponds to a power - law index of @xmath93 , and @xmath94 , for the old and young samples , respectively , i.e. both values are significantly flatter as compared to those of bound systems in virial equilibrium . on the other hand , our values for the disk clusters are in excellent agreement with that found ( @xmath95 ) for star clusters in nearby galaxies by @xcite . @xcite have obtained an index of @xmath96 for clusters losing mass adiabatically over time . thus , mass - loss from clusters could yield a flatter radius - mass relationship . the similar observed range of sizes for high - mass clusters independent of whether they are young or old suggests that these clusters have reached their equilibrium sizes and that the evolution hasnt played a role in changing their sizes . the observed radius - mass relationship then implies that the sizes of lower mass clusters are well above their equilibrium values , or in other words they are distended ones . on an average , younger clusters are farther from equilibrium values as compared to older ones . together , these results imply that many of the young low - mass clusters in our sample are loosely bound or unbound , or probably are expanding ob stellar associations . the extended low - mass clusters are being selectively destroyed as a function of time , whereas compact clusters of all masses have survived . however , the sizes of even the most compact low - mass ( mass@xmath97 @xmath4 ) clusters are larger than their virial expected values , implying that they are expanding and may not survive over the hubble time . the observed differences in the csf for young and older clusters are consistent with the expected evolutionary effects . both , the disruption of the loose ob associations and the dynamical trend towards relaxation would diminish the number of large low - mass systems . thus , the destruction process is both mass and size dependent , with the most extended clusters in each mass bin being the first ones to be dispersed . the near equality of the mean size for young and old clusters of masses higher than @xmath13 @xmath4 implies that all high mass clusters survive for ages more than @xmath98 yr or that the destruction process is independent of the size for these high mass clusters . in the next section , we will discuss these observational results in the context of results obtained by other studies , and propose the most likely scenario of cluster disruption in m82 . the presence of two populations of star clusters in m82 , one young ( age@xmath99 yr ) , and the other relatively old ( age@xmath98 yr ) allows us to interpret the observed differences of their distribution functions in terms of evolutionary effects . we start this section by comparing our cmf for the entire disk , with that published for the region b in m82 . the bright blue region known as m82 b about a kiloparsec distance to the northeast of the nucleus has been the subject of study in a series of articles ( * ? ? ? * ; * ? ? ? these authors were able to detect more than 200 cluster candidates using the hst / wfpc2 images , which were used to construct luminosity and mass functions . they found that the mass function is log - normal with a turn - over in the mass function at @xmath100(mass)@xmath101 , well above their completeness limit for cluster detection . adopting an estimated age of 1 gyr for this system , they argued that the log - normal mass distribution can not be the result of selective destruction of smaller clusters as the clusters evolve , and instead should have been born that way . in essence , they favored a log - normal initial cmf for m82 clusters , and proposed that the gcs probably were also born with log - normal initial cmf . a log - normal initial cmf is not consistent with our dataset for younger clusters in m82 . in fact , both the younger and older clusters of our sample follow power - law forms , with the index being marginally different for the two populations . more importantly , the apparent turn - over masses for the two samples are similar , and are at least a factor of five lower than that reported by @xcite . bright clusters of m82 b make part of our sample , and hence these differences are intriguing . the explanation for the different conclusions lies in the different adopted ages : 100 myr ( ours ) versus 1 gyr @xcite . available spectroscopic data favor ages between 50350 myr : @xcite derived an age of @xmath75 myr for two knots ( b2 - 1 and b1 - 1 ) in region m82 b , and 5065 myr for the knots f and l at similar galactocentric distance as region b , but to the southwest of the nucleus . recent determination of ages for 7 more clusters by @xcite also support these relatively younger ages . thus , an age of 1 gyr seems extremely high for the cluster population in m82 , and the ages are likely to be closer to the one we have adopted . the @xmath20-band luminosity drops by a factor of 8.5 due to passive evolution of ssps between 100 myr and 1 gyr @xcite . thus , the photometrically derived masses by @xcite could have been overestimated by a factor as large as this . if we take into account this factor , their turn - over mass values would be similar to the ones we find in the present study . a lower turn - over mass dilutes the case for a log - normal initial cmf . whether low - mass clusters are preferentially disrupted with respect to the high mass ones is the question waiting to be answered by the ever growing data on extragalactic star clusters . so far , the best studied cases are m51 and the large magellanic cloud ( lmc ) . for m51 clusters , @xcite concluded that the disruption rate at early times ( @xmath8830 myr ) is independent of cluster mass . in the lmc , @xcite report evidence for mass - dependent disruption for masses below a few @xmath102 @xmath4 . our result showing a tendency for a flatter slope of the mass function at older ages is an indication that the disruption processes acting in the disk of m82 are mass - dependent . for a given mass , clusters of larger radii are loosely bound , and it is easier to disrupt them . hence , one would expect those clusters to be missing in older cluster samples . @xcite found size - dependent evolution for the m51 clusters . in the antennae galaxy , @xcite have found a decrease of the average cluster size with age , as well . in the present study , we find that the average size of the clusters is smaller for older clusters of mass @xmath103 @xmath4 . thus , evolutionary processes selectively disrupt loosely bound clusters . in the introduction , we have mentioned several physical processes that are capable of disrupting a star cluster . we refer the readers to the works of @xcite , @xcite and @xcite , for details on these physical processes . at early times ( @xmath104 myr ) , disruption is caused mainly due to the expulsion of the intra - cluster gas through supernova explosions . this process is independent of cluster mass , as long as the stellar imf , or the mass fraction of the gas , are not systematically different for low and high mass clusters . the other process that can change the slope of the cluster mass function is the tidal shock experienced by the clusters as they move in the gravitational field of a galaxy . according to @xcite , this process becomes important after @xmath105 myr in normal galaxies . however , in the case of m82 , @xcite have estimated a disruption timescale as short as 30 myr for a cluster of mass @xmath106 @xmath4 at 1 kpc away from the center , with a dependence on mass that varies as @xmath107 . hence , the derived slope of @xmath108 for the nuclear clusters , which are younger than 10 myr , likely represents the initial slope of the mass function . on the other hand , clusters in the disk of m82 are expected to suffer from the dynamical processes of cluster disruption . this kind of disruption process destroys first the larger clusters . the over - abundance of compact clusters in our sample of the older clusters , as compared to the sample of younger clusters , is quite consistent with this idea . thus our dataset support the relatively short disruption timescale estimated for m82 . very little is known regarding the distribution of sizes of sscs . for clusters in m51 , @xcite found a power - law relationship with an index of @xmath109 , which agrees well with that for the galactic globular clusters ( @xmath110 ) . using more recent hst / acs data of m51 , @xcite also find a power - law form for the csf , albeit over a much smaller range in sizes . in the case of m82 , the power - law form is not evident in the observed distribution of sizes of bright clusters considered in the present study ( figure [ fig_size_fun ] ) : we should have detected large numbers of compact clusters , which is not the case . as compared to our sample in m82 , the low - mass limit in the cluster samples of m51 @xcite is an order of magnitude lower . if the low - mass clusters are systematically more compact than their high - mass counterparts , then the observed difference in the functional forms of the two galaxies can be understood in terms of the difference in the low - mass limits . however , no clear mass - radius relation was found for m51 clusters . hence , the csf for the bright cluster in m82 is different from that found in m51 . it is likely that this difference might be related to the relative youth of the disk of m82 . in a galaxy with clusters forming continuously over time , the selective disruption of large clusters is expected to induce an accumulation of clusters of the smaller sizes at any given cluster mass . this cumulative process seems to have taken place in m51 , while in m82 we are likely witnessing the initial distribution of sizes . in this study , we have carried out an objective search for star clusters on the hst / acs images of m82 in filters f435w(b ) , f555w(v ) , and f814w(i ) . the search has led to the discovery of 393 clusters in the disk and 260 clusters in the nuclear region . the magnitude and fwhm of these clusters were used to construct luminosity and size distribution functions . we find that the luminosity function follows a power - law with an index of 2 , with an apparent turn - over at the faint end . monte carlo simulations carried out by us show that this turn - over is a consequence of incompleteness in the detection of faint clusters rather than a turn - over in the intrinsic luminosity of the population . we used simple stellar population synthesis models to derive visual extinction values and photometric masses for the clusters , adopting a uniform age of 8 myr for the nuclear clusters and 100 myr for the disk ones . the resultant mass distribution functions for the nuclear and disk regions follow power - law functions , with a marginally steeper index value of 1.8 for the younger nuclear regions as compared to 1.5 for the older disk regions . the cluster size distribution function , constructed for the clusters brighter than the turn - over magnitude , follows a log - normal function with its center at 10 pc fwhm ( @xmath111 pc ) for the most massive clusters ( mass @xmath90 @xmath4 ) . for lower masses ( mass=(0.21.0)@xmath112 @xmath4 ) , the center is marginally shifted to larger values for the younger , and smaller values for the older clusters . this tendency implies that the extended low - mass clusters are selectively destroyed during their dynamical evolution . the marginally flatter slope of the mass function , and an over - abundance of compact clusters in the older sample , are pointing towards a mass and size dependent cluster disruption process at work in the disk of m82 . we would like to thank the hubble heritage team at the space telescope science institute for making the reduced fits files available to us . we also thank an anonymous referee for many useful comments that have lead to an improvement of the original manuscript . additionally , we thank prof . alessandro bressan for useful comments on the manuscript . this work is partly supported by conacyt ( mexico ) research grants 42609-f , 49942-f and 58956-f . bastian , n. , gieles , m. , lamers , h. j. g. l. m. , scheepmaker , r. a. , & de grijs , r. 2005 , , 431 , 905 bastian , n. , konstantopoulos , i. , smith , l. j. , trancho , g. , westmoquette , m. s. , & gallagher , j. s. , iii , 2007 , , 379 , 1333 bertin , e. , & arnouts , s. 1996 , , 117 , 393 cardelli , j. a. , clayton , g. c. , & mathis , j. s. 1989 , , 345 , 245 de grijs , r. , oconnell , r. w. , & gallagher , j. s. 2001 , , 121 , 768 de grijs , r. , bastian , n. , & lamers , h. j. g. l. m. 2003a , , 583 , l17 de grijs , r. , bastian , n. , & lamers , h. j. g. l. m. 2003b , , 340 , 197 de grijs , r. , anders , p. , bastian , n. , lynds , r. , lamers , h. j. g. l. m. , & oneil , e. j. 2003c , , 343 , 1285 de grijs , r. , parmentier , g. , & lamers , h. j. g. l. m. 2005 , , 364 , 1054 de grijs , r. , & parmentier , g. 2007 , chinese journal of astronomy and astrophysics , 7 , 155 fall , s. m. , & zhang , q. 2001 , , 561 , 751 frster schreiber , n. m. , genzel , r. , lutz , d. , kunze , d. , & sternberg , a. 2001 , , 552 , 544 frster schreiber , n. m. , genzel , r. , lutz , d. , & sternberg , a. 2003 , , 599 , 193 freedman , w. l. , et al . 1994 , , 427 , 628 gieles , m. , bastian , n. , lamers , h. j. g. l. m. , & mout , j. n. 2005 , , 441 , 949 gieles , m. , larsen , s. s. , scheepmaker , r. a. , bastian , n. , haas , m. r. , & lamers , h. j. g. l. m. 2006a , , 446 , l9 gieles , m. , portegies zwart , s. f. , baumgardt , h. , athanassoula , e. , lamers , h. j. g. l. m. , sipior , m. , & leenaarts , j. 2006b , , 371 , 793 girardi , l. , bertelli , g. , bressan , a. , chiosi , c. , groenewegen , m. a. t. , marigo , p. , salasnich , b. , & weiss , a. 2002 , , 391 , 195 harris , w. e. 1991 , , 29 , 543 holtzman , j. a. , et al . 1992 , , 103 , 691 king , i. r. 1981 , , 22 , 227 konstantopoulos , i. , bastian , n. , smith , l. j. , trancho , g. , westmoquette , m. s. , & gallagher , j. s. , iii , 2007 , astro - ph/0710.1028 kroupa p. , 2001 , mnras 322 , 231 larsen , s. s. 2004 , , 416 , 537 maz - apellniz , j. 2004 , astrophysics and space science library , 315 , 231 mayya , y. d. , carrasco , l. , & luna , a. 2005 , , 628 , l33 mayya , y. d. , bressan , a. , carrasco , l. , & hernandez , l. 2006 , , 649 , 172 mccrady , n. & graham , j.r . 2007 , , 663 , 844 melo , v. p. , muoz - tun , c. , maz - apellniz , j. , & tenorio - tagle , g. 2005 , , 619 , 270 mengel , s. , lehnert , m. d. , thatte , n. , & genzel , r. 2005 , , 443 , 41 mutchler , m. , bond , h. e. , christian , c. a. , frattare , l. m. , hamilton , f. , januszewski , w. , levay , z. g. , mountain , m. , noll , k. s. , royle , p. , gallagher , j. s. , & puxley , p. 2007 , , 119 , 1 oconnell , r. w. , & mangano , j. j. 1978 , , 221 , 62 oconnell , r. w. , gallagher , j. s. , iii , hunter , d. a. , & colley , w. n. 1995 , , 446 , l1 rieke , g. h. , loken , k. , rieke , m. j. , & tamblyn , p. 1993 , , 412 , 99 satyapal , s. , watson , d. m. , pipher , j. l. , forrest , w. j. , greenhouse , m. a. , smith , h. a. , fischer , j. , & woodward , c. e. 1997 , , 483 , 148 scheepmaker , r. a. , haas , m. r. , gieles , m. , bastian , n. , larsen , s. s. , & lamers , h. j. g. l. m. 2007 , , 469 , 925 smith , l. j. , westmoquette , m. s. , gallagher , j. s. , oconnell , r. w. , rosario , d. j. , & de grijs , r. 2006 , , 370 , 513 smith , l. j. , et al . 2007 , , 667 , l145 spitzer , l. 1987 , princeton , nj , princeton university press , 1987 , 191 whitmore , b. c. , schweizer , f. , leitherer , c. , borne , k. , & robert , c. 1993 , , 106 , 1354 whitmore , b. c. , zhang , q. , leitherer , c. , fall , s. m. , schweizer , f. , & miller , b. w. 1999 , , 118 , 1551 lcccl all & 44274 & 82515 & 151565 & 510 @xmath25/pixel + extended & 83% & 77% & 60% & @xmath113 + candidates & 7632 & 7688 & 6307 & @xmath113 & @xmath114 pix + clusters & 421 ( 263 ) & 456 ( 306 ) & 390 ( 208 ) & see @xmath1152.1 + lccccccrcccl 1n & 18.39 & 0.01 & 1.14 & 1.57 & 2.38 & 0.00 & @xmath116 & 0.20 & 5.42 & 0.08 & h + 2n & 18.43 & 0.01 & 0.69 & 0.93 & 1.01 & 0.22 & @xmath117 & 0.36 & 4.67 & 0.14 & m35se + 3n & 18.68 & 0.01 & 1.27 & 2.09 & 3.19 & 0.62 & @xmath118 & 0.86 & 5.75 & 0.35 & a1,m86se + 4n & 18.95 & 0.01 & 1.11 & 1.82 & 2.65 & 0.47 & @xmath119 & 0.66 & 5.34 & 0.27 & + 5n & 18.96 & 0.01 & 1.20 & 1.75 & 2.68 & 0.18 & @xmath120 & 0.32 & 5.36 & 0.13 & m6nw + 6n & 18.99 & 0.01 & 1.14 & 1.63 & 2.46 & 0.11 & @xmath121 & 0.25 & 5.23 & 0.10 & m82se + 7n & 19.16 & 0.01 & 1.05 & 1.79 & 2.54 & 0.55 & @xmath122 & 0.77 & 5.20 & 0.31 & m83se + 8n & 19.34 & 0.02 & 0.60 & 0.97 & 0.93 & 0.03 & @xmath123 & 0.20 & 4.26 & 0.08 & + 9n & 19.35 & 0.01 & 0.73 & 1.42 & 1.65 & 0.53 & @xmath124 & 0.74 & 4.64 & 0.30 & + 10n & 19.45 & 0.00 & 1.30 & 2.16 & 3.32 & 0.68 & @xmath125 & 0.94 & 5.51 & 0.38 & + 11n & 19.47 & 0.01 & 0.92 & 1.38 & 1.86 & 0.09 & @xmath126 & 0.24 & 4.71 & 0.09 & m20se + 12n & 19.50 & 0.01 & 1.35 & 2.02 & 3.21 & 0.35 & @xmath127 & 0.51 & 5.43 & 0.20 & m77se + 13n & 19.57 & 0.01 & 0.73 & 1.07 & 1.24 & 0.06 & @xmath128 & 0.21 & 4.33 & 0.09 & m12se + 14n & 19.66 & 0.02 & 1.53 & 2.40 & 3.92 & 0.65 & @xmath129 & 0.90 & 5.75 & 0.36 & m84se + 15n & 19.72 & 0.02 & 0.60 & 1.02 & 0.99 & 0.11 & @xmath130 & 0.25 & 4.14 & 0.10 & +

given a closed orientable euclidean cone 3-manifold with cone angles and which is not almost product , we describe the space of constant curvature cone structures on with cone angles . we establish a regeneration result for such euclidean cone manifolds into spherical or hyperbolic ones and we also deduce global rigidity for euclidean cone structures .

math0510432

let @xmath0 be a closed orientable euclidean cone 3-manifold with cone angles @xmath4 . its singular locus @xmath5 is a trivalent graph consisting of @xmath6 circles and edges , that we enumerate from @xmath7 to @xmath6 . to those edges one can associate a multiangle @xmath8 , where @xmath9 is the angle of the @xmath10-th circle or edge of @xmath5 . let further @xmath11 denote the vector of lengths of singular circles and edges . we are interested in the space of cone structures on @xmath0 obtained by changing the multiangle . a _ hyperbolic cone structure _ on @xmath0 is a hyperbolic cone manifold @xmath12 with an embedding @xmath13 such that @xmath14 is a retract of @xmath15 . a euclidean cone 3-manifold @xmath0 is _ almost product _ if @xmath16 where @xmath17 is a two - dimensional cone manifold and @xmath18 is a finite subgroup . [ thm : main ] let @xmath0 be a closed orientable euclidean cone 3-manifold with cone angles @xmath19 . if @xmath0 is not almost product , then for every multiangle @xmath20 there exists a unique cone manifold structure of constant curvature in @xmath21 on @xmath0 with those cone angles . if all cone angles of @xmath0 are @xmath22 , then every point in @xmath23 is the multiangle of a hyperbolic cone structure on @xmath0 . if some of the cone angles is @xmath3 , then the subset @xmath24 of multiangles of euclidean cone structures is a smooth , properly embedded hypersurface that splits @xmath25 into 2 connected components @xmath26 and @xmath27 , corresponding to multiangles of spherical and hyperbolic cone structures respectively . furthermore , for each @xmath28 the tangent space of @xmath29 at @xmath30 is orthogonal to the vector of singular lengths @xmath31 . this theorem describes the structures corresponding to multiangles in the open cube . we can also describe the structures corresponding to multiangles contained in some parts of the boundary . for instance , we show that the multiangles contained in @xmath32 ( i.e. none of the angles is @xmath22 and at least one of them is zero ) are angles of hyperbolic cone structures . however , we do not give a description of structures in all points of @xmath33^q$ ] , as this would involve studying collapses at cone angle @xmath22 and deformations and regenerations of seifert fibered geometries . [ cor : sphericity ] let @xmath34 be an irreducible closed orientable @xmath35-orbifold . if there exists a euclidean cone structure @xmath0 on @xmath34 with cone angles strictly less than the orbifold angles of @xmath34 , then @xmath34 is spherical . to prove this corollary , we show in lemma [ lem : cnotap ] that the euclidean cone structure is not almost product , using the irreducibility of @xmath34 . by theorem [ thm : main ] we can obtain a spherical cone structure on @xmath0 by increasing any of the cone angles . we are using here that @xmath29 is orthogonal to the vector of singular lengths @xmath36 . by proposition [ prop : largerspherical ] the orbifold angles can be realized by a spherical cone structure . this corollary gives an alternative argument to the last step in the proof of the orbifold theorem in @xcite , which is more natural from the point of view of cone manifolds . d. cooper and s.p . kerckhoff have announced a different approach to the spherical uniformisation . a special case of this theorem was proved in @xcite , assuming that the singular locus was a knot and a technical hypothesis involving cohomology . even if some of the techniques of @xcite are used , this paper does not rely on it . now we describe the organization of the paper : we are interested in the rotational part of the holonomy of the euclidean cone manifold in @xmath37 ( in fact its lift to @xmath38 ) , that we denote by @xmath39 . some basic properties of this representation are studied in section [ sec : almostproduct ] . in order to deform the structure , we view @xmath37 as the stabilizer of a point in @xmath40 or @xmath41 , thus we study the local properties of the varieties of representations in the isometries of @xmath40 and @xmath41 around @xmath42 . this is done in section [ sec : varofrep ] , using the cohomological results of section [ sec : cohomologytangentbundle ] . in section [ section : regenerationstructures ] we give the conditions for a deformation of the representation @xmath39 to correspond to a path of hyperbolic or spherical cone manifolds . this is applied in section [ sec : pathsofstructures ] to construct paths of hyperbolic and spherical structures by deforming one of the cone angles . the analysis of the local parametrization of the variety of representations is completed in section [ sec : foldlocus ] , where all deformations of the structure are constructed . the global results are established in section [ sec : global ] . finally , in section [ sec : example ] we illustrate the main theorem for cone manifold structures on the 3-sphere with singular locus given by the whitehead link . [ [ acknowledgments ] ] acknowledgments + + + + + + + + + + + + + + + we thank bernhard leeb for helpful conversations on the contents of this paper and for sharing his ideas with us . given a euclidean cone manifold @xmath0 , its smooth part is denoted by @xmath43 . then @xmath44 has a non - complete and non - singular flat riemannian metric . in particular it has a developing map @xmath45 and a holonomy representation @xmath46 we consider the composition of @xmath47 with the projection @xmath48 . [ lem : almostproduct ] @xmath0 is almost product if and only if the image of @xmath49 is contained in @xmath50 . if @xmath0 is almost product then the lemma is straightforward . to prove the converse , we assume that the image of @xmath51 is contained in @xmath52 . since @xmath53 has two components , composition with @xmath54 defines a morphism @xmath55 and , if it is non - trivial , we take the corresponding covering . this induces a branched covering of @xmath0 , with branching locus contained in the edges and circles of @xmath5 that have cone angle @xmath22 . hence we may assume that the image of @xmath51 is contained in @xmath56 . the vertical foliation of @xmath57 invariant under @xmath56 induces an isometric flow of @xmath44 . it extends to a flow on the whole of @xmath0 , since this restriction of the holonomy implies that the components of @xmath5 are parallel to the leaves of the flow . if @xmath58 , then @xmath0 is a smooth flat manifold and the lemma is well known . hence we assume that @xmath59 . since the components of @xmath5 are leaves of the flow , either all leaves are closed or they accumulate on tori at constant distance from the components of @xmath5 . the latter is not possible , since the concentric tori of accumulation of fibers have principal curvatures of constant sign . hence the leaves of the flow are closed , and they are leaves of a seifert fibration , whose basis has a natural structure of a euclidean cone 2-manifold . one can construct a horizontal surface @xmath27 perpendicular to the flow , and we can enlarge it as much as possible to be complete . since @xmath27 has cone angles @xmath60 , the singular locus of @xmath27 is finite . thus @xmath27 is compact and therefore the seifert fibration is virtually a product . the euclidean cone manifold @xmath0 is _ seifert fibered _ if it admits a seifert fibration such that @xmath5 is a union of fibers . it is _ almost seifert fibered _ if it is the quotient of a seifert fibered cone manifold by a finite subgroup of isometries that preserve the fibration . a cone manifold which is almost seifert fibered but not seifert fibered admits a partition by circles and intervals , so that the end - points of the intervals lie in singular edges with cone angle @xmath22 . this corresponds to an orbifold seifert fibration . in particular , if cone angles are @xmath3 , then almost seifert fibered implies seifert fibered . [ cor : almostseifert ] the euclidean cone manifold @xmath0 is almost product iff it is almost seifert fibered . in particular , if all cone angles of @xmath0 are @xmath3 and its smooth part is not seifert fibered , then @xmath0 is not almost product . almost product implies trivially almost seifert fibered . assume now that @xmath0 is seifert fibered . then the smooth part of @xmath0 is a product . let @xmath61 represent the fiber of this fibration . since the components of @xmath5 are also fibers , the holonomy of @xmath62 is either a nontrivial translation or a screw motion with non - trivial translation length . since @xmath62 is central , the direction of this vector must be preserved . thus the image of @xmath51 is contained in @xmath56 . it follows that when @xmath0 is almost seifert fibered , then the image of @xmath51 is contained in @xmath53 , and by lemma [ lem : almostproduct ] @xmath0 is almost product . we consider the action by conjugation of @xmath37 on @xmath63 , the vector space of @xmath64 matrices with real entries . [ lem : invariantsubspace ] @xmath0 is not almost product if and only if the only subspace of @xmath65 invariant under @xmath66 is the space of diagonal matrices . as @xmath37-module acting by conjugation , we have the splitting : @xmath67 where @xmath68 is the set of diagonal matrices , @xmath69 is the subspace of antisymmetric ones and @xmath70 is the subspace of traceless symmetric matrices . the action on @xmath68 is trivial and this space is always invariant . the action on @xmath71 is equivalent to the usual action of @xmath37 on @xmath57 , and having a nonzero invariant vector in @xmath57 is equivalent for @xmath51 to be contained in @xmath56 . since symmetric matrices diagonalize orthogonally , matrices in @xmath72 have an orthogonal basis of eigenvectors . furthermore the sum of eigenvalues is zero , and therefore every non - trivial element has at least a one dimensional eigenspace . thus having a non - trivial invariant element of @xmath72 is equivalent for @xmath51 to have an invariant line , i.e. to be contained in @xmath53 . [ lem : cnotap ] let @xmath0 and @xmath34 be as in the situation of corollary [ cor : sphericity ] . then @xmath0 is not almost product . by hypothesis the cone angles of @xmath0 are @xmath3 . therefore if @xmath0 is almost product , then @xmath0 is a quotient of @xmath73 where @xmath74 is a euclidean turnover ( a cone manifold structure on @xmath75 with three cone points of cone angles satisfying @xmath76 ) . this cone manifold @xmath17 embeds in @xmath0 and induces an essential spherical 2-orbifold in @xmath34 , contradicting the irreducibility of @xmath34 . we assume from now on that @xmath44 is the smooth part of a closed euclidean cone manifold @xmath0 which is not almost product . the latter assumption may be viewed as a nondegeneracy condition and is in that respect similar to the assumption of not being seifert fibered in the deformation theory of spherical cone - manifolds , cf . @xcite , @xcite . in the euclidean case the flat bundle of infinitesimal isometries @xmath77 contains the bundle of infinitesimal translations @xmath78 as a parallel subbundle . the inclusion @xmath79 fits into a short exact sequence of flat vectorbundles and connection - preserving maps @xmath80 where @xmath81 carries the _ quotient _ connection ( @xmath82 is _ not _ parallel ! ) . in fact @xmath83 is again isomorphic to @xmath84 as a flat bundle . this can for instance be seen if we look at the corresponding short exact sequence of @xmath85-modules @xmath86 now we have isomorphisms of @xmath85-modules @xmath87 , which in turn yield isomorphisms of flat bundles @xmath88 . note that the first isomorphism is particular to our 3-dimensional situation . we will freely identify @xmath84 with either @xmath89 or @xmath83 from now on . + the translational part of a euclidean isometry @xmath90 w.r.t . a base point @xmath91 @xmath92 gives rise to a cocycle @xmath93 twisted by @xmath94 , i.e. @xmath95 for @xmath96 . the _ holonomy cocycle _ is the composition @xmath97 for some fixed @xmath98 . this is a cocycle twisted by @xmath51 , i.e. @xmath99 for @xmath100 . the cohomology class of this cocycle is independent of @xmath101 and is denoted by @xmath102\in h^1(\pi_1 m ; \r^3_{rot \circ hol})\,.\ ] ] let @xmath103 be closed and @xmath104 a loop based at @xmath105 . we define the integral of @xmath106 along @xmath104 as @xmath107 where @xmath108 denotes parallel transport along @xmath104 from @xmath109 to @xmath110 and we identify @xmath111 with @xmath112 by developing @xmath44 on @xmath112 . for @xmath113 closed the assignment @xmath114 defines a group cocycle @xmath115 , the so - called period cocycle of @xmath106 . the period map @xmath116 descends to an isomorphism in cohomology @xmath117 & \lmt [ z_{\omega}]\,.\end{aligned}\ ] ] observe that on each side there is a canonically defined cohomology class . on the left hand side we have @xmath118 . @xmath119 is parallel and gives rise to a class @xmath120 \in h^1(m;tm)$ ] . on the right hand side we have the class of the holonomy cocycle @xmath121 . in fact , these classes coincide modulo the period map and correspond geometrically to a global rescaling of @xmath44 : [ lem : pid = v ] @xmath122)=v$ ] and the image of @xmath121 in @xmath123 is tangent to a deformation , which globally rescales the euclidean metric on @xmath44 . let @xmath104 be a loop based at @xmath105 and @xmath124 a lift of @xmath104 to @xmath125 . we compute @xmath126 which shows that the periods of @xmath119 reproduce the holonomy cocycle defined with respect to @xmath127 . furthermore for @xmath128 let @xmath129 then @xmath130 and @xmath131 is the holonomy representation of @xmath44 with rescaled euclidean metric @xmath132 . we obtain @xmath133 and consequently @xmath134 . this means the corresponding path of characters is precisely tangent to @xmath121 , cf . the later discussion in section [ sec : varofrep ] . let @xmath135 denote the @xmath136-tube around @xmath5 in @xmath0 minus the singular locus itself . let @xmath137 denote a compact core of @xmath44 . @xmath138 is a disjoint union of tori and surfaces of higher genus . note that the pair @xmath139 is homotopy equivalent to the pair @xmath140 , by homotopy invariance we may identify the corresponding cohomology groups . + if @xmath141 are cylindrical coordinates around the @xmath10-th edge and @xmath142 is a function with @xmath143 near @xmath144 and @xmath145 near @xmath146 , we define as in @xcite @xmath147 as in @xcite we obtain : [ l2 ] the forms @xmath148 are bounded on @xmath135 , in particular @xmath149 . geometrically , the classes @xmath150 $ ] correspond to deformations of @xmath135 changing the length of the @xmath10-th edge . these deformations are independent . therefore the following does not come as a surprise : [ lem : nontriv ] the classes @xmath150 $ ] are linearly independent in @xmath151 , furthermore the relation @xmath120|_{\partial \overline{m}}=\sum_i [ \omega^i_{len}]$ ] holds . the first assertion follows as in @xcite . the periods of both forms @xmath119 and @xmath152 reproduce the holonomy cocycle restricted to @xmath153 . since the period map descends to an isomorphism in cohomology , the result follows . @xmath154 . this group is isomorphic to the subspace of @xmath155 invariant under @xmath51 , which is trivial since we assume that @xmath0 is not almost product ( see the proof of lemma [ lem : invariantsubspace ] ) . from the long exact cohomology sequence we obtain : @xmath156 and the map @xmath157 is an injection . to compute @xmath158 we use the @xmath149-cohomology : @xmath159 . our starting point is the following theorem of @xcite , where the hypothesis about cone angles is used : [ thm : l2 ] if @xmath44 is the smooth part of a closed euclidean cone manifold with cone angles @xmath60 , then @xmath160 [ cor : l2 ] @xmath161 . since every element of @xmath159 is represented by a parallel form , this cohomology group is isomorphic to the space of elements in @xmath162 which are invariant under the action of @xmath51 by conjugation . by lemma [ lem : invariantsubspace ] , this subspace is one dimensional . let @xmath163 denote a compact core of @xmath44 . by homotopy invariance @xmath164 . [ lem : inj ] the map @xmath165 is an injection . looking at the exact sequence of the pair @xmath166 , it suffices to show that the image of the map @xmath167 is zero . to prove it , we observe that @xmath168 factors through @xmath159 . by corollary [ cor : l2 ] , @xmath159 is one - dimensional and its generator is represented by the identity of @xmath84 . therefore the image of @xmath168 is contained in the span of @xmath120 $ ] . by lemma [ lem : nontriv ] , @xmath120 $ ] restricts to a nontrivial class in @xmath151 . on the other hand , by exactness the image of @xmath168 is contained in the kernel of @xmath169 , hence it is zero . as a corollary of the proof above we notice : [ l2:inj ] the map @xmath170 is an injection . we may thus view @xmath149-cohomology as a subspace of ordinary cohomology in degree 1 . [ lem : h1q ] @xmath171 . by lemma [ lem : inj ] , the long exact sequence of the pair and poincar duality , we have a short exact sequence : @xmath172 since @xmath173 and @xmath174 are poincar dual , @xmath175 in addition , @xmath176 ( see @xcite or @xcite ) . continuing the long exact sequence we get : [ lem : h2isomorphic ] the inclusion induces an isomorphism @xmath177 . let @xmath178 denote the cohomology of the @xmath179-module @xmath57 with the action induced by @xmath51 . notice that the induced flat bundle over @xmath44 is isomorphic to @xmath84 . [ lem : cohomologies ] there is a natural map @xmath180 which is an isomorphism for @xmath181 and an injection for @xmath182 . when @xmath7 , this map is the inverse of the isomorphism @xmath183 in lemma [ lem : pid = v ] . there is a natural isomorphism @xmath184 , where @xmath185 denotes any aspherical cw - complex with the same fundamental group as @xmath44 , and @xmath186 also denotes the flat bundle over @xmath185 with fiber @xmath187 and holonomy @xmath51 . the space @xmath185 can be constructed from @xmath44 itself by adding cells of dimension @xmath188 . thus the cohomology of the pair @xmath189 is trivial in dimension @xmath190 , and the lemma follows easily from the long exact sequence in cohomology for the pair @xmath189 . [ cor : injectiontoboundary ] let @xmath191 denote the components of @xmath192 . the restriction induces an injection : @xmath193 the result follows from corollary [ lem : h2isomorphic ] , lemma [ lem : cohomologies ] and the asphericity of the surfaces @xmath194 . since @xmath44 is three dimensional by a theorem of culler , cf . @xcite , @xmath47 lifts to a representation @xmath195 and similarly @xmath51 lifts to @xmath196 where @xmath197 is a lift of @xmath94 . we will drop the notational distinction between @xmath198 and @xmath47 , resp . between @xmath199 and @xmath94 from now on . let @xmath200 and @xmath201 denote the varieties of representations in @xmath202 and @xmath203 respectively . let @xmath204 and @xmath205 be the varieties of characters . let @xmath206 denote the character of @xmath39 . we recall some general facts about spaces of representations : let @xmath207 be a finitely generated discrete group and @xmath208 a lie group . we equip @xmath209 with the compact - open topology . once and for all we fix a presentation @xmath210 . without loss of generality we may assume that the set of generators contains with any of its elements its inverse . the relations @xmath211 determine functions @xmath212 which identify @xmath213 with the subset @xmath214 . if @xmath208 is a real algebraic group ( as for example @xmath215 or @xmath216 ) , the @xmath217 are polynomial functions and @xmath213 acquires the structure of a real algebraic set . similarly , if @xmath208 is a complex algebraic group ( for example @xmath218 ) , then @xmath213 is a complex algebraic set . let @xmath219 be the semidirect product of @xmath208 and @xmath220 defined via @xmath221 . let @xmath222 be given . a simple but important observation is that @xmath223 if and only if @xmath224 . the relations therefore determine functions @xmath225 such that @xmath226 and @xmath227 . note that for @xmath228 fixed the functions @xmath229 are linear . if we identify @xmath230 with @xmath231 via right translation , then we have : @xmath232 . the differential of the group multiplication @xmath233 at @xmath234 considered as a map @xmath235 via right translation is given by @xmath236 on the other hand the group multiplication law in @xmath237 reads @xmath238 which proves the claim in this case . the general case is easily reduced to this one . in particular if the @xmath217 cut out @xmath213 transversely , then @xmath239 is identified with @xmath240 ) . in general , if @xmath241 is a smooth point , there is a linear injection @xmath242 which identifies @xmath239 with a subspace of @xmath243 . this injection is known as _ weil s construction _ , and it extends naturally to the non - smooth case by using the zariski tangent space . a cocycle @xmath244 is a coboundary , i.e. @xmath245 , if and only if @xmath244 is tangent to the orbit of the action of @xmath208 on @xmath213 via conjugation . [ rep_var_smooth ] @xmath246 is a smooth point with tangent space @xmath247 . similarly , @xmath248 is a smooth point with tangent space @xmath249 . there is an infinite sequence of obstructions for a cocycle @xmath250 to be integrated into an actual path of representations tangent to @xmath244 , cf . these obstructions live in @xmath251 . using the injection of corollary [ cor : injectiontoboundary ] and the fact that the obstructions are natural , we conclude that they vanish , because they vanish in @xmath252 . then using artin s theorem , cf . @xcite , they are in fact integrable . thus every element in @xmath247 is integrable , and therefore @xmath39 is a smooth point of @xmath253 with tangent space @xmath247 . for @xmath254 the same argument applies since @xmath255 is the complexification of the module @xmath256 . @xmath257 is a smooth point of local real dimension @xmath6 and tangent space isomorphic to @xmath258 . similarly , @xmath259 is a smooth point of local complex dimension @xmath6 and tangent space isomorphic to @xmath260 . since @xmath39 is irreducible , @xmath261 in each case . since @xmath202 is compact ( resp . the action of @xmath262 is proper on the irreducible part of @xmath263 , cf . * lemma 6.24 ) ) and the tangent space to the orbit is given by @xmath247 ( resp . by @xmath249 ) , the result follows with lemma [ rep_var_smooth ] . the fiber of the map @xmath264 at @xmath265 is the the space of cocycles @xmath266 , which is given by a system of linear equations whose coefficients depend continuously on @xmath228 . therefore its dimension is an upper semi - continuous function of @xmath228 in general . if @xmath246 is a smooth point with @xmath267 ( which is the case if the @xmath217 cut out @xmath253 transversely or for @xmath268 according to lemma [ rep_var_smooth ] ) , then @xmath269 injects into @xmath266 for @xmath228 in a neighbourhood of @xmath39 via weil s construction ( [ eqn : weil ] ) : @xmath270 and hence @xmath271 is locally the projection of a vector bundle . more precisely we have : @xmath272 is isomorphic to the tangent bundle @xmath273 near @xmath39 . the map @xmath274 is a vector bundle isomorphism near @xmath39 . @xmath275 is a smooth point . let @xmath276 denote the character of @xmath275 . [ vectorbundle ] @xmath276 is a smooth point of local real dimension @xmath277 . the induced map @xmath278 is locally the projection of a vector bundle isomorphic to the tangent bundle @xmath279 near @xmath39 . let us digress into a more general situation first : let @xmath208 be a compact lie group , @xmath44 a smooth manifold ( not necessarily compact ) and @xmath280 a smooth free action . the associated infinitesimal action is the lie algebra homomorphism @xmath281 defined by @xmath282 . since the original action was free , this homomorphism is injective and we will identify @xmath231 with its image in @xmath283 . let @xmath219 be the semidirect product of @xmath208 and @xmath284 defined via @xmath285 . we extend the natural action of @xmath208 on @xmath84 via the differential to an action of the group @xmath286 in the following way : @xmath287 the relation @xmath288 ensures that this defines a group action . this action is free since the original action was free , it is clearly proper . [ extendedaction ] @xmath289 . the fibers of the map @xmath290 are precisely the orbits of the action of the group @xmath237 on @xmath84 . we return to the situation of proposition [ vectorbundle ] : clearly @xmath291 via @xmath292 as @xmath202-modules . for a representation @xmath293 and @xmath294 we have @xmath295 therefore the map @xmath296 is @xmath216-equivariant and lemma [ extendedaction ] yields the proposition . from the long exact cohomology sequence and proposition [ vectorbundle ] we obtain : [ shortexact ] there is a short exact sequence @xmath297 modulo the choice of a splitting of the exact sequence in corollary [ shortexact ] , an infinitesimal deformation of the holonomy of the euclidean structure on @xmath44 is therefore determined by the infinitesimal deformation of its rotational part and of its translational part , and both can be independently prescribed . let @xmath298 a system of meridians for @xmath179 ( i.e. one for each component of @xmath5 ) . we define the angle function @xmath299 in a neighborhood @xmath300 of @xmath301 , so that @xmath302 . it is related to the trace by the equality : @xmath303 in particular @xmath304 is analytic . we extend @xmath304 to a neighborhood in @xmath305 as a complex analytic function : @xmath306 so that the complex length of @xmath307 is @xmath308 ( i.e. a translation of length @xmath309 plus a rotation of angle @xmath310 ) . + the differentials @xmath311 live in the cotangent space to the varieties of characters , thus @xmath312 [ prop : kerdmu ] @xmath313 , where @xmath121 is the class of the holonomy cocycle . to check that @xmath314 , we observe that @xmath315 maps @xmath121 to zero , because @xmath316 is a rotation with a fixed axis . in particular @xmath311 evaluated at @xmath121 is zero . to see the other inclusion , we use that @xmath317 , \ldots , [ \omega^q_{len } ] \right\rangle \,,\ ] ] where we view @xmath158 as a subspace of @xmath318 via lemma [ lem : inj ] . since the forms @xmath319 are @xmath149 near the singular locus according to lemma [ l2 ] , we conclude that @xmath320 , which we view as a subspace of @xmath158 via corollary [ l2:inj ] . since @xmath159 is spanned by @xmath120 $ ] the result follows . [ cor : dimdmu ] @xmath321 along this section @xmath322 denotes either @xmath41 or @xmath40 . the stabilizer of a point @xmath323 is isomorphic to @xmath324 . thus we view @xmath39 as a representation of @xmath179 in this stabilizer . we consider a path of characters @xmath325 with @xmath326 . assume that this path is differentiable to the right at @xmath327 . the derivative @xmath328 is an element of @xmath329 , where @xmath284 is the lie algebra of @xmath330 ( using weil s construction ) . the lie algebra @xmath284 decomposes into rotational and translational part with respect to @xmath101 : @xmath331 where @xmath332 is the lie algebra of the stabilizer of @xmath323 . thus @xmath333 [ thm : regeneration ] let @xmath334 be a path in @xmath335 with @xmath326 and differentiable to the right at @xmath327 . if @xmath336 , then @xmath337 is the holonomy character of a @xmath322-structure on @xmath44 , for @xmath338 and some @xmath339 . if in addition @xmath340 is a rotation for each meridian @xmath341 , then the structure on @xmath44 completes to a cone manifold structure . consider @xmath342 , the space of constant sectional curvature @xmath343 , where @xmath344 is the curvature of @xmath322 , for @xmath345 . equivalently , @xmath342 is @xmath322 with the metric tensor scaled by @xmath346 . we fix a base point @xmath101 in @xmath342 independently of @xmath347 , so that the pointed euclidean space @xmath348 is the limit of @xmath349 when @xmath350 , and consider the union @xmath351}\mathbf x_t.\ ] ] we equip @xmath352 with a manifold structure , using the local charts given by the exponential maps at @xmath101 and the parameter @xmath345 , after identifying isometrically @xmath353 for every @xmath354 . notice that the choice of @xmath101 is relevant for the topology of @xmath352 . choose a smooth path of representations @xmath131 with character @xmath337 so that @xmath355 and @xmath356 . for every @xmath357 , the action of @xmath358 on @xmath342 for @xmath359 extends continuously to the action of @xmath360 on @xmath361 for the @xmath362-topology . notice that @xmath131 acts on @xmath342 isometrically , since rescaling the metric does not change the isometry group . the previous lemma provides an action on @xmath352 , which is isometric on each @xmath342 . we start describing the local coordinates . let @xmath363 denote the riemannian exponential . using the isometric identification @xmath364 , we have @xmath365 where @xmath366 . thus , the ( inverse of ) the local charts is given as follows . given an open set @xmath367 , @xmath368 parametrizes a subset of @xmath352 via the map : @xmath369 to prove the lemma it suffices to show that @xmath370 uniformly for @xmath371 in a compact set for the @xmath362-topology . the lie algebra of @xmath372 decomposes as a sum @xmath373 , where @xmath374 is the subalgebra of infinitesimal rotations around @xmath101 and @xmath375 the subspace of infinitesimal translations with respect to @xmath101 . there is an isometric identification @xmath376 , so that for any @xmath377 : @xmath378 here we view @xmath379 when we write @xmath380 and @xmath381 for @xmath382 . according to global cartan s decomposition , we write @xmath383 , where @xmath384 and @xmath385 belongs to the stabilizer of @xmath101 in @xmath386 . notice that @xmath387 as @xmath350 by hypothesis . using this notation : @xmath388 where @xmath389 is an analytic function . here we use that @xmath390 together with the campbell - hausdorff formula . using again this formula and the fact that @xmath391 , we get : @xmath392 where @xmath393 is also analytic . hence @xmath394 when @xmath350 this converges to @xmath395 uniformly for @xmath121 in a compact subset of @xmath187 , because @xmath396 and @xmath397 is a lift of @xmath398 , by hypothesis . to prove convergence in the @xmath362-topology , we write @xmath399 with @xmath400 , and @xmath401 small . the previous calculation yields easily : @xmath402 we can compute the derivative with respect to @xmath136 when @xmath403 and we get @xmath362-convergence . we modify slightly goldman s construction to deform the structure . let @xmath163 be a compact core of @xmath404 . consider @xmath405 , which is a bundle over @xmath163 with fiber @xmath406 . it is in fact a union of bundles @xmath407 with fiber @xmath408 . each @xmath409 has a natural flat connection that varies continuously with @xmath347 . the developing map of the euclidean structure induces a section @xmath410 with values in @xmath361 transverse to the flat connection . since the image of @xmath411 is compact , we compose it with the flow @xmath412 of the vector field tangent to the direction of @xmath347 ( this is defined globally when @xmath413 but not when @xmath414 , because @xmath361 is not homeomorphic to @xmath342 ) . for small values of @xmath359 , @xmath415 is a section , still transverse to the flat connection by @xmath362-continuity , hence inducing a developing map . this provides a structure on @xmath163 , that can be completed by controlling its behavior on @xmath416 , using the hypothesis about the meridians . the volume of these cone manifolds is increasing with @xmath347 , because the starting euclidean structure is viewed as totally degenerate with volume zero . in fact we get more precisely : [ prop : schlafli ] let @xmath417 be the family of cone manifolds constructed in theorem [ thm : regeneration ] of constant curvature @xmath418 . assume moreover that the path of characters is analytic . then @xmath419 where the sum runs over the edges and circles @xmath420 of @xmath5 . _ sketch of proof . _ the construction of developing maps in the proof of theorem [ thm : regeneration ] can be made so that we have a set of points @xmath421 such that the balls @xmath422 cover @xmath0 , where @xmath211 is much smaller that the injectivity radius , and @xmath423 varies analytically ( we assume that the path of characters is analytic ) . let @xmath424 the cone structure with holonomy @xmath337 and define @xmath425 by construction , this @xmath426 is an analytic family of polyhedra in the space @xmath322 . analyticity implies that the topological type of @xmath426 changes in a discrete subset of times @xmath347 . thus we may apply schlfli s formula to them . adding all the terms , we get the formula of the proposition , because the contribution of nonsingular edges is trivial . see ( * ? ? ? 4.2 ) for further details . [ rem : trans ] in the hyperbolic case , @xmath427 and @xmath428 . therefore in this case @xmath429 can be viewed as the imaginary part . in the spherical case , @xmath430 and the lie algebra of the stabilizer of a point @xmath431 is conjugate to the diagonal subalgebra of @xmath432 . therefore in this case @xmath433 can be viewed as the difference between factors . [ rem : rotation ] for every @xmath228 with character @xmath434 , @xmath435 for every @xmath436 with character @xmath437 , @xmath438 up to changing the indices , by corollary [ cor : dimdmu ] we may assume that @xmath439 are linearly independent . thus the set @xmath440 is a smooth complex curve in a neighborhood @xmath300 of @xmath301 . notice that the class of the holonomy cocycle @xmath121 is the tangent vector to @xmath441 at @xmath301 , by proposition [ prop : kerdmu ] . thus , using remark [ rem : trans ] we have : to every path @xmath442 with @xmath443 and @xmath444 , theorem [ thm : regeneration ] applies . [ lem : deg2 ] the restricted map @xmath445 is a branched covering of degree two . by proposition [ prop : kerdmu ] , @xmath446 , thus @xmath447 is either constant or a branched covering of degree @xmath448 . seeking a contradiction , assume that @xmath447 is constant , then we have a path of characters in the curve @xmath449 to which we apply theorem [ thm : regeneration ] . since each @xmath450 stays constant on this curve , this path corresponds to hyperbolic cone manifolds with constant cone angle . thus by schlfli s formula ( proposition [ prop : schlafli ] ) they have constant volume , contradicting the fact that the volume increases from zero . thus @xmath447 is a branched covering of degree @xmath451 . to prove that the degree @xmath452 is precisely @xmath453 , we assume that @xmath454 and seek again a contradiction . we look at the inverse image or real points @xmath455 , because for characters here , the image of @xmath456 is a rotation . this inverse image is a graph with @xmath457 branches starting at @xmath301 , and the angle between the branches is @xmath458 . thus two of the branches are real and the remaining @xmath459 have nontrivial imaginary part . hence theorem [ thm : regeneration ] applies to those @xmath459 branches . in addition , if @xmath460 , then there are branches for which @xmath456 is strictly larger than @xmath461 and branches for which @xmath456 is strictly less than @xmath461 . since @xmath456 is the cone angle , we have constructed regenerating families of hyperbolic cone manifolds with both increasing and decreasing cone angles , contradicting schlfli s formula . [ cor : hyperbolicregeneration ] there is a family of hyperbolic cone structures obtained by decreasing @xmath462 . since @xmath447 is a branched covering of degree 2 , @xmath455 has 4 branches starting at @xmath301 . two of them are real , and to the other two one applies theorem [ thm : regeneration ] . the cone angle @xmath462 of those branches must decrease by schlfli s formula . these two branches correspond to complex conjugate representations , i.e. with opposite orientations , because changing the sign of @xmath121 corresponds to changing the orientation . there is a family of spherical cone structures obtained by increasing @xmath462 . to construct the spherical structure we consider @xmath463 this is a real analytic curve in a neighborhood @xmath464 of @xmath301 . it is the set of real points of @xmath441 . by lemma [ lem : deg2 ] , the map @xmath465 is locally equivalent to the map @xmath466 in a neighborhood of @xmath467 . thus the inverse fiber of the map @xmath468 consists of pairs of points , except for @xmath461 , which consists of a single point . let @xmath469 be two subintervals of @xmath470 such that @xmath471 and @xmath472 . we consider the set @xmath473 because @xmath474 is the condition that guarantees that @xmath475 is the character of a representation @xmath228 in @xmath476 so that @xmath477 is a rotation , by remark [ rem : rotation ] . theorem [ thm : regeneration ] applies to the latter two branches , because the difference of tangent vectors gives the translational part , that is necessarily a non - zero multiple of @xmath121 ( see remark [ rem : trans ] ) . if we replace @xmath478 by @xmath479 ( i.e. we change the order of the factors ) then we get the same structures with different orientation . [ rem : furtherregenerations ] we have constructed regenerations by deforming the cone angle @xmath462 because we assumed that the forms @xmath480 are linearly independent . below in corollary [ cor : linearrelation ] we prove that the linear relation satisfied by these forms is : @xmath481 thus we can regenerate by deforming any of the cone angles . now we analyze the behavior of the real analytic map @xmath482 in a neighborhood of @xmath301 . [ lem : equivalencex2 ] in a neighborhood of @xmath301 , @xmath483 is equivalent to the map @xmath484 in a neighborhood of the origin via a real analytic change of coordinates at both the source and at the target . by corollary [ cor : dimdmu ] and up to permuting coordinates , we may assume that the forms @xmath485 are linearly independent . thus the set @xmath486 is a smooth ( real analytic ) curve in a neighborhood @xmath300 of @xmath301 . by lemma [ lem : deg2 ] we can choose real analytic coordinates @xmath487 in a neighborhood of @xmath301 with @xmath488 such that @xmath489 and @xmath490 for @xmath491 . note that the curve @xmath492 corresponds to the @xmath493-axis in these coordinates . since @xmath494 and @xmath495 , the implicit function theorem yields that the set @xmath496 is locally around @xmath327 the graph of a real analytic function @xmath497 . let @xmath498 be the real analytic function defined by @xmath499 let further @xmath500 be the unique real analytic function satisfying @xmath501 and @xmath502 . then the map @xmath90 defined by @xmath503 is a local diffeomorphism . we obtain @xmath504 and @xmath505 for @xmath491 . with the local diffeomorphism @xmath506 defined by @xmath507 we obtain @xmath508 . the same result holds true for the complex analytic extension @xmath509 just by composing with the complex analytic extensions of the coordinate changes . the _ fold locus _ @xmath510 is the set of points where @xmath483 is not a local diffeomorphism . in the coordinate system of lemma [ lem : equivalencex2 ] the fold locus is the set @xmath511 , hence : [ cor : jmanifold ] there exists a neighborhood @xmath300 of @xmath301 such that both @xmath512 and @xmath513 are codimension one submanifolds . in addition , for each @xmath514 , @xmath515 . each @xmath516 is the rotational part of the holonomy of a euclidean cone manifold structure . the translational part lies in @xmath517 . when we deform a character @xmath518 in @xmath519 we can also deform continuously the vector @xmath121 in @xmath520 , which is a one dimensional subspace that varies continuously , by lemma [ lem : equivalencex2 ] . this corresponds to deforming continuously the holonomy @xmath47 by a family of representations of @xmath179 in @xmath216 . the condition @xmath521 ensures that the image of meridians are rotations ( cf . * prop . 9.6 ) ) . since those representations map the meridians to rotations , it follows that they correspond to holonomy representations of euclidean cone manifolds . [ cor : eeucl ] points of @xmath522 are angles of euclidean cone manifold structures on @xmath0 . let @xmath523 denote the lengths of the singular circles and components of @xmath0 . those are unique up to homothety . [ prop : lnormal ] the vector @xmath36 is normal to @xmath522 at @xmath524 . this is a consequence of schlfli s formula ( prop . [ prop : schlafli ] ) . [ cor : linearrelation ] on @xmath525 , @xmath526 . it holds on @xmath527 by proposition [ prop : lnormal ] . in addition , it also holds on @xmath528 $ ] . by lemma [ lem : equivalencex2 ] and proposition [ prop : kerdmu ] , those spaces generate the whole @xmath529 . it follows from this corollary that we can regenerate into hyperbolic or spherical cone manifold structures by decreasing or increasing any of the cone angles , the geometry depends on the sign of @xmath530 . ( see remark [ rem : furtherregenerations ] . ) [ cor : lrecm ] let @xmath0 be a closed euclidean cone manifold with cone angles @xmath60 which is not almost product . then deformations up to dilations of @xmath0 into euclidean cone manifold structures are parametrized by the @xmath6 cone angles in a manifold @xmath29 of dimension @xmath531 and transverse to the vector of singular lengths @xmath532 . the previous analysis determines all representations of @xmath179 into @xmath216 up to conjugation that map meridians to rotations , i.e. pairs @xmath533 such that @xmath534 and further @xmath535 , see figure [ fig : foldlocus ] . [ cor : lrr ] let @xmath0 be as above . then deformations of @xmath0 into constant curvature structures up to dilations are parametrized by the @xmath6 cone angles , and the type of geometry depends on the side of @xmath29 . the analysis of this section and previous ones determine the structure of the spaces of characters of @xmath179 in @xmath330 . namely , spherical regenerations correspond to pairs of @xmath202-characters @xmath475 such that @xmath536 . according to lemma [ lem : equivalencex2 ] , these pairs are determined by the value of @xmath483 , which is the multiangle of the structure , cf . figure [ fig : foldlocus ] . hyperbolic regenerations correspond to @xmath262-characters @xmath518 with @xmath537 , which are not @xmath202-characters , cf . figure [ fig : hypreg ] . again such a character is determined up to complex conjugation by the value of @xmath483 . next , we shall show that the deformations of the structures are uniquely determined by the deformations of the characters , adapting an argument from @xcite . let @xmath163 be a compact core of @xmath404 , so that we will consider deformations of structures on @xmath163 , knowing that the structure on the surface @xmath416 is determined by the holonomy , and the structure on @xmath538 is given by taking the cone . using the notation of section [ section : regenerationstructures ] , @xmath539 is the union of flat bundles @xmath409 with fiber @xmath342 the space of constant curvature @xmath540 , and holonomy the deformation of the euclidean holonomy representation . since @xmath409 is flat , there is a horizontal foliation , which is also used to define locally a projection onto the fibers @xmath342 . the euclidean structure is induced by a section @xmath541 transverse to the horizontal foliation . in section [ section : regenerationstructures ] , @xmath542 is deformed to @xmath543 , following the flow in the time direction on @xmath29 . let @xmath544 be another deformation , i.e. a section of class @xmath362 such that @xmath545 and @xmath546 . we claim that , for @xmath547 sufficiently small , there exist a diffeomorphism @xmath548 of @xmath44 isotopic to the identity so that the developing maps corresponding to @xmath549 and @xmath550 are equivalent ( namely , the sections @xmath549 and @xmath550 composed with the local projections to @xmath342 differ by isometries of @xmath551 ) . following @xcite , we take a foliated neighborhood @xmath552 of @xmath553 in @xmath554 , so that the intersection of leaves with @xmath552 are balls , and consider @xmath555 in @xmath29 , via the flow as in section [ section : regenerationstructures ] . for @xmath347 sufficiently small , closeness and transversality imply that the image of both @xmath556 and @xmath549 meet each horizontal leave in @xmath557 precisely once . the intersection with the horizontal leaves can be used to define the map @xmath548 as in @xcite . in this section we finish the proof of theorem [ thm : main ] . we first establish some properties of the set of euclidean angles and recall some results about the set of hyperbolic and spherical ones . let @xmath29 denote the set of multiangles with a euclidean cone manifold structure . [ prop : completesubmfld ] @xmath29 is a properly embedded hypersurface of @xmath23 . first we show that every euclidean cone manifold structure on @xmath0 with cone angles in @xmath558 is not almost product . this will imply that @xmath29 is a @xmath531 submanifold , by corollary [ cor : eeucl ] . by contradiction , we assume that there exists such a euclidean cone structure on @xmath0 which is almost product . since all cone angles are in @xmath558 , this implies that @xmath0 is seifert fibered and @xmath5 is a union of fibers . by corollary [ cor : almostseifert ] , in this case @xmath0 is almost product for any cone angle , contradicting the hypothesis of theorem [ thm : main ] . to prove properness , we show that if @xmath559 is a sequence in @xmath29 converging to @xmath560 , then @xmath561 . let @xmath562 be the corresponding sequence of cone structures on @xmath0 with multiangles @xmath559 . we can rescale them so that @xmath563 . since the cone angles are uniformly @xmath3 , by theorem 5.3 and corollary 5.4 of @xcite the cone manifolds @xmath562 are uniformly thick : they have a base point @xmath564 with @xmath565 . thus they have a convergent subsequence to a euclidean cone manifold of diameter 1 , hence a cone structure on @xmath0 with cone angles @xmath566 . in ( * ? ? 3 ) the following proposition is proved , which applies to @xmath0 under the assumptions of the paper . [ prop : largerspherical ] if @xmath567^q$ ] is a multiangle of a spherical cone structure on @xmath0 which is not seifert fibered , then every @xmath568^q$ ] with @xmath569 is the multiangle of a spherical cone structure with cone angles @xmath570 . if @xmath567^q$ ] is a multiangle of a hyperbolic cone structure on @xmath0 , then every @xmath568^q$ ] with @xmath571 is the multiangle of a hyperbolic cone structure with cone angles @xmath570 . [ rem : smallness ] when all cone angles are @xmath22 , the orbifold is spherical and hence small ( without essential 2-suborbifolds ) . in particular all turnovers in @xmath0 are compressible or boundary parallel . before proving theorem [ thm : main ] , we need to establish incompatibility between structures of different sign . we normalize the constant curvature to be @xmath572 , @xmath327 or @xmath573 . [ lem : incompatibility ] given @xmath574^q$ ] , @xmath575 can not be the multiangle of two cone structures with different normalized constant curvatures ( @xmath572 , @xmath327 or @xmath573 ) . seeking a contradiction , assume for instance that @xmath576^q$ ] is the multiangle of a hyperbolic and euclidean cone structure simultaneously . using corollary [ cor : hyperbolicregeneration ] , the euclidean cone structure can be regenerated to hyperbolic ones , which , by construction , have arbitrarily small diameter , and also cone multiangle arbitrarily close to @xmath575 . on the other hand , when we perturb the hyperbolic cone structure with cone multiangle @xmath575 , we find hyperbolic cone manifolds with the same multiangles but diameter bounded below away from zero . this contradicts the global rigidity of hyperbolic cone manifolds proved in ( * ? ? ? the same argument works for spherical and euclidean ( global rigidity in the spherical case is the content of ( * ? ? ? * thm . 1.7 ) ) . finally we deal with the case that @xmath575 is simultaneously the multiangle of a hyperbolic and spherical structure on @xmath0 . consider @xmath577 sufficiently large so that @xmath578 for @xmath579 . by proposition [ prop : largerspherical ] , @xmath580 is the multiangle of hyperbolic cone manifold , hence a hyperbolic orbifold . notice that this orbifold can not be spherical , because its fundamental group is infinite . consider now a path of cone multiangles between @xmath575 and @xmath580 which is decreasing along each component . by proposition [ prop : largerspherical ] every multiangle in this path corresponds to a hyperbolic cone structure . however , sphericity must fail at some multiangle of this path , because @xmath580 is not the multiangle of a spherical structure . by the arguments in the proof of proposition [ prop : completesubmfld ] , when sphericity fails we find precisely a euclidean structure , and we have reduced to the first case of incompatibility between hyperbolic and euclidean structures . the following proposition is also used in the proof of theorem [ thm : main ] . [ prop : angle0 ] every multiangle in @xmath581 with at least one angle @xmath327 is the multiangle of a hyperbolic cone structure . by ( 1.5 ) , the smooth part of @xmath0 is hyperbolic . applying thurston s dehn filling theorem , we have the proposition for a neighborhood of the origin . to cover the rest of the multiangles , we just have to enlarge some of the angles , keeping the other ones equal to zero ( i.e. complete cusps ) . thus we have a lower bound on the cone angles , and in addition the diameter is infinity . applying the results of @xcite , we can enlarge each one of the cone angles up to @xmath3 . notice that some convergence results of @xcite apply to cone manifolds without essential turnovers . in our case there are no turnovers by remark [ rem : smallness ] except if we allow cusped turnovers ( i.e.turnovers with some cone angle 0 ) . however cusped turnovers are not a problem in those arguments , because all cone angles are @xmath3 , and therefore they can not converge to a euclidean cone 2-manifold . [ rem : opennessat0 ] the hyperbolic structures of the previous proposition can be deformed in a neighborhood in @xmath581 . this remark can be easily proved adapting the arguments of the proof of hyperbolic dehn filling for orbifolds in ( * ? ? ? b ) and using the infinitesimal rigidity results established in @xcite . _ proof of theorem [ thm : main ] : _ assume first that @xmath0 is a euclidean cone manifold as in the statement and all cone angles are @xmath22 . then using the regeneration results of section [ sec : pathsofstructures ] and proposition [ prop : largerspherical ] , every point in @xmath23 is the multiangle of a hyperbolic cone structure on @xmath0 . assuming that at least one of the angles of @xmath0 is @xmath3 , then there exists a spherical cone structure on @xmath0 with all cone angles @xmath22 , again by section [ sec : pathsofstructures ] and proposition [ prop : largerspherical ] . we consider all segments in @xmath581 starting at @xmath582 and ending at some point with at least one coordinate zero . the first point of the segment is the multiangle of a spherical cone structure and the last one is the multiangle of a hyperbolic one , by proposition [ prop : angle0 ] . by remark [ rem : opennessat0 ] , we can assume that the multiangle of the hyperbolic structure lies in the open cube @xmath23 . starting at @xmath583 , we move along the segment by decreasing the cone angles and obtaining a family of spherical cone manifolds . this family can not be spherical all the time because the endpoint of the segment corresponds to a hyperbolic structure and they are incompatible by lemma [ lem : incompatibility ] . since all cone angles are @xmath3 , then the end of the spherical subsegment is the multiangle of a euclidean structure , by the same argument as in proposition [ prop : completesubmfld ] . by the regeneration results of section [ sec : pathsofstructures ] and proposition [ prop : largerspherical ] , we connect the multiangle of this euclidean structure to the endpoint of the segment by multiangles of hyperbolic cone structures . the previous argument shows that every point in @xmath23 is the multiangle of a constant curvature cone structure on @xmath0 , and that @xmath29 is a hypersurface that divides @xmath23 in two components @xmath27 and @xmath26 corresponding respectively to hyperbolic and spherical multiangles . for the uniqueness , notice first that in lemma [ lem : incompatibility ] we establish incompatibility between hyperbolic , spherical and euclidean for a given multiangle . global rigidity for hyperbolic and spherical structures is proved in @xcite . using local rigidity of the regenerations ( corollary [ cor : lrr ] ) and global rigidity for the hyperbolic structures , we get global rigidity for the euclidean ones . in this section we illustrate the main theorem for cone manifold structures on the 3-sphere with singular locus given by the whitehead link , which is the 2-component link depicted in figure [ fig : whitehead ] . furthermore we discuss geometric structures corresponding to multiangles contained in the boundary of @xmath584 ^ 2 $ ] to some extent . shmatkov has computed the curve of multiangles of euclidean structures by constructing the corresponding cone manifolds explicitly , cf . these structures are not almost product . the hyperbolic region had earlier been computed by a. mednykh in a similar way , cf . @xcite and references therein . alternatively , one can proceed as follows : let @xmath44 denote the complement of the whitehead link in @xmath585 . if @xmath586 are meridians around the two components of @xmath587 , the fundamental group has the following presentation , cf . @xcite : @xmath588 with @xmath589 . the @xmath254-character variety of @xmath44 has been computed in @xcite . namely , after identifying @xmath590 with the image of the map @xmath591 in @xmath592 , it is given by @xmath593 with @xmath594 and @xmath595 holonomy representations of hyperbolic cone manifold structures lift to irreducible representations @xmath228 with @xmath598 and @xmath599 . we may write @xmath600 and @xmath601 with @xmath602 and @xmath603 the cone angles around the two components of @xmath587 . rotational parts of euclidean holonomies correspond to representations as above where in addition @xmath604 and @xmath605 fail to be local coordinates . this is precisely the locus where the discriminant of @xmath101 computed with respect to the variable @xmath244 vanishes . the discriminant is given by the following polynomial : @xmath606 for the resulting curve of multiangles of euclidean cone manifold structures see figure [ fig : euclideanregion ] . for multiangles contained in some parts of the boundary of @xmath607 ^ 2 $ ] we can construct cone manifold structures as well using our main theorem : for multiangles of the form @xmath608 , @xmath609 , resp . @xmath610 , @xmath611 , we obtain hyperbolic cone manifold structures , whereas for multiangles of the form @xmath612 , @xmath613 , resp . @xmath614 , @xmath615 , we obtain spherical ones . let us suppose there existed such a structure . then we consider the double branched cover of @xmath585 branched along the component of the whitehead link with cone angle @xmath22 . this branched cover is again @xmath585 since the components of the whitehead link are unknotted . the other component of the whitehead link lifts to the torus link @xmath622 , whose complement is known to be non - hyperbolic . on the other hand the lift of the intital hyperbolic cone metric will be a hyperbolic cone metric on @xmath585 with singular locus @xmath622 , which is a contradiction in view of the results in @xcite . , _ notes on notes of thurston _ , in _ analytical and geometric aspects of hyperbolic space _ ( ed . by d. b. a. epstein ) , london math . lecture notes , ser . * 111 * , 1987 , cambridge univ . press , cambridge , pp . 392 . , _ geometric structures on manifolds and varieties of representations _ , in _ geometry of group representations _ ( boulder , co , 1987 ) , pp . 169198 , contemp . math * 74 * , amer . soc . , providence , ri , 1988 .

we present a comprehensive study of the thermodynamic properties of the three - dimensional fermionic hubbard model , with application to cold fermionic atoms subject to an optical lattice and a trapping potential . our study is focused on the temperature range of current experimental interest . we employ two theoretical methods - dynamical mean - field theory and high - temperature series - and perform comparative benchmarks to delimit their respective range of validity . special attention is devoted to understand the implications that thermodynamic properties of this system have on cooling . considering the distribution function of local occupancies in the inhomogeneous lattice , we show that , under adiabatic evolution , the variation of any observable ( e.g. , temperature ) can be conveniently disentangled into two distinct contributions . the first contribution is due to the redistribution of atoms in the trap during the evolution , while the second one comes from the intrinsic change of the observable . finally , we provide a simplified picture of a recently proposed cooling procedure , based on spatial entropy separation , by applying this method to an idealized model .

1009.2761

one of the main ongoing efforts in cold atom gases is the investigation of strongly correlated phases . our theoretical understanding of strongly correlated phases is , in general , far from complete . for example , computational studies , of fermions in particular , are severely limited . as a result , the possibility of performing analog simulations of model hamiltonians , using cold atoms in optical lattices@xcite , has raised great hopes . the remarkable controllability of cold atom systems , allowing , for example , the application of a specific time - dependent perturbation , has also opened the possibility of studying strongly correlated systems in regimes inaccessible to solid state materials , especially away from equilibrium . among all model hamiltonians relevant to the physics of strong correlations , the hubbard model has attracted the greatest attention . on one hand , this model is the simplest to have important competition between kinetic and potential energies . it also plays in the field of strong correlations a somewhat analogous role to the one played by the ising model in classical statistical mechanics . on the other hand , its physics is of direct relevance to high - temperature superconductivity , a phenomenon still veiled in mystery . in fact , this new field often dubbed ` condensed matter of light and atoms ' was pioneered by the theoretical prediction@xcite and experimental observation of an incompressible regime , characteristic of a mott insulator , and of the transition between this phase and itinerant ones , first for bosons@xcite and recently also for fermions@xcite . for recent reviews , see , e.g. , refs . . in order to make progress towards the ultimate goal of performing analog simulations of model hamiltonians , a good synergy between experimental efforts and theoretical investigations is crucial . for example , theoretical inputs are useful in establishing maps , in parameter space , of the location of the different phases present in a realistic setup which takes into account the trap confining potential . moreover , experiments are currently confronted with the great difficulty of cooling fermions in optical lattices to sufficiently low temperatures to reach many interesting strongly correlated phases . this relevant temperature range is significantly lower than the one corresponding to mere quantum degeneracy . theoretical control over these issues is greatly needed and requires a quantitative understanding of the thermodynamic properties of the hubbard model , both for the homogeneous system and in the presence of a trap . in this article , we perform a comprehensive study of the thermodynamic properties of the homogeneous three - dimensional fermionic hubbard model , and of cold fermionic atoms in a three - dimensional optical lattice subjected to a trapping potential . we focus on the range of temperature which is of direct interest for current experiments as well as for the next generation of experiments . particular emphasis is put on aspects related to cooling of the system . the main theoretical technique used in the present study is dynamical mean - field theory ( dmft ) . this approach ( reviewed , e.g. , in ref . ) is a controlled approximation which is able to capture the competition between the kinetic energy , that tends to delocalize atoms over the whole lattice , and the repulsive potential energy that prevents atoms from occupying the same site , hence promoting localization . we also use another theoretical approach , namely high order high - temperature series expansions@xcite . one of the goals of the present article is to delimit , in parameter space , the respective range of validity of each of these approaches , and to provide a benchmark for their use through quantitative comparisons . this article is organized as follows . in sec . [ sec : model_methods ] , we define the model considered in this article , specify notations and conventions and briefly outline the theoretical methods used . in sec . [ sec : hom ] , we provide detailed results for the thermodynamics of the homogeneous hubbard model in three dimensions . finally , the effect of the trapping potential is considered in sec . [ sec : trap ] , with several applications geared towards cooling fermionic cold atom systems . cold fermionic gases with two hyperfine states loaded in an optical lattice potential can be described in a wide range of parameters by the fermionic hubbard model@xcite . the model hamiltonian is given by @xmath0 where @xmath1 is the creation operator of a fermion with spin @xmath2 at site @xmath3 and @xmath4 is the density operator at site @xmath3 . @xmath5 is measured in units of the lattice spacing . the spin @xmath6 labels the two hyperfine states of the atoms , and @xmath7 labels neighboring sites . the first term in the hamiltonian describes the kinetic energy of the atoms with @xmath8 the hopping amplitude between nearest - neighbor sites . in this paper , we mainly consider a three dimensional cubic lattice and use the half - bandwidth @xmath9 as our energy units . the second term represents the @xmath10-wave scattering between fermions in different hyperfine states , and its strength @xmath11 is proportional to the s - wave scattering length which can be tuned over a wide range using a feshbach resonance . we often use a grand - canonical description of the system in which the filling can be adjusted by the chemical potential @xmath12 . the last term in the hamiltonian represents the external trapping potential . aspects related to this trapping term will be discussed in more detail later . we investigate the properties of the system in a moderately high temperature regime , typically @xmath13 , or @xmath14 where @xmath15 is the inverse temperature . this is the temperature regime of interest in current ultracold quantum gases experiments@xcite . also one should note that we often use @xmath16 . at moderate temperature and commensurate filling ( one atom per site ) the homogeneous model displays a crossover between a liquid phase at weak interaction and an incompressible regime at larger interaction . the latter is characterized by suppressed density fluctuations and is a mott insulator . the properties of these two phases will be discussed in detail in the next sections . a simple picture of these two regimes can already be obtained by considering the extreme limits of vanishing interaction and vanishing hopping . in the first case , the system only consists of a mixture of non - interacting free fermions . the fermions are delocalized and form a fermi sea . in the case of vanishing hopping , the so - called atomic limit , the system consists of disconnected sites , and the problem of a single site can be solved . the atoms localize and a mott insulator forms at half filling . the atomic limit can be seen as the zeroth order term in a high - temperature expansion of the grand potential , @xmath17 , in @xmath18 , where @xmath19 is the grand partition function . the series for the grand potential in a uniform system ( @xmath20 ) can be written : @xmath21 where @xmath22 is the fugacity , @xmath23 , @xmath24 and @xmath25 is the partition function of a single site in the atomic limit . the atomic limit can be improved by systematically calculating@xcite higher order series coefficients , @xmath26 . thermodynamic quantities at high temperatures can be accurately derived from leading terms in eq . [ grand ] . for intermediate couplings and temperatures , the solution of the hubbard model is highly non trivial . in this article , we mainly use dynamical mean - field theory ( dmft)@xcite to explore this regime . dmft is an approximate method in which the lattice model is replaced by a self - consistent impurity model . this model can be solved using highly accurate numerical algorithms such as the strong - coupling ct - qmc algorithm@xcite which we use in the present work . although approximate , dmft is a controlled approximation : it is exact in both non - interacting and atomic limits , and bridges the gap between the two limits . furthermore , it becomes mathematically exact in the formal limit of infinite lattice coordination . physically , dmft neglects spatial correlations , but treats accurately local quantum fluctuations . besides dmft , we also use high - order high - temperature series expansions @xcite . in fact , one of the aims of this article is to provide benchmarking of dmft in the high - temperature domain from comparison with these series . more details on the regions of validity of the two methods are given in sec . [ sec : theory_validity ] . finally , at low temperature ( @xmath27 ) , a phase transition into an antiferromagnetically order phase occurs@xcite , this regime will not be considered in the rest of this paper . in this section we focus on the properties of the homogeneous hubbard model . we discuss in particular the density , the double occupancy and the entropy in this model . these quantities are very important in the characterization of ultra cold quantum gases . the effect of the trap will be considered in sec . [ sec : trap ] . the physics of the competition between kinetic and potential energy in the hubbard model is exemplified by the behavior of the density per site @xmath28 as a function of the chemical potential @xmath12 . for simplicity , since we are considering the homogeneous system , we drop the site index in this section . at low enough temperature , this quantity has a qualitatively different behavior for weak and strong interaction ( fig . [ fg : n_mu_low_t ] ) . at weak coupling ( cf . @xmath29 in fig . [ fg : n_mu_low_t ] ) , @xmath30 has a smooth evolution from @xmath31 for @xmath32 to @xmath33 for @xmath34 . in this case , the compressibility @xmath35 is always finite when @xmath36 . in the opposite limit of strong coupling ( cf . @xmath37 in fig . [ fg : n_mu_low_t ] ) , the potential energy disfavors the presence of two fermions on a single site and @xmath38 develops a plateau around @xmath39 . on this plateau , characteristic of the mott insulating state , the compressibility @xmath40 vanishes in the zero temperature limit . as the chemical potential must be of the order of @xmath11 to overcome the potential interaction and to generate doubly occupied sites , the central plateau widens with increasing interaction . versus chemical potential @xmath12 for different interaction strength @xmath11 and at fixed inverse temperature @xmath41 ( dmft ) . for large interaction strength a clear mott plateau of filling @xmath42 develops . [ fg : n_mu_low_t],width=264 ] increasing the temperature causes a softening of these features . [ fg : n_mu_high_t ] shows the density plotted at @xmath43 . versus chemical potential @xmath12 for increasing interaction @xmath11 and fixed inverse temperature @xmath44 ( dmft ) . only a reminiscent behavior of the mott plateau is left for large interaction strength . , width=264 ] at weak interaction strength , thermal excitations reduce the slope which characterizes the variation of the density with chemical potential . for strong interactions , large temperatures of the order of @xmath11 ( the mott gap ) are needed to generate thermal excitations and to destroy the mott plateau . one can see comparing figs . [ fg : n_mu_low_t ] and [ fg : n_mu_high_t ] how the plateau found at @xmath45 for @xmath46 disappears at high temperature only leaving behind an inflection in the @xmath47 curve . the double occupancy @xmath48 ( figs . [ fg : d_mu_beta10 ] , [ fg : d_n_beta10 ] ) offers another perspective on the mott phenomenon . this quantity is particularly interesting as in cold atom systems it can be directly measured @xcite . hence , understanding how double occupancy varies is very useful to characterize the state of an experimental system . furthermore , this quantity is directly related to the potential energy of the system which is given for a site by @xmath49 . in fig . [ fg : d_mu_beta10 ] , we display the behavior of @xmath50 as a function of the chemical potential @xmath12 . at weak coupling , @xmath50 tracks the behavior of the density ( for the non - interacting system , @xmath51 ) and the two quantities contain essentially the same information . however , for large coupling , @xmath50 is strongly suppressed when the chemical potential is lower than the potential energy @xmath11 , and increases steeply in the region @xmath52 , precisely the region where the density increases above unity . this qualitative difference between weak and strong coupling is well apparent on fig . [ fg : d_n_beta10 ] where we plot @xmath50 versus @xmath38 . for @xmath53 , the curve is nearly quadratic while for @xmath47 the double occupancy @xmath50 is close to zero for @xmath54 and linear for @xmath55 . the behavior of @xmath50 at strong coupling is to be contrasted with the behavior of the density . for strong coupling , @xmath38 presents a plateau at @xmath56 , and hence can be used to determine the chemical potentials for which @xmath38 drops below unity as well as for which @xmath38 exceeds unity . instead , @xmath50 is suppressed in the whole region @xmath57 , and is only sensitive to the point at which @xmath38 exceeds unity . this is advantageous when it comes to experimentally detecting the mott state@xcite . in the inset of fig . [ fg : d_n_beta10 ] , we also display a plot of @xmath58 as a function of @xmath38 . versus chemical potential @xmath12 at fixed inverse temperature @xmath59 and different values of the interaction strength @xmath11 ( dmft ) . , width=264 ] versus density @xmath38 at fixed inverse temperature @xmath59 and different values of the interaction strength @xmath11 ( dmft ) . inset : @xmath60 at the inverse temperature @xmath59 and different values of the interaction strength @xmath11 ( dmft ) . , width=264 ] another quantity of fundamental importance is the entropy as preparation of cold atom systems is often done almost adiabatically . therefore , the entropy of the system , rather than the temperature , is a conserved quantity which can be used as a constant to characterize the system evolution . we calculate the entropy by integrating a fit to the dmft energy data starting from infinite temperature . the high temperature regime ( @xmath61 ) is approximated using a second order high temperature expansion . versus chemical potential @xmath12 at the weak interaction strength @xmath53 for different values of the inverse temperature @xmath62 ( dmft ) . , width=264 ] in the weakly interacting regime ( fig . [ fg : s_mu_u0.5 ] ) , the entropy per site @xmath10 has a simple evolution with temperature and chemical potential : it is maximal at @xmath63 and decreases monotonically with increasing @xmath64 . with decreasing temperatures , @xmath65 decreases uniformly over the whole chemical potential range . versus chemical potential @xmath12 at the strong interaction strength @xmath47 for different values of the inverse temperature @xmath62 ( dmft ) . the curves legend is the same as in fig . [ fg : s_mu_u0.5 ] . , width=264 ] in the strongly correlated regime ( fig . [ fg : s_mu_u5 ] ) , @xmath65 presents non - trivial features . already at relatively high temperatures , the entropy has two maxima located approximately at @xmath66 and @xmath67 , while it has a minimum at the particle - hole symmetric point @xmath68 . from an atomic limit point of view , the presence of these two peaks in the entropy at @xmath66 and @xmath67 is easily interpreted . these two chemical potential values correspond to the regions where the density crosses over rapidly between @xmath69 and @xmath56 ( @xmath70 ) , and between @xmath71 and @xmath72 ( @xmath73 ) . the charge fluctuations in these regions are maximal and contribute significantly to the entropy . the temperature evolution is also interesting ( fig . [ fg : s_mu_u5 ] ) : the entropy for @xmath74 and @xmath52 decreases quickly with lowering @xmath75 , while in the region @xmath76 the temperature - dependence slows down and the entropy approaches a finite value , @xmath77 . this value reflects the fact that , in the mott insulator regime , the system gets frozen into a local moment state ( with two spin - states per site ) in a rather extended temperature range below the mott gap . naturally , we expect that , as the system is cooled down further , the entropy will eventually decrease again below @xmath78 as magnetic correlations develop between local moments . as dmft neglects spatial correlations in the paramagnetic phase , within this theory , the decrease in entropy only happens right at the nel transition where long - range antiferromagnetic order sets in ( this nel transition occurs at a lower temperature than the ones studied in the present paper ) . however , in reality , the entropy will start deviating significantly from @xmath78 above the nel transition , as short - range correlations develop @xcite . nevertheless , the single - site dmft description can be regarded as a good description of the paramagnetic mott insulator down to a characteristic temperature . we will see in sec . [ sec : theory_validity ] how it is possible to evaluate this temperature with the help of high temperature series expansions . in fig . [ fig : s_n_hom ] , a similar behavior is observed by plotting the entropy versus the density . for low interaction strength , a maximum of the entropy is found at half filling . in contrast , at large interaction strength and intermediate temperatures , a dip in the entropy arises at half filling , which is due to the freezing of the density degree of freedom in the mott insulator . at low temperatures , the entropy decreases linearly in temperature for the liquid regimes away from half - filling , while the value at half filling remains fixed to @xmath78 ( for a range of temperatures , see above ) . in fig . [ fig : s_over_n_n_hom ] , we additionally plot the entropy per particle @xmath79 as this quantity will be useful later on to better understand the behavior of the entropy in the presence of a trapping potential . we note in particular that upon dividing the entropy by the particle number , the entropy per particle decreases with increasing particle number . [ bt ] versus density @xmath38 at different interaction strengths @xmath11 and inverse temperatures @xmath62 ( dmft ) . the curves legend is the same as in fig . [ fg : s_mu_u0.5 ] . , width=264 ] [ bt ] at different interaction strengths @xmath11 and inverse temperatures @xmath62 ( dmft ) . the curves legend is the same as in fig . [ fg : s_mu_u0.5 ] . , width=264 ] the double occupancy provides a wealth of information on the physics of cold atom systems . for example , this quantity can be used as a thermometer in a certain regime of temperatures and interactions@xcite . to obtain a good understanding of how this quantity behaves in a trapped system , we first show here the dependence of the double occupancy on temperature in an homogeneous system ( fig . [ fig : pomeranchuk ] ) . for moderate interaction strength , the double occupancy presents a non - monotonous behavior at low temperatures . the appearance of the initial decrease in the double occupancy with increasing temperature ( from @xmath80 ) is analogous to the pomeranchuk effect observed in liquid helium-3 and has been discussed previously in the half filled case @xcite . the reason behind this decrease of the double occupancy with increasing temperature is that the system prefers , when heated , to localize the atoms . for this localized state , the ( spin ) entropy is larger than for a state where fermions form a fermi liquid . in this regime , the minimum of @xmath50 is determined by the quasi - particle coherence temperature . since , in the particle - hole symmetric case , the coherence temperature decreases with @xmath11 , the ` pomeranchuk ' effect occurs if @xmath81 is not too large and disappears in the mott insulator . away from particle - hole symmetry , the system behavior is quite different . in that case , the ` pomeranchuk ' temperature at which @xmath82 has a minimum increases with @xmath11 . we also note that the double occupancy has recently been studied below and close to the nel transition to the antiferromagnetic order@xcite , a regime that we do not consider here . in the antiferromagnetic phase , the coherent alignment of spins causes an increase of the double occupancy . as the entropy and double occupancy are related by the thermodynamic ( maxwell ) relation @xcite @xmath83 the pomeranchuk minimum in @xmath82 translates into a non - monotonous behavior of the entropy as a function of the interaction strength . for this quantity , a maximum is found at sufficiently low temperatures . as in the case of the double occupancy , this effect persists away from half filling ( fig.[fig : pomeranchuk ] ) . we will see later in sec . [ sec : isentropic_trap ] the consequences of these considerations for systems subjected to a trapping potential . in the strong interaction limit , the low - energy properties become similar to the properties of the heisenberg model . the entropy for this model has been recently studied in ref . . [ bt ] as a function of temperature @xmath75 for different interactions strength @xmath11 ( dmft ) . ( from top to bottom : @xmath42 , @xmath84 and @xmath85 ) . right column : entropy @xmath10 as a function of @xmath11 for different temperatures ( dmft ) . ( from top to bottom : @xmath42 , @xmath84 and @xmath85 ) . the inverse temperature @xmath62 is measured in units of @xmath86 . [ fig : pomeranchuk],width=264 ] in this section , we perform a comparative study of the validity of the two theoretical approximations used in the present paper : dmft and high - temperature series . the motivation for doing this is twofold . on the one hand , in regimes where a converged result can be reliably extracted from the series expansion ( which is an exact technique implying no further approximation ) , we can use this comparison to benchmark the validity of dmft as an approximation to the thermodynamic properties of the three - dimensional hubbard model . on the other hand , by comparing different orders of the series expansion , we can delineate the regime in which this method can be used reliably and in which regime dmft is a better option . in fig . [ fig : dvsbeta_high_mu ] and fig . [ fig : dvsbeta ] , we display the results obtained by the two methods for the entropy and the double occupancy as a function of temperature , at intermediate interaction strength , for two different chemical potentials . different orders ( up to order @xmath87 ) of the series expansions are also compared in these figures . it is immediately clear from these two figures that the temperature down to which the series expansion can be safely used , and consequently down to which a reliable assessment of dmft can be made , depends considerably on the value of the chemical potential ( or density ) . for @xmath88 ( fig . [ fig : dvsbeta_high_mu ] ) , corresponding to a quite high density , the agreement between the different orders of the series expansion , and the overall agreement with dmft is essentially perfect over the whole range of temperatures considered . in fact , only very small deviations of the second order expansion can actually be detected . this agreement provides strong evidence for the correctness of both methods down to quite low temperatures in the high - density regime ( and by particle hole - symmetry , also in the low density regime ) . in contrast for @xmath89 ( corresponding to an intermediate filling @xmath90 ) , it is clear from fig . [ fig : dvsbeta ] that the different orders of the series expansion start to deviate from one another already at a rather high temperature @xmath91 . below this temperature , the series expansion method becomes unreliable . this breakdown happens approximately at the same temperatures for the entropy and the double occupancy . as shown below , this regime of intermediate densities is the hardest one for the series expansion method . by comparing the different quantities as a function of the chemical potential at fixed temperature , we identify more clearly these different regimes of density . in fig . [ fig : svsmu ] , we show the results for the entropy per site @xmath10 as a function of chemical potential at a fixed temperature @xmath92 . here one can clearly identify three different regions . at very low ( and very high , due to particle - hole symmetry ) chemical potential @xmath12 , where the density of the system is close to zero ( close to two particles per site ) , both dmft and the series expansion give reliable results down to fairly low temperatures ( at least of the order of @xmath93 ) . in contrast , in the intermediate region around @xmath70 and @xmath73 where the density crosses over from 0 to 1 and from 1 to 2 respectively , the series expansion breaks down at a rather high temperature . already at @xmath94 it is not reliable anymore . very close to half filling , the opposite trend is found . in this region , the series expansion gives reliable results down to quite low temperatures , actually lower ( as further detailed below ) than the temperature at which dmft ceases to be an accurate approximation@xcite . in fig . [ fig : diff_series_series ] , we summarize the validity of the high temperature series expansion by showing the difference between different orders . for this plot , it appears that the breakdown temperature for the series expansion depends sensitively on the chemical potential and that the method works best close to very low and very high filling , as well as exactly at half filling . [ bt ] and double occupancy @xmath50 at fixed chemical potential @xmath95 and intermediate interaction strength @xmath96 obtained by high temperature series and dmft . , title="fig:",width=264 ] [ bt ] and double occupancy @xmath50 at fixed chemical potential @xmath89 and intermediate interaction strength @xmath96 obtained by high temperature series and dmft method . , title="fig:",width=264 ] [ bt ] versus the chemical potential at intermediate interaction strength @xmath96 obtained by high temperature series and dmft method . , title="fig:",width=264 ] [ bt ] . , title="fig:",width=264 ] in fig . [ fig : svsbeta_pade ] , we focus in more detail on the comparison of the entropy at half filling ( @xmath97 ) . as mentioned previously , the single site dmft approximation overestimates the entropy at half filling , yielding a value @xmath78 in the large-@xmath11 limit because it neglects short - range magnetic correlations . this value also corresponds to the mean field value for the entropy of the heisenberg model in its paramagnetic phase . in reality , the value of the entropy is reduced by short - range antiferromagnetic correlations@xcite . in fig . [ fig : svsbeta_pade ] , this limitation of single - site dmft is exposed as we compare entropy obtained with dmft to the ones given by the series expansion and the heisenberg model . while the dmft curve saturates at @xmath78 at low temperature , the different orders of the series expansion reach a lower value of the entropy before diverging . the pade expansion of the high temperature series expansion lies slightly above the entropy of the heisenberg model calculated in ref . by qmc simulations this situation represents a case in which the high temperature expansion can be used down to a lower temperature than single - site dmft . finally , we note that further extensions of dmft ( especially cluster - dmft methods@xcite ) exist which overcome these limitations of single - site dmft , and restore the physical effects of short - range magnetic correlations . [ bt ] obtained by series expansion , dmft and qmc ( for the heisenberg model)@xcite . [ fig : svsbeta_pade],title="fig:",width=264 ] in present day experimental setups , an external potential is usually present to confine the atomic cloud . this confinement can be due to different sources , for example to the focusing of the lattice laser beams or a magnetic or dipole trap . in most cases , a parabolic form is a good approximation for the shape of the confining potential at the location of the atom cloud . however , many refinements can be considered to obtain a more precise spatial dependence of the confining potential @xmath98 ( last term of eq . [ eq : hubb_ham ] ) . within the local density approximation ( lda ) , the properties of the trapped system at a certain position @xmath5 are assumed to be those of the homogeneous system with the chemical potential set to the value @xmath99 . lda has been found to be a good approximation for local quantities in a three dimensional fermionic gas in an optical lattice@xcite , and was also validated in the high temperature regime applicable to ongoing experiments @xcite . however , as lda neglects the influence of surrounding sites with different densities , the main inaccuracies occur close to phase boundaries , where proximity effects occur , and also when one computes non - local longer range physical observables@xcite . using lda and taking the continuum limit , it is possible to describe the system using rescaled variables that do not depend explicitly on the strength of the confining potential . to illustrate this we consider a radial symmetrical form for the trapping potential @xmath100 in @xmath50 dimensions ( @xmath101 is the lattice spacing ) . thus , the position - dependent chemical potential becomes @xmath102 with @xmath5 the @xmath50-dimensional position vector labeling each lattice site and @xmath103 the chemical potential at the center of the trap . within lda , any local observable @xmath104 in the inhomogeneous system is related to its homogeneous counterpart by @xmath105 , where @xmath106 denotes the corresponding quantity in the homogeneous system ( in the following we will drop the label ` h ' ) . the number of atoms , which corresponds to the sum of the local occupancy over the whole system , can be expressed in the following way : @xmath107 with @xmath108 the surface of a sphere in @xmath50 dimensions . changing variables to an integration over the chemical potential using ( [ eq : mu_r_general_trap ] ) finally leads to : @xmath109 in this expression @xmath110 is the dimensionless chemical potential ( similarly @xmath111 ) . the lda approximation enters in the assumption that @xmath38 is a function of @xmath112 only through its dependence on @xmath12 . this formula can be easily generalized to the case for which the strength of the confining potential is different along the different cartesian directions ( which is usually the case in experiments ) by replacing @xmath113 with a proper averaged quantity @xmath114 . for example , in three dimensions @xmath115 . [ eq : lda_n ] shows that the dimensionless combination @xmath116 is a quantity that does not depend on the strength of the confining potential , and hence can be used to describe properties of experimental systems regardless of the particular realization of the trap . for the one - dimensional case , see ref . . it also shows that in this approximation the key quantity is the observable in the homogeneous system , e.g. , @xmath30 , and that everything can be derived from it . obviously the same holds true for all the other quantities that can be expressed as a sum of a local quantity over the whole system . another very useful way to express averages of observables over the trap is to introduce the distribution function of site occupancies in the system , defined for @xmath117 as : @xmath118 and the related quantity in which each site is weighted by its occupancy : @xmath119 using again a continuous - space integration over the system and changing variables in favor of the chemical potential , one obtains : @xmath120^{d/\alpha-1}. \label{eq : q_n_def}\ ] ] in this expression , @xmath121 is the ( dimensionless ) compressibility of the homogeneous system : @xmath122 hence , the local chemical potential and local compressibility entirely determine the distribution of local site occupancies in the trap . the distribution @xmath123 and rescaled distribution @xmath124 obey the sum - rules : @xmath125 with @xmath126 the total atom number and @xmath127 its rescaled form . these distribution functions allow to express the average of any observable over the trap , within the lda approximation , as : @xmath128 where @xmath129 is the local operator corresponding to the observable . because @xmath123 is a normalized distribution obeying the sum - rule ( [ eq : sumrule ] ) , the last expression is particularly useful . as we shall see later , when varying an external parameter , it allows to separate in a simple manner the changes in @xmath130 which are due to a redistribution of the particles in the trap ( reshaping of @xmath123 ) from the contribution due to the intrinsic dependence of the local observable on the parameter , already present in the homogeneous system . in the following , we make use of this description , in combination with thermodynamic relations , in order to discuss cooling or heating of the trapped system as the coupling is changed . we will concentrate on the case of a three - dimensional lattice in a harmonic potential i.e. , @xmath131 , @xmath132 . one consequence of the presence of an inhomogeneous trapping potential is that different quantum phases can spatially coexist in the gas . this can actually be seen as a favorable situation , in which several different physical regimes can be studied in a single experiment . in an optical lattice realizing the three - dimensional hubbard model , coexistence between liquid and mott insulating regions in the trap were for example documented in theoretical studies @xcite . in fig . [ fig : state_diagram ] , we display the different regimes expected in a three - dimensional optical lattice confined into a parabolic trap , as a function of the coupling @xmath81 and of the scaled particle number @xmath133 . different temperatures in the currently accessible range are considered . at still lower temperature ( not displayed ) , antiferromagnetic long - range order @xcite will occur in the regimes with a commensurate mott plateau . the state diagram of fig . [ fig : state_diagram ] , which generalizes to different temperatures the results of ref . , was obtained on the basis of the theoretical calculations described in the previous section for the homogeneous model , using lda approximation . in addition , in appendix [ sec : simpleapprox ] , we introduce a simple approximation which allows to obtain analytical expressions for the various crossover lines of the state diagram in the low - temperature regime . the state diagram displays four characteristic regimes ( labeled l , b , mc and ms ) , which are illustrated by the four corresponding density profiles @xmath134 and local occupancy distribution functions @xmath124 calculated at four representative points and displayed in fig . [ fig : profiles ] . for low interaction strength ( regime ` l ' , fig . [ fig : profiles ] a ) the density profile adjusts to the trapping profile and the system remains a fermi liquid everywhere in the trap . with increasing temperature the density distribution broadens . the weighted particle number distribution @xmath124 displays a maximum at filling unity which , according to eq . ( [ eq : q_n_def ] ) reflects the smaller compressibility at that filling . a rather sharp drop is seen towards larger occupancies which represents the center of the trap , whereas a slower decay occurs towards smaller particle numbers , due to the tails of the density distribution . increasing the temperature shifts weight from larger occupancies towards smaller occupancies . for very large values of the scaled particle number @xmath133 , a band insulator with @xmath33 forms in the center of the trap ( regime ` b ' , fig . [ fig : profiles ] c ) . the pinning to @xmath33 , and hence the band insulator , is destroyed by increasing the temperature . in the presence of a large band insulating region , the corresponding distribution @xmath124 displays a sharp peak at filling @xmath33 . increasing the temperature , this peak decreases and the weight moves to lower occupancies . for larger interaction strength ( regime ` mc ' , fig . [ fig : profiles ] e ) a mott - insulating region appears , in which the density is pinned to @xmath42 particle per site . close to the boundary of the trap , the mott insulating region is surrounded by a liquid region . the mott - insulating region shows up in @xmath124 as a large and narrow peak at filling @xmath42 with a sharp edge on the large occupation numbers side . the peak reflects the essentially vanishing compressibility of the mott insulator . increasing the temperature decreases the size of the mott insulating plateau , and results in a shift of the weight from filling one to low densities . increasing the number of atoms in the trap at large interaction strength can increase the pressure exerted on the atoms , and can cause the occurrence of a liquid region with filling larger than one in the center , surrounded by a shell of mott insulator with @xmath42 ( regime ` ms ' , fig . [ fig : profiles ] g ) . correspondingly , the sharp peak in @xmath124 broadens somewhat . recently , experimental evidence of the mott insulating region has been reported@xcite . this has been achieved by observing the suppression of the double occupancy in the mott - insulating region@xcite and the compression of the cloud as a response to the variation of the external trapping potential@xcite . we note that in bosonic two - dimensional gases the density profiles @xmath134 and therefore the occupancy number distributions @xmath124 can nowadays be measured with a very high spatial resolution@xcite . in three dimensional gases , the integrated column density can be measured for example by using an electron microscope@xcite . furthermore , new techniques such as the immersion of a single trapped ion into the atomic gas are being developed to locally measure the density in three dimensional systems as well@xcite . [ bt ] values at which the central density takes the values 0.995 , 1.005 and 1.995 . the gray dashed line marks the crossover from the liquid to the mott state with increasing interaction . the crosses indicate the points for which the density profiles are plotted in fig . [ fig : profiles ] . , title="fig:",width=264 ] [ bt ] , @xmath135 . c and d ( in ` b ' for low-@xmath75 ) : @xmath136 , @xmath137 . e and f ( in ` mc ' ) : @xmath138 , @xmath135 . g and h ( in ` ms ' ) : @xmath138 , @xmath137 . , width=321 ] cold atom clouds are almost perfectly isolated from their environment . therefore , assuming that manipulations can be performed adiabatically , the quantity that is conserved during the evolution of the system is the entropy . however , the temperature will in general change . consequently , studying the effects of an isentropic change of parameters on the system temperature is important . here we focus our attention on the increase of the interaction strength , @xmath11 , and identify if such a change can help to reach lower temperatures in a trapped system , i.e. whether cooling occurs . [ bt ] along isentropic lines , in a case where cooling occurs as @xmath139 is increased . the continuous line corresponds to a lower entropy per particle than the dashed line , @xmath140 . the figure illustrates that for this situation , increasing @xmath139 while keeping @xmath75 constant will increase the entropy ( eq . [ eq : cooling_rate]).,title="fig:",width=226 ] the change in temperature induced by a change in a certain parameter @xmath139 at constant entropy is @xmath141 . if this quantity is negative ( positive ) , an increase of the parameter is associated with cooling ( heating ) . as we now show , one can consider alternatively the change in entropy at constant temperature @xmath142 . indeed , the location @xmath143 of isentropic lines in the @xmath144 plane is defined by the equation @xmath145=\rm{const.}$ ] . taking a derivative of this equation , we obtain : @xmath146 denoting by @xmath147 the specific heat of the system , we finally obtain the expression of the relative cooling rate in the form : @xmath148 since @xmath149 is a positive quantity , we see that there is cooling ( heating ) when @xmath150 is positive ( negative ) . this is illustrated schematically in fig . [ fig : temperature_change ] . in the following , we consider the temperature change under an adiabatic increase of the coupling @xmath11 . we observe that the derivative @xmath151 is related to a derivative of the total double occupancy @xmath152 through the maxwell relation @xcite : @xmath153 so that the relative cooling rate reads : @xmath154 hence , when the derivative of @xmath155 with respect to temperature is negative ( positive ) there will be cooling ( heating ) upon an isentropic increase of the interaction strength . one advantage to use the change of @xmath155 is that this quantity can be measured quite accurately in present experiments@xcite . in fig . [ fg : dvst_u1]-[fg : dvst_u3 ] , we plot @xmath155 as a function of temperature for different interaction strengths and particle numbers . for @xmath136 ( fig . [ fg : dvst_u1 ] ) , @xmath156 is a decreasing function of temperature for all particle numbers . this implies that in the weak - coupling regime an increase of @xmath11 generates cooling . increasing the interaction the situation gradually changes . for @xmath157 ( fig . [ fg : dvst_u2 ] ) , @xmath156 becomes much flatter and cooling is restricted to large particle numbers or low temperatures . finally , for @xmath138 ( fig . [ fg : dvst_u3 ] ) , the tendency inverts and at high temperature , heating occurs . it has to be stressed that the absolute value of the derivative drops with increasing interaction and that at large @xmath11 the heating or cooling is essentially negligible . for the interaction strength @xmath136 and at different scaled particle numbers @xmath158 ( dmft ) . , width=264 ] for the interaction strength @xmath157 and at different scaled particle numbers @xmath158 ( dmft ) . , width=264 ] for the interaction strength @xmath138 and at different scaled particle numbers @xmath158 ( dmft ) . , width=264 ] in order to better understand the origin of the cooling or heating , we write the derivative @xmath159 using eq . ( [ eq : q_n_def ] ) , under the form : @xmath160.\ ] ] the usefulness of this expression resides in the clear separation of two different contributions . the first term in the integrand takes into account changes to the total double occupancy in the trap due to the intrinsic temperature dependence of the double occupancy @xmath161 in the homogeneous system . cooling can occur from this term whenever the pomeranchuk mechanism discussed above applies . the second term instead represents the contribution due to the redistribution of the atoms in the trap upon a temperature change . the two terms can be calculated separately in order to determine the most relevant mechanism behind the cooling observed for low interaction . in fig . [ fg : pomer_contributions_lowu ] , we plot the first ( `` intrinsic '' , dotted lines ) and second ( `` redistribution '' , continuous lines ) terms of eq . [ eq : dddt_qfunc ] at @xmath136 for different atom numbers @xmath133 . note that , for readability , these quantities are plotted versus _ temperature @xmath162 . we notice that the `` redistribution '' term is always negative , and hence the reshaping of the density profile always induces a cooling effect . in contrast , the `` intrinsic '' term is positive at high temperature and becomes negative only for @xmath163 ( @xmath164 ) . hence , we conclude that at high temperature the cooling is dominated by the redistribution of atoms in the trap , while at lower temperatures , both the redistribution and the intrinsic ` pomeranchuk ' effect contribute on comparable footing . the latter may even dominate at still lower temperatures ( e.g. , @xmath165 in fig . [ fg : pomer_contributions_lowu ] ) . a qualitative understanding of the behavior of each term in eq . ( [ eq : dddt_qfunc ] ) can be achieved from the inspection of the properties of @xmath60 and @xmath124 . the first observation is that @xmath60 is a monotonically increasing function of @xmath38 ( cf . inset of fig . [ fg : d_n_beta10 ] ) . secondly , from fig . [ fig : profiles]b and d we notice that @xmath166 ( possibly with the exception of the region around @xmath42 ) is always negative for @xmath38 larger than a certain value @xmath167 and always positive for @xmath38 smaller than @xmath167 . furthermore @xmath168 implies that @xmath169 . combining these two observations , we conclude that the second term in eq.([eq : dddt_qfunc ] ) is generally negative . hence , the redistribution of the atoms indeed produces cooling in general , as observed above . the presence of the peak in @xmath124 around @xmath42 might undermine the reasoning but this is never the case for the parameters considered here ( fig . [ fg : pomer_contributions_lowu ] ) . the behavior of the first term in eq . [ eq : dddt_qfunc ] is closely connected to the pomeranchuk effect in homogeneous systems . as we saw in fig . [ fig : pomeranchuk ] in the homogeneous system the pomeranchuk effect is active only at sufficiently low temperature . therefore , the intrinsic contribution in the trap can only lead to cooling at low temperatures . this is in good agreement with the results in fig . [ fg : pomer_contributions_lowu ] . the situation is only quantitatively different at stronger coupling . for @xmath138 ( fig . [ fg : pomer_contributions_highu ] ) , the redistribution of atoms still corresponds to cooling , although the absolute value of the contribution is roughly an order of magnitude smaller than for weaker interaction strength . however , in this case , the ` intrinsic ' contribution becomes dominant and its sign corresponds to heating ( opposite to the pomeranchuk effect ) . at lower temperature , the intrinsic term becomes negative again ( cooling ) but anyhow the cooling rate in this regime is fairly small . for @xmath136 ( dmft ) . continuous line ( left side , top to bottom : @xmath170 ) : redistribution of atoms in the trap ( reshaping , second term in eq . ( [ eq : dddt_qfunc ] ) ) . dotted line ( right side , top to bottom : @xmath171 ) : intrinsic change of the double occupancy with temperature ( first term in eq.([eq : dddt_qfunc ] ) ) . , width=264 ] for @xmath138 ( dmft ) . continuous line ( left side , top to bottom : @xmath170 ) is the reshaping , dotted ( left side , top to bottom : @xmath171 ) is the intrinsic double occupancy change with temperature . , width=264 ] the conclusion of this section is that in trapped systems an increase of the interaction is accompanied by cooling for interactions weaker than the interaction needed to have a mott insulator . on the other hand , at larger interactions , there is no substantial cooling associated with an increase of @xmath11 , and even slight heating can occur at high temperature . on the whole , the dominant contribution to cooling is usually the redistribution of the atoms in the trap , although the pomeranchuk effect ( intrinsic contribution ) can become operative at low temperature . reaching sufficiently low temperatures to observe complex quantum phases is one of the main challenges currently faced by cold atom physics experimentalists . in this section , we show that by adiabatically reshaping the confining trap , to divide the system into entropy rich and poor regions , a gas can be cooled down by one order of magnitude lower than currently achievable using state - of - the - art techniques @xcite . before presenting our method , we would like to point out that employing adiabatic changes to cool down gases was proposed in related contexts . for example , a bose - einstein condensate was experimentally produced by adiabatically deforming the external trapping potential to increase the gas phase space density @xcite . furthermore , reshaping the underlying trap to create entropy rich regions that are later isolated from the remaining system was proposed for bosons loaded into an optical lattice @xcite . for fermions confined to an optical lattice , cooling could be achieved by immersion into a bosonic bath @xcite . in ref . , it was cleverly proposed to reduce the entropy of lattice fermions in contact with a bosonic reservoir by compressing them into a band insulator or more generally a gapped phase . this last cooling method requires a transfer of entropy between two distinguishable quantum gases , a process experimentally demonstrated in ref . . our cooling method does not rely on immersing the atoms into a reservoir but on creating spatially distinct regions of high and low entropy that can be isolated from one another , and on subsequently removing the high entropy region . once this procedure is completed the remaining system has a much lower average entropy per particle allowing for the study of interesting phenomena requiring lower temperatures than previously attainable . to demonstrate our cooling procedure , we use a twofold approach . we first present our method using an idealized setup , and in a second time , we revisit with a new perspective the experimental setup presented in ref . . we further highlight the differences between our cooling method and another one recently proposed in ref . . [ bt ] ( solid line ( full system ) , dashed line ( core region ) , right upper panel ) as a function of the density . the core region is taken to be cylindrical with @xmath172 , the storage region is a rectangular box of size @xmath173 . each region has a homogeneous density @xmath174 , @xmath175 . the average entropy per particle for the total system is @xmath176 and @xmath177 for the core region . the ratio of particles in the core region versus the total particle number is @xmath178.,width=302 ] as explained above , we first present our cooling method using an idealized setup uncluttered with experimental details . in the following , we show how , given an idealized two - fluid model , one can start with a gas , loaded in an optical lattice and confined to an harmonic trap , having an entropy per particle , e.g.,around @xmath179 and obtain a final system characterized by a near zero entropy per particle while keeping about @xmath180 of the atoms . as we have seen in fig . [ fig : s_over_n_n_hom ] , at high temperatures , the entropy per particle @xmath181 is largest for low densities . therefore to segregate the entropy in our system , we would like to create two distinct regions ( fig . [ fig : idealized_trap ] ) : ( i ) a `` core region '' with a deep trapping potential in which the density @xmath182 is close to two particles per site , i.e. @xmath183 and ( ii ) a `` storage region '' with a very flat trapping potential in which the density @xmath184 is very low , i.e. @xmath185 . in the language of the @xmath123 distribution introduced in the previous sections , this idealized setup results in two sharp peaks : one at a very low density due to the storage region and another one very close to @xmath33 due to the core . to obtain these two regions of unequal densities , we start from a gas confined to the usual harmonic trapping potential with total entropy @xmath186 , and adiabatically deform the trap to reach the trapping profile presented in fig . [ fig : idealized_trap ] . while the total entropy remains constant under this deformation , the entropy is now inhomogeneously distributed . the entropy per particle in the core region will be very low , ideally zero , whereas in the storage region the entropy per particle will be quite large . in contrast , the temperature @xmath187 remains equal throughout the system and is set by the constraint requiring entropy conservation , i.e. @xmath188 where @xmath189 and @xmath190 . for the idealized two - fluid model presented here , eq . [ eq : entropy_shaping ] simplifies to @xmath191 where @xmath192 and @xmath193 are the number of particles in the core and storage regions respectively . in this idealized situation an infinitely narrow barrier is used to separate the two regions adiabatically . after the separation , the high entropy region can be removed and the core region can be used to perform the experiment . as the density is uniform through out the core region , the average entropy per particle at the time of separation is given by @xmath194 . therefore , the final average entropy per particle in the core region , @xmath195 , can be much lower than the initial average entropy per particle . it is important to note that after this separation further adiabatic changes of the system parameters can lead to temperature changes , but the average entropy per particle will remain constant . the minimal entropy which can be reached depends on different quantities as seen from eq . [ eq : entropy_shaping2 ] . as expected the initial entropy @xmath186 is an important factor . furthermore , the cooling procedure is more efficient if the entropy per particle in the storage region is much larger than the entropy per particle in the core region , i.e. @xmath196 . for fixed temperatures , a typical behavior of the entropy per particle with the density is shown in fig . [ fig : s_over_n_n_hom ] . as one can see the decrease in entropy per particle with increasing density is more pronounced at larger temperatures . hence the procedure will become less efficient at lower initial temperature . we also find that the differences in entropy per particle are largest between very low and very high densities , such that the procedure would be most efficient if such densities were used for the storage and core regions respectively . however , it is important to note that it is not essential that the state stabilized in the core region is gapped as the band insulator is . what really matters is that a sizable difference is achieved between the entropies per particle characterizing the densities of the core and storage regions . by tuning the trap shape the number of atoms in the core and storage regions can be adjusted . ideally one would like to create a very large storage region with a lot of atoms at very low density , since there the entropy per atom would be maximal . however , this situation can only be achieved within a certain range due to experimental limitations . in particular the trap can only be engineered within a certain spatial extension and the final number of particle ( @xmath197 ) should be reasonably large in order to perform the actual experiment afterward . before turning our attention to the experimentally realizable setup , one important comment is in order . we want to emphasize that the adiabatic isolation of the core region from the storage region is favorable or even necessary to achieve cooling . without proper isolation , the entropy of the storage region can flow back into the experimentally relevant region . at best this backflow of entropy may simply render the cooling scheme less efficient , but in the worst scenario it may heat up the region of interest . at low temperature , heating may occur because the experimentally relevant phase may accommodate more easily the entropy than its parent high density state ( in our case the band insulator ) . therefore , while changing back the trap shape to generate the experimentally relevant phase , the entropy may flow back into the region of interest if the core and storage regions are not properly isolated from each other . [ bt ] ( solid line ( full system ) , dashed line ( core region ) , right upper panel ) as a function of the density . we chose the following experimentally realistic parameters : @xmath198 , @xmath199 , @xmath200 , @xmath201 , @xmath202 , @xmath203 , @xmath204 , @xmath205 , and @xmath206 atoms . the average entropy per particle in the total system is @xmath207 and in the core region @xmath208 . the ratio of particles in the core region versus the total particle number is @xmath209 . obtained with dmft . see also fig . 2 of ref . .,width=302 ] let us now turn our attention to the experimentally realizable trapping potential . to be truly relevant , this trapping profile should be achievable using present technology and should not require fine - tuning of the system parameters . such an experimentally realistic trapping potential has been described in ref . . here we only briefly summarize the setup . in order to achieve the required entropy modulation , the potential has to allow for a tight trapping in the core region , for a wide shallow ring in the storage region and for high potential barriers isolating these two regions from each other . to produce this profile , we envision to use three elements , ( i ) a shallow harmonic trap ( either magnetic or optical ) , ( ii ) a dimple which confines atoms to a small region around the trap symmetry axis and helps to create the band insulator , and ( iii ) a cylindrically - symmetric potential barrier to isolate high and low entropy regions . the dimple ( ii ) and potential barrier ( iii ) are produced by red- and blue - detuned laser beams respectively , creating attractive or repulsive dipole potentials . the dimple has a gaussian profile , while the barrier needs to be a narrow annulus . experimentally this can be realized either by setting a tightly focused laser beam in rapid rotation , or by engineering the beam profile using phase plates or other diffractive optics@xcite . consequently , in addition to the lattice potential , the trapping profile is given by @xmath210 where @xmath211 are the potential amplitudes , @xmath212 is a measure of the anisotropy of the harmonic trap , @xmath213 are the waists of the gaussian laser beams forming the dimple and barrier , @xmath214 is the radius of the cylindrical barrier , and @xmath101 the lattice spacing . an example of the resulting trapping profile is shown in fig . [ fig : realistic_trap ] . the gain in entropy per particle for this particular case is approximately one order of magnitude while keeping about 50@xmath215 of the atoms . in fig . [ fig : realistic_trap ] , in addition to the density distribution and entropy per particle for the corresponding trapping profile , one can see on the right panel from the @xmath123 distribution that the two sharp peaks of the idealized setup have been replaced with broader features . in particular , the lower peak is not well defined and removing the storage region results in an almost complete extinction of its weight . however , experimentally removing the storage region is not straightforward due to the presence of the optical lattice potential . in ref . , several possible methods have been proposed . fortunately , since the publication of this article , different techniques aiming at locally addressing the atoms have been developed and could be employed to remove the storage region . one of these methods makes use of an electron microscope and could be used to blast away the storage region in a very controlled manner @xcite . another possibility would be to use locally trapped ions to remove the atoms from the storage region @xcite . all previous considerations assumed the system evolution to be perfectly adiabatic . in a real experiment this will not be the case as the deformation of the trap needs to be performed within a finite time . in ref . , we showed that , for a one - dimensional system , reshaping the trapping potential can be performed within an experimentally realistic timescale while inducing heat on a scale more than ten times smaller than the antiferromagnetic exchange coupling . consequently , using the scheme presented above , experimentalists could cool down quantum gases to one order of magnitude lower than presently achievable while keeping about half the atoms in the system . cooling into highly sought - after quantum phases could thus be achieved . in this article , we performed a detailed study of the thermodynamics of the three - dimensional fermionic hubbard model , for a rather wide range of couplings . we mainly focused on the temperature regime @xmath216 , of current interest for experiments on cold fermionic atoms in optical lattices . our theoretical study is mainly based on single - site dynamical mean - field theory ( dmft ) , a well established theoretical method based on a controlled approximation in which non - local correlations are neglected but local quantum fluctuations are treated accurately . in addition , we used high - temperature series expansion . a comparative study between these two methods allowed for a precise assessment of their respective range of validity . dmft is found to be accurate down to fairly low temperatures when not too close to half - filling ( one particle per site ) . because of their convergence properties , the series expansion are most useful at low density or exactly at half - filling . at or close to half - filling , single - site dmft can be trusted only down to @xmath217 . below this temperature , short - range magnetic correlations set in , which require the use of cluster extensions of dmft . hence , our study validates the use of single - site dmft for understanding experimental results on cold fermionic atoms in a 3d lattice in the currently accessible temperature range , while future experiments at lower temperature will require extensions of the method . we considered the implications of the thermodynamic properties of the homogeneous system for fermionic atoms in an optical lattice confined in a parabolic potential , within the local density approximation . a state diagram was established , with different regimes for the density profile in the trap . special emphasis was devoted to the distribution function of site occupancies in the trap , a quantity which can be experimentally measured by imaging techniques with single lattice - site resolution . such techniques have recently become available for two - dimensional systems . this distribution function proved to be of particular use when discussing how a given observable ( e.g. , temperature ) changes in the trap as the system evolves in an adiabatic manner when a parameter is varied . indeed , it allows for a clear separation of two contributions , one corresponding to the redistribution of atoms in the trap during the evolution , and another from the intrinsic change of the observable . we applied these considerations to the temperature change in the trap under an adiabatic variation of the coupling , and identified the regimes where a significant cooling can be expected . however , we would like to point out that for these results to be completely applicable to trapped cold atoms confined to optical lattices , one would need to take into account several processes that could prevent a fully adiabatic evolution . considering the effects of processes such as lattice heating will be the subject of future works . finally , we elaborated on a previously proposed cooling mechanism , based on the separation of regions of small and large entropy . this procedure is promising for cooling further cold fermionic systems by approximately one order of magnitude , a necessary step and major future challenge for accessing and studying experimentally strongly correlated phases in those systems . _ note added : _ when the present work was near completion we became aware that another study , conducted by fuchs _ et al . _ , on the thermodynamics of the 3d hubbard model was also in its final stage . the focus of ref . is on the characterization of the low temperature region near the antiferromagnetic transition . in the temperature range where our two works overlap , the results are in agreement . we are grateful to m. ferrero and o. parcollet for discussions and for sharing with us their ct - qmc code . we thank s. wessel for providing us with the entropy data of the heisenberg model from ref . . we acknowledge discussions and previous collaborations on related subjects with i. bloch , t. esslinger , f. gerbier , e. gull , r. jrdens , m. khl , j. oitmaa , l. pollet , a. rosch , c. salomon , and m. troyer . we acknowledge funding from the agence nationale de la recherche , the darpa - ole program , the ` triangle de la physique ' and the fonds qubcois de recherche sur la nature et les technologies . in this appendix , we present a very simple approximation that allows us to obtain the state diagram ( fig . [ fig : state_diagram ] and ref . ) of trapped fermionic atoms in an optical lattice with very little computational effort . it also allows us to obtain approximate analytical expressions for the crossover lines in this state diagram , hence providing qualitative understanding into the numerical results presented above in this article . this approximation is based on the following approximate form of the one - particle spectral function of the homogeneous system : @xmath218 in this expression , @xmath219 is the half - bandwidth and @xmath220 is a parameter that plays the role of the mott gap ( see below ) . the rationale behind this expression is the following . when @xmath221 , ( weakly interacting regime ) it describes a density of states broadened by interaction effects . when @xmath222 , it describes two hubbard bands ( fig . [ fig : analytic_dos ] ) separated by a mott gap . the contribution of quasiparticle states appearing in between the two hubbard bands are neglected , because the temperatures considered in this article are typically larger than the quasiparticle coherence scale . compared to the atomic limit , this approximation has a better behavior in the low temperature limit while still retaining a simplicity that allows for a completely analytic solution . the disadvantage of the atomic limit is that it models the zero temperature spectral function as an unrealistic pair of delta functions located at @xmath223 and @xmath11 . the atomic limit is indeed the first term of the expansion in @xmath18 and as such is valid at high temperature . our approximation instead takes into account the fact that the hubbard bands are broadened by the kinetic term . the result is in better agreement with the dmft data at temperatures lower than the limit of validity of the atomic limit . our approximation eventually breaks down due to the fact that at low temperature the details of the shape of the hubbard bands become more relevant and the box - like model of the bands shows its limits . in this sense , our approximation should only be regarded as acceptable at high temperature . this form of the spectral function leads to the following expression of the dependence of the particle number on chemical potential : @xmath224 in the zero - temperature limit , this expression reduces to : @xmath225 i.e. a function with steps at @xmath31 , @xmath42 and @xmath33 ( zero compressibility ) , while @xmath30 in between the steps ( liquid regions ) is approximated by a linear dependence on @xmath12 ( constant compressibility approximation ) . we regard the parameter @xmath220 as a fitting parameter , function of coupling and temperature , and perform a least square minimization of eq . [ eq : n_analytic_model ] to the dmft data to determine its value . the result of this procedure is displayed in fig . [ fig : gap_vs_t ] . the fit looses in quality as we reduce the temperature and the value of @xmath11 . restricting to the high temperature and large @xmath11 region , the dependence of @xmath220 on @xmath11 and @xmath75 can be separated to a good approximation and @xmath227 can be fit remarkably well by the law : @xmath228 where the parameter with the largest deviation is @xmath229 which shows a slight increase as a function of @xmath11 . using the approximate expression ( [ eq : n_analytic_model ] ) of @xmath230 for the homogeneous system , it is straightforward to obtain the state diagram of the trapped system by using the relation between the scaled density @xmath133 and the chemical potential @xmath103 at the center of the trap , which reads , in the lda approximation : @xmath231 the calculations can be performed analytically in the @xmath80 limit ( we warn the reader that @xmath80 is considered here only as a formal limit in order to make an analytical calculation possible , since the approximations made in this paper are no longer valid physically in this limit ) . let us for example focus on the crossover lines which delimit the ` mc ' region in fig . [ fig : state_diagram ] ) within which the central occupancy is @xmath42 . in view of our approximate form of @xmath30 , the lower boundary of this regime will correspond to @xmath232 and the upper boundary to @xmath233 . inserting these values into ( [ eq : lda_bis ] ) and performing the integrations at @xmath80 yields the following expressions for the lower and upper boundaries of the mc region : @xmath234 @xmath235 using the above determination of @xmath220 , this provides an explicit form of the boundaries . analytical expressions can be similarly obtained for all crossover lines in the state diagram . in fig . [ fig : state_diagram_app ] we compare these approximate analytical expressions to the actual lines obtained for a dmft calculation at a low enough temperature @xmath236 and find very satisfactory agreement . finally , we note that thermodynamic quantities such as the double occupancy and the entropy can in principle be reconstructed from a given approximate expression for @xmath237 thanks to the thermodynamic ( maxwell ) relations relating their derivatives : @xmath238

we explore the potential of a formulation of the navier - stokes equations incorporating a random description of the small - scale velocity component . this model , established from a version of the reynolds transport theorem adapted to a stochastic representation of the flow , gives rise to a large - scale description of the flow dynamics in which emerges an anisotropic subgrid tensor , reminiscent to the reynolds stress tensor , together with a drift correction due to an inhomogeneous turbulence . the corresponding subgrid model , which depends on the small scales velocity variance , generalizes the boussinesq eddy viscosity assumption . however , it is not anymore obtained from an analogy with molecular dissipation but ensues rigorously from the random modeling of the flow . this principle allows us to propose several subgrid models defined directly on the resolved flow component . we assess and compare numerically those models on a standard green - taylor vortex flow at reynolds 1600 . the numerical simulations , carried out with an accurate divergence - free scheme , outperform classical large - eddies formulations and provides a simple demonstration of the pertinence of the proposed large - scale modeling . large - scale fluid flow dynamics , stochastic transport , subgrid model , turbulence , taylor - green flow

1611.03413

the large - scale modeling of fluid flow dynamics remains nowadays a major research issue in fluid mechanics or in geophysics despite an enormous research effort since the first investigations on the subject 150 years ago @xcite . the research themes behind this topic cover fundamental issues such as turbulence modeling and the analysis of fully developed turbulent flows , but also more applicative research problems related to the definition of practical numerical methods for the simulation of complex flows . in this latter case the difficulty consists in setting up a reliable modeling of the large - scale dynamics in which the contribution of unresolved small - scale processes are explicitly taken into account . for the navier - stokes equations , the problem is all the more difficult that the spatial and temporal scales are tightly interacting together . the neglected processes include , among others things , the action of the unresolved motion scales , complex partially - known forcing , an incomplete knowledge of the boundary conditions and eventual numerical artifacts . such unresolved processes must be properly taken into account to describe accurately the energy transfers and to construct stable numerical simulations . in real world situations , the complexity of the involved phenomenon prevents the use of an accurate but inescapably expensive deterministic modeling . we advocate instead the use of a stochastic modeling . within this prospect , we aim at describing the missing contributions through random fields encoding a flow component only in a probability distribution sense . those variables correspond to the discrepancies or errors between the dynamical model and the actual dynamics . their modeling is of the utmost importance in geophysics , either for data assimilation or forecasting issues . in both cases , an accurate modeling of the flow errors dynamics enables to maintain an ensemble of flow configurations with a sufficient but also meaningful diversity . small - scale processes are responsible both for an energy dissipation but also for local backscattering of energy @xcite . the introduction of random variables in the flow dynamics has been considered by several authors , as it constitutes an appealing mechanism for the phenomenological modeling of intermittent processes associated to the inverse energy cascade @xcite . recently those models have regained a great interest for the modeling of geophysical flows @xcite in climate sciences ( see also the thematic issue @xcite or the review @xcite ) . numerous turbulence models proposed in the context of large eddies simulations ( les ) and reynolds average simulations ( rans ) introduce _ de facto _ an eddy viscosity assumption to model the energy dissipation due to unresolved scales @xcite . this concept dates back to the work of boussinesq @xcite and prandtl @xcite . it relies on the hypothesis that the energy transfer from the resolved scales to the subgrid scales can be described in a similar way to the molecular viscosity mechanism . it is therefore not at all related to any uncertainty or error quantities . in models dealing explicitly with a statistical modeling of the turbulent fluctuations there is thus some incoherency in representing directly the dissipative mechanism attached to random terms through an eddy viscosity assumption . in this work we will not make use of such an hypothesis . instead , we will rely on a general diffusion expression that emerges naturally from our formalism . this subgrid model is properly derived from a general lagrangian stochastic model of the fluid motion in which the fluid parcels displacement is decomposed in two components : a smooth differentiable ( possibly random ) function and a random field , uncorrelated in time but correlated in space . such a decomposition consists in separating or `` filtering '' a rough velocity in a smooth slow time - scale component and a fast oscillating velocity field representing the unresolved flow . though there is , in general , no sharp time - scale separation in turbulent flows , the resolved velocity can be interpreted as a temporally coarse - grained component whereas the time - uncorrelated component stands for the small time - scale unresolved velocity . as a temporal smoothing imposes implicitly a spacial smoothing , this separation can be thus interpreted in terms of a les filtering technique . yet , the corresponding eulerian formulation does not ensue from a filtering procedure . it is thus not prone to errors associated to the violation of the commutation assumption between the filter and the spatial derivatives @xcite . besides , those equations introduce an effective advection related to the small - scale velocity inhomogeneity . this modified advection , empirically introduced in langevin models of particle dispersion @xcite , corresponds exactly to a phenomenon , termed _ turbophoresis _ , related to the migration of inertial particles in regions of lower turbulent diffusivity @xcite . the large - scale representation of the navier - stokes equations on which we rely in this study are built from a stochastic version of the reynolds transport theorem @xcite . this modified reynolds transport theorem , which constitutes here the cornerstone of our large - scale dynamics representation , is presented in the following section . general invariance properties of the corresponding large - scale dynamics such as scale and galilean invariances are detailed in a comprehensive apppendix . in section [ experiments ] several novel subgrid tensors will be devised and compared on a standard green - taylor vortex flow @xcite . we will show that all the proposed schemes outperform the usual dynamic smagorinsky subgrid formulation @xcite . numerous methodological choices can be envisaged to devise stochastic representations of the navier - stokes equations . the simplest method considers additional random forcing to the dynamics . this is the choice that has been the most often performed since the work of benssoussan @xcite . another choice , in the wake of kraichnan s work @xcite , consists in closing the large - scale flow representation in the fourier space by relying on a langevin equation @xcite . obviously the frontiers between those two methodologies are sometimes fuzzy , and numerous works rely on both strategies in order to devise the shape that should take the random variables evolution law @xcite . lagrangian stochastic models based on langevin equations have been also intensively used for turbulent dispersion @xcite or in probability density function ( pdf ) modeling of turbulent flows @xcite . those lagrangian models , which require to model the drift and diffusion functions , lead to very attractive particle based representations of complex flows @xcite . they are nevertheless not adapted to global large - scale eulerian representations of the flow dynamics . in this work , we will rely on a different framework in specifying the stochastic nature of the velocity from the very beginning as proposed in @xcite . the basic idea is built on the assumption that the lagrangian fluid particles displacement results from a smooth velocity component and a highly oscillating stochastic velocity component uncorrelated in time , @xmath0 with the velocity components : @xmath1 in this decomposition the first right - hand term is a smooth function of time associated to the large - scale velocity component . the second term stands for the small - scale velocity field . it is a white noise velocity component defined from the ( formal ) time - derivative of the random field : @xmath2 . this random field is a three - dimensional centered wiener process ; it is thus uncorrelated in time but can be anisotropic and inhomogeneous in space . since we focus in this study only on incompressible flows , the small - scale component is defined as a divergence - free random field ; it is hence associated to a divergence - free diffusion tensor : @xmath3 analogously to the standard deterministic case , the derivation procedure from the physical conservation laws of the navier - stokes equations is based primarily on the reynolds transport theorem ( rtt ) . the rtt provides the expression of the rate of change of a scalar function , @xmath4 , within a material volume , @xmath5 . for a stochastic flow ( [ sto - velocity ] ) with an incompressible small - scale velocity component ( @xmath6 ) , this expression derived in @xcite ( see also appendix [ appendix : srt ] where a comprehensive derivation is provided for readers convenience ) , is given by : @xmath7 { { \mathrm{d}}}t + { \boldsymbol{\nabla}}q { \boldsymbol{\cdot}}{\boldsymbol{\sigma}}{{\mathrm{d}}}{\boldsymbol{b}}_t\biggr){{\mathrm{d}}}\x . \label{sto - r}~\end{gathered}\ ] ] this modified rtt involves the time increment of the random scalar quantity @xmath4 ( the differential of @xmath4 at a fixed point ) instead of the time derivative . a diffusion operator emerges also naturally . for clarity s sake , this term is designated as `` subgrid stress tensor '' following the protocols of large eddies simulation ( les ) . however , its construction is quite different . it is not based on boussinesq s eddy viscosity assumption nor on any structural turbulence models @xcite but arises directly from stochastic calculus rules . it expresses the mixing process exerted on the scalar quantity by the fast oscillating velocity component . this diffusion term is directly related to the small - scale component through the _ variance tensor _ , @xmath8 , defined from the diagonal of the small - scale velocity covariance : @xmath9 it can be checked that the variance tensor corresponds to an eddy viscosity term ( with units in @xmath10 . this term plays thus a role similar to the eddy viscosity models introduced in classical large scale representations @xcite or to the spectral vanishing viscosity @xcite . it is also akin to numerical regularization models considered in implicit models @xcite . our approach is nevertheless more general as it does not rely on _ a priori _ fixed shape of the reynold stress ( e.g. boussinesq assumption ) nor does it presuppose a given numerical discrete scheme ( e.g. implicit models ) . a corrective advection term , @xmath11 , appears also in the stochastic rtt formulation . this correction expresses the influence of the small scales inhomogeneity on the large ones . a drift is engendered from the regions associated with maximal variance ( maximal turbulent kinetic energy - tke ) toward the area of minimum variance ( e.g. minimal tke ) . this corresponds to the _ turbophoresis _ phenomenon associated with turbulence inhomogeneity , which drives inertial particles toward the regions of lower diffusivity @xcite . for homogeneous noise , the variance tensor is constant and this corrective advection does not come into play . it can be noted that this advection correction is of the same form as the one proposed in @xcite . through this modified rtt , stochastic versions of the mass and momentum conservation equations can be ( almost ) directly derived . incompressibility conditions can for instance be immediately deduced from the rtt applied to @xmath12 and the flow jacobian ( @xmath13 ) : @xmath14 together with the incompressibility of the random term , the incompressibility condition reads thus : @xmath15 in the case of an incompressible large - scale flow component , @xmath16 , this reduces to : @xmath17 note that for a divergence - free isotropic random field such as the kraichnan model @xcite the last condition is naturally satisfied , since this unresolved velocity component is associated with a constant variance tensor . applying the rtt to the fluid density , @xmath18 , and canceling this expression for arbitrary volumes , we get the following mass conservation constraint : @xmath19 for an incompressible fluid with constant density , together with a volume - preserving ( isochoric ) condition on the large - scale velocity component , we retrieve the incompressibility conditions ( [ incompressibility ] ) . it can be noted also that equation ( [ mass - conservation-1 ] ) still constitutes a transport equation since it preserves energy . as a matter of fact , it can be shown that the energy intake brought by the small - scale component is exactly compensated by the energy dissipation associated to the diffusion term @xcite . the application of the stochastic version of the rtt on the stochastic momentum and the introduction of the forces acting on the flow enables to derive from the second newton law the following navier - stokes equations @xcite : [ sto - ns ] @xmath20 in this expression @xmath21 is the dynamic viscosity , @xmath22 denotes the large - scale ( slow ) pressure contribution and @xmath23 is a zero - mean turbulent pressure related to the small - scale velocity component . similarly to the classical reynolds decomposition , the dynamics of the resolved component includes an additional stress term that depends here on the resolved velocity component . a correction of the advection velocity also occurs . both terms depend on the variance tensor which gathers the action of the turbulent fluctuations on the large - scale velocity . it can be observed that the large - scale energy evolution is dissipative . this generalizes thus the boussinesq 1877 assumption , which conjectures a dissipative effect of the turbulent fluctuations on the mean flow . in the case of a divergence - free isotropic random field ( with a constant diagonal variance tensor ) , this system boils down to an intuitive constant eddy viscosity diffusivity model : @xmath24 where the laplacian dissipation term is augmented by the random field variance . the use of constant eddy viscosity thus finds here its justification as a direct consequence of an isotropic turbulence assumption . the subgrid stress tensor involved in our formalism constitutes an anisotropic diffusion whose preferential directions of diffusion are given by the small - scale velocity variance tensor . setting the diffusion tensor , @xmath25 , or its associated variance tensor allows us to define directly the subgrid diffusion term and the effective advection . for instance imposing to the small - scale random velocity to live on the iso - density surfaces provides immediately a clear justification of the isopycnal diffusion employed in oceanic circulation models @xcite . the specification of the turbulent fluctuations in terms of a stochastic process provides a means to interpret different subgrid models but also to devise new ones either through _ a priori _ specifications or data - driven strategies . general invariance properties of the proposed large - scale representation are listed in the appendix [ prop - nsu ] . we briefly summarize them here . it is in particular shown that the distribution of the velocity anomaly is in the general case not gaussian and does not consequently correspond to normal or quasi - normal approximations @xcite . we show that this stochastic representation has remarkable invariance properties ; it is galilean invariant and preserves ( in the absence of molecular viscosity ) the euler equations scale invariance properties . otherwise , a useful scaling for the variance tensor is derived from the kolmogorov - richardson scaling and a dimensionless number relating the large - scale kinetic energy to a characteristic value of the velocity variance at the resolved scale . interested readers may also consult @xcite for the derivation of stochastic representation of several geophysical flows . in this section we assessed , through numerical simulations , the performances of the proposed large - scale dynamics for different variance tensor models . those simple models have been defined from local statistics of the resolved component and compared to the classical smagorinsky subgrid model associated with a dynamical procedure @xcite . those numerical experiments have been performed on the taylor - green flow @xcite . taylor - green vortex flow is a critical test for numerical schemes , as both the convective term and viscous term play important roles . due to the energy cascade generated by the convective term , the flow becomes rapidly turbulent with the creation of small - scale structures up to a dissipation peak . this stage is followed by a decay phase similar to a decaying homogeneous turbulence . as a consequence , a precise and high - order representation of the viscous and convective terms is needed to get an accurate numerical simulation . this flow is considered as a prototypical system to study the production of small - scale eddies due to vorticity increase and vortex stretching mechanism @xcite . in cartesian coordinates , this flow is defined by the following initial conditions : @xmath26 and @xmath27 the computation domain is defined as a cubic box with periodic boundary conditions on all the faces . the length of the domain is set to @xmath28 in each of the axis direction , which gives a characteristic length @xmath29 and a reynolds number @xmath30 . in the literature , several high - order numerical methods have been tested for the direct numerical simulation ( dns ) of the taylor - green vortex flow , see @xcite and references therein . in this study we used a discrete scheme built on a divergence - free wavelet basis @xcite . this scheme presents several computational advantages . first of all , it guarantees a divergence - free solution in the physical domain with a good numerical complexity . besides , as the spatial filters considered corresponds to a multi - resolution projection , two successive filtering operations can be switched together . this property reveals useful within the germano dynamic strategy enabling to estimate the subgrid tensor weight factor . this numerical scheme achieves similar performances to a pseudo - spectral method . the different numerical simulations performed are mainly assessed and compared according to the evolution along time of the following criterion : * the mean kinetic energy ( ke ) @xmath31 * the mean kinetic energy dissipation rate @xmath32 where @xmath33 is the rate of strain tensor : @xmath34 . the mean kinetic energy is linked to the dissipation rate by : @xmath35 to clearly separate those two forms of the energy dissipation we will denote @xmath36 the expression computed from the kinetic energy differentiation and @xmath37 the dissipation computed from the rate of strain norm . besides , in the sequel , all those averaged quantities are computed in their dimensionless form : @xmath38 where @xmath39 denotes the convective time . the temporal evolution of those mean energy quantities enable to monitor the quality of the solution over time and to assess the accuracy of the discrete scheme used for the velocity gradients . in addition to those criterion , we will plot the energy spectrum of the resolved velocity at several distinct times . we evaluated first the ability of the divergence - free wavelet based method to reproduce faithfully the main features of the taylor - green flow . for this purpose , two different direct numerical simulations have been conducted . the first experiment concerns a simulation at a low reynolds number : @xmath40 . for this case the divergence - free wavelet has been run on a regular grid of @xmath41 points . for comparison purpose we performed a classical pseudo - spectral simulation with @xmath41 fourier modes . let us note that at this low reynolds number ( @xmath40 ) , only @xmath42 fourier modes are required to represent accurately all the hydrodynamics scales in the limit of @xmath43 , where @xmath44 is the kolmogorov scale @xcite . the time evolution of the mean kinetic energy and the mean dissipation rate obtained for both methods are plotted on the left and right panels of figure [ energyfig ] respectively . as can be observed on those two graphics the solutions superimpose almost perfectly . for both methods the energy dissipation computed from the rate of strain tensor norm , @xmath45 , and from the kinetic energy differentiation , @xmath46 , coincide perfectly . we plotted therefore only the rate of strain norm . ( left ) and its dissipation rate @xmath47 ( right ) as a function of the dimensionless time . comparison between the divergence - free wavelet method and the fourier pseudo - spectral method for @xmath40.,title="fig:",scaledwidth=55.0% ] ( left ) and its dissipation rate @xmath47 ( right ) as a function of the dimensionless time . comparison between the divergence - free wavelet method and the fourier pseudo - spectral method for @xmath40.,title="fig:",scaledwidth=55.0% ] in a second experiment , both simulations have been then carried out for a moderate reynolds number fixed at @xmath48 . since the study of @xcite , numerous numerical experiences have been conducted to understand accurately the dynamic of the taylor - green vortex flow at this reynolds number . as mentioned in @xcite , all the scales of the flow are captured with @xmath49 degrees of freedom but @xmath50 degrees of freedom are sufficient to represent its main characteristics . we therefore run the divergence - free wavelet based simulation on a @xmath50 mesh grid . we also performed a fourier pseudo - spectral simulation with a dealiasing procedure at the same resolution . on figure [ energyfig1600 ] we displayed for both simulations the kinetic energy and dissipation rate time evolutions . on the same figure , for comparison purpose , we also plotted the curves corresponding to a @xmath49 fourier modes pseudo - spectral solution ( with dealiasing ) available from @xcite . as in the previous case the dissipation @xmath45 matches @xmath46 . we therefore only show the rate of strain tensor norm in the right panel of figure [ energyfig1600 ] . ( left ) and its dissipation rate @xmath45 ( right ) as a function of the dimensionless time . comparison between the divergence - free wavelet method and two fourier pseudo - spectral simulations for @xmath51.,title="fig:",scaledwidth=55.0% ] ( left ) and its dissipation rate @xmath45 ( right ) as a function of the dimensionless time . comparison between the divergence - free wavelet method and two fourier pseudo - spectral simulations for @xmath51.,title="fig:",scaledwidth=55.0% ] it can be observed that the divergence - free wavelet simulation is in good agreement with the spectral reference solution , especially before the dissipation peak when the convective phenomena is predominant . slight discrepancies appear in the decaying phase starting after the dissipation peak ( @xmath52 ) . those experiments show that the divergence - free wavelet based method provides results of comparable quality to those obtained from pseudo - spectral simulations at identical resolutions . the wavelet simulation performed on the @xmath50 mesh grid will serve as a reference " dns for comparison purpose . we relied on this method to carry out all the numerical simulation performed in this study . the performances of various subgrid modeling are discussed in the next section . one of the main advantages of the stochastic formalism we propose lies in the great flexibility of the anisotropic diffusion specification . the variance tensor , @xmath53 , can be fixed from _ a priori _ shape imposed either directly on the small scale variance or on the diffusion tensor . in some cases , such knowledge could probably be inferred from the physical approximations considered to constitute the model . aspect ratio simplifications and/or boundary conditions that are not perfectly known could be used as well to constrain the small - scale velocities to live on specific iso - surface . another route would consist in specifying the turbulent velocity components from small - scale measurements statistics . despite all those directions are worth exploring as they open new strategies to design sub - grid tensors , in this study we choose to focus on several simple models of tensor @xmath54 to explore the potential of the proposed modeling . the first model consists in assuming ergodic assumption to compute the variance tensor from local statistics of the resolved velocity component . the variance tensor is here defined from empirical velocity covariance computed on a local neighborhood : @xmath55 where @xmath56 denotes the empirical mean on a spatial or temporal window @xmath57 , @xmath58 is the spatial scale considered for the simulation , which corresponds to a mesh grid composed of @xmath59 grid points , and @xmath44 is the kolmogorov scale . the empirical averaging , @xmath60 , is computed either spatially over a small ( @xmath61 ) window centered around point @xmath62 or temporally , at point @xmath62 , over the time interval @xmath63 $ ] . in the following , they are referred to as the spatial and temporal local covariances respectively . both models are weighted by the variance tensor scale ratio derived in section [ section - scaling ] . the second model is defined from two successive filtering of the resolved velocity component and a scale similarity assumption . the filtering is defined through the associated wavelet multi - scale projector . more precisely , since the velocity @xmath64 can be decomposed as : @xmath65+\sum_{|\boldsymbol j| \ge |{\boldsymbol \ell}|}{\cal q}_{\boldsymbol j}[{\boldsymbol{w}}(\x , t)],~~{\cal q}_{\boldsymbol j}={\cal p}_{\boldsymbol j+1}-{\cal p}_{\boldsymbol j},\ ] ] with @xmath66 the projector onto the scaling functions basis @xcite , the resolved filtered velocity @xmath67 is defined by : @xmath68.\ ] ] the second , so called `` test '' , filtering is here defined by the projection of @xmath69 onto resolution @xmath70 , which is the immediate coarser resolution following the simulation scale : @xmath71.\ ] ] the stochastic navier - stokes equations ( [ eqinitialsyst1 ] ) rewritten respectively for @xmath72 and @xmath73 reads : @xmath74 and @xmath75 taking the difference of the momentum equations and for the two subsequent levels @xmath76 and @xmath77 provides the residual dynamics : @xmath78 where @xmath79 , @xmath80 and @xmath81 . note that a similarity assumption @xmath82 has been imposed for the variance tensor . it can be noticed that , if the projector @xmath83 commutes with differentiation , due to the filtering projection property , the stokes equation in the left - hand side cancels after filtering . then , instead of solving the right - hand expression at level @xmath77 and then projecting back at a finer level the estimated variance tensor , we rather propose to solve it at scale , @xmath84 , in a least squares sense . introducing the variance tensor incompressibility constraint we seek the minimizer of the following nonlinear functional : @xmath85 with the positivity constraint : @xmath86 in practice the minimization of the functional @xmath87 has been carried out using a quasi - newton method combined with the broyden - fletcher - goldfarb - shanno ( bfgs ) method @xcite to approximate the hessian matrix . the variance tensor has been assumed to be diagonal at all points . in order to impose the positivity constraint , instead of computing @xmath54 we preferred to compute its square root @xmath88 . for details on the computation of the gradient of @xmath87 , we refer to appendix [ gradientcomputation ] . the third model evaluated corresponds to the classical smagorinsky eddy viscosity formulation , @xmath89 coupled with the germano estimation procedure @xcite to fix dynamically the eddy viscosity constant from a least squares estimation @xcite and a filtering of the velocity field at two consecutive scales . the `` test '' filtering is as previously defined by the projection of @xmath90 onto the immediate coarser resolution @xmath91 . denoting the filtering operation @xmath92 , the constant @xmath93 is given by : @xmath94 with @xmath95 and @xmath96 denotes a spatial averaging . the performances of the different sub - grid models have been compared for a @xmath42 mesh grid , which corresponds to a resolution scale @xmath97 . figure [ energyfigles ] shows for the different models the time evolution of the mean kinetic energy ( left panel ) and the mean energy dissipation rate ( right panel ) . those curves can be compared to the reference solution computed on a @xmath50 grid ( @xmath98 ) . an important remark can be raised here . the filtered velocity fields do not correspond to the solution of a navier - stokes equation started from the large - scale ( filtered ) initial conditions . it is thus not relevant to correlate the large - scale solutions with the filtered velocity . it is more meaningful to compare them directly with the dns . from the rate of strain tensor norm displayed on the right side of figure [ energyfigles ] , two groups of subgrid tensor can be recognized . the first one gathers the dynamic smagorinsky tensor ( dyn - s ) and the variance tensor defined from the spatial local velocity covariance ( spatial ) . the second group is composed of the optimal variance tensor with a scale similarity assumption ( opt - ss ) and the variance tensor built from the velocity temporal covariance ( temporal ) . for both groups , we displayed on the right - hand side of figure [ energyfigles2 ] and figure [ energyfigles3 ] the energy dissipation rate computed from the kinetic energy rate of change . on the left - hand side we plotted the kinetic energy associated to each model . those curves can be compared to the dns reference curve . we observe from those figures that all the large - scale models achieved to reproduce the right amount of kinetic energy and energy dissipation rate . the dynamic smagorinsky subgrid tensor is the model which provides the lower kinetic energy and the lower rate of strain norm . we note the first group of subgrid tensors exhibits in a general way a too fast dissipation in the first phase of the flow ( @xmath99 $ ] ) , characterized by the domination of the advection mechanism . an undue smoothing of the resolved velocity gradient results from this too high dissipation rate . this is confirmed by looking at the rate of strain norm plotted in figure [ energyfigles ] . within this first group , the model build from local spatial velocity covariance performs better than the original smagorinsky tensor . the second group of variance tensor outperforms clearly the results of the first group . both the temporal covariance model and the optimal variance tensor show strikingly very close results . the optimal variance tensor has a slightly higher dissipation rate at the dissipation peak whereas the temporal covariance model fits almost perfectly the dns results around the dissipation peak ( @xmath100 $ ] ) . both models slightly differ from the dns in the decaying phase . in a general way , we note that all the models have a too slow dissipation rate at the end of the decaying phase ( @xmath101 ) . they all smooth too much the velocity gradients in that regime . it must be outlined that the good behavior of the temporal covariance confirms the relevance of the variance tensor scaling since no dynamics strategy has been performed in that case . ( left ) and of the energy dissipation rate @xmath102 ( right ) as a function of the dimensionless time . comparison between the three variance tensor models , the dynamic smagorinsky model and the dns wavelet - based solution for the taylor - green vortex at @xmath51.,title="fig:",width=283,height=226 ] ( left ) and of the energy dissipation rate @xmath102 ( right ) as a function of the dimensionless time . comparison between the three variance tensor models , the dynamic smagorinsky model and the dns wavelet - based solution for the taylor - green vortex at @xmath51.,title="fig:",width=283,height=226 ] ( left ) and of the energy dissipation rate @xmath103 ( right ) as a function of the dimensionless time . comparison between the spatial covariance model , the dynamic smagorinsky model and the dns wavelet - based solution for the taylor - green vortex at @xmath51.,title="fig:",width=283,height=226 ] ( left ) and of the energy dissipation rate @xmath103 ( right ) as a function of the dimensionless time . comparison between the spatial covariance model , the dynamic smagorinsky model and the dns wavelet - based solution for the taylor - green vortex at @xmath51.,title="fig:",width=283,height=226 ] ( left ) and of the energy dissipation rate @xmath104 ( right ) as a function of the dimensionless time . comparison between the temporal local covariance model , the optimal scale similarity model and the dns wavelet - based solution for the taylor - green vortex at @xmath51.,title="fig:",width=283,height=226 ] ( left ) and of the energy dissipation rate @xmath104 ( right ) as a function of the dimensionless time . comparison between the temporal local covariance model , the optimal scale similarity model and the dns wavelet - based solution for the taylor - green vortex at @xmath51.,title="fig:",width=283,height=226 ] ( left ) and @xmath105 ( right ) . comparison between the three variance tensor models , the dynamic smagorinsky model and the dns solution . taylor - green vortex at @xmath51.,title="fig:",scaledwidth=55.0% ] ( left ) and @xmath105 ( right ) . comparison between the three variance tensor models , the dynamic smagorinsky model and the dns solution . taylor - green vortex at @xmath51.,title="fig:",scaledwidth=55.0% ] ( left ) and @xmath106 ( right ) . comparison between the three variance tensor models , the dynamic smagorinsky model and the dns solution . taylor - green vortex at @xmath51.,title="fig:",scaledwidth=55.0% ] ( left ) and @xmath106 ( right ) . comparison between the three variance tensor models , the dynamic smagorinsky model and the dns solution . taylor - green vortex at @xmath51.,title="fig:",scaledwidth=55.0% ] we display next on figure [ spectreles1 ] and on figure [ spectreles2 ] the energy spectrum associated to the different models at four different instants . all the models provide satisfying solutions with similar spectrum . compared to the other models , the dynamic smagorinsky subgrid stress produces a noticeable energy bump at the cutoff scale . the temporal and optimal variance tensor have spectrum which are in general closer to the dns spectrum . at the end of the turbulence decay phase ( @xmath107 ) all the models provide close results . as we rely on a wavelet scheme for the numerical simulations , it is insightful to inspect the discrete power spectra computed from the wavelet coefficients . they are plotted on figures [ wspectreles1 ] and [ wspectreles2 ] . wavelet power spectrum corresponds to an averaged version of the fourier spectrum @xcite and exhibits the same slopes as the fourier spectrum @xcite . a discrete version of the wavelet spectrum as plotted here provides one energy measure per scale level . it can be observed on the four spectra that the dynamic smagorinsky tensor exhibits an undue energy intake at the cutoff scale . this amplification of energy is likely due to noisy velocity fields at the cutoff . the three different model ensuing from our statistical representation of the small - scale velocity component does not show such a deficiency . ( left ) and @xmath105 ( right ) . comparison between the three variance tensor models , the dynamic smagorinsky model and the dns solution . taylor - green vortex at @xmath51.,title="fig:",scaledwidth=55.0% ] ( left ) and @xmath105 ( right ) . comparison between the three variance tensor models , the dynamic smagorinsky model and the dns solution . taylor - green vortex at @xmath51.,title="fig:",scaledwidth=55.0% ] ( left ) and @xmath106 ( right ) . comparison between the three variance tensor models , the dynamic smagorinsky model and the dns solution . taylor - green vortex at @xmath51.,title="fig:",scaledwidth=55.0% ] ( left ) and @xmath106 ( right ) . comparison between the three variance tensor models , the dynamic smagorinsky model and the dns solution . taylor - green vortex at @xmath51.,title="fig:",scaledwidth=55.0% ] for information purpose we draw on figure [ smgctfig1600 ] the value along time of the constant weighting the dynamic smagorinsky model . the obtained maximum value is about @xmath108 and the mean value is @xmath109 , this is in good agreement with the predicted values , see @xcite . and @xmath97.,scaledwidth=55.0% ] regarding the different criteria explored , we see that the proposed large - scale subgrid tensors based on local velocity covariances perform better than the dynamic smagorinsky subgrid model . among all the models proposed , the optimal variance tensor estimation based on a two - scale similarity assumption and the variance tensor constructed from a local temporal variance of the resolved velocity field outperform significantly the others . they lead to very good results that are strikingly close to each other . to our knowledge they both outperform the state - of - the - art solutions for this flow ( * ? ? ? * ; * ? ? ? * and references therein ) . though very simple , the temporal covariance model weighted by a unique constant fixed through the quadratic scaling rule presented section [ section - scaling ] gives very good results . this demonstrates the pertinence of this scaling . the choices operated here are among the simplest that can be devised . more involved schemes could be easily imagined . for instance , variance tensors based on vorticity statistics might be very interesting to explore . another route would be to elaborate this tensor from statistics extracted from measurements or dns data . large - eddies simulation models could likely be proposed in this spirit for non - homogeneous turbulence or boundary layer flows . note also that both correlation schemes rely on a constant fixed through a rough dimensional scaling . a dynamical procedure in the same vein as the one used for the dynamics smagorinsky subgrid model could be beneficial to get a finer estimate of the constant . as our purpose was here to bring a simple demonstration of the wide potentiality offered by the proposed stochastic modeling , we leave such potential improvements to future works . in this paper we have described a decomposition of the navier - stokes equations in terms of a temporally smooth velocity component and a fast oscillating random field associated to the unresolved flow component . this decomposition leads to a new large - scale representation paradigm , which can be interpreted as a large eddies simulation formalized through a time - scale separation . an advection correction and a subgrid diffusion term both emerge from this formalism . they encode respectively _ turbophoresis _ phenomenon and anisotropic mixing effect due to turbulence . the corresponding subgrid tensor enables generalizing the boussinesq eddy viscosity concept on solid a theoretical ground . such large - scale representation have been assessed on a taylor - green vortex flow . we compared different models of the variance tensor built from local averaging and a scale similarity least squares estimation procedure . the different numerical simulations outperform a standard dynamic smagorinsky model based on boussinesq s eddy viscosity assumption . we believe those first results constitute are very encouraging as more advanced models could yet yield substantial improvements . as a matter of fact this formalism , built from physical conservation laws , paves the way to new possibilities to design efficient subgrid schemes . one could for instance explore models where the diffusion tensor is learned from dns data or from small - scale observations . another route of investigation consists in defining adapted basis for the small - scale random field from the fluctuations observed on two consecutive scales of the resolved tensor . 10 a. aspden , n. nikiforakis , s. dalziel , and j.b . analysis of implicit les methods . , 3(1):103126 , 2008 . j. bardina , j.h . ferziger , and w.c . reynolds . improved subgrid scale models for large eddy simulation . volume 80 , page 1357 . a. bensoussan and r. temam . equations stochastique du type navier - stokes . , 13 , 1973 . boris , f.f . grinstein , e.s . oran , and r.l . new insights into large - eddy simulation . , 10:199228 , 1992 . j. boussinesq . essai sur la thorie des eaux courantes . mmoires prsents par divers savants lacadmie des sciences , 23 ( 1 ) : 1680 , 1877 . m. brachet , d. meiron , s. orszag , b. nickel , r. morf , and u.frisch . small - scale structure of the taylor - green vortex . , 130(6):411452 , 1983 . m. brooke , k. kontomaris , t. hanratty , and j. mclaughlin . turbulent deposition and trapping of aerosols at a wall , a 4 , 825834,1992 . ch - h . bruneau , p. fischer , z. peter , and a. yger comparison of numerical methods for the computation of energy spectra in 2d turbulence . part i : direct methods , 1(1):0:50 , 2002 . r. buizza , m. miller , and t.n . stochastic representation of model uncertainties in the ecmwf ensemble prediction system . , 125:28872908 , 1999 . t. chacon - rebollo and r. lewandowski . . springer new - york , 2014 . chapelier , m. de la llave plata , f. renac , and e. martin . final abstract for onera taylor- green dg participation . in _ 1st international workshop on high - order cfd methods at the 50th aiaa aerospace sciences meeting _ , nashville , tn , 2013 . m. caporaloni , f. tampieri , f. trombetti , and o. vittori transfer of particles in non- isotropic air turbulence . , 32:565568,1975 . t. dairay , e. lamballais , s. laizet , and j.c . numerical dissipation vs. subgrid - scale modeling for large eddy simulation . to be published , 2016 . j. deardorff . a numerical study of three - dimensional turbulent channel flow at large reynolds numbers . , 1970 . e. deriaz and v. perrier divergence - free and curl - free wavelets in 2d and 3d , application to turbulent flows . , 7(3):137 , 2006 . c. franzke , t. okane , j. berner , p. williams , and v. lucarini . stochastic climate theory and modeling . , 6(1):6378 , 2015 . p. gent and j. mcwilliams . isopycnal mixing in ocean circulation models . , 20:150155 , 1990 . m. germano . turbulence : the filtering approach . , 1992 . m. germano , u. piomelli , p. moin , and w. h. cabot . a dynamic subgrid - scale eddy viscosity model . , 3:17601765 , 1991 . s. ghosal . an analysis of numerical errors in large - eddy simulations of turbulence . , 125:187206 , 1996 . s. ghosal and p. moin . the basic equations for the large eddy simulation of turbulent flows in complex geometry . , 1995 . a. gronskis , d. heitz , and e. mmin . inflow and initial conditions for direct numerical simulation based on adjoint data assimilation . , 242(6):480497 , 2013 . i. grooms and a. majda . efficient stochastic superparameterization for geophysical turbulence . , 110(12 ) , 2013 . d. haworth and s. pope . a generalized langevin model for turbulent flows . 29:387405 , 1986 . b. dubrulle j .- p laval and j.c . langevin models of turbulence : renormalization group , distant interaction algorithms or rapid distortion theory ? , 15(5):13271339 , 2006 . s. kadri - harouna , p. drian , p. has , and e. mmin . divergence - free wavelets and high order regularization . , 103(1):8099 , 2013 . s. kadri - harouna and v. perrier . effective construction of divergence - free wavelets on the square . , 2013 . g. karamanos and g. karniadakis . a spectral vanishing viscosity method for large - eddy simulations . , 163(1):2250 , 2000 . r. kraichnan . the structure of isotropic turbulence at very high reynolds numbers . , 5:477543 , 1959 . r. kraichnan . small - scale structure of a scalar field convected by turbulence . , 11:945963 , 1968 . r. kraichnan . convergents to turbulence functions . , 41:189217 , 1970 . h. kunita . . cambridge university press , 1990 . e. lamballais , v. fortun , and s. laizet . straightforward high - order numerical dissipation via the viscous term for direct and large eddy simulation . , 230:32703275 , 2011 . w. layton . approximating the larger eddies in fluid motion v : kinetic energy balance of scale similarity models . , 2000 . c. leith . atmospheric predictability and two - dimensional turbulence . , 28 , 1971 . c. leith . stochastic backscatter in a subgrid - scale model : plane shear mixing layer . , 2(3):15211530 , 1990 . d. lilly . on the application of the eddy viscosity concept in the inertial subrange of turbulence . technical report 123 , ncar , 1966 . d. lilly . a proposed modification of the germano subgrid - scale closure . , 3:27462757 , 1992 . j. macinnes and f. bracco . stochastic particles dispersion modelling and the tracer - particle limit . , 4(12):28092824 , 1992 . a. majda , i. timofeyev , and e. vanden eijnden . models for stochastic climate prediction . , 1999 . a. majda , i. timofeyev , and e. vanden eijnden . a systematic strategies for stochastic mode reduction in climate . , 60:17051722 , 2003 . p.j . mason and d.j . stochastic backscatter in large - eddy simulations of boundary layers . , 242:5178 , 1992 . e. mmin . fluid flow dynamics under location uncertainty . , 108(2):119146 , 2014 . c. meneveau and j. katz . scale - invariance and turbulence models for large - eddy simulation . , 32:132 , 2000 . minier s. chibbaro and s. pope . guidelines for the formulation of lagrangian stochastic models for particle simulations of single - phase and dispersed two - phase turbulent flows . , 26,1113303 , 2014 . a.s . monin and a.m. yaglom . , volume ii . mit press , 1975 . j. nocedal and s.j . springer series in operations research . sringer - verlag , new - york , 1999 . s. orszag . analytical theories of turbulence . , 41:363386 , 1970 . s. orszag . numerical simulation of the taylor - green vortex . in springer - verlag , editor , _ international symposium on computing methods in applied sciences and engineering _ , volume 2 , pages 5064 , 1974 . t. palmer and p. williams . theme issue stochastic physics and climate modelling. , 366(1875 ) , 2008 . r. pasquetti . spectral vanishing viscosity method for large - eddy simulation of turbulent flows . , 27(1 - 3):365375 , 2006 . v perrier , t. philipovitch , and c. basdevant , . wavelet spectra compared to fourier spectra , 36:15061519 , 1995 . u. piomelli , w. cabot , p. moin , and s. lee . subgrid - scale backscatter in turbulent and transitional flows . , 3(7):17661771 , 1991 . s. pope . lagrangian pdf methods for turbulent flows . , 26:23 - 63 , 1994 . s. pope . . cambridge university press , 2000 . l. prandtl . bericht uber untersuchungen zur ausgebildeten turbulenz . , 5:136139 , 1925 . g. da prato and j. zabczyk . . cambridge university press , 1992 . v. resseguier , e. mmin , and b. chapron . geophysical flows under location uncertainty , part i random transport and general models https://hal.inria.fr/hal-01391420 , 2016 . v. resseguier , e. mmin , and b. chapron . geophysical flows under location uncertainty , part ii quasi - geostrophy and efficient ensemble spreading https://hal.inria.fr/hal-01391476 [ https://hal.inria.fr/hal-01391476 ] , 2016 . v. resseguier , e. mmin , and b. chapron . geophysical flows under location uncertainty , part iii sqg and frontal dynamics under strong turbulence conditions https://hal.inria.fr/hal-01391484 , 2016 . p. sagaut . . springer - verlag , scientic computation series , 2005 . b. sawford . generalized random forcing in random - walk models of turbulent dispersion models . , 29:35823585 , 1986 . u. schumann . stochastic backscatter of turbulence energy and scalar variance by random subgrid - scale fluxes . , 451:293318 , 1995 . g. sehmel . particle deposition from turbulent air flow . , 75:17661781 , 1970 . g. shutts . a kinetic energy backscatter algorithm for use in ensemble prediction systems . , 612:30793012 , 2005 . j. smagorinsky . general circulation experiments with the primitive equation : i. the basic experiment . , 91:99165 , 1963 . e. tadmor . convergence of spectral methods for nonlinear conservation laws . , 26(1):3044 , 1989 . g. taylor and a. green . mechanisms of the production of small eddies from large ones . , 1938 . van rees , d.i . leonard , a.and pullin , and p. koumoutsakos . a comparison of vortex and pseudo - spectral methods for the simulation of periodic vortical flows at high reynolds numbers . , 230:27942805 , 2011 . we derive in this appendix the expression of the reynolds transport theorem for the drift plus noise decomposition : in a lagrangian stochastic picture , the infinitesimal displacement associated to a trajectory @xmath110 of a particle is noted : @xmath111 the random field involved in this equation is defined for all points of the fluid domain , @xmath112 through the kernel @xmath113 associated to the diffusion operator @xmath114 @xmath115 this operator is assumed to be of finite norm for any orthonormal basis of the associated hilbert space @xmath116 ) and to have a null boundary condition on the domain frontier : @xmath117 function @xmath118 denotes a @xmath119-dimentional brownian function ( see @xcite for more information on infinite dimensional wiener process ) and the resulting @xmath119-dimensional random field , @xmath120 , is a centered vectorial gaussian function correlated in space and uncorrelated in time with covariance tensor : @xmath121 we observe that those random fields have a ( mean ) bounded norm : @xmath122 , where the trace of the covariance tensor is given for any basis @xmath123 of @xmath116 as @xmath124 . in the following we will note the diagonal of the covariance tensor as : @xmath125 . tensor , @xmath8 , that will be referred to as the variance tensor @xmath9 is by definition a symmetric positive definite matrix at all spacial points , @xmath126 . this quantity corresponds to the time derivative of the so - called quadratic variation process : @xmath127 the notation @xmath128 stands for the quadratic cross - variation process of @xmath129 and @xmath130 . this central object of stochastic calculus , can be interpreted as the covariance along time of the increments of @xmath129 and @xmath130 . the quadratic variation process is briefly presented in appendix [ quadvar ] . the drift term , @xmath16 , of lagrangian expression ( [ particle_dx ] ) , represents the `` smooth '' differentiable part of the flow whereas the random part , @xmath131 figures the small - scale velocity component associated to a much thinner time - scale . this component is modeled as a delta - correlated random field in time as it represents a highly non regular process at the resolved time scale . in this work , we assume that this small - scale random component is incompressible and therefore associated to a divergence free diffusion tensor : @xmath132 . this assumption , which obviously does not prevent the drift component ( and therefore the whole field ) to be compressible , leads to much simpler modeling and remains realistic for the models considered in this study . let us consider now a spatially regular enough scalar function @xmath133 of compact support , transported by the stochastic flow ( [ particle_dx ] ) and that vanishes outside volume @xmath5 and on its boundary @xmath134 . as this function is assumed to be transported by the stochastic flow , it constitutes a stochastic process defined from its initial time value @xmath130 : @xmath135 we will assume that both functions have bounded spatial gradients . besides , the initial function @xmath136 vanishes outside the initial volume @xmath137 and on its boundary . let us point out that in this construction , function @xmath133 can not be a deterministic function . as a matter of fact , if it was the case , its differential would be given by a standard ito formula @xmath138 which here cancels as @xmath133 is transported by the flow . a separation between the slow deterministic terms and the fast brownian term would yield a specific noise term orthogonal to @xmath139 . this would boil down to the deterministic case , which is not the general goal followed here . as @xmath133 is a random function , the differential of @xmath140 involves the composition of two stochastic processes . its evaluation requires the use of a generalized ito formula usually referred in the literature to as the ito - wentzell formula ( see theorem 3.3.1 , * ? ? ? this extended ito formula incorporates in the same way as the classical ito formula a quadratic variation terms related to process @xmath141 but also co - variation terms between @xmath141 and the gradient of the random function @xmath142 . its expression is in our case given by @xmath143 it can be immediately checked that for a deterministic function , the standard ito formula is recovered since the co - variations terms between @xmath141 and @xmath144 cancel . it follows from [ dxi=0 ] that for a fixed grid point , function @xmath145 is solution of a stochastic differential equation of the form @xmath146 where the brownian term must compensate the brownian term of ( [ dxi=0 ] ) . the quadratic variation term involved in ( [ dxi=0 ] ) is given through ( appendix [ quadvar ] ) as @xmath147 and the covariation term reads @xmath148 in these expressions @xmath149 ( resp . @xmath150 ) designates the kernel associated to operator @xmath151 ( resp . @xmath152 ) . now , identifying first the brownian term and next the deterministic terms of equations ( [ dxi=0 ] ) and ( [ dxi_t ] ) , we infer @xmath153 and finally get @xmath154 this differential at a fixed point , @xmath126 , defines the equivalent in the deterministic case of the material derivative of a function transported by the flow . the differential of the integral over a material volume of the product @xmath155 is given by @xmath156 where the first line comes from @xmath157 and the second one from the integration by part formula of two ito processes . hence , from ( [ smd ] ) this differential is @xmath158 introducing @xmath159 the ( formal ) adjoint of the operator @xmath160 in the space @xmath161 with dirichlet boundary conditions , this expression can be written as @xmath162 with the complete expression of @xmath159 ( remarking that the second right - hand term of [ defl ] is self - adjoint ) and if @xmath163 , where @xmath164 stands for the characteristic function , we get the sought form of this differential : @xmath165{{\mathrm{d}}}{\boldsymbol{x}},\nonumber\ ] ] we recall here the notion of quadratic variation and co - variation , which play a central role in stochastic calculus . we will here restrict ourselves to standard ito processes . quadratic variation and co - variation correspond respectively to the variance and covariance of the process increments along time . the quadratic co - variation process denoted as @xmath166 , ( respectively the quadratic variation for @xmath167 ) is defined as the limit in probability over a partition @xmath168 of @xmath169 $ ] with @xmath170 , and a partition spacing @xmath171 , noted as @xmath172 and such that @xmath173 when @xmath174 : @xmath175 for brownian motion these covariations can be easily computed and are given by the following rules @xmath176 , @xmath177 , where @xmath178 is a deterministic function and @xmath179 a scalar brownian motion . first of all , it is important to emphasize that the distribution of the velocity anomaly , @xmath180 , with the eulerian velocity field @xmath181 is , in the general case , _ not _ gaussian . as a matter of fact , due to the multiplicative noise ( the diffusion tensor depends _ a priori _ on the flow ) and as the resolved dynamics is nonlinear , the ensemble mean of the flow velocity is in general not given by @xmath182 . hence , this construction does not correspond to a turbulence model with a gaussian closure hypothesis of the fourth - order correlation functions as considered in the millionschikov hypothesis @xcite . nor is it based on quasi - normal approximations such as the edqnm approaches @xcite , which opposite to the previous hypothesis ensures the positivity of the energy spectrum . in the proposed decomposition , the velocity is clearly a markovian stochastic process , but its distribution is not gaussian . note also that in our approach no isotropic assumption on the increments nor on the random component are considered to define the subgrid tensor . as already mentioned , the proposed stochastic representation can be interpreted as a temporal decomposition of the original navier - stokes equations . in the following sections we list several properties of this model . we begin first by a useful scaling relation of the variance tensor . [ section - scaling ] in the conditions of kolmogorov - richardson scaling , at scale @xmath183 , the velocity increments and the eddies turn - over time scale as @xmath184 and @xmath185 respectively , with @xmath186 denoting a constant energy dissipation rate across the inertial scales range . the turn - over time ratio for two different scales in this range , @xmath187 , exhibits a direct relation between a change of time scale and a change of spacial resolution . a coarsening in time yields thus a space dilation . besides , at scale @xmath58 , we have : @xmath188 at the smallest scale , this quantity corresponds to the variance tensor @xmath189 . we have thus : @xmath190 within a given cell , @xmath191 , at scale @xmath58 , the energy of the small - scale random field is given as : @xmath192 which from ( [ eq1 ] ) gives us : @xmath193 this relation provides us an expression of the scaling between the subgrid variance tensor at scale @xmath58 and the resolved velocity anomalies . it will serve us to impose a proper tuning of the subgrid term when the variance tensor is defined from the resolved velocity . considering the scaled coordinates @xmath194 , @xmath195 , with velocity @xmath196 , pressure @xmath197 and a variance tensor @xmath198 , we have : @xmath199 @xmath200 assuming , as in the previous section , that the characteristic value of the resolved variance tensor , @xmath201 , depends linearly on the variance tensor at the smallest scale , @xmath202 , the dimensionless form of the large - scale stochastic navier - stokes equations reads : @xmath203 where we introduced a dimensionless number @xmath204 which relates the typical kinetic energy at scale @xmath58 to a characteristic value of the velocity variance , @xmath205 , at the same level . in this expression we considered from ( [ scaling ] ) a characteristic value of the resolved variance tensor @xmath206 where we introduced a scale factor ratio , @xmath207 ( with k(1)=1 ) . as previously inferred , under the kolmogorov scaling hypothesis this ratio scales as @xmath208 . let us finally observe that for a typical variance tensor , @xmath202 , tending to zero , ( [ d - nsu ] ) tends to the original navier - stokes equations . it is the situation occurring when the variance tensor is negligible in front of the drift energy . in the next section we will see that the large - scale model proposed conserves the invariance properties of the original navier - stokes equations . let us consider in the following the classical symmetry groups of the incompressible navier - stokes equations in a periodic domain and in the absence of external forcing . space translation invariance is achieved if the whole velocity field @xmath209 is still a solution of the navier - stokes equations . in our setting we have @xmath210 separating the brownian component from the smooth term yields : @xmath211 , @xmath212 and hence @xmath213 , @xmath214 , @xmath215 , @xmath216 , @xmath217 , and translational invariance follows immediately . time - shift invariance is obtained recalling that brownian motion has itself a time - shift invariance property . this property allows us to write ( in law ) for @xmath218 : @xmath219 which leads straightforwardly to time - shift invariance . these two symmetry groups correspond to a constant rotation or to a reflection of the coordinates system . the transformed coordinates are @xmath220 and @xmath221 . note that @xmath222 , which yields @xmath223 and hence @xmath224 . the invariance of the transformed navier - stokes equations follows immediately : @xmath225 a model is said to be galilean invariant if it is invariant with respect to an inertial transformation of the representation frames : @xmath226 . the velocity in this translated non - accelerating system of reference reads @xmath227 with @xmath228 from those equations we obtain : @xmath229 as the pressure and viscous force are galilean invariant and since @xmath230 , this shows the system is galillean invariant . as with the original navier - stokes equation , the stochastic model is not invariant to a time reversal transformation : @xmath231 . this property is respected only for the euler equation . despite the time reversibility property of brownian motion this loss of symmetry is due to the even nature of the quadratic tensor . considering the scaled transformation : @xmath232 we get : @xmath233 the two last relations highlight a very interesting property of the modified navier - stokes equations . when the viscosity can be neglected in comparison to the variance tensor , the large - scale stochastic navier - stokes equations become scale - invariant for any values of the scale exponent . they share this property with the euler equations . when the friction term is considered the equations are invariant only for @xmath234 . to sum up , the large - scale stochastic representation of the navier - stokes equations have the remarkable feature of keeping all the symmetries of the original equations and to follow the same scale invariance property as the euler equation when neglecting the friction term . this last property , which inherits from the time scale invariance of brownian motion , is often not verified for large - scale eddy viscosity representations of the euler equations . using the condition @xmath235 , we can write : @xmath236 where @xmath237 is the multi - scale residual term . then , it is easy to see that : @xmath238\xrightarrow[\epsilon\to 0 ] { } \\ -\frac{1}{2}\langle \partial_{x_m } \partial_{x_n}[a_{mn}({\boldsymbol{w } } _ { { \boldsymbol \ell}}-\lambda { \boldsymbol{w } } _ { { \boldsymbol \ell ' } } ) ] , \bar{\boldsymbol{w } } _ { { \boldsymbol \ell}}{{\boldsymbol{\nabla}}}{^{\scriptscriptstyle t}}{\boldsymbol{w } } _ { { \boldsymbol \ell}}+{\boldsymbol{w } } _ { { \boldsymbol \ell'}}{{\boldsymbol{\nabla}}}{^{\scriptscriptstyle t}}\bar{\boldsymbol{w } } _ { { \boldsymbol \ell}}-\frac{1}{2}\sum_{i , j=1}^d\partial_{x_i } \partial_{x_j}[a_{ij}({\boldsymbol{w } } _ { { \boldsymbol \ell}}-\lambda { \boldsymbol{w } } _ { { \boldsymbol \ell'}})]\rangle,\end{gathered}\ ] ] and integration by part gives : @xmath239\biggr).\end{gathered}\ ] ] similarly , the derivative @xmath240 with respect to the constant @xmath241 is given by :

the support vector machine ( svm ) is one of the most successful learning methods for solving classification problems . despite its popularity , svm has a serious drawback , that is sensitivity to outliers in training samples . the penalty on misclassification is defined by a convex loss called the hinge loss , and the unboundedness of the convex loss causes the sensitivity to outliers . to deal with outliers , robust variants of svm have been proposed , such as the robust outlier detection algorithm and an svm with a bounded loss called the ramp loss . in this paper , we propose a robust variant of svm and investigate its robustness in terms of the breakdown point . the breakdown point is a robustness measure that is the largest amount of contamination such that the estimated classifier still gives information about the non - contaminated data . the main contribution of this paper is to show an exact evaluation of the breakdown point for the robust svm . for learning parameters such as the regularization parameter in our algorithm , we derive a simple formula that guarantees the robustness of the classifier . when the learning parameters are determined with a grid search using cross validation , our formula works to reduce the number of candidate search points . the robustness of the proposed method is confirmed in numerical experiments . we show that the statistical properties of the robust svm are well explained by a theoretical analysis of the breakdown point .

1409.0934

support vector machine ( svm ) is a highly developed classification method that is widely used in real - world data analysis @xcite . the most popular implementation is called @xmath0-svm , which uses the maximum margin criterion with a penalty for misclassification . the positive parameter @xmath0 tunes the balance between the maximum margin and penalty . as a result , the classification problem can be formulated as a convex quadratic problem based on training data . a separating hyper - plane for classification is obtained from the optimal solution of the problem . furthermore , complex non - linear classifiers are obtained by using the reproducing kernel hilbert space ( rkhs ) as a statistical model of the classifiers @xcite . there are many variants of svm for solving binary classification problems , such as @xmath1-svm , e@xmath1-svm , least square svm @xcite . moreover , the generalization ability of svm has been analyzed in many studies @xcite . in practical situations , however , svm has drawbacks . the remarkable feature of the svm is that the separating hyperplane is determined mainly from misclassified samples . thus , the most misclassified samples significantly affect the classifier , meaning that the standard svm is extremely fragile to the presence of outliers . in @xmath0-svm , the penalties of sample points are measured in terms of the hinge loss , which is a convex surrogate of the 0 - 1 loss for misclassification . the convexity of the hinge loss causes svm to be unstable in the presence of outliers , since the convex function is unbounded and puts an extremely large penalty on outliers . one way to remedy the instability is to replace the convex loss with a non - convex bounded loss to suppress outliers . loss clipping is a simple method to obtain a bounded loss from a convex loss @xcite . for example , clipping the hinge loss leads to the ramp loss @xcite . in mathematical statistics , robust statistical inference has been studied for a long time . a number of robust estimators have been proposed for many kinds of statistical problems @xcite . in mathematical analysis , one needs to quantify the influence of samples on estimators . here , the influence function , change of variance , and breakdown point are often used as measures of robustness . in machine learning literature , these measures are used to analyze the theoretical properties of svm and its robust variants . in @xcite , the robustness of a learning algorithm using a convex loss function was investigated on the basis of an influence function defined over an rkhs . when the influence function is uniformly bounded on the rkhs , the learning algorithm is regarded to be robust against outliers . it was proved that the quadratic loss function provides a robust learning algorithm for classification problems in this sense @xcite . from the standpoint of the breakdown point , however , convex loss functions do not provide robust estimators , as shown in ( * ? ? ? * chap . 5.16 ) . in @xcite , yu et al . showed a convex loss clipping that yields a non - convex loss function and proposed a convex relaxation of the resulting non - convex optimization problem to obtain a computationally efficient learning algorithm . they also studied the robustness of the learning algorithm using the clipped loss . in this paper , we provide a detailed analysis on the robustness of svms . in particular , we deal with a robust variant of kernel - based @xmath1-svm . the standard @xmath1-svm @xcite has a regularization parameter @xmath1 , and it is equivalent with @xmath0-svm ; i.e. , both methods provide the same classifier for the same training data , if the regularization parameters , @xmath1 and @xmath0 , are properly tuned . we also introduce a new robust variant called robust @xmath2-svm that has another learning parameter @xmath3 . the parameter @xmath4 denotes the ratio of samples to be removed from the training dataset as outliers . when the ratio of outliers in the training dataset is bounded above by @xmath4 , robust @xmath2-svm is expected to provide a robust classifier . robust @xmath2-svm is closely related to the robust outlier detection ( rod ) algorithm @xcite . indeed , rod is to robust @xmath2-svm what @xmath0-svm is to @xmath1-svm @xcite . our main contribution is to derive the _ exact _ finite - sample breakdown point of robust @xmath2-svm . the finite - sample breakdown point indicates the largest amount of contamination such that the estimator still gives information about the non - contaminated data ( * ? ? ? * chap.3.2 ) . we show that the finite - sample breakdown point of robust @xmath2-svm is equal to @xmath4 , if @xmath1 and @xmath4 satisfy simple inequalities . conversely , we prove that the finite - sample breakdown point is strictly less than @xmath4 , if these key inequalities are violated . the theoretical analysis partly depends on the boundedness of the kernel function used in the statistical model . as a result , one can specify the region of the learning parameters @xmath2 such that robust @xmath2-svm has the desired robustness property . this property will be of great help to reduce the number of candidate learning parameters @xmath2 , when the grid search of learning parameters is conducted with cross validation . some of previous studies are related to ours . in particular , the breakdown point was used to assess the robustness of kernel - based estimators in @xcite . in that paper , the influence of a single outlier is considered for a general class of robust estimators . in contrast , we focus on a variant of svm and provide a detailed analysis of the robustness property based on the breakdown point . in our analysis , an arbitrary number of outliers is taken into account . the paper is organized as follows . in section [ sec : brief_introduction ] , we introduce the problem setup and briefly review the topic of learning algorithms using the standard @xmath1-svm . section [ sec : robust_nu_svm ] is devoted to the robust variant of @xmath1-svm . we show that the dual representation of robust @xmath2-svm has an intuitive interpretation , that is of great help to compute the breakdown point . an optimization algorithm is also presented . in section [ sec : breakdown_point ] , we introduce a finite - sample breakdown point as a measure of robustness . then , we evaluate the breakdown point of robust @xmath2-svm . in section [ sec : asymptotic_properties ] , we investigate the statistical asymptotic properties of the proposed method on the basis of order statistics . section [ sec : numerical_experiments ] examines the generalization performance of robust @xmath2-svm via numerical experiments . the conclusion is in section [ sec : concluding_remarks ] . detailed proofs of the theoretical results are presented in the appendix . let us summarize the notations used throughout this paper . let @xmath5 be the set of natural numbers , and let @xmath6 $ ] for @xmath7 denote a finite set of @xmath5 defined as @xmath8 . the set of all real numbers is denoted as @xmath9 . the function @xmath10_+$ ] is defined as @xmath11 for @xmath12 . for a finite set @xmath13 , the size of @xmath13 is expressed as @xmath14 . for a reproducing kernel hilbert space ( rkhs ) @xmath15 , the norm on @xmath15 is denoted as @xmath16 . see @xcite for a description of rkhs . let @xmath17 ( resp . @xmath18 ) be an @xmath19-dimensional vector of all ones ( resp . all zeros ) . let us introduce the classification problem with an input space @xmath20 and binary output labels @xmath21 . given i.i.d . training samples @xmath22\}\subset\mathcal{x}\times\{+1,-1\}$ ] drawn from a probability distribution over @xmath23 , a learning algorithm produces a decision function @xmath24 such that its sign provides a prediction of output labels for input points over test samples . the decision function @xmath25 predicts the correct label on the sample @xmath26 if and only if the inequality @xmath27 holds . the product @xmath28 is called the margin of the sample @xmath26 for the decision function @xmath29 @xcite . to make an accurate decision function , the margins on the training dataset should take large positive values . in kernel - based @xmath1-svm @xcite , an rkhs @xmath15 endowed with a kernel function @xmath30 is used to estimate the decision function @xmath31 , where @xmath32 and @xmath33 . the misclassification penalty is measured by the hinge loss . more precisely , @xmath1-svm produces a decision function @xmath31 as the optimal solution of the convex problem , @xmath34_+\\ \displaystyle \st\ f\in\mathcal{h},\ b,\rho\in\rbb , \end{array}\end{aligned}\ ] ] where @xmath35_+$ ] is the hinge loss of the margin with the threshold @xmath36 . the second term @xmath37 is the penalty for the threshold parameter @xmath36 . the parameter @xmath1 in the interval @xmath38 is the regularization parameter . usually , the range of @xmath1 that yields a meaningful classifier is narrower than the interval @xmath38 , as shown in @xcite . the first term in is a regularization term to avoid overfitting to the training data . a large positive margin is preferable for each training data . the representer theorem @xcite indicates that the optimal decision function of is of the form , @xmath39 for @xmath40 . the input point @xmath41 with a non - zero coefficient @xmath42 is called a support vector . the regularization parameter @xmath1 provides a lower bound on the fraction of support vectors . thanks to the representer theorem , even when @xmath15 is an infinite dimensional space , the above optimization problem can be reduced to a finite dimensional quadratic convex problem . this is the great advantage of using rkhs for non - parametric statistical inference @xcite . as pointed out in @xcite , @xmath1-svm is closely related to a financial risk measure called conditional value at risk ( cvar ) @xcite . roughly speaking , the cvar of samples @xmath43 at level @xmath44 such that @xmath45 is defined as the average of its @xmath1-tail , i.e. , @xmath46 , where @xmath47 is a permutation on @xmath6 $ ] such that @xmath48 holds . in the literature , @xmath49 is defined as the negative margin @xmath50 . for a regularization parameter @xmath1 satisfying @xmath45 and a fixed decision function @xmath31 , the objective function in is expressed as @xmath51_+ \\ & = \frac{1}{2}\|f\|_{\mathcal{h}}^2+\nu\cdot\frac{1}{\nu{m}}\sum_{i=1}^{\nu{m}}r_{\sigma(i)}. \end{aligned}\ ] ] details are presented in theorem 10 of @xcite . hence , @xmath1-svm yields a decision function that minimizes the sum of the regularization term and the cvar of the negative margins at level @xmath1 . in @xmath0-svm , the decision function is obtained by solving @xmath52_+\\ \displaystyle \st\ f\in\mathcal{h},\ b\in\rbb . \end{array}\end{aligned}\ ] ] note that the threshold in the hinge loss is fixed to one in @xmath0-svm , whereas @xmath1-svm determines the threshold with the optimal solution @xmath36 . a positive regularization parameter @xmath53 is used instead of @xmath1 . for each training data , @xmath1-svm and @xmath0-svm can be made to provide the same decision function by appropriately tuning @xmath1 and @xmath0 . in this paper , we focus on @xmath1-svm and its robust variants rather than @xmath0-svm . the parameter @xmath1 has the explicit meaning shown above , and this interpretation will be significant when we derive the robustness property of our method . in the robust @xmath0-svm proposed in @xcite , the hinge loss @xmath54_{+}$ ] in is replaced with the so - called ramp loss @xmath55_{+}\}$ ] . by truncating the hinge loss , the influence of outliers is suppressed , and the estimated classifier is expected to be robust against outliers included in the training data . here , we propose a robust @xmath2-svm that is a robust variant of @xmath1-svm . to remove the influence of outliers , we introduce the outlier indicator , @xmath56 $ ] , for each training sample , where @xmath57 is intended to indicate that the sample @xmath58 is an outlier . the same idea is used in @xcite . assume that the ratio of outliers is less than or equal to @xmath4 , and define the finite set @xmath59 as @xmath60 for @xmath1 and @xmath4 such that @xmath61 , robust @xmath2-svm is formalized using rkhs @xmath15 as @xmath62_+ , \\ \displaystyle\ \st\ \ f\in\mathcal{h},\ \ { \eta}=(\eta_1,\ldots,\eta_m)^t\in{}e_\mu,\ \ b,\rho\in\rbb . \end{array}\end{aligned}\ ] ] the optimal solution , @xmath32 and @xmath33 , provides the decision function @xmath31 for classification . influence from samples with large negative margins is removed by setting @xmath63 to zero . throughout the paper , we will assume that @xmath64 and @xmath65 are natural numbers to avoid technical difficulties . robust @xmath2-svm is closely related to the robust outlier detection ( rod ) algorithm @xcite . about modified algorithms of rod and robust @xmath2-svm , the equivalence is shown in @xcite . in rod , the classifier is given by the optimal solution of @xmath66_{+ } , \\ \displaystyle \st\ f\in\mathcal{h},\ \ b\in\rbb,\\ \phantom{\st}\ { \eta}=(\eta_1,\ldots,\eta_m)^t\in{}[0,1]^m,\,\ \sum_{i=1}^{m}\eta_i\geq{}m(1-\mu ) , \end{array}\end{aligned}\ ] ] where @xmath67 is a regularization parameter . in the original rod , the linear kernel is used . to obtain the classifier , the rod algorithm solves a semidefinite relaxation of the above problem . furthermore , robust @xmath2-svm is related to cvar at levels @xmath1 and @xmath4 . indeed , for the parameters , @xmath1 and @xmath4 , and a fixed decision function @xmath31 , the objective function in is represented as @xmath68_+ \nonumber\\ & = \min_{\rho\in\rbb}\ \frac{1}{2}\|f\|_{\mathcal{h}}^2 -\nu\rho + \frac{1}{m}\sum_{i=1}^{m}\big[\rho+r_i\big]_+ -\max_{{\eta}\in{e_\mu}}\frac{1}{m}\sum_{i=1}^m(1-\eta_i)r_i \label{eqn : dc - representation } \\ & = \frac{1}{2}\|f\|_{\mathcal{h}}^2 + ( \nu-\mu)\cdot\frac{1}{(\nu-\mu)m}\sum_{i=\mu{m}+1}^{\nu{m}}\!\!r_{\sigma(i ) } , \label{eqn : diff_cvar}\end{aligned}\ ] ] where @xmath69 is the negative margin and @xmath70 is its sort in the descending order defined in section [ sec : brief_introduction ] . the second term in is the average of the negative margins included in the middle interval presented in figure [ fig : margin_hist ] , and it is expressed by the difference of cvars at levels @xmath1 and @xmath4 . the learning algorithm based on this interpretation is proposed in @xcite under the name cvar-@xmath71-svm . the two methods can be shown to be equivalent by setting @xmath72 and @xmath73 . in this paper , the learning algorithm based on is referred to as robust @xmath2-svm to emphasize that it is a robust variant of @xmath1-svm . $ ] for a fixed decision function @xmath31 . ] the representer theorem ensures that the optimal decision function of is represented by @xmath74 when the kernel function of the rkhs @xmath15 is given by @xmath75 . as in the case of the standard @xmath1-svm , the number of support vectors , i.e. , the input points @xmath76 such that @xmath77 , is bounded below by @xmath78 . in addition , the kkt condition of leads to the fact that any support vector @xmath76 satisfies @xmath79 . it is hard to obtain a global optimal solution of , since the objective function is non - convex . as shown in @xcite , the objective function in is expressed as a difference of convex functions ( dc ) by using a cvar representation . hence , the dc algorithm @xcite and convex - concave programming ( cccp ) @xcite are available to efficiently obtain a stationary point of . the same approach is taken by robust @xmath0-svm using the ramp loss @xcite . in algorithm [ alg : dc_alg_robust_numusvm ] , the dc algorithm for robust @xmath2-svm based on the expression is presented . the derivation of the dc algorithm is presented in appendix [ appendix : dca ] . algorithm [ alg : dc_alg_robust_numusvm ] is guaranteed to converge in a finite number of iterations . in the dc algorithm , a monotone decrease of the objective value is generally guaranteed , and in algorithm [ alg : dc_alg_robust_numusvm ] , the objective value in each iteration is determined by @xmath80 , which can take only a finite number of distinct values . a stationary point is obtained when the objective value is unchanged . the above argument is based on a convergence analysis of robust @xmath0-svm using the ramp loss @xcite . in addition , an argument based on polyhedral dc programming shows that the algorithm converges after a finite number of iterations @xcite . one can use another stopping rule such that the algorithm terminates when the same @xmath80 is obtained in two consecutive iterations . if the cyclic phenomenon of @xmath81 is prohibited in some way , convergence in a finite number of iterations is guaranteed . gram matrix @xmath82 defined as @xmath83 $ ] , and training labels @xmath84 . the matrix @xmath85 is defined as @xmath86 . let @xmath31 be an initial decision function . compute the sort @xmath87 of the negative margin @xmath50 , and set @xmath88 for @xmath89 $ ] . let @xmath81 be @xmath90 . set @xmath91 and @xmath92 . compute the optimal solution @xmath93 of the problem @xmath94 set @xmath95 , where @xmath96 denotes component - wise multiplication of two vectors . compute @xmath36 and @xmath97 using @xmath98 , where @xmath99 . the decision function @xmath100 . the partial dual problem of with a fixed outlier indicator @xmath101 has an intuitive geometric picture . some variants of @xmath1-svm can be geometrically interpreted on the basis of the dual form @xcite . substituting into the objective function in , we obtain the lagrangian of problem with a fixed @xmath102 as @xmath103 where non - negative slack variables @xmath104 $ ] are introduced to represent the hinge loss . here , the parameters @xmath105 and @xmath106 for @xmath89 $ ] are non - negative lagrange multipliers . for a fixed @xmath80 , the lagrangian is convex in the parameters @xmath107 , and @xmath108 and concave in @xmath109 and @xmath110 . hence , the min - max theorem ( * ? ? ? * proposition 6.4.3 ) yields @xmath111 let us give a geometric interpretation of the above expression . for the training data @xmath22\}$ ] , the convex sets , @xmath112 $ ] and @xmath113 $ ] , are defined as the reduced convex hulls of data points for each label , i.e. , @xmath114\\ & = \bigg\ { \sum_{i : y_i=\pm1}\gamma_i'k(\cdot , x_i)\in\mathcal{h}:\!\!\ ! \sum_{i : y_i=\pm1}\gamma_i'=1,\ 0\leq \gamma_i'\leq \frac{2\eta_i}{(\nu-\mu)m } \ \text{for $ i$ such that $ y_i=\pm1 $ } \bigg\}. \end{aligned}\ ] ] the coefficients @xmath115 $ ] in @xmath116 $ ] are bounded above by a non - negative real number that is usually less than one . hence , the reduced convex hull is a subset of the convex hull of the data points in the rkhs @xmath15 . each reduced convex hull is regarded as the domain of the input samples of each label . accordingly , let @xmath117 $ ] be the minkowski difference of two subsets , @xmath118= \mathcal{u}_{\eta}^{+}[\nu,\mu;d ] \ominus \mathcal{u}_{\eta}^{-}[\nu,\mu;d ] , \end{aligned}\ ] ] where @xmath119 of subsets @xmath13 and @xmath120 denotes @xmath121 . eventually , for each @xmath80 , the optimal value in the above is represented by @xmath122\right\}. \end{aligned}\ ] ] hence , the optimal value of is @xmath123 , where @xmath124}\|f\|_{\mathcal{h}}^2 . \end{aligned}\ ] ] therefore , the dual form of robust @xmath2-svm is expressed as the maximization of the minimum distance between two reduced convex hulls , @xmath112 $ ] and @xmath113 $ ] . the estimated decision function in robust @xmath2-svm is provided by the optimal solution of up to a scaling factor depending on @xmath125 . moreover , the optimal value is proportional to the squared rkhs norm of the function @xmath126 in the decision function @xmath31 . let us describe how to evaluate the robustness of learning algorithms . there are a number of robustness measures for evaluating the stability of estimators . for example , the influence function evaluates the infinitesimal bias of the estimator caused by a few outliers included in the training samples . the gross error sensitivity is the worst - case infinitesimal bias defined with the influence function @xcite . in this paper , we use the _ finite - sample breakdown point _ , and it will be referred to as the breakdown point for short . the breakdown point quantifies the degree of impact that the outliers have on the estimators when the contamination ratio is not necessarily infinitesimal @xcite . in this section , we present an exact evaluation of the breakdown point of robust @xmath2-svm . the breakdown point indicates the largest amount of contamination such that the estimator still gives information about the non - contaminated data ( * ? ? ? * chap.3.2 ) . more precisely , for an estimator @xmath127 based on a dataset @xmath128 of size @xmath19 that takes a value in a normed space , the finite - sample breakdown point is defined as @xmath129 where @xmath130 is the family of datasets of size @xmath19 including at least @xmath131 elements in common with the non - contaminated dataset @xmath128 , i.e. , @xmath132\}\subset\mathcal{x}\times\{+1,-1\}\,:\,|d'\cap{d}|\geq{m-\kappa}\,\big\}. \end{aligned}\ ] ] for simplicity , the dependency of @xmath133 on the data set @xmath128 is dropped . the condition of the breakdown point @xmath134 can be rephrased as @xmath135 where @xmath136 is the norm on the normed space . in most cases of interest , @xmath134 does not depend on the dataset @xmath128 . for example , the breakdown point of the one - dimensional median estimator is @xmath137 . to start with , let us derive a lower bound of the breakdown point for the optimal value of problem that is expressed as @xmath138 up to a constant factor . as shown in section [ subsec : dual_problem ] , the boundedness of @xmath138 is equivalent to the boundedness of the rkhs norm of @xmath32 in the estimated decision function @xmath31 . given a labeled dataset @xmath22\}$ ] , let us define the label ratio @xmath139 as @xmath140 [ theorem : breakdown - point - optvalue ] let @xmath128 be a labeled dataset of size @xmath19 with a positive label ratio @xmath139 . for the parameters @xmath141 such that @xmath142 and @xmath143 , we assume @xmath144 . then , the following two conditions are equivalent . ( i ) : : the inequality @xmath145 holds . ( ii ) : : uniform boundedness , @xmath146 holds , where @xmath147 is the family of contaminated datasets defined from @xmath128 . the proof is given in appendix [ appendix : proof_breakdown_obj ] . the inequality @xmath144 is a requisite condition . if this inequality is violated , the majority of , say , positive labeled samples in the non - contaminated training dataset can be replaced with outliers . in such a situation , the statistical features in the original dataset will not be retained . indeed , if @xmath148 holds , @xmath149 is unbounded over @xmath150 regardless of @xmath1 . since it is proved by a rigorous description of the above intuitive interpretation , the proof is omitted . theorem [ theorem : breakdown - point - optvalue ] indicates that the breakdown point of the rkhs element in the estimated decision function is greater than or equal to @xmath4 , if @xmath4 and @xmath1 satisfy inequality . conversely , if the inequality @xmath151 is violated , the breakdown point of robust @xmath2-svm does not reach @xmath4 , even though @xmath65 samples are removed from the training data . in addition , the inequality indicates the trade - off between the ratio of outliers @xmath4 and the ratio of support vectors @xmath125 . this result is reasonable . the number of support vectors corresponds to the dimension of the statistical model . when the ratio of outliers is large , a simple statistical model should be used to obtain robust estimators . if there is no outlier in training data , i.e. , @xmath152 , inequality reduces to @xmath153 . for the standard @xmath1-svm , this is a necessary and sufficient condition for the optimization problem to be bounded @xcite . when the contamination ratio in a training dataset is greater than @xmath4 , the estimated decision function is not necessarily bounded . [ theorem : breakdown - point_upper_bound ] suppose that @xmath1 and @xmath4 are rational numbers such that @xmath154 and @xmath155 . then , there exists a dataset @xmath128 of size @xmath19 with the label ratio @xmath139 such that @xmath144 and @xmath156 hold , where @xmath157 is defined from @xmath128 . the proof is given in appendix [ appendix : upperbound_breakdown_opt ] . theorems [ theorem : breakdown - point - optvalue ] and [ theorem : breakdown - point_upper_bound ] lead to the fact that the breakdown point of the function part @xmath32 in the estimated decision function @xmath158 is exactly equal to @xmath159 , when the learning parameters of the robust @xmath2-svm satisfy @xmath144 and @xmath151 . otherwise , the breakdown point of @xmath160 is strictly less than @xmath4 . we show the robustness of the bias term @xmath97 . let @xmath161 be the estimated bias parameter obtained by robust @xmath2-svm from the training dataset @xmath128 . we will derive a lower bound of the breakdown point of the bias term . then , we will show that the breakdown point of robust @xmath2-svm with a bounded kernel is given by a simple formula . [ theorem : bounded_bias ] let @xmath128 be an arbitrary dataset of size @xmath19 with a positive label ratio @xmath139 . suppose that @xmath1 and @xmath4 satisfy @xmath61 , @xmath143 , and @xmath144 . for a non - negative integer @xmath162 , we assume @xmath163 then , uniform boundedness @xmath164 holds , where @xmath165 is defined from @xmath128 . the proof is given in appendix [ appendix : proof_breakdown_bias ] . note that the inequality is a sufficient condition of inequality . theorem [ theorem : bounded_bias ] guarantees that the breakdown point of the estimated decision function @xmath166 is not less than @xmath167 when holds . when the kernel function is bounded , the boundedness of the function part @xmath32 in the decision function @xmath166 almost guarantees the boundedness of the bias term @xmath97 . [ theorem : bounded_kernel_bounded_bias ] let @xmath128 be an arbitrary dataset of size @xmath19 with a positive label ratio @xmath139 . for the parameters @xmath141 such that @xmath61 and @xmath143 , suppose that @xmath144 and @xmath168 hold . in addition , assume that the kernel function @xmath75 of the rkhs @xmath15 is bounded , i.e. , @xmath169 . then , uniform boundedness , @xmath170 holds , where @xmath147 is defined from @xmath128 . the proof is given in appendix [ appendix : proof_bounded_kernel_breakdown_bias ] . compared with theorem [ theorem : bounded_bias ] in which arbitrary kernel functions are treated , theorem [ theorem : bounded_kernel_bounded_bias ] ensures that a tighter lower bound of the breakdown point is obtained for bounded kernels . the above result agrees with those of other studies . the authors of @xcite proved that bounded kernels produce robust estimators for regression problems in the sense of bounded response , i.e. , robustness against a single outlier . [ cols="^,^ " , ] figure [ fig : estimator_and_testerror ] shows the results of the numerical experiments . the maximum norm of the estimated decision function is plotted for the parameter @xmath171 on the same axis as fig . [ fig : breakdown - ponit ] . the top ( bottom ) panels show the results for a gaussian ( linear ) kernel . the left and middle columns show the maximum norm of @xmath160 and @xmath97 , respectively . the maximum test errors are presented in the right column . in all panels , the red points denote the top 50 percent of values , and the asterisk ( @xmath172 ) is the point that violates the inequality @xmath151 . in this example , the numerical results agree with the theoretical analysis in section [ sec : breakdown_point ] ; i.e. , the norm becomes large when the inequality @xmath173 is violated . accordingly , the test error gets close to @xmath174no information for classification . even when the unbounded linear kernel is used , robustness is confirmed for the parameters in the left lower region in the right panel of fig . [ fig : breakdown - ponit ] . in the bottom right panel , the test error gets large when the inequality @xmath151 holds . this result comes from the problem setup . even with non - contaminated data , the test error of the standard @xmath1-svm is approximately @xmath174 , because the linear kernel works poorly for spiral data . thus , the worst - case test error can go beyond @xmath174 . for the parameter at which is violated , the test error is always close to @xmath174 . thus , a learning method with such parameters does not provide any useful information for classification . ccc + ) is the point that violates the inequality @xmath173 . , title="fig : " ] & ) is the point that violates the inequality @xmath173 . , title="fig : " ] & ) is the point that violates the inequality @xmath173 . , title="fig : " ] + plot of @xmath175 & plot of @xmath176 & plot of test error + + + ) is the point that violates the inequality @xmath173 . , title="fig : " ] & ) is the point that violates the inequality @xmath173 . , title="fig : " ] & ) is the point that violates the inequality @xmath173 . , title="fig : " ] + plot of @xmath175 & plot of @xmath176 & plot of test error + we compared the generalization ability of the robust @xmath2-svm with existing classifiers such as standard @xmath1-svm and robust @xmath0-svm using the ramp loss . the datasets are presented in table [ table : sim_result ] . all the datasets are provided in the mlbench and kernlab libraries of the r language @xcite . in all the datasets , the number of positive samples is less than or equal to that of negative samples . before running the learning algorithms , we standardized each input variable with mean zero and standard deviation one . we randomly split the dataset into training and test sets . to evaluate the robustness , the training data was contaminated by outliers . more precisely , we randomly chose positive labeled samples in the training data and changed their labels to negative ; i.e. , we added outliers by flipping the labels . after that , robust @xmath2-svm , robust @xmath0-svm using the ramp loss , and the standard @xmath1-svm were used to obtain classifiers from the contaminated training dataset . the prediction accuracy of each classifier was then evaluated over test data that had no outliers . linear and gaussian kernels were employed for each learning algorithm . the learning parameters , such as @xmath177 , and @xmath0 , were determined by conducting a grid search based on five - fold cross validation over the training data . for robust @xmath2-svm , the parameter @xmath178 was selected from the region @xmath179 in . for standard @xmath1-svm , the candidate of the regularization parameter @xmath1 was selected from the interval @xmath180 , where @xmath181 is the label ratio of the contaminated training data . for robust @xmath0-svm , the regularization parameter @xmath0 was selected from the interval @xmath182 $ ] . in the grid search of the parameters , 24 or 25 candidates were examined for each learning method . thus , we needed to solve convex or non - convex optimization problems more than @xmath183 times in order to obtain a classifier . the above process was repeated 30 times , and the average test error was calculated . the results are presented in table [ table : sim_result ] . for non - contaminated training data , robust @xmath2-svm and robust @xmath0-svm were comparable to the standard @xmath1-svm . when the outlier ratio is high , we can conclude that robust @xmath2-svm and robust @xmath0-svm tend to work better than the standard @xmath1-svm . in this experiment , the kernel function does not affect the relative prediction performance of these learning methods . in large datasets such as spam and satellite , robust @xmath2-svm tends to outperform robust @xmath0-svm . when learning parameters , such as @xmath141 , and @xmath0 , are appropriately chosen by using a large dataset , learning algorithms with plural learning parameters clearly work better than those with a single learning parameter . in addition , in robust @xmath0-svm , there is a difficulty in choosing the regularization parameter . indeed , the parameter @xmath0 does not have a clear meaning , and thus , it is not straightforward to determine the candidates of @xmath0 in the grid search optimization . in contrast , the parameter @xmath1 in @xmath1-svm and its robust variant has a clear meaning , i.e. , a lower bound of the ratio of support vectors and an upper bound of the margin error on the training data @xcite . such clear meaning is of great help to choose candidate points of regularization parameters . we conducted another experiment in which the learning parameters @xmath141 and @xmath0 were determined using only one validation set , i.e. , non - cross validation ( the details are not presented here ) . the dataset was split into training , validation and test sets . the learning parameters , @xmath184 , and @xmath0 , that minimized the prediction error on the validation set were selected . this method greatly reduced the computational cost of the cross validation . however , robust @xmath2-svm did not necessarily produce a better classifier compared with the other methods . since robust @xmath2-svm has two learning parameters , we need to carefully select them using cross validations rather than simple validations in order to achieve high prediction accuracy . rccccccc + & & & + outlier & & & @xmath1-svm & & & & @xmath1-svm + 0%&.258(.032)&.270(.038 ) & * .256(.051 ) & & * .179(.038 ) & .188(.043 ) & .181(.039 ) + 5% & * .256(.039)&.273(.047)&.258(.046)&&.225(.042 ) & .229(.051 ) & * .224(.061 ) + 10% & * .297(.060)&.306(.067)&.314(.060)&&.249(.059 ) & * * .230(.046 ) & .259(.062 ) + 15% & * .329(.061)&.339(.064)&.345(.062)&&.280(.053 ) & * .280(.050 ) & .294(.064 ) + + + & & & + outlier & & & @xmath1-svm & & & & @xmath1-svm + 0%&.033(.010 ) & .035(.008 ) & * .033(.006)&&*.032(.008)&.035(.012 ) & .033(.010 ) + 5%&.034(.009 ) & * .034(.010 ) & .043(.015)&&*.032(.005)&.033(.007 ) & .033(.006 ) + 10%&.055(.015 ) & * .051(.026 ) & .076(.036 ) & & * * .035(.008)&.043(.025 ) & .038(.008 ) + 15%&.136(.058 ) & * .120(.050 ) & .148(.058)&&.160(.083 ) & * .145(.070 ) & .150(.110 ) + + + & & & + outlier & & & @xmath1-svm & & & & @xmath1-svm + 0%&.237(.018 ) & * .232(.014 ) & .246(.018)&&*.238(.021 ) & .240(.019 ) & .243(.022 ) + 5%&.239(.019 ) & * .237(.016 ) & .269(.036)&&*.264(.025 ) & .267(.024 ) & .273(.024 ) + 10%&**.280(.046 ) & .299(.042 ) & .330(.030)&&.302(.039 ) & * .293(.036 ) & .315(.038 ) + 15%&**.338(.042 ) & .349(.030 ) & .351(.026)&&*.344(.028 ) & .344(.031 ) & .353(.016 ) + + + & & & + outlier & & & @xmath1-svm & & & & @xmath1-svm + 0%&.083(.005 ) & .088(.006 ) & * .083(.005 ) & & .081(.005 ) & .086(.006 ) & * .081(.006 ) + 5% & * * .094(.008 ) & .104(.013 ) & .109(.010 ) & & .095(.008 ) & .097(.009 ) & * .095(.008 ) + 10% & * * .129(.022 ) & .152(.020 ) & .166(.067 ) & & * .129(.015 ) & .133(.017 ) & .141(.030 ) + 15% & * * .201(.029 ) & .240(.030 ) & .256(.091 ) & & * * .206(.018 ) & .223(.030 ) & .240(.055 ) + + + & & & + outlier & & & @xmath1-svm & & & & @xmath1-svm + 0%&.097(.004 ) & .096(.003 ) & * * .094(.003 ) & & .069(.031 ) & .067(.004 ) & * * .063(.004 ) + 5%&.101(.003 ) & * .100(.005 ) & .100(.004 ) & & * .072(.015 ) & .078(.007 ) & .078(.043 ) + 10% & * * .148(.020 ) & .161(.026 ) & .161(.019 ) & & * .117(.034 ) & .126(.040 ) & .137(.027 ) + we presented robust @xmath2-svm and studied its statistical properties . the robustness property was analyzed by computing the exact breakdown point . as a result , we obtained inequalities for the learning parameters @xmath1 and @xmath4 that guarantee the robustness of the learning algorithm . the statistical theory of the l - estimator was then used to investigate the asymptotic behavior of the classifier . numerical experiments showed that the inequalities are critical to obtaining a robust classifier . the prediction accuracy of the proposed method was numerically compared with those of other methods , and it was found that the proposed method with carefully chosen learning parameters delivers more robust classifiers than those of other methods such as standard @xmath1-svm and robust @xmath0-svm using the ramp loss . in the future , we will explore the robustness properties of more general learning methods . another important issue is to develop efficient optimization algorithms . although the dc algorithm @xcite and convex relaxation @xcite are promising methods , more scalable algorithms will be required to deal with massive datasets that are often contaminated by outliers . according to , the objective function of the robust @xmath2-svm is expressed as @xmath185 using the convex functions @xmath186 and @xmath187 defined as @xmath188,\\ \psi_1({\alpha},b)&=\max_{{\eta}\in{e_\mu}}\frac{1}{m}\sum_{i=1}^{m}(1-\eta_i)r_i , \end{aligned}\ ] ] where @xmath49 is the negative margin @xmath189 and @xmath82 is the gram matrix defined by @xmath190 $ ] . let @xmath191 be the solution obtained after @xmath192 iterations of the dc algorithm . then , the solution is updated to the optimal solution of @xmath193 where @xmath194 with @xmath195 is an element of the subgradient of @xmath187 at @xmath196 . the subgradient of @xmath187 is given as @xmath197 where @xmath198 denotes the convex hull of the set @xmath199 . as shown in algorithm [ alg : dc_alg_robust_numusvm ] , a parameter @xmath81 that meets the condition in the above subgradient is obtained from the sort of the negative margin of the decision function defined from @xmath196 . the dual problem of is presented in , up to a constant term that is independent of the optimization parameter . the proof is decomposed into two lemmas . lemma [ lemma : breakdown - point - er - svm ] shows that condition ( i ) is sufficient for condition ( ii ) , and lemma [ lemma : unbounded - breakdown - point - er - svm-1 ] shows that condition ( ii ) does not hold if inequality is violated . for the dataset @xmath22\}$ ] , let @xmath200 and @xmath201 be the index sets defined as @xmath202 . when the parameter @xmath4 is equal to zero , the theorem holds according to the argument on the standard @xmath1-svm @xcite . below , we assume @xmath203 . [ proof of lemma [ lemma : breakdown - point - er - svm ] ] we show that @xmath204 $ ] is not empty for any @xmath150 . for a parameter @xmath4 such that @xmath144 , let @xmath205 be a positive constant satisfying @xmath206 . then , is expressed as @xmath207 . for a contaminated dataset @xmath208\}\in\mathcal{d}_{\mu{m}}$ ] , let us define @xmath209 as an index set such that @xmath210 for @xmath211 is replaced with @xmath212 as an outlier . in the same way , @xmath213 is defined for negative samples in @xmath128 . therefore , for any index @xmath214 in @xmath215 or @xmath216 , we have @xmath217 . the assumptions of the theorem ensure @xmath218 . from @xmath219 , we obtain @xmath220 given @xmath221 , the sets @xmath222 $ ] and @xmath223 $ ] are defined by @xmath224\\ & = \bigg\ { \sum_{i : y_i'=\pm1}\gamma_i'k(\cdot , x_i')\in\mathcal{h}\,:\,\sum_{i : y_i'=\pm1}\!\gamma_i'=1,\ 0\leq \gamma_i'\leq \frac{2\eta_i}{(\nu-\mu)m } , \ \ \text{$\gamma_i'=0 $ for $ i\not\in{}i_{\pm}\setminus\widetilde{i}_{\pm}$ } \bigg\}. \end{aligned}\ ] ] note that @xmath225\subset\mathcal{u}_{\eta}^{\pm}[\nu,\mu;d']$ ] holds because of the additional constraint , @xmath226 for @xmath227 . in addition , we have @xmath225 \subset \mathrm{conv}\{k(\cdot , x_i)\,:\,i\in{}i_{\pm}\}$ ] , since only the element @xmath228 with @xmath229 , i.e. @xmath230 , can have a non - zero coefficient @xmath231 in @xmath225 $ ] . we prove that @xmath222 $ ] and @xmath223 $ ] are not empty . the size of the index sets @xmath232 is bounded below by @xmath233 the inequality @xmath151 is equivalent to @xmath234 . hence , we have @xmath235 implying that the sets @xmath225 $ ] are not empty . indeed , the coefficients defined by @xmath236 for @xmath237 and otherwise @xmath238 admit all the constraints in @xmath225 $ ] . since @xmath239\subset\mathcal{u}_{\eta}^{\pm}[\nu,\mu;d']$ ] holds , we have @xmath240\subset\mathcal{v}_{\eta}[\nu,\mu;d']$ ] , where @xmath241 = \widetilde{\mathcal{u}}_{\eta}^{+}[\nu,\mu;d']\ominus\widetilde{\mathcal{u}}_{\eta}^{-}[\nu,\mu;d']$ ] . now , let us prove the inequality @xmath242 } \|f\|_\mathcal{h}^2 < \infty . \end{aligned}\ ] ] the above argument leads to @xmath243 } \|f\|_{\mathcal{h}}^2 \leq \min_{f\in\widetilde{\mathcal{v}}_{\eta}[\nu,\mu;d']}\|f\|_{\mathcal{h}}^2<\infty \end{aligned}\ ] ] for any @xmath244 . let us define @xmath245=\mathrm{conv}\{k(\cdot , x_i)\,:\,i\in{i}_{+}\}\ominus\mathrm{conv}\{k(\cdot , x_i)\,:\,i\in{i}_{-}\}$ ] for the original dataset @xmath128 . then , the inclusion relation @xmath225\subset\mathrm{conv}\{k(\cdot , x_i)\,:\,i\in{}i_{\pm}\}$ ] leads to @xmath241\subset\mathcal{c}[d]$ ] . hence , we obtain @xmath246 } \|f\|_{\mathcal{h}}^2 \\ & \leq \max_{\eta\in{e}_\mu}\min_{f\in\widetilde{\mathcal{v}}_{\eta}[\nu,\mu;d ' ] } \|f\|_{\mathcal{h}}^2 \\ & \leq\ \max_{\eta\in{e}_\mu}\max_{f\in\widetilde{\mathcal{v}}_{\eta}[\nu,\mu;d ' ] } \|f\|_{\mathcal{h}}^2 \\ & \leq\ \max_{f\in\mathcal{c}[d]}\|f\|_{\mathcal{h}}^2 \\ & < \infty . \end{aligned}\ ] ] the boundedness of @xmath247}\|f\|_{\mathcal{h}}^2 $ ] comes from the compactness of @xmath245 $ ] and the continuity of the norm . more precisely , it is bounded above by twice the maximum eigenvalue of the gram matrix defined from the non - contaminated data @xmath128 . the upper bound does not depend on the contaminated dataset @xmath150 . thus , the first inequality of holds . [ proof of lemma [ lemma : unbounded - breakdown - point - er - svm-1 ] ] we use the same notation as in the proof of lemma [ lemma : breakdown - point - er - svm ] . without loss of generality , we assume @xmath250 . the parameter @xmath4 is expressed as @xmath206 for @xmath205 . we prove that there exists a feasible parameter @xmath244 and a contaminated training set @xmath251\}\in\mathcal{d}_{\mu{m}}$ ] such that @xmath252=\emptyset$ ] . the construction of the dataset @xmath253 is illustrated in figure [ fig : plot - lemma_opt_infinity ] . suppose that @xmath254 and @xmath255 and that @xmath256 holds for all @xmath257 , meaning that all outliers in @xmath253 are made by flipping the labels of the negative samples in @xmath128 . this is possible , because @xmath258 holds . the inequality @xmath248 leads to @xmath259 the outlier indicator @xmath260 is defined by @xmath261 for @xmath65 samples in @xmath216 , and @xmath262 otherwise . this assignment is possible because @xmath263 then , we have @xmath264 from and , we have @xmath265 in addition , @xmath266 holds only when @xmath267 . therefore , we have @xmath268=\emptyset$ ] . the infeasibility of the dual problem means the unboundedness of the primal problem . hence , there exists a contaminated dataset @xmath150 and an outlier indicator @xmath269 such that @xmath270}\|f\|_{\mathcal{h}}^2=\infty \end{aligned}\ ] ] holds . for a rational number @xmath272 , there exists an @xmath7 such that @xmath273 and @xmath274 hold . for such @xmath19 , let @xmath22\}$ ] be a training data such that @xmath275 and @xmath276 , where the index sets @xmath277 are defined in the proof of appendix [ appendix : proof_breakdown_obj ] . since the label ratio of @xmath128 is @xmath278 , and we have @xmath144 . for @xmath157 defined from @xmath128 , let @xmath251\}\in\mathcal{d}_{\mu{m}+1}$ ] be a contaminated dataset of @xmath128 such that @xmath279 outliers are made by flipping the labels of the negative samples in @xmath128 . thus , there are @xmath65 negative samples in @xmath253 . let us define the outlier indicator @xmath260 such that @xmath261 for @xmath65 negative samples in @xmath253 . then , any sample in @xmath253 with @xmath262 should be a positive one . hence , we have @xmath268=\emptyset$ ] . the infeasibility of the dual problem means that the primal problem is unbounded . thus , we obtain @xmath280 . the non - contaminated dataset is denoted as @xmath22\}$ ] . for the dataset @xmath128 , let @xmath283 and @xmath201 be the index sets defined by @xmath202 . under the conditions of theorem [ theorem : bounded_bias ] , inequality holds . given a contaminated dataset @xmath208\}\in\mathcal{d}_{\mu{m}-\ell}$ ] , let @xmath284 be the negative margin of @xmath285 , i.e. , @xmath286 for @xmath212 . for @xmath33 , the function @xmath287 is defined as @xmath288 where the index set @xmath289 is given by @xmath290\,:\,\mu{m}+1\leq{j}\leq\nu{m}\ } \end{aligned}\ ] ] for the sorted negative margins , @xmath291 . for simplicity , we drop the dependency of the permutation @xmath47 on @xmath97 . the estimated bias term @xmath292 is the optimal solution of @xmath287 because of . the function @xmath287 is continuous . in addition , @xmath287 is linear on the interval such that @xmath289 is unchanged . hence , @xmath287 is a continuous piecewise linear function . below , we prove that the minimum solution of @xmath287 is bounded regardless of the contaminated dataset @xmath293 . for the non - contaminated data @xmath128 , let @xmath294 be a positive real number such that @xmath295 the existence of @xmath294 is guaranteed . indeed , one can choose @xmath296 because the rkhs norm of @xmath297 is uniformly bounded above for @xmath298 and @xmath128 is a finite set . for the contaminated dataset @xmath299\}\in\mathcal{d}_{\mu{m}-\ell}$ ] , let us define the index sets @xmath300 and @xmath301 for each label by @xmath302\,:\,y_i'=\pm1\},\\ i_{\mathrm{in},\pm}'&=\{i\in{}i_{\pm } ' \,:\ , |f_{d'}(x_i')|\leq{}r\ } , \\ i_{\mathrm{out},\pm}'&=\{i\in{}i_{\pm}'\,:\ , |f_{d'}(x_i')|>r\}. \end{aligned}\ ] ] for any non - contaminated sample @xmath210 , we have @xmath303 . hence , @xmath212 for @xmath304 should be an outlier that is not included in @xmath128 . this fact leads to @xmath305 let us prove the first statement . if @xmath306 holds , we have @xmath310 from the definition of the index set @xmath311 . let us consider two cases : ( i ) for all @xmath312 , @xmath313 holds , and ( ii ) there exists an index @xmath314 such that @xmath315 . for a fixed @xmath97 such that @xmath306 , let us assume ( i ) above . then , for any index @xmath214 in @xmath316 , we have @xmath317 , meaning that @xmath318 . hence , the size of the set @xmath316 is less than or equal to @xmath319 . therefore , the size of the set @xmath320 is greater than or equal to @xmath321 . the first inequality of leads to @xmath322 . therefore , in the set @xmath289 , the number of negative samples is more than the number of positive samples . for a fixed @xmath97 such that @xmath306 , let us assume ( ii ) above . due to the inequality , for any index @xmath323 , the negative margin @xmath284 is at the top @xmath64 of those ranked in the descending order . hence , the size of the set @xmath320 is greater than or equal to @xmath324 . therefore , the size of the set @xmath316 is less than or equal to @xmath325 . the second inequality of leads to @xmath326 . also in the case of ( ii ) , the negative label dominates the positive label in the set @xmath289 . for negative ( resp . positive ) samples , the negative margin is expressed as @xmath327 ( resp . @xmath328 ) with a constant @xmath329 . thus , the continuous piecewise linear function @xmath287 is expressed as @xmath330 where @xmath331 are constants as long as @xmath289 is unchanged . as proved above , @xmath332 is a positive integer , since negative samples are more than positive samples in @xmath289 , when @xmath306 . as a result , the optimal solution of the bias term should satisfy @xmath333 in the same manner , one can prove the second statement by using the fact that @xmath307 is a sufficient condition of @xmath334 then , we have @xmath335 in summary , we obtain @xmath336 we use the same notation as in the proof of theorem [ theorem : bounded_bias ] in appendix [ appendix : proof_breakdown_bias ] . note that inequality holds under the assumption of theorem [ theorem : bounded_kernel_bounded_bias ] . the reproducing property of the rkhs inner product yields @xmath337 for any @xmath299\}\in\mathcal{d}_{\mu{m}}$ ] due to the boundedness of the kernel function and inequality . hence , for a sufficiently large @xmath338 , the sets @xmath339 and @xmath340 become empty for any @xmath150 . under inequality , suppose that @xmath313 holds for all @xmath312 . then , for @xmath341 , we have @xmath317 . thus , @xmath318 holds . since @xmath339 is the empty set , @xmath316 is also the empty set . therefore , @xmath342 has only negative samples . let us consider the other case ; i.e. , there exists an index @xmath314 such that @xmath315 . assuming that @xmath168 , one can prove that the negative labels dominate the positive labels in @xmath289 in the same manner as the proof of theorem [ theorem : bounded_bias ] . eventually , for any @xmath150 , the function @xmath287 is strictly increasing for @xmath306 . in the same way , one can prove that @xmath287 is strictly decreasing for @xmath307 . moreover , for any @xmath150 and for @xmath343 , one can prove that the absolute value of the slope of @xmath287 is bounded below by @xmath308 according to the argument in the proof of theorem [ theorem : bounded_bias ] . as a result , we obtain @xmath344 . d. j. crisp and c. j. c. burges . a geometric interpretation of @xmath1-svm classifiers . in s. a. solla , t. k. leen , and k .- mller , editors , _ advances in neural information processing systems 12 _ , pages 244250 . mit press , 2000 . f. perez - cruz , j. weston , d. j. l. hermann , and b. schlkopf . extension of the @xmath1-svm range for classification . in _ advances in learning theory : methods , models and applications 190 _ , pages 179196 , amsterdam , 2003 . ios press . a. takeda , h. mitsugi , and t. kanamori . a unified robust classification model . in john langford and joelle pineau , editors , _ proceedings of the 29th international conference on machine learning ( icml-12 ) _ , icml 12 , pages 129136 , new york , ny , usa , july 2012 . omnipress . a. takeda and m. sugiyama . -support vector machine as conditional value - at - risk minimization . in william w. cohen , andrew mccallum , and sam t. roweis , editors , _ icml _ , volume 307 of _ acm international conference proceeding series _ , pages 10561063 . acm , 2008 . p. tsyurmasto , s. uryasev , and j. gotoh . support vector classification with positive homogeneous risk functionals . technical report , research report 2013 - 4 , department of industrial and systems engineering , university of florida.(downloadable from www . edu / uryasev / publications/ ) , 2013 . y. yu , m. yang , l. xu , m. white , and d. schuurmans . relaxed clipping : a global training method for robust regression and classification . in _ neural information processing systems _ , pages 25322540 . mit press , 2010 .

the paper considers a bidirectional power flow model of the electric vehicles ( evs ) in a charging station . the evs can inject energies by discharging via a vehicle - to - grid ( v2 g ) service which can enhance the profits of the charging station . however , frequent charging and discharging degrade battery life . a proper compensation needs to be paid to the users to participate in the v2 g service . we propose a menu - based pricing scheme , where the charging station selects a price for each arriving user for the amount of battery utilization , the total energy , and the time ( deadline ) that the ev will stay . the user can accept one of the contracts or rejects all depending on their utilities . the charging station can serve users using a combination of the renewable energy and the conventional energy bought from the grid . we show that though there exists a profit maximizing price which maximizes the social welfare , it provides no surplus to the users if the charging station is aware of the utilities of the users . if the charging station is not aware of the exact utilities , the social welfare maximizing price may not maximize the expected profit . in fact , it can give a zero profit . we propose a pricing strategy which provides a guaranteed fixed profit to the charging station and it also maximizes the expected profit for a wide range of utility functions . our analysis shows that when the harvested renewable energy is small the users have higher incentives for the v2 g service . we , numerically , show that the charging station s profit and the user s surplus both increase as v2 g service is efficiently utilized by the pricing mechanism .

1612.00106

electric vehicles ( evs ) have several advantages over the traditional gasoline powered vehicles . for example , evs are more environment friendly and more energy efficient . realizing the above , regulators ( e.g. federal energy regulator commission ( ferc ) ) are providing incentives to the consumers to switch to electric vehicles . manufacturers ( e.g. tesla , nissan ) are increasingly developing evs equipped with superior technologies . as a result , electric vehicles are increasingly become popular . however , a wide deployment of evs requires an extensive network of charging stations which can be capable of charging large number of vehicles . vehicle - to - grid ( v2 g ) service has been proposed @xcite to enhance the profitability of the evs . in the v2 g service , evs can inject energies to the grid by discharging from their batteries . thus , this bidirectional power flow where evs can both charge and discharge has a lot of potential . hence , a lot of effort is going on for developing bidirectional evs@xcite . however , the higher amount of charging and discharging cycles will degrade the battery life . hence , the owners of the evs have to be compensated adequately for the v2 g service . thus , though a charging station can gain an additional profit using the v2 g service of the evs , however without a proper pricing mechanism , the owners of the evs will not prefer the v2 g service in the first place which may nullify the profit of the charging station . without a proper control mechanism , the cost of the charging station and the peak energy consumption may increase . without a profitable charging station , the wide - scale deployment of the evs will remain a distant dream . the charging station needs to select prices in order to earn profits by allocating resources in an intelligent manner among the evs . the charging station also needs to provide adequate compensation to the owners if the ev is used for the v2 g service . however , high prices or low compensation may not provide incentives to the owners which may reduce the profit . hence , a proper pricing mechanism for charging the evs and the v2 g service is imperative for a charging station . we propose a menu based pricing scheme for charging an ev . whenever an ev arrives at the charging station , the charging station offers a variety of contracts @xmath0 to the ev s owner ( or , user ) at a price @xmath1 to the user where the user will be able to charge at least @xmath2 units of energy within the deadline @xmath3 for completion , and the battery usage will be limited to @xmath4 amount . the battery usage is the total amount of charging and discharging of the ev . the user either accepts one of the contracts by paying the specified price or rejects all of those based on its payoff . we assume that the user gets a utility for consuming @xmath2 amount of energy within the deadline @xmath3 . however , the user also has to incur a cost for battery utilization @xmath5 . the payoff of the user ( or , user s surplus ) for a contract is the difference between the utility and the sum of the cost incurred , and the price to be paid for the contract . the user will select the option which fetches the highest payoff . the various advantages of the above pricing scheme should be noted . first , it is an online pricing scheme . it can be adapted for each arriving user . second , since the charging station offers prices for different levels of charging required and the deadline , the charging station can prioritize one contract over the others depending on the energy resources available . third , the charging station also provides options of the maximum battery utilization to the users . the user who is not interested in the v2 g service is entitled to do that by selecting the contract with @xmath5 as @xmath6 . finally , the user s decision is much simplified . she only needs to select one of the contracts ( or , reject all ) . we consider that the charging station is equipped with renewable energy harvesting devices and a storage device for storing energies . the charging station may also buy conventional energy from the market to fulfill the contract of the user if required . hence , if a new user accepts the contract @xmath0 , a cost is incurred to the charging station . this cost may also depend on the existing evs and their resource requirements . a contract also specifies the maximum battery utilization which restricts the v2 g service generated from an ev . hence , the charging station needs to find the optimal cost for each contract . we show that obtaining the cost of fulfilling a contract is a _ linear programming _ problem . we also show that if a user accepts a contract with a higher battery utilization , the cost of the contract is lower ( lemma 1 ) . we consider two optimization problems i ) social welfare maximization , and ii ) the ev charging station s profit maximization . we investigate the existence of a pricing mechanism which maximizes the ex - post social welfare , _ i.e. _ , maximizes the social welfare for every possible realization of the utility function . we show that there exists such a pricing strategy . the pricing scheme is simple to compute , as the charging station selects a price which is equal to the marginal cost for fulfilling a certain contract for a new user ( theorem [ thm : max_socv2 g ] ) . however , the above pricing scheme only provides _ zero _ profit to the charging stations . thus , such a pricing scheme may not be useful to the charging station . we show that when a charging station is _ clairvoyant _ ( i.e. , the primary knows the utilities of the users ) , there exists a pricing scheme which satisfies both the objectives ( lemma [ thm : profit_max ] ) . though in the above pricing mechanism , the user s surplus becomes @xmath6 . a _ clairvoyant _ charging station may not be beneficial for the user s surplus . in the scenario where the charging station does not know the exact utilities of the users , we show that _ there may not exist a pricing strategy which simultaneously maximizes the ex - post social welfare and the expected profit_. one has to give away the ex - post social welfare maximization in order to achieve expected profit maximization . _ however , the user s surplus becomes higher compared to the clairvoyant scenario_. hence , an uncertainty of the utility enhances the user s surplus . we propose a pricing strategy which can fetch the highest possible profit to the charging station under the condition that it maximizes the ex - post social welfare ( theorem [ thm : profitmax_uncertainty ] ) . above pricing strategy provides a _ worst case _ maximum profit to the charging station . since the above pricing strategy may not yield the _ maximum expected _ profit to the charging station , we have to relax the constraint the social welfare to be maximized in order to yield a higher profit to the charging station . whether a contract will be selected by the user does not depend on the price of the contract , but also , the prices of other contracts . thus , achieving a pricing scheme which maximizes the expected profit is difficult because of the discontinuous nature of the profits . we propose a pricing strategy which yields a fixed ( say @xmath7 ) amount of profit to the charging station . further , we show that a suitable choice of @xmath7 can maximize the profit of the charging station for a class of utility functions ( theorem [ thm : aclassutility ] ) . in section [ sec : v2g_profitability ] we characterize the conditions which will yield higher profits to the charging station in the v2 g service . we show that if the conventional energy price is high , the charging station selects more preferable prices for enticing v2 g services if the renewable energy harvested is low . when the charging station s storage capacity is low and the renewable energy harvesting is low , the user s incentives for the v2 g service also increases . the v2 g service also increases the profit of the charging station . finally , we , empirically provide insights how a trade - off between the profit of the charging station and the social welfare can be achieved for various pricing schemes ( section [ sec : simulation_results ] ) . our numerical analysis shows that both the user s surplus and the charging station s profit increase with the v2 g service . the energy consumption during the peak period decreases as users provide more v2 g services during the peak period . charging schedule for evs using price signals for _ unidirectional _ service have been considered@xcite . the above papers did not consider the optimal discharging schedule of the evs as these papers only considered unidirectional power flow . as a result , these paper did not consider the battery degradation cost incurred in the v2 g service . few papers considered the _ bidirectional _ power flow @xcite . however , their focus was scheduling of the charging and discharging pattern of the evs , rather than the pricing aspects . naturally , these papers did not consider whether users will prefer the amount of battery degradation found in the optimal scheduling process . the social welfare maximizing and profit maximizing prices are also not considered . @xcite considered the optimal pricing to the evs in a day - ahead setting for residential charging . the users control the charging and discharging pattern at each instance for the price selected by the aggregator in a pre - specified manner . however , the users can arrive randomly in the charging station , the charging station needs to select a price for each arriving user using an online algorithm . the users can not control the charging and discharging schedule at each instance in the charging station . compared to _ all _ the above papers , in our approach , the charging station specifies the amount of battery utilization for each contract . the charging station also selects prices for different deadlines in our approach . hence , the user can now choose its own deadline and can specify the v2 g service it is willing to commit based on its own need . deadline differentiated prices have been considered @xcite . in our previous work @xcite , we also considered a menu - based pricing scheme where the charging station provides prices for different deadlines and prices to the new user . however , the above papers did not consider the _ bidirectional _ power flow problem . thus , in this setting the charging station now also needs to select different prices for different amounts of battery utilization . _ to the best of our knowledge , this is the first attempt to consider menu - based pricing which considered the bidirectional power flow from the evs . _ further , this is the first work that incentivizes the users to participate in the v2 g service by finding optimal prices for different amounts of battery degradation while maximizing the social welfare or profit of the charging station . we consider that evs arrive throughout a day at the charging station for charging . suppose that the user @xmath8 arrives at time @xmath9 . the job of the charging station is to select a price for charging in order to maximize the profit . we consider a vehicle - to - grid ( v2 g ) service where electric vehicles ( evs ) can feed back stored power to the grid . specifically , the energy can be discharged from the batteries of the evs . however , the ev batteries have fixed number of charging - discharging cycles @xcite . when the battery is used for feeding back energy to the grid , the battery wear cost may be significant . thus , the users may want to limit the battery utilization as low as possible . we consider that a charging station wants to maximize its profit over a certain time period ( e.g. over a day ) . it will offer a menu - based price contract @xmath10 for contract @xmath0 to the user @xmath8 which arrives at time @xmath9 in the charging station ( fig . [ fig : ev_charging ] ) . if the user selects the menu @xmath0 , then , the user @xmath8 will be able to charge @xmath2 amount of energy at most within the deadline @xmath3 and the maximum amount of _ additional _ battery utilization ( definition 1 ) is restricted to @xmath5 . the deadline @xmath3 denotes that the user has to leave by time @xmath3 , after which the ev will not be able to charge . first , we describe the constraints the charging station has to satisfy the contract @xmath11 for the user @xmath8 . * charging / discharging rate * : let @xmath12 be the amount of energy provided to ( or , discharged from ) the ev @xmath8 during time @xmath13 . @xmath14 indicates that the ev @xmath8 is charged and @xmath15 indicates that energy is discharged from the ev @xmath8 during time @xmath13 . also note that there is an initial set @xmath16 of existing evs . vehicle @xmath17 requires additional @xmath18 amount of energy within the deadline @xmath19 . the charging and discharging efficiency of the ev @xmath8 is denoted as @xmath20 and @xmath21 respectively . note that @xmath22 where @xmath23 is the positive part of @xmath12 ( it denotes that the electric vehicle @xmath8 is charged during time @xmath13 ) and @xmath24 denotes the negative part of @xmath12 ( it denotes the amount discharged from ev @xmath8 during time @xmath13 ) . hence , the following set of constraints must be satisfied @xmath25 @xmath26 indicates the total amount of energy required for charging and @xmath27 indicates the total amount of energy used when evs discharge during time @xmath13 . note that an ev can not simultaneously charge and discharge , hence , we must have @xmath28 for all @xmath29 and @xmath3 . _ later we will show in theorem [ thm : noposinegative ] that in an optimal solution @xmath28 though we have not explicitly considered the above constraint in the system . * ev s battery limit * : let the battery level of the ev @xmath30 at time @xmath3 be @xmath31 . let the battery level at the start of the time @xmath9 be @xmath32 . since the battery of the ev can be charged or discharged in the v2 g service , the total amount of charging and discharging must satisfy the limits on the battery levels : the battery level must be between a lower value @xmath33 ( it can be @xmath6 ) and the high value @xmath34 ( the highest capacity ) . hence , we must have @xmath35 note that @xmath36 is @xmath37 . * charging and discharging rate limit * : there is also a charging and discharging limit . hence , we have @xmath38 @xmath39 , and @xmath40 are respectively the charging and discharging rate limits . * hybrid source * : the charging station is equipped with renewable energy harvesting devices ( fig . [ fig : hybrid_v2 g ] ) . the charging station also has a storage device with capacity @xmath41 . the charging station has the forecast of harvested energy as @xmath42 for time @xmath13 . the charging station also can buy energies from the conventional market . suppose that the amount of energy used from the storage device during time @xmath13 be @xmath43 and let the energy be stored from the electric vehicles and the conventional energy in the storage device be @xmath44 during time @xmath13 ( fig . [ fig : hybrid_v2 g ] ) . let @xmath45 be the level of battery at time @xmath3 . the charging efficiency and discharge efficiency of the battery of the charging station is considered to be @xmath46 and @xmath47 respectively . thus , @xmath48 note that our model can also incorporate the scenario where the batteries have some static leakage rate i.e. the battery level decreases with time . * energy to and from the grid * : let the amount the charging station buys from the conventional market and sells to the grid as @xmath49 and @xmath50 respectively for time @xmath13 ( fig : [ fig : hybrid_v2 g ] ) . hence , we have @xmath51 since @xmath44 is the negative of @xmath43 , thus , we can represent the above constraint in the following @xmath52 we also must have @xmath53 since the battery of the charging station should not charge or discharge at the same time . though we have not explicitly considered the above constraint , however , we show that in an optimal solution in theorem [ thm : noposinegative ] , we always have @xmath53 . note that the constraints in ( [ eq : battery_capacity ] ) and ( [ eq : storage ] ) also specify the bound on the amount of charging and discharging from the evs . the amount of charging can not exceed the total amount of stored energy in the battery and the energy bought from the grid . the stored energy depends on the harvested renewable energy , energy charged to the evs and discharged from the evs , energy bought and energy sold to the grid . [ fig : hybrid_v2 g ] depicts various system parameters . * maximum battery utilization * in the vehicle - to - grid ( v2 g ) service , frequent charging and discharging may reduce the lifecycle of the battery of an ev . thus , the users may not like the evs be charged and discharged very often . we define a metric which will model the total maximum utilization of the battery of an ev . battery utilization is defined as the absolute value of the difference between the battery levels at two subsequent time intervals . hence , if the deadline is @xmath54 for user @xmath8 , then the total battery utilization for user @xmath8 is @xmath55 . the battery utilization defines the total level of charging and discharging has been done . if the ev @xmath8 is used only for charging , then @xmath56 . in a contract @xmath57 , the battery of ev needs a charging amount of @xmath2 , and thus , the total battery utilization has to be at least @xmath2 . thus , the contract @xmath11 where @xmath58 denotes the _ additional battery utilization _ apart from the charging amount @xmath2 specified by the contract . we denote @xmath5 as the maximum _ additional _ battery utilization with slight abuse of notation . specifically , in the contract @xmath11 , the maximum utilization is restricted to @xmath4 for user @xmath8 . note that @xmath2 is the energy that the user @xmath8 will receive in the contract @xmath11 . suppose the maximum utilization remaining for an existing user @xmath17 at time @xmath9 is @xmath59 . thus , if the user @xmath8 selects the contract @xmath11 , the constraint that the charging station has to satisfy is @xmath60 the above constraint is not linear . in the following we reduce it in a linear form . note from ( [ eq : dischargelimit ] ) that @xmath61since @xmath62 and @xmath63 are the positive and negative parts of @xmath37 , thus , @xmath64 . hence , the expression in ( [ eq : bu ] ) becomes @xmath65 note that we must have @xmath28 . we show in theorem [ thm : noposinegative ] that this is indeed true in an optimal solution . user s utility for the contract @xmath0 is denoted as @xmath66 which is a random variable . the realized value @xmath67 is only known to the user @xmath8 , but not known to the charging station in general . the payoff of user @xmath8 or _ user s surplus _ if she selects the contract @xmath0 is @xmath68 ( fig . [ fig : ev_charging ] ) . if she rejects all her payoff is @xmath6 . the user will select the contract that fetches the maximum payoff to her . thus , for a menu of prices @xmath10 , the user @xmath8 selects @xmath69 $ ] such that it maximizes the following @xmath70 note that @xmath71 only if @xmath72 . if such a solution is not unique , any convex combination of these solutions is also optimal since a user can select any of the maximum payoff contracts . we denote the decision as @xmath73 . note that the decision whether to accept the menu price @xmath10 depends not only on the price @xmath10 but also other price menus i.e. @xmath74 where @xmath75 and @xmath76 and @xmath77 . this is because the user only selects the price menu which is the most favorable . note that if the maximum payoff that user gets among all the price menus ( or , contracts ) is negative , then the user will not charge i.e. , @xmath78 for all @xmath2 , @xmath3 and @xmath79 . we also assume that if there is a tie between charging and not charging , then the user will decide to charge i.e. , if the maximum payoff that user can get is @xmath6 , then the user will decide to charge . amount . ] since the users arrive to the charging station at any time throughout the day , the charging station does not know the exact arrival times for the future vehicles . we consider that the charging station is myopic or near - sighted i.e. , it selects its price for user @xmath8 without considering the future arrival process of the vehicles . however , it will consider the cost incurred to charge the existing evs . note that as the number of existing users increases , the marginal cost can increase to fulfill a contract for an arriving user , hence , such a pricing strategy may not maximize the payoff in a long run . note that , _ a myopic pricing strategy is optimal in the case the marginal cost of fulfilling a demand of a new user is independent of the number of existing users . _ in practice , the charging station often has fixed number of charging spots . thus , the charging station may want to select high prices for user @xmath8 , in order to make the charging spots available for the users who can pay more but only will arrive in future . hence , considering the future arrival process of the vehicles may increase the profit of the charging station . however , such a pricing strategy is against the first come first serve basis which is the current norm for charging vehicles . our approach can be seen as a fair allocation process , where the charging station serves users based on the first come first serve basis . the optimal cost for the charging station to fulfill the contract @xmath57 to the user @xmath8 is thus @xmath80 where @xmath81 is the per unit cost of buying energy from the grid and @xmath82 is the per unit cost of selling energy to the grid . note that our model can also incorporate time varying , strictly increasing convex costs @xmath83 or concave prices @xmath84.we assume that @xmath85 in order to avoid arbitrage opportunity . however , the above cost structures will destroy the properties of the linear programming . our model can be easily extended to the setting where there are upper bounds on @xmath49 and @xmath50 . the problem will still be a linear programming problem . the cost @xmath81 and the price @xmath82 are assumed to be known . in case they change in a dynamic manner i.e. , they are real time prices , we consider that @xmath81 is the expected cost , and @xmath82 is the expected price . note that the above problem is a linear programing problem ( similar to the setting considered in @xcite ) and thus , it is easy to solve . however , the number of constraints are hugher compared to @xcite as we consider the v2 g service . in the following theorem , we show that the optimal cost to the charging station given in ( 12 ) , @xmath86 and @xmath53 for all @xmath3 , and @xmath87 . [ thm : noposinegative ] in an optimal solution of the problem @xmath88 , we must have the decision variables @xmath28 @xmath89 and @xmath3 . further , in an optimal solution , we also have @xmath53 for all @xmath3 . _ proof _ : see appendix [ sec : proof_thmnoposinegative ] . the above theorem shows that in an optimal solution , neither the ev nor the battery of the charging station simultaneously charge and discharge . now we calculate the _ additional _ cost imposed to the charging station for fulfilling the new contract . first , we introduce some notations which we use throughout . [ defn : vlt ] the charging station has to incur the cost @xmath90 for serving the existing users and the contract @xmath11 for the new user @xmath8 where @xmath90 is the value of the linear optimization problem @xmath88 . since @xmath88 is a linear optimization problem , it is easy to compute @xmath91 . further , note that if the above problem is infeasible for some @xmath2 , @xmath3 and @xmath5 , then we consider @xmath91 as @xmath92 . [ defn : v - k ] let @xmath93 be the amount that the charging station has to incur to satisfy the requirements of the existing evs if the new user does not opt for any of the price menus . if user @xmath8 does not accept any price menu , then the charging station still needs to satisfy the demand of existing users i.e. , the charging station must solve the problem @xmath88 with @xmath94 . @xmath93 is the value of that optimization problem . from definitions [ defn : vlt ] and [ defn : v - k ] we can visualize @xmath95 as the additional cost or marginal cost to the charging station when the user @xmath8 accepts the price menu @xmath96 . we assume that the prediction @xmath97 is perfect for all future times and is known to the charging station . however , menu - based pricing approach can be extended to the setting where the estimated generation does not match the exact amount . first , we can consider a conservative approach where @xmath42 can be treated as the worst possible renewable energy generation . as a second approach , we can accumulate various possible scenarios of the renewable energy generations , and try to find the cost to fulfill a contract for each such scenario . for example , if there are @xmath98 number of possible instances of the renewable energy generation amount in future . then , we can find the optimal cost for each such instance of renewable energy generation @xmath99 where @xmath100 instead of @xmath42 . we then can compute the average ( or , the weighted average , if some instance has greater probability ) of the optimal costs , and that cost can be taken as the cost of fulfilling a certain contract . the following result entails that the v2 g service indeed reduces the cost of the charging station . [ lm : v2 g ] @xmath101 for any @xmath102 . thus , @xmath103 for any @xmath104 . _ outline of the proof _ : note from ( [ eq : battery_utilization ] ) that the decision space is more restricted for @xmath105 compared to @xmath106 if @xmath102 . every feasible solution with @xmath105 is also feasible with @xmath106 since @xmath102 . thus , the above result follows . since the charging station can sell some energies taken from the evs , the cost of fulfilling the contract @xmath0 will be lower as @xmath5 will increase . if the user @xmath8 selects the menu @xmath0 , then the additional cost incurred by the charging station is @xmath107 . hence , the profit of the charging station is @xmath108 . if the user does not select any of the menus , then the charging station gets @xmath6 profit . thus , the expected profit of the charging station is @xmath109 where @xmath110 is the probability of the event that the menu @xmath0 is accepted by the user @xmath8 . note from ( [ eq : profitmax_expec ] ) that the profit maximization is a difficult problem as the menu selected by an user inherently depends on the prices selected for other menus . for example , if the price selected for a particular contract is high , the user will be reluctant to take that as compared to a lower price one . the profit is a discontinuous function of the prices and thus , the problem may not be convex even when the marginal distribution of the utilities are concave . we consider that the charging station decides the price menus in order to fulfill one of the two objectives ( or , both) i ) social welfare maximization and ii ) its profit maximization . the social welfare is the sum of user surplus and the profit of the charging station . as discussed in section [ sec : userutility ] for a certain realized values @xmath67 if the user @xmath8 selects the price menu @xmath10 , then its surplus is @xmath68 , otherwise it is @xmath6 . as discussed in section [ sec : profit ] the profit of the charging station is @xmath111 for a given price @xmath10 if the user selects the menu , and it is @xmath6 otherwise . hence , the social welfare maximization problem is to select the price menu @xmath112 which will maximize the following @xmath113 recall that in order to find @xmath91 we have to solve @xmath114(cf . ( [ eq : vlt ] ) ) which is a constrained optimization problem . since the charging station is unaware of the utilities of the users , the charging station has two choices - i ) decides a price and hopes that it will maximize the social welfare for the realized values of utilities ( _ ex - post _ maximization ) , or ii ) decides a price and hopes that it will maximize the social welfare in an expected sense ( _ ex - ante _ maximization ) . thus , the _ ex - ante _ maximization does not guarantee that the social welfare will be maximized for every realization of the random variables @xmath115 . however , in the _ ex - post _ maximization , the social welfare is maximized for each possible realization of the random variables . thus , _ ex - post _ maximization is a stronger concept of maximization ( and thus , also more desirable ) and it is not necessary that there exist pricing strategies which maximize the ex - post social welfare . however , we show that in our setting there exist pricing strategies which maximize the ex - post social welfare . note that _ ex - post _ social welfare maximization is the same as ( [ eq : socialwelfare_maxproblem ] ) . social welfare maximization does not guarantee that the charging station may get a positive profit . it is important for the wide scale deployment of the charging stations that the charging station must have some profit . the charging station needs to select @xmath10 in order to maximize the expected profit ( given in ( [ eq : profitmax_expec ] ) ) . note that in order to select optimal @xmath10 , the charging station has to obtain @xmath91 i.e. , it has to solve the problem @xmath116 ( cf . ( [ eq : vlt ] ) ) for each choice of @xmath2 , @xmath3 and @xmath5 . note that in order to select optimal @xmath10 , the charging station has to obtain @xmath91 and @xmath93 ( definitions [ defn : vlt ] & [ defn : v - k ] ) . however , @xmath117 and @xmath93 do not depend on @xmath10 . hence , we can separate the problem first the charging station finds @xmath91 and @xmath118 , and then it will select @xmath10 to fulfill the objective . _ we now focus on finding optimal _ @xmath10 . first , we state the optimal values of the social welfare for any given realization of the user s utilites . next , we state a pricing strategy which attains the above optimal value . note that if @xmath119 for each @xmath2,@xmath3 , and @xmath5 , then the social welfare is maximized when the user @xmath8 does not charge . in this case , the optimal value of social welfare is @xmath6 . on the other hand if @xmath120 for some @xmath2 , @xmath3 , and @xmath5 then the social welfare is maximized when the user @xmath8 charges its car . if the user accepts the price menu @xmath10 , then the social welfare is @xmath121 . thus , the maximum social welfare in the above scenario is @xmath122 as described in the following theorem [ thm : max_socialwelfare ] the maximum value of social welfare is @xmath123 . we obtain [ thm : max_socv2 g ] if @xmath124 , then it will maximize the ex - post social welfare . _ outline of the proof _ : if the price is set according to the above , the user will not select any contract if @xmath125 which gives @xmath6 social welfare . the user select the contract which maximizes the payoff if @xmath126 . in this case , the ex - post social welfare is @xmath122 . hence , the result follows . however , the above pricing mechanism does not give any non - zero profit to the charging station . the above pricing strategy is distribution independent , it holds for _ any arbitrary distribution . _ also note that the pricing strategy also maximizes the social welfare in the long run when the additional cost of fulfilling a contract ( i.e. @xmath107 ) does not depend on the existing users in the charging station . the condition that @xmath127 is independent of the existing evs in the charging station is satisfied if either all demand can be fulfilled using renewable energy or there is no renewable energy generation . hence , in the _ two above extreme cases , the myopic pricing strategy is also optimal in the long run_. we have already seen a pricing strategy which maximizes the ex - post social welfare in theorem [ thm : max_socv2 g ] , however , this pricing strategy does not give any positive profit . naturally the question arises what is the pricing strategy that maximizes the expected profit of the charging station which will also maximize the ex - post social welfare . we show that there exists a pricing strategy which may provide better profit to the charging station while maximizing the ex - post social welfare . first , we introduce a notation which we use throughout . [ defn : lower_endpoint ] let @xmath128 be the lowest end - point of the marginal distribution of the utility @xmath66 . [ thm : profitmax_uncertainty ] consider the pricing strategy : @xmath129 the pricing strategy maximizes the ex - post social welfare . the profit is @xmath130 . _ outline of proof _ : first , note that adding a constant does not change the optimal solution . hence , if @xmath131 , then @xmath132 is also optimal for price strategy in ( [ eq : pricesocialmax ] ) . now , if the price is set according to ( [ eq : pricesocialmax ] ) , when @xmath133 , the user always select the contract . if the condition is not satisfied , the user does not select the contract . hence , the ex - post social welfare is always maximized . note that if @xmath134 , then such a pricing strategy will provide a positive profit to the charging station . hence , if @xmath107 is smaller than the lowest end - point of the distribution function , then the profit will be positive . also note that the users which have higher utilities i.e. , higher @xmath128 will give more profits to the charging station . the charging station needs to know the lowest end - points of the support set of the utilities unlike in theorem [ thm : max_socv2 g ] . however , the charging station does not need to know the exact distribution functions of the utilities similar to theorem [ thm : max_socv2 g ] . the lowest end - point can be easily obtained from the historical data . the pricing strategy maximizes the ex - post social welfare similar to theorem [ thm : max_socv2 g ] . this is also the _ maximum possible profit that the charging station can have under the condition that it maximizes the ex - post social welfare with probability @xmath135_. however , it may not maximize the expected profit of the charging station . in other words , the pricing strategy which maximizes the expected profit needs not maximize the ex - post social welfare . * when the charging station is clairvoyant * : we have seen that if the charging station is unaware of the realized values of the utilities , there is no pricing strategy which maximizes both the expected profit and the ex - post social welfare . however , we show that if the charging station is clairvoyant _ i.e. _ , it is aware of the realized values of the utilities of the users , then there exists a pricing strategy which both maximizes the social welfare and the profit of the charging station . first , we introduce a notation . [ defn : optimal ] let @xmath136 . [ thm : profit_max ] let @xmath137 where @xmath138 is given in definition [ defn : optimal ] . such a pricing strategy maximizes the profit as well as the social welfare . _ there can be other pricing strategies which simultaneously maximize the social welfare and the profit . _ though the joint profit and social welfare maximizing pricing strategy may not be unique , the profit of the charging station is the unique and is given by @xmath139 the above pricing strategy is an example of _ value - based _ pricing strategy where prices are set depending on the valuation or the utility of the users @xcite . in contrast , the price strategy stated in theorem [ thm : max_socv2 g ] is an example of _ cost - based _ pricing strategy where the prices only depend on the costs . in the value - based pricing strategy , the user surplus decreases , in fact it is amount . in that case , it will be @xmath140 optimal profit maximizing strategy . ] @xmath6 in our case . thus all the user surplus is transferred as the profit of the charging station . _ thus , uncertainty regarding the utilities enhance the user s surplus . _ theorem [ thm : profitmax_uncertainty ] entails that the charging station only has a positive profit if @xmath141 . if the above condition is not satisfied , the charging station s profit will be @xmath6 . in the following we consider a pricing strategy which will give a guaranteed positive profit to the charging station . consider the pricing strategy @xmath142 where @xmath143 . note that the pricing strategy stated in ( [ eq : pricesocialmax ] ) is a variant of the pricing strategy stated in ( [ eq : guranteed_profit ] ) where @xmath144 if we allow @xmath7 can also be @xmath6 . the pricing strategy stated in ( [ eq : guranteed_profit ] ) gives the same positive profit irrespective of the menu selected by the user . the regulator such as ferc can select a @xmath7 judiciously to trade off between the profit of the charging station and the social welfare . from lemma [ lm : v2 g ] , the pricing strategy stated in ( [ eq : guranteed_profit ] ) selects _ lower price for higher battery utilization . _ this is desirable , as the user needs to be given incentive to battery utilization . higher @xmath7 will deter the user s surplus . also note that when @xmath143 , it may not maximize the ex - post social welfare from theorem [ thm : profitmax_uncertainty ] . very high value of @xmath7 also decreases the profit of the charging station , as users will be reluctant to accept any of the menus . the expected profit of the charging station for the above pricing strategy is [ thm : expected_payoff ] the expected profit of the charging station when it selects price according to ( [ eq : guranteed_profit ] ) is @xmath145 . _ outline of the proof _ : note that if a user selects any of the contracts , then the charging station s profit is @xmath7 . hence , the charging station s expected profit is @xmath7 times the probability that at least one of the contracts will be accepted . now , we provide an example where the pricing strategy in ( [ eq : guranteed_profit ] ) can also maximize the expected profit for a suitable choice of @xmath7 . first , we introduce a notation [ defn : alpha ] let @xmath146 . note that since @xmath66 is bounded and the probability distribution is continuous , thus , @xmath147 exists . note from theorem [ thm : expected_payoff ] that @xmath147 corresponds to @xmath7 the charging station can get the maximum possible expected profit when the prices are of the form ( [ eq : guranteed_profit ] ) . if the utilities are drawn from a strictly increasing continuous distribution , then the set of @xmath148 would be singleton and we do not need to specify the maximum . now , consider the pricing strategy @xmath149 where @xmath147 is as given in definition [ defn : alpha ] . the above pricing strategy maximizes the profit for a class of utility functions which we describe below . [ assum : utility ] suppose that the utility function @xmath150 for all @xmath2 , @xmath3 & @xmath5 ; @xmath151 is a constant and known to the charging station , and @xmath152 is a random variable whose realized value is not known to the charging station . in the above class of utility function , the uncertainty is only regarding the realized value of the random variable @xmath152 . note that @xmath152 is independent of @xmath2 , @xmath5 and @xmath3 , hence,@xmath152 is considered to be an additive noise . it is important to note that we do not put any assumption whether _ @xmath152 should be drawn from a continuous or discrete distribution . _ however , if the distribution is discrete , we need the condition that @xmath147 must exist . [ thm : aclassutility ] the pricing strategy stated in ( [ eq : price_alpha ] ) maximizes the expected profit of the charging station ( given in ( [ eq : profitmax_expec ] ) ) when the utility functions are of the form given in assumption [ assum : utility ] . _ the above result is surprising . it shows that a simple pricing mechanism such as the fixed profit can maximize the expected payoff for a large class of utility functions . _ however , if the utilities do not satisfy assumption [ assum : utility ] the above pricing strategy may not be optimal . the v2 g service will proliferate only if the users participate in that service . the charging station can attain extra profits through the v2 g service . however , the users will only select the menu with positive battery utilization if they get enough compensation . thus , the charging station s profit inherently depends on whether the users have incentives to participate in the v2 g services . in this section , we will analyze the conditions under which the users will be willing to participate in the v2 g service , and the profit of the charging station will increase . first , we discuss the cost of battery utilization . users will strictly prefer lower utilization as lower @xmath5 will increase the battery life . a higher battery utilization may increase the battery degradation cost @xcite . we denote the cost associated with the utilization @xmath5 for user @xmath8 is @xmath153 where @xmath154 is a strictly increasing function . the cost @xmath154 depends on the the state of the battery @xcite . we assume that the user s utility @xmath66 is @xmath155 we consider that the cost function @xmath153 for the battery utilization as a linear function i.e. @xmath156 . recently , @xcite shows that the per unit degradation cost for discharging remains almost constant for a wide range of values . hence , a linear cost model can be a good approximation of the cost function . however , our analysis can be easily extended to other cost models . the charging station and even the user may not know the exact value of @xmath157 . but , the ev manufacturer can easily provide the _ pessimistic _ approximation of @xmath157 such as the worst possible battery degradation cost for per unit of energy . cents and @xmath158 cents . in this example @xmath157 can be taken as @xmath158 cents per kwh . ] the realized value of the utility function of user @xmath8 is now @xmath159 . note that if @xmath104 provides a positive payoff and a higher payoff to the user compared to @xmath160 , then the user will opt for v2 g service . the following result formalizes the condition . [ thm : condn ] user @xmath8 opts for battery degradation @xmath104 , if @xmath161 for some @xmath104 . now , we consider the pricing strategy stated in ( [ eq : guranteed_profit ] ) i.e. , @xmath162 . note that with a linear _ battery degradation cost _ discussed in the last section @xmath163 our next result characterizes the condition for user s participation in the v2 g service for a linear battery degradation cost . from theorem [ thm : condn ] if the following is satisfied : @xmath164 for some @xmath104,@xmath2,and @xmath3 ; then the user @xmath8 will have any incentive for v2 g service when the pricing strategy is as given in ( [ eq : guranteed_profit ] ) . the expected profit of the charging station also increases under the above condition . _ outline of the proof _ : note that @xmath165 . thus , the user will never select the contract with @xmath160 if @xmath164 for some @xmath104 . in the fixed profit scheme , the charging station always gets a profit of @xmath7 for if a contract is selected . when @xmath164 for some @xmath104 , the user s probability of selecting any contract increases . hence , the expected profit of the charging station increases from theorem 5 . thus , if @xmath164 for some @xmath104 , the user s surplus and the profit of the charging station both increase . if the harvested renewable energy is large enough , then the charging station can fulfill the demand using the renewable energy . hence , the difference between @xmath91 and @xmath166 will be not enough for the user to participate in the v2 g service . on the other hand , if the harvested renewable energy is small , then the charging station may have to buy expensive conventional energy from the grid to fulfill the demand . hence , the user will have a higher incentive to participate in the v2 g service as the difference between @xmath91 and @xmath166 may be significant . the difference is more significant when the conventional energy is more expensive ( e.g. peak period ) . _ thus , the user will have a higher incentive to participate in the v2 g service when the renewable energy generation is small and the cost of the conventional energy is high . _ if the storage capacity of the charging station is large , the charging station may buy energy from the grid during the off - peak period and use it during the peak period . this also reduces the difference between @xmath166 and @xmath91 . however , if the storage capacity is low , the difference between @xmath166 and @xmath91 again increases . hence , _ the user s participation towards the v2 g service is more likely when the storage capacity of the charging station is small and the renewable energy harvesting is low . _ _ this shows the necessity of the v2 g service for better profitability as the high storage capacity is very costly to procure and the penetration of the renewable energy is still low . _ so far , we assumed that evs only come for charging where @xmath167 for menu - pricing . however , once the v2 g service proliferates , the users with their fully charged batteries may come during the peak hours to the charging station in order to _ only _ discharge . the users can charge their batteries again in their homes during the off - peak times . in this manner , the users can gain some profits . though , we have not explicitly considered this scenario , our model can be easily extended to the above scenario . for example , the price @xmath168 where @xmath169 will denote that the user @xmath8 s ev will be discharged _ at most _ @xmath2 amount within deadline @xmath3 and additional battery utilization @xmath170 . the negative price indicates that the charging station will pay to the user , as the user is delivering energy to the grid . note that the _ only _ constraint needs to be changed is to replace @xmath2 with @xmath171 in ( [ eq : battery_utilization ] ) . since @xmath172 still remains a linear programming problem ( definition [ defn : vlt ] ) , the charging station can still find @xmath107 and can select prices according to the strategies discussed before . hence , our results also holds in this scenario . we numerically study and compare various pricing strategies presented in this paper . we evaluate the profit of the charging station and the user s surplus achieved in those pricing strategies . we also analyze the impact of the v2 g service . similar to @xcite , the user s utility for energy @xmath173 is taken to be of the form @xmath174 thus , the user s utility is a strictly increasing and concave function in the amount energy consumed @xmath173 . the quadratic utility functions for ev charging have also been considered in @xcite . note that the user s desired level of charging is @xmath175 . _ we assume that @xmath175 is a random variable_. @xcite shows that in a commercial charging station , the average amount of energy consumed per ev is @xmath176kwh with standard deviation @xmath177kwh . we assume that @xmath175 is a truncated gaussian random variable with mean @xmath176kwh and standard deviation @xmath177kwh , where the truncation is to the interval @xmath178 $ ] . we assume that the maximum battery capacity is @xmath179 , and the minimum capacity as @xmath180 . the initial battery level of a new user is assumed to be uniformly distributed in the interval @xmath181 $ ] . note that the upper bound is @xmath182 , since the user s desired level of charging is @xmath175 . following @xcite , the deadline or the time spent by an electric vehicle in a commercial charging is distributed with an exponential distribution with mean @xmath183 hours . thus , the preferred deadline ( @xmath184 ) of the user to be an exponentially distributed random variable with mean @xmath183 . users strictly prefer a lower deadline for a given energy . we assume the utility as a convex decreasing function of the deadline . the utility of the user after the preferred deadline is assumed to be @xmath6 . hence , the user s utility for energy @xmath2 and deadline @xmath3 is chosen as the following @xmath185 the cost for the battery utilization is taken as @xmath186 . we assume that @xmath187 . thus , the total utility for the user for contract @xmath0 is @xmath188 the arrival process of electric vehicles is assumed to be a non - homogeneous poisson arrival process since the arrival rates vary over time . for example , during the peak - hours ( 8 am to 5 pm ) the arrival rate is higher compared to the off - peak hours . the arrival rates are chosen as @xmath189 ( @xmath190 , resp . ) vehicles per hour during the peak period ( off - peak period , resp . ) . we also assume that the maximum charging rate @xmath39 is @xmath191 kw . we also assume @xmath40 as @xmath192 kw . the energy harvesting device is assumed to harvest energy which is a truncated gaussian random variable with mean @xmath193 and standard deviation @xmath135 per hour . initial battery level is assumed to be @xmath6 _ i.e. _ it is fully discharged . the prices @xmath81 for the conventional energy is assumed to be governed by time - of - use ( tou ) time scale . thus , the cost of buying conventional energy varies over time . we assume that the selling price , @xmath194 . we assume that the charging station is unaware of the exact utilities . however , it knows the lower endpoints of the support sets . the charging station also knows the value of @xmath195 and the structure of the utility function , but does not know the realized values . we consider the pricing strategy that we have introduced in section [ sec : price_uncertainty ] , which is given here @xmath196 recall from definition [ defn : lower_endpoint ] that @xmath197 is the lowest end - point of the utility @xmath198 we study the impact of @xmath7 . we also study the impact of the maximum storage capacity of the charging station @xmath41 . towards this end , we consider two different values of @xmath41 ; a low value , @xmath199kwh , and a high value , @xmath200kwh . fig . [ fig : betavsadmittedusers ] shows that as @xmath7 increases the number of admitted users decreases . however , the decrement is slow initially . when @xmath7 becomes larger than a threshold , the price selected to the users becomes very large , and thus , fewer number of evs are admitted . note that v2 g service enables the charging station to admit more users . [ fig : consumer_surplus ] shows that as @xmath7 increases , the user s surplus decreases , as the users pay more . the decrement is not significant for small values of @xmath7 . also , note that v2 g enhances the user s surplus even though the users incur costs for the v2 g service . as @xmath7 increases the profit increases initially . however , when @xmath201 , the number of users served decreases rapidly , hence , the profit also drops . v2 g service significantly increases the profit of the charging station . at high values of @xmath7 both users surpluses and the profit decrease significantly . low values of @xmath7 gives high users surpluses , however , the profit is low . @xmath202 $ ] is the best candidate for the balance between profit and users surpluses . [ fig : soc_wellowb ] shows that if the charging station s storage capacity is low , the user s surplus and the profit decrease . this is because when the storage capacity is low , the charging station can only store lower amount of energy for the future use . hence , the charging station has to pay more to buy conventional energy during the peak period . hence , the charging station s profit and the user s surplus both decrease . we subsequently study the effect of @xmath7 on the time spent by the users in the charging station ( fig . [ fig : deadline ] ) . our analysis shows that users spend more time in the charging station with the increase in @xmath7 . as @xmath7 increases , the users which have preferences for lower deadlines have to pay more , since the cost of fulfilling lower deadline contracts is high . hence , those users are reluctant to accept the contract . thus , the accepted users spend more time in the charging station . [ fig : deadline ] also shows that the evs spend lower amount of time when @xmath41 is high . this is because when @xmath41 is high , the charging station s cost to fulfill contracts with a lower deadline is lower compared to low @xmath41 . [ fig : bu ] shows the average amount of battery utilization of each ev as a function of @xmath7 . as @xmath7 increases the average battery utilization increases . when @xmath7 is high , the user has to pay a larger amount . however , if the user agrees for higher battery utilization , then the price may be low because the charging station may incur a lower cost . hence , the average battery utilization increases as @xmath7 becomes high . since the charging station can only store a lower amount of renewable energy and a lower amount of energy bought from the grid during the off - peak periods , hence , the average battery utilization is higher for lower @xmath41 . the batteries of evs are used more for lower storage capacity . [ fig : active_max ] shows the impact of @xmath7 on the _ maximum _ number of evs that can simultaneously charge ( or , the number of active users ) along time . this given an idea of how much charging spots we need , in order to implement the pricing strategy . since the average time spent by users in the charging station increases with @xmath7 and the number of admitted users are almost the same for @xmath203 , the number of active users increases as @xmath7 increases . since when @xmath41 is low the average time spent by an ev increases , thus , the number of active users is higher when @xmath41 is low . [ fig : conven ] shows the energy bought from the grid over the day . the figure shows that the peak is significantly reduced during the on - peak period . in fact , the peak is shifted to the off - peak period even though more users come during the on peak period . this is because during the on - peak periods , the evs are mainly used for discharging or the charging station uses more energy from its storage . during the off - peak period , the evs are charged , and energy is stored in the battery of the charging station . [ fig : diffbuy ] shows that the peak energy used from the grid is higher when @xmath41 is high . this is because when @xmath41 is high , more energy is bought and stored during the off - peak period which results in a higher peak . the reduction of the peak decreases very slowly with @xmath7 initially . however , as @xmath7 becomes very high , very few users are admitted which results in the drop of peak energy . as @xmath7 increases the users with higher utilities will accept the contracts . thus , users with higher energy demand will be more likely to accept the contract and hence , the average charging amount for each ev should increase with @xmath7 . however , fig . [ fig : energy_mean ] shows that for @xmath203 , the average energy consumed by each ev does not increase significantly with the increase in @xmath7 . the apparent anomaly is due to the fact that the users with higher demand but with smaller deadline preferences , may have to pay more because of the increase in the price to fulfill the contract as @xmath7 increases . hence , such users will not accept the offers which results in a decrease of the average energy consumption with the increase in @xmath7 initially . however , as @xmath7 increases only the users with higher demand accept offers increasing the average energy consumption . since the user spends more time for lower @xmath41 , the average energy consumed by each ev is slightly higher for low @xmath41 . [ fig : price ] shows that as @xmath7 increases the price increases . interestingly , price is not higher during the peak - period , since more evs are involved in the v2 g service during the peak period . during the off - peak period , the price from the conventional energy is low which does not provide any incentive for the v2 g service . a new price is selected every - time an ev is admitted . however , our study shows that the variation of prices within a time interval is nominal . this paper proposed a menu - based pricing approach for v2 g service where the charging station selects prices to the arriving users for different charging amounts , additional amount of battery utilization , and the deadline . the user either selects one of the contracts or rejects all based on her utility function . when the charging station is aware of the utilities , there exists a profit maximizing pricing strategy which can also maximize the social welfare , however , it provides zero user s surplus . however , when the charging station is not aware of the utilities there is _ no _ profit maximizer pricing strategy which can maximize the ex - post social welfare . this paper also considered a pricing strategy which gives a fixed positive profit to the charging station irrespective of the contract chosen . the conditions where the users have incentives to participate in the v2 g service are characterized , and it is seen that the v2 g service is more preferable when the harvested energy is low , and the storage capacity of the charging station is low . this work can be extended in several directions . for example , the characterization of the optimal price in the non - myopic setting remain open . this paper considered a fixed price for selling the energy to the grid . the optimal price that the grid should set to enhance the v2 g service or reduce the peak consumption constitutes a future research direction . e. sortomme , m. m. hindi , s. d. j. macpherson , and s. s. venkata , `` coordinated charging of plug - in hybrid electric vehicles to minimize distribution system losses , '' _ ieee transactions on smart grid _ , vol . 2 , no . 1 , pp . 198205 , march 2011 . m. alizadeh , h. t. wai , m. chowdhury , a. goldsmith , a. scaglione , and t. javidi , `` optimal pricing to manage electric vehicles in coupled power and transportation networks , '' _ ieee transactions on control of network systems _ , vol . pp , no . 99 , pp . 11 , 2016 . j. xu and v. w. s. wong , `` an approximate dynamic programming approach for coordinated charging control at vehicle - to - grid aggregator , '' in _ smart grid communications ( smartgridcomm ) , 2011 ieee international conference on _ , oct 2011 , pp . 279284 . h. xing , m. fu , z. lin , and y. mou , `` decentralized optimal scheduling for charging and discharging of plug - in electric vehicles in smart grids , '' _ ieee transactions on power systems _ , vol . 31 , no . 5 , pp . 41184127 , sept 2016 . b. g. kim , s. ren , m. van der schaar , and j. w. lee , `` bidirectional energy trading and residential load scheduling with electric vehicles in the smart grid , '' _ ieee journal on selected areas in communications _ , vol . 31 , no . 7 , pp . 12191234 , july 2013 . a. nayyar , m. negrete - pincetic , k. poolla , and p. varaiya , `` duration - differentiated energy services with a continuum of loads , '' _ corr _ , vol . abs/1408.5825 , 2014 . [ online ] . available : http://arxiv.org/abs/1408.5825 a.ghosh and v.aggarwal , `` control of charging of electric vehicles through menu based pricing , '' 2016 . [ online ] . available : http://arxiv.org/abs/1609.09037 m. a. delucchi and t. e. lipman , `` an analysis of the retail and lifecycle cost of battery - powered electric vehicles , '' _ transportation research part d : transport and environment _ , vol . 6 , no . 6 , pp . 371 404 , 2001 . c. zhou , k. qian , m. allan , and w. zhou , `` modeling of the cost of ev battery wear due to v2 g application in power systems , '' _ ieee transactions on energy conversion _ , 26 , no . 4 , pp . 10411050 , dec 2011 . w. shuai , p. maill , and a. pelov , `` charging electric vehicles in the smart city : a survey of economy - driven approaches , '' _ corr _ , vol . abs/1601.03925 , 2016 . [ online ] . available : http://arxiv.org/abs/1601.03925 l. gan , u. topcu , and s. h. low , `` optimal decentralized protocol for electric vehicle charging , '' _ ieee transactions on power systems _ , vol . 28 , no . 2 , pp . 940951 , may 2013 . `` evaluating electric vehicle charging impacts and customer charging behaviors- experiences from six smart grid investment grant projects , '' https://www.smartgrid.gov/files/b3_revised_master-12-17-2014_report.pdf , december , 2014 . now , since @xmath209 , and @xmath210 , thus , @xmath211 . thus , @xmath212 . hence , @xmath213 the other constraints are independent of @xmath214 , hence , it will be feasible . thus , with the modified @xmath215 , @xmath216 , and @xmath217 , with the other unchanged solution the value of the optimization problem is _ lower_. hence , this leads to a contradiction . now , we show that @xmath28 in an optimal solution of ( [ eq : vlt ] ) . suppose that in an optimal solution we have @xmath218 and @xmath219 for some @xmath29 and @xmath3 . now , consider the following modification for each @xmath29 . @xmath220 note that because of the above modification @xmath221 hence , the constraints in ( [ eq : charged ] ) , ( [ eq : chargeeachev ] ) , ( [ eq : dischargelimit ] ) , ( [ eq : limit ] ) , ( [ eq : battery_utilization ] ) are satisfied for this modified solution . now , consider the following @xmath222 since @xmath223 , and @xmath224 for all @xmath29 and @xmath225 for at least index @xmath30 , thus , @xmath226 . thus , if @xmath227 and @xmath228 are modified as follows @xmath229 then the constraints in ( [ eq : battery_capacity ] ) , ( [ eq : battery_level ] ) are also satisfied when the rest of the solution are unchanged . however , the cost decreases as @xmath230 with the modified values of @xmath231 , @xmath232 , @xmath233 . this contradicts that @xmath234 are optimal . hence , the result follows . the above theorem also shows the transformation of the optimal solution requires to get @xmath53 ( cf.([eq : transformse ] ) , ( [ eq : lowervaluee ] ) ) and @xmath28 ( cf.([eq : transformr ] ) , ( [ eq : transformrtch ] ) , ( [ eq : lowervaluer ] ) ) in case if @xmath235 are bounded . the modified solution will still be bounded . suppose the statement is false . without loss of generality , assume that @xmath236 where @xmath237 for some @xmath5 , @xmath2 and @xmath3 achieves a strictly higher expected payoff than the pricing strategy @xmath238 . the expected profit of the charging station for pricing strategy @xmath236 is given by @xmath239 now , we evaluate the expression @xmath240 . the user @xmath8 will select the menu @xmath10 with a positive probability if @xmath241 and for every @xmath242 , @xmath243 since @xmath244 are fixed , hence , the above inequality is either satisfied or not satisfied with probability @xmath135 . more specifically , the user selects the menu @xmath0 if @xmath241 and @xmath245 without loss of generality , assume that @xmath246 be the maximum value for which the above inequality is satisfied i.e. @xmath247 the random variable @xmath152 only affects the probability whether @xmath248 or not . hence , the charging station s expected profit is upper bounded by @xmath249 note that by the definition of @xmath147 ( definition [ defn : alpha ] ) , @xmath250 however by theorem [ thm : expected_payoff ] the expected payoff of the charging station when it selects the price @xmath251 is given by the expression in the left - hand side of the inequality in ( [ eq : compare ] ) . hence , this leads to a contradiction . thus , the result follows .

we study free - free absorption of radio emission by winds of massive stars . we derive formulae for the optical depth through the wind measured from a point of emission along a jet , taking into account compton and photoionization heating and compton , recombination , line and advection cooling . we apply the developed formalism to radio monitoring data for cyg x-1 , which allows us to obtain strong constraints on the structure of its inner jet . with the data at 15 ghz , and taking into account an anisotropy of the stellar wind in cyg x-1 , we estimate the location of the peak of that emission along the jet at about one orbital separation , i.e. , cm . given a previous determination of the turnover frequency in cyg x-1 , this implies the location of the base of the jet at gravitational radii . we also obtain corresponding results at 8.3 ghz and 2.25 ghz , which roughly follow the standard conical partially self - absorbed jet model . furthermore , we find that the level of the orbital modulation depends on the radio flux , with the modulation being substantially stronger when the radio flux is lower . this is explained by the height of the radio emission along the jet decreasing with the decreasing radio flux , as predicted by jet models . based on the finding of the flux - dependent orbital modulation , we are able to estimate a range of the possible changes of the form of the radio / x - ray correlation in cyg x-1 due to free - free absorption . we also derive predictions for the orbital modulation and flux attenuation at frequencies beyond the 2.2515 ghz range . [ firstpage ] accretion , accretion discs radio continuum : stars stars : individual : cyg x-1 stars : individual : hde 226868 x - rays : binaries x - rays : stars .

1105.4291

cyg x-1 is an archetypical and well studied persistent black - hole binary . its companion is the ob supergiant hde 226868 . cyg x-1 is both a radio and x - ray source , and the fluxes in these two bands are strongly correlated ( e.g. , @xcite , hereafter paper i ) . the likely cause of the correlation is the jet in this system ( @xcite , hereafter s01 ) being formed by the matter of an inner hot accretion flow @xcite . the radio emission is modulated at the orbital period @xcite , which is caused by free - free absorption in the stellar wind from the massive companion . as shown by @xcite ( hereafter sz07 ) , the observed modulation levels imply that the average radio fluxes are substantially absorbed , which then , as discussed in paper i , modifies the form of the observed radio / x - ray correlation . in this work , we study in detail the free - free absorption of radio emission in winds of massive stars in high - mass x - ray binaries ( hereafter hmxbs ) . we derive formulae for the optical depth measured from the point of emission along the jet . we also consider emission of counter - jets and compton scattering by the wind . then , we apply our results to cyg x-1 , taking into account the strong anisotropy of its stellar wind , found by @xcite , as well as the radio emission being extended ( s01 ) . this yields strong constraints on the structure of the jet in cyg x-1 , in particular it implies that a major part of this emission takes place at distances comparable to the binary orbit . based on the results of this application , we estimate the effect of the free - free absorption on the form of the radio / x - ray correlation in cyg x-1 . a standard wind velocity profile from a massive star is ( e.g. , @xcite ) , @xmath2 where @xmath3 is the distance from the centre of the donor of radius @xmath4 , @xmath5 is the terminal velocity , and @xmath6 parameterizes the wind acceleration . here , a small correction due to velocity reaching the sound speed rather than being null at the stellar surface has been neglected . the ion density is , @xmath7 where @xmath8 is the direction - dependent factor accounting for the wind asymmetry due to its focusing ( see below ) , @xmath9 is the total mass - loss rate , @xmath10 and @xmath11 ( measured from the axis connecting the stars ) are the polar and azimuthal angle , respectively , @xmath12 is the proton mass , @xmath13 is the mean ion molecular weight , and @xmath14 is the h fraction . for simplicity , we hereafter assume @xmath15 . the local wind temperature follows from the balance of radiative heating and cooling and advection . in an optically - thin approximation @xcite , applicable here , the radiative rates depend only on the shape of the irradiating spectrum and on the ionization parameter , @xmath16 @xcite , where @xmath17 is the distance to the irradiating source with the ionizing luminosity , @xmath18 . the net rate of x - ray photoionization heating and recombination cooling per ion has been fitted as @xcite , @xmath19 ( in units of erg s@xmath20 ) , where @xmath21 is in k and @xmath22 kev is the characteristic energy of the ionizing spectrum . given that the wind temperature in cyg x-1 has @xmath23 kev , we can neglect the second term in parentheses above . the formulae of @xcite have been derived for the case of a 10-kev bremsstrahlung x - ray spectrum in a high - mass x - ray binary . the rate of bremsstrahlung and line cooling has been fitted as @xcite , @xmath24n(r ) } \label{c_bl}\end{aligned}\ ] ] where @xmath25 k. in the non - relativistic regime , the rate of compton energy exchange between electrons and photons is proportional to @xmath26 . the rate per electron including klein - nishina corrections but assuming @xmath27 can be written as , @xmath28 where @xmath29 is the differential energy flux , @xmath30 is the thomson cross section , @xmath31 is the electron mass , @xmath32 , and @xmath33 and @xmath34 give the klein - nishina corrections to compton heating and cooling , respectively . the former is given by @xcite , @xmath35 separating @xmath29 into the stellar and x - ray component , we have , @xmath36 , \label{compton2}\ ] ] where @xmath37 is the stellar luminosity , @xmath38 is the stellar temperature , @xmath39 is the differential luminosity originating at the black - hole location , @xmath3 and @xmath17 is the distance to the centre of the donor and to the compact object , respectively , @xmath40 , and @xmath41 is the riemann zeta function . the x - ray heating term in equation ( [ compton2 ] ) is valid at any values of @xmath42 , provided @xmath43 . equation ( [ compton2 ] ) also assumes that cooling is dominated by photons at @xmath44 , which , e.g. , is the case when @xmath45 . if it is not satisfied , @xmath34 needs to be included as well , see eq . ( 17 ) in @xcite . the function @xmath33 is plotted in fig . [ f : g_e ] . we see it decreases rather fast at @xmath46 kev . at a point @xmath47 , @xmath48 where @xmath49 is the separation between the stars . then , assuming the velocity profile of equation ( [ velocity ] ) , we need to solve the energy equation , which , for an ionized gas , can be written as , @xmath50 = c_{\rm x}+c_{\rm b , l}+c_{\rm c } , \label{energy}\ ] ] this can be solved for @xmath51 adopting a boundary condition close to the stellar surface , where the wind is optically thick and our assumption of optical thinness breaks down . in an optically thick wind , @xcite find @xmath52 . we have found that the solution above the assumed boundary , @xmath53 , is virtually independent of the value of @xmath54 , and we adopt @xmath55@xmath56 . at low radii , @xmath57 , and cooler particles are brought to a given point from radii closer to the star . at all radii , adiabatic expansion cools the wind . thus , the dominant effect of the left - hand side of equation ( [ energy ] ) is cooling . we have found that solutions of this equation are rather well approximated by replacing the differential terms ( i.e. , its left - hand side ) by an algebraic advection cooling term , @xmath58 , see , e.g. , @xcite for a similar approach . we have found that we achieve the highest accuracy in this approach defining @xmath59 as , @xmath60 we thus solve an algebraic equation for @xmath51 , @xmath61 see section [ cygx1 ] for an example comparison of the two solutions . , by which the compton heating of electrons with @xmath27 by photons with energy @xmath62 is reduced due to the klein - nishina effects . , width=245 ] . the point of radio emission is at the height @xmath63 along the jet ( shown by the red line ) , which originates at the compact - object location . a point along the photon path through the wind towards the observed is at the distance , @xmath64 , from the emission point , at the distance , @xmath3 , from the centre of the donor , and at the distance , @xmath17 , from the compact object , which is surrounded by the source of x - rays . , width=238 ] we then assume the bulk of the radio emission at a given @xmath65 occurs at a distance , @xmath63 , along a jet from its origin at the compact object , see fig . [ f : geo ] . this @xmath63 corresponds to the place along the jet where it becomes optically thin to synchrotron self - absorption ( @xcite , hereafter bk79 ) . the jet is assumed to be perpendicular to the circular orbital plane , see fig . [ f : geo ] . we note that the assumption of perpendicularity is not a crucial ; the jet can be inclined ( as suggested for cyg x-1 by @xcite ) or curved ( as in the cyg x-1 model of sz07 ) , which introduces a phase lag to an orbital modulation . then , @xmath66 is to be interpreted as the height above the orbital plane in units of the distance to the rotation axis of the donor . in this case , @xmath49 would be somewhat higher than the orbital separation . we neglect here this complication . we also neglect any effects of the jet opening angle , as a second - order effect . in the case of cyg x-1 , it has been constrained to be @xmath67 ( s01 ) , which is negligibly small for our calculations . as shown in fig . [ f : geo ] , @xmath64 is the path length from the emission location along the jet to a point at a radius , @xmath3 , from the donor centre . we find , @xmath68 , \label{r2}\ ] ] where @xmath69 is the binary inclination and @xmath70 is the orbital phase , which equals @xmath71 , @xmath72 at the superior and inferior conjunction , respectively . then , @xmath73 the free - free absorption coefficient at a frequency , @xmath65 , is , @xmath74 where @xmath21 is in units of k and @xmath65 is in ghz . at a reference temperature , @xmath75 , we can write @xmath76 , where @xmath77 is the absorption coefficient at an ion density of @xmath78 . we define @xmath78 as the density at @xmath79 under the assumption of @xmath80 . we also define a normalization optical depth of @xmath81 . then , the free - free optical depth is , @xmath82^{-3/2 } { \rm d}(l / a ) . \label{tau_v}\ ] ] this can be calculated numerically with @xmath51 as the solution of either equation ( [ energy ] ) or ( [ t_eq ] ) . on the other hand , we also consider the case of an isothermal wind with constant velocity ( which is approached at at @xmath83 ) . in that case , @xmath84 in this approximation , we obtain an analytical formula for @xmath85 , @xmath86 we find @xmath87 everywhere except for @xmath88 , @xmath89 , when @xmath90 and equation ( [ tau_ff ] ) appears to have a singularity . however , @xmath85 is still finite there , @xmath91 . for @xmath92 , @xmath93 where @xmath94 is the difference in the optical depths at the superior and inferior conjunctions . for large @xmath63 , the leading order of @xmath85 is independent of @xmath70 , @xmath95 , \quad z / a\gg 1 , \label{tau_asymp}\ ] ] and @xmath94 is lower by one order of @xmath96 , @xmath97,\quad z / a\gg 1 . \label{dtau_asymp}\ ] ] potential constraints from eclipses of the radio emission from both the jet and counter - jet are discussed in appendix [ eclipses ] . appendix [ scattering ] presents results for compton scattering ( or another process with the cross section proportional to the density ) analogous to the above ones for free - free absorption . we define the full modulation depth due to absorption as , @xmath98 where @xmath99 and @xmath100 are the maximum and minimum fluxes in the model averaged and folded light curves , observed at @xmath101 and @xmath102 , respectively , and @xmath102 represents the possible phase lag . in our model , we note that @xmath104 ( or , equivalently , @xmath105 ) in the model is uniquely determined by @xmath66 , @xmath69 and @xmath106 . the attenuation averaged over the orbit and the average @xmath85 are defined as , @xmath107 where the integration can be done over only 0@xmath72 due to the planar symmetry . as a rough approximation for @xmath108 , @xmath109 may be used . we note that @xmath104 , @xmath110 , @xmath108 , @xmath102 are functions of frequency . the ephemeris of cyg x-1 has been determined , e.g. , by @xcite , @xcite . it can be written as , @xmath111=50077.995 + p e,\ ] ] where @xmath112 d , @xmath113 is the time of a superior conjunction ( the black hole furthest from the observer ) , and @xmath62 is an integer . observational determinations of the mass of the donor , @xmath114 , and of the black hole , @xmath115 , @xmath4 , @xmath69 , and the distance , @xmath116 , are all mutually connected . we use here the constraints of @xcite , and of the evolutionary models of @xcite . @xcite obtained the most likely ranges of @xmath117@xmath118 , @xmath119@xmath120 , @xmath121@xmath122 , @xmath123@xmath124 , @xmath1252.0 ) kpc . they also measured the donor temperature as @xmath126 k , and derived , @xmath127 , \label{r1_i}\ ] ] where @xmath116 is in kpc . here , we choose their solution that also agrees with the detailed evolutionary modelling of @xcite , namely , @xmath128 , @xmath129 kpc , @xmath130 , @xmath131 , @xmath132 , which yields @xmath133 kpc ( with @xmath134 uncertainties ) , which value is compatible with adopted here within @xmath135 . then , @xcite determined @xmath136 , again within @xmath134 from our value . they also found @xmath137 , @xmath138 , and @xmath139 . the effect of these difference on our results have been found to be relatively minor , and less important than the uncertainty regarding the structure of the stellar wind . ] the eccentricity is @xmath140 @xcite , and we here assume the orbit to be circular . the total mass loss rate in the hard spectral state was found by @xcite as @xmath141 yr@xmath20 . however , the mass loss can be highly asymmetric in binary systems , especially those in which the donor is close to filling its roche lobe . this is the case in cyg x-1 , where @xmath142 of the roche - lobe radius @xcite . in this situation , the wind is focused towards the black hole @xcite , with the wind density inside the roche lobe being substantially higher than that outside it . the presence of the focused wind component was found to fit very well the he ii @xmath1434686 emission profile of cyg x-1 @xcite . evidence for the focused component based on soft x - ray lines was also found by @xcite and @xcite . further complexity of the wind geometry was found by @xcite . namely , they found the unfocused , locally spherically symmetric wind component emitted by the x - ray irradiated hemisphere of the donor to be undetectable . this weakness is apparently due to wind ionization by x - rays , which strongly reduces the wind radiative acceleration by the lines emitted by the donor . on the other hand , the other hemisphere of the donor , shaded from the x - rays , emits a normal wind , similar to that of an isolated massive star @xcite . summarizing , the wind of hde 226868 has three main components : the wind focused along the axis joining the stars , which carries most of the mass loss ; the standard radiatively driven wind from the hemisphere in the shadow , and a weak wind component from the donor hemisphere facing the black hole . as found by @xcite , this third component is weak in both spectral states , hard and soft , of cyg x-1 . radio photons are emitted by the jet relatively high above the orbital plane and may be attenuated mainly by this third weak component , i.e. , the quasi - spherical wind from the hemisphere facing the x - ray source . given its weakness , it has @xmath144 , but no specific constraints are currently available . for the wind velocity profile , we use here that of equation ( [ velocity ] ) with @xmath145 cm s@xmath20 and @xmath146 @xcite . the stellar luminosity for our assumed parameters is @xmath147 erg s@xmath20 . we then use the average hard x - ray / soft @xmath148-ray spectrum of cyg x-1 as observed by the osse and batse detectors on board of _ compton gamma ray observatory _ @xcite . we fit this spectrum with an e - folded power law , @xmath149 , obtaining @xmath150 cm s@xmath20 ( normalized to the osse flux ) , @xmath151 , @xmath152 kev , yielding @xmath153 erg s@xmath20 . the fit is based on the data at @xmath154 kev , and it quite severely underestimates the actual spectrum in softer x - rays . however , the main contribution to compton heating is from hard x - rays , see equation ( [ compton2 ] ) , and thus this underestimate has a negligible effect . we note that the klein - nishina correction is very important for this spectrum , with the compton heating rate calculated non - relativistically , with @xmath155 , being 2.7 times higher than the correct rate calculated using @xmath33 of equation ( [ g_e ] ) . on the other hand , photoionization is mostly due to soft x - rays , which spectrum in cyg x-1 is relatively uncertain due to absorption . we assume here @xmath156 erg s@xmath20 in equation ( [ xi ] ) . then , we can express the characteristic free - free optical depth of the wind , @xmath106 in equation ( [ tau_v ] ) , as , @xmath157 where we used @xmath158 k as the reference temperature . given the lack of a measurement of the wind in the polar regions @xcite , we treat @xmath8 as a free parameter of the model . we study here radio monitoring data at 15 ghz from the ryle telescope , which cover mjd 5022653902 , and the arcminute microkelvin imager ( ami ) , mjd 5457355540 . also , we use the 2.25 ghz and 8.3 ghz monitoring data from the green bank interferometer , for mjd 5040951823 . these data are the same as in paper i , and we refer to it for details . we fit only data corresponding to the hard spectral state , which corresponds to mjd 5035050590 , 5066050995 , 5102551400 , 5164051840 , 5196052100 , 5256552770 , 5288052975 , 5311553174 , 5354053800 , 5388055375 ( which intervals contain 12408 ryle / ami measurements and 1160 gbi ones ) , see @xcite and paper i. an important issue here is that these telescopes do not spatially resolve the radio emission . on the other hand , the emission in the hard state is known to be extended . s01 have found that @xmath1591/2 of emission at 8.4 ghz observed by vlba is within the beam of a longitudinal extent of @xmath160 , whereas the remainder forms an extended jet visible up to @xmath161 . this has been confirmed by further vlba observations at 8.4 ghz reported in @xcite , in which the resolved fraction was @xmath1620.4 . the resolution of @xmath163 corresponds to the length of @xmath164 cm , which is @xmath165 . the resolved 8 ghz emission can thus have only a tiny orbital modulation . the observational situation at 15 ghz is more complex . the vlba beam size of the observations reported in @xcite is @xmath166 , which corresponds to @xmath167 cm @xmath168 . the unresolved and total fluxes for the three observations taken on the same days as the 8.4 ghz ones of s01 ( i.e. , mjd 51035 , 51037 , 51039 ) are ( 7.3 , 7.7 ) mjy , ( 10.9 , 12.2 ) mjy , ( 6.1 , 7.8 ) mjy . thus , the unresolved fractions are relatively small , 0.050.22 . however , cyg x-1 was also monitored by the ryle telescope three times on each of those days , yielding the total fluxes of ( 13.2 , 14.9 , 16.0 mjy ) , ( 13.2 , 12.2 , 10.4 mjy ) , ( 18.9 , 19.4 , 18.1 mjy ) , respectively , i.e. , substantially more than the total vlba fluxes for the 1st and 3rd observation . although all of the ryle observations were taken @xmath169 d before the start of the vlba observations , cyg x-1 was in a relatively stable state at that time with no large - amplitude variability detected . it thus appears that the vlba measurement missed an extended part of the resolved emission . then , the resolved fractions are @xmath1700.7 . as discussed in @xcite , it is unlikely that the radio jet in cyg x-1 follows a single set of power - law dependencies from its base to the resolved parts . it appears that it contains a compact part within @xmath171@xmath172 , and then a secondary dissipation event leads to formation of a more extended part of the jet , resolved by vlba . the observed strong orbital modulation ( see below ) has to occur in the compact part . the division of the fluxes between the two parts as a function of time remains unknown . based on the above data , we can only roughly estimate that about a half of the flux is emitted by the compact part . given that the radio - ir spectrum of cyg x-1 appears to be a single power law @xcite , both components have approximately the same spectra . with the model developed in section [ wind ] , we can fit the orbital modulation profiles , which are given by the light curves averaged and folded over the orbital period . in this way , we obtain the location of the bulk of the emission , @xmath66 . we neglect the emission of the counter - jet , as its emission is beamed away from the observer and it passes through a much higher column density than that of the jet , including the dense focused component close to the binary plane . given the results discussed in section [ radio ] , we split the radio light curves into unmodulated and modulated parts , @xmath173 where @xmath174 it time . the second term on the right - hand side corresponds to emission of the inner part of the jet , within @xmath175 or so , which is orbitally modulated . the first term corresponds to the larger - scale emission , which is not modulated . then , @xmath176 , \label{fintr}\ ] ] where @xmath177 , @xmath178 denote the light curves averaged over time intervals of the length @xmath179 . given that the observed emission comes from a region of the size @xmath180 cm ( s01 ; @xcite ) and a relativistic speed of the jet ( s01 ; @xcite ) , we assume that both parts of @xmath181 vary in the same way ( apart of the orbital modulation ) on a @xmath182-s time scale , below which scale the radio variability is weak . equation ( [ fintr ] ) implies the average attenuation of @xmath183 the modulation depth becomes , @xmath184 where @xmath185 is the minimum optical depth through the wind ( corresponding to the maximum flux , at @xmath101 ) . in the calculations below , we assume either @xmath186 ( see section [ radio ] ) or @xmath187 , which provide us with an estimate of the sensitivity of the calculations to taking into account the unmodulated part of the flux . the assumed @xmath186 is , given the available data , only approximate , as well as it is unlikely that it was constant during the course of the monitoring . in figs . [ f : orb_mod][f : orb_mod15_var ] below , we show the results for @xmath187 , which directly correspond to the observations of the modulation . after obtaining folded and averaged light curves of @xmath188 , we subtract from them @xmath189 , as implied by equation ( [ split ] ) , and then recalculate the folded average modulation . we need to properly average the folded light curve . we are interested in variability governed by the orbital time scale , and not in that on much shorter time scales . therefore , we rebin the observed light curve with a bin size equal @xmath190 , where @xmath191 is the number of bins per period . in this way , we also partly remove a bias on the folded averages due to non - uniform coverage , e.g. , due to a large numbers of points in some orbital bins and low number in other bins . the errors on the resulting averages are then calculated from dispersion of fluxes averaged within individual bins of the same phase range and the size @xmath192 . this method was applied to periodic variability of cyg x-1 by @xcite . we fit @xmath193 , where @xmath194 is the phase bin width , instead of using the value at a bin centre . also , finite phase lags with respect to the superior conjunction are observed in cyg x-1 . then , the phase lag is a free parameter of the model , and @xmath195 of equations ( [ tau_v][tau_ff ] ) is replaced by @xmath196 . we first compare our results with those of sz07 ( whose results are for @xmath187 ) . in that work , the atomic heating and cooling was treated more accurately than in this work . on the other hand , adiabatic cooling was not included , which resulted in a constant temperature , @xmath197 , at large radii . thus , we first compare their results with those using our analytical , constant @xmath21 , model , equation ( [ tau_ff ] ) . using the parameters from sz07 , namely , @xmath198 , @xmath199 k and @xmath200 , we obtain @xmath201 , fitted to the 15 ghz data , whereas sz07 found @xmath202 ; thus there is a good agreement . on the other hand , we find @xmath203 using the full model of equation ( [ tau_v ] ) . the higher value found by us is due to our inclusion of adiabatic cooling , which reduces the temperature at large radii and thus increases the wind opacity there , requiring in turn the emission place at a higher height . however , although sz07 used the same total @xmath9 as that used here , they assumed isotropy , without taking into account focusing and the reduction of the wind from the irradiated surface of the donor . this resulted in a relatively high density of the wind . as a result , the 15 ghz emission of cyg x-1 was found to be attenuated to @xmath204 , see fig . 13 in sz07 . as also shown there , the average attenuation decreased with decreasing frequency at @xmath205 ghz , to @xmath206 at 2.25 ghz , as well as also decreased with increasing frequency at @xmath207 ghz . however , as argued in paper i , this is not compatible with the 2220 ghz spectrum being rather straight with @xmath208 @xcite , and with no substantial dip at 15 ghz , see also figs . 1011 in paper i. this supports the wind density in the polar regions being substantially lower than that for an isotropic wind , see section [ pars ] . thus , the free - free optical depth of the wind is most likely substantially lower than that modelled by sz07 . furthermore , the model of sz07 implies a very strong orbital modulation , by @xmath20990 per cent in the 150220 ghz range . this range was observed by @xcite , and no apparent orbital modulation was seen . in particular , there was one 146 ghz observation at @xmath210 , at which the measured flux was about equal to the average one . ) orbital modulation ( data points with errros ) at ( a ) 15 ghz , ( b ) 8.3 ghz , ( c ) 2.25 ghz fitted by the model ( solid curve ) of equation ( [ tau_v ] ) with @xmath128 . the flux averages are @xmath211 mjy , 14.2 mjy , 14.5 mjy , respectively . , width=279 ] ) orbital modulation ( data points with errros ) at ( a ) 15 ghz , ( b ) 8.3 ghz , ( c ) 2.25 ghz fitted by the model ( solid curve ) of equation ( [ tau_v ] ) with @xmath128 . the flux averages are @xmath211 mjy , 14.2 mjy , 14.5 mjy , respectively . , width=279 ] ) orbital modulation ( data points with errros ) at ( a ) 15 ghz , ( b ) 8.3 ghz , ( c ) 2.25 ghz fitted by the model ( solid curve ) of equation ( [ tau_v ] ) with @xmath128 . the flux averages are @xmath211 mjy , 14.2 mjy , 14.5 mjy , respectively . , width=279 ] , @xmath212 , @xmath128 ) . the solid , dashed and dotted curves correspond to the orbital phase of @xmath213 , 0.5 and 1 , respectively . , width=279 ] , @xmath212 , @xmath128 ) at the direction of @xmath214 , @xmath215 . the dashed and dotted curves correspond to the solution of the differential energy equation ( [ energy ] ) with the boundary condition imposed at @xmath216 and @xmath217 , respectively , and the solid curve corresponds to the algebraic solution of equation ( [ t_eq ] ) . , width=279 ] ) orbital modulation at 15 ghz fitted by the model of equation ( [ tau_v ] ) with @xmath128 for the fluxes ( a ) @xmath218 and ( b ) @xmath219 , where @xmath220 mjy is the average for the entire sample . , width=279 ] ) orbital modulation at 15 ghz fitted by the model of equation ( [ tau_v ] ) with @xmath128 for the fluxes ( a ) @xmath218 and ( b ) @xmath219 , where @xmath220 mjy is the average for the entire sample . , width=279 ] ( in units of @xmath49 ) on the 15 ghz flux , for the case @xmath186 , @xmath221 . the line shows the best fit . , width=264 ] our best data are those at 15 ghz , and we fit them with the model of equation ( [ tau_v ] ) . we first give the results neglecting the extended character of the radio emission , i.e. , for @xmath187 . we obtain @xmath222 , @xmath223 , @xmath224 , at @xmath225 . [ f : orb_mod](a ) presents a fit of this model to the folded and averaged light curve . the model modulation depth is @xmath226 , and the orbit - averaged attenuation is @xmath227 . the values of @xmath228 and @xmath104 are very similar to those obtained by @xcite , who used a smaller data set and a different model . [ f : dtau ] shows variation of the absorption coefficient along the photon path , for our best - fit model . we see the opacity drops fast at @xmath229 . the region @xmath230 is at radii of @xmath231 . [ f : radial ] shows an example of the radial temperature profile for the best - fit model . we see that indeed the solution is virtually independent of the adopted inner boundary condition and the differences between the solution of the differential energy equation and the approximate one with the advection term are relatively small , @xmath232 per cent . we then consider the effect of taking into account the effect of a part of the radio flux being emitted well beyond the orbit , assuming @xmath186 . we obtain the same values for @xmath8 and @xmath102 as for @xmath187 , and @xmath233 at @xmath234 , and with @xmath235 , @xmath236 . we note , however , that the obtained @xmath63 is relatively close to the binary plane , which may no more be in the polar region , and thus not to be consistent with the fitted low value of wind - density scaling factor , @xmath8 . thus , we also consider models with fixed larger values of @xmath8 , but still @xmath140 . at @xmath221 , @xmath237 , @xmath226 , @xmath238 , @xmath239 , and at @xmath240 , @xmath241 , @xmath242 , @xmath243 , @xmath244 . since @xmath245 in both cases , the emission region is likely to be in the low - density polar region . although these values of @xmath246 are rather high , the bad fit quality may be due to our lack of knowledge of the actual structure of the stellar wind and of the spatial distribution of the radio emission along the jet . hereafter , we adopt @xmath221 , where both @xmath144 and @xmath245 , but also use @xmath240 for comparison in some cases . based on general features of the jet emission , e.g. , bk79 , @xcite , we expect that the height at which the bulk of emission at a given frequency emerges increases with increasing radio flux . this is due to the associated increase of the density of the relativistic electrons in the jet , which increases the synchrotron self - absorption optical depth . assuming the magnetic field strength to be proportional to the equipartition value with the relativistic electrons , we derive ( using the formalism of @xcite ) , @xmath247 where @xmath248 is the index of the electron distribution ( @xmath249 , where @xmath148 is the lorentz factor ) , which gives @xmath250 for @xmath251 . [ equation ( 5 ) of @xcite gives this result for @xmath252 . ] the increased height of the point where the bulk of emission occurs results in less wind absorption and a lower modulation depth . also , a higher x - ray flux ( which is positively correlated with the radio flux , e.g. , paper i ) leads to a higher temperature , equation ( [ t_eq ] ) , and thus less wind absorption of radio photons . given the good statistical quality of the 15 ghz light curve , we can test the above theoretical prediction with the data . for this purpose , we have divided the observed range of the 15 ghz flux in the hard state , @xmath253 , into three parts , satisfying ( 1 ) @xmath254 , ( 3 ) @xmath255 , and ( 2 ) that in between . as the criterion , we have used the radio fluxes averaged over a single orbital period , @xmath256 , instead of individual measurements , @xmath253 , since the latter are themselves affected by the orbital modulation . the unweighted geometric average fluxes in these ranges , @xmath257 , 2 and 3 are @xmath258 mjy , 11.6 mjy and 15.6 mjy , respectively . in order to account for a part of the radio flux emitted at large distances , equation ( [ split ] ) , we subtract from each of the three folded light curves a part of the local average , i.e. , @xmath259 . we take into account that the plasma temperature of the absorbing wind changes in response to the changing flux within the hard state . given that the x - ray flux in the hard state of cyg x-1 is correlated with the 15 ghz flux as @xmath260 , we adjust accordingly the normalization of @xmath261 and @xmath18 . then , fitting the model of equation ( [ tau_ff ] ) for @xmath187 , @xmath128 , @xmath262 , we have obtained the location of the emission at @xmath263 , @xmath264 , @xmath265 , which imply @xmath266 , @xmath267 and @xmath268 , and @xmath269 , @xmath270 , @xmath271 , for @xmath257 , 2 , 3 , respectively . the folded light curves and the fits are shown in fig . [ f : orb_mod15_var ] . at @xmath186 , @xmath221 , @xmath272 , @xmath273 , @xmath274 , which imply @xmath275 , @xmath270 and @xmath271 , and @xmath276 , @xmath277 , @xmath278 , for @xmath257 , 2 , 3 , respectively . this @xmath279 dependence is shown in fig . [ f : z_f ] . we fit it as a power law , @xmath280 , and obtain @xmath281 , i.e. , the statistical significance of an increase of @xmath63 with @xmath253 is @xmath282 . we see that the orbital modulation depth strongly decreases with increasing @xmath253 , also at a high statistical significance . thus , these calculations confirm the dependence of the radio emission region along the jet on the jet power in relativistic electrons with a high statistical significance . as discussed above , we assume that the long term variability occurs in the same way for the modulated and unmodulated parts of the radio flux . we can then write the dependence of equation ( [ zf ] ) for the intrinsic flux , equation ( [ fintr ] ) , and estimate , @xmath283 where @xmath284 , @xmath285 $ ] , and @xmath286 we obtain @xmath287 , @xmath288 for ( @xmath187 , @xmath212 ) , ( @xmath186 , @xmath221 ) , respectively . in the latter case , we also fit the @xmath289 dependence , and obtain @xmath290 , i.e. , an almost identical value to that of equation ( [ q ] ) . these values are somewhat higher than the theoretical prediction of equation ( [ zf ] ) . this may reflect either the approximate character of the assumed disc model of bk79 , the fitted model being too approximate , or both . we then fit the 8.3 ghz and 2.25 ghz data , see fig . [ f : orb_mod](b c ) . the obtained parameters at ( @xmath187 , @xmath212 ) are @xmath291 , @xmath292 ( @xmath293 , 3.9/9 ) , which result in @xmath294 , @xmath295 , and @xmath296 , @xmath297 , respectively . the phase lags are @xmath298 , @xmath299 , respectively . at ( @xmath186 , @xmath221 ) , @xmath300 , @xmath301 ( @xmath302 , 7.6/9 ) , which result in @xmath294 , @xmath295 , and @xmath303 , @xmath304 , respectively . ( at @xmath240 , @xmath305 , @xmath306 , respectively . ) the phase lags are unchanged . these data cover a much shorter interval in the hard state than the ryle / ami 15 ghz data do , and have been taken with a substantially lower sensitivity . thus , the resulting folded light curves are much more noisy , especially that for 2.25 ghz , which has the parameters only weakly constrained . fitting in the log space at 2.25 ghz , 8.3 ghz and 15 ghz with @xmath307 , we find @xmath308 , @xmath309 for @xmath221 , 0.2 , respectively . this is somewhat more than the prediction of the model of bk79 , which may be due to either the departures of the jet in cyg x-1 from that model and/or our fitting model being too approximate . cyg x-1 in all spectral states shows a power - law correlation between the radio and x - ray fluxes , @xmath310 , where @xmath311 is either the flux in an x - ray energy interval , the bolometric flux , @xmath312 , or the flux in the comptonization x - ray component , see paper i. in particular , the correlation of the 15 ghz flux with @xmath312 in the hard state has the index of @xmath313 for the simultaneous asm and bat data . qualitatively , the hard - state correlation in cyg x-1 is similar to that seen in the hard state of low - mass x - ray binaries ( lmxbs ) containing black - holes , see , e.g. , @xcite , @xcite , @xcite . however , the correlation in cyg x-1 is significantly steeper than that in lmxbs , e.g. , @xmath314 in gx 3394 @xcite . a part or whole of this difference is due to the effect of free - free absorption of the radio emission in the stellar wind . an observed correlation index , @xmath315 , where @xmath311 is either the flux in an x - ray band , the bolometric flux or the flux emitted by hot electrons ( see paper i ) , is related to its corresponding intrinsic index , @xmath316 , by @xmath317 \over { \rm d}\ln f}<1 . \label{index}\ ] ] so , the effect of removing free - free absorption of the radio fluxes is to reduce their variability range , without affecting the variability range of the x - ray flux . note that the factor @xmath3 is the same regardless of the choice of the energy range for the x - ray flux . the coefficient @xmath3 can be determined from the results shown in figs . [ f : orb_mod15_var][f : z_f ] . we estimate it as @xmath318/\delta \ln f$ ] . we obtain @xmath319 , @xmath320 for ( @xmath187 , @xmath212 ) , ( @xmath186 , @xmath221 ) , respectively . this implies @xmath321 , @xmath322 , respectively , for the correlation with the bolometric flux . we note that although the wind absorption is significantly stronger for the @xmath186 case , it acts only on a 1/2 of the radio flux . we can also directly calculate the effect of the wind absorption on an individual measurement of the 15 ghz flux , @xmath253 , performed at an orbital phase , @xmath70 . assuming @xmath8 and @xmath69 and using the dependence of @xmath289 of equation ( [ z_f ] ) allows us to write @xmath181 as , @xmath323\right\ } = f_0\left(z \over z_0\right)^{1/q } , \label{f_intr}\ ] ] which we numerically solve for @xmath63 , which yields @xmath181 . the resulting @xmath181 points for @xmath186 , @xmath221 are shown together with the 39 kev fluxes in fig . [ f : correlation](a ) . [ f : correlation](a ) also compares the radio / x - ray correlation for cyg x-1 with that in three lmxbs ( from @xcite ) . this figure is similar to one shown in paper i , which , however , showed the cyg x-1 radio fluxes as observed . we see that the cyg x-1 points have now the slope of @xmath324 , relatively similar to that of the lmxb h1743322 @xcite . the 8.3 ghz and 2.25 ghz data do not allow us to directly make similar estimates , given the poor statistics of the folded light curves , see fig . [ f : orb_mod](b c ) . since the radio spectrum is approximately flat , @xmath325 @xcite , we assume that the @xmath326 estimated for 15 ghz is accompanied by changes of @xmath66 at the other frequencies proportional to those inferred at 15 ghz , i.e. , we use equation ( [ z_f ] ) with the value of @xmath327 obtained for 15 ghz , but @xmath328 and @xmath329 are from the fits to the 2.25 ghz and 8.3 ghz data . figs . [ f : correlation](b c ) show the resulting @xmath181 points for @xmath186 , @xmath221 . given the strong noise in those data , it is difficult to accurately determine the slope formed by those points , but it appears similar to that found for the 15 ghz data . however , given the observational uncertainty regarding the actual wind density in the polar regions and possible systematic uncertainties of the fitted model , the case with a higher wind density is also possible , as discussed in section [ cygx1 ] . for example , for ( @xmath186 , @xmath240 ) , @xmath330 at 15 ghz , which corresponds to @xmath331 . this would make cyg x-1 similar to objects in the main branch of the radio / x - ray correlation , e.g. , gx 3394 . thus , our results do not allow us to unambiguously determine the slope of the intrinsic radio / x - ray correlation in cyg x-1 . on the other hand , if we assume that cyg x-1 in not a unique object , the above constrains the wind density in the polar region to @xmath332 . our results for the 2.2515 ghz data allow us to make some predictions for the orbital modulation and attenuation by the wind of jet emission at other frequencies , especially higher ones . as estimated in section [ correlation ] , given the implied low values of @xmath66 for high - frequency radio emission , it may originate within the dense , focused , wind , which would result in much higher wind density than in the polar regions , and hence much higher depths of orbital modulation than those we estimate here . , solid curve ) and the average attenuation ( dashed curve ) as a function of frequency predicted by the model fitted to the 15 ghz data with @xmath186 , @xmath240 , @xmath128 , and @xmath335 based on the 2.2515 ghz data . , width=257 ] still , including the focused wind would introduce additional free parameters to the model , in particular , the photon path going through different wind components . instead , we assume @xmath240 , which , as discussed in sections [ cygx1][correlation ] , approximately corresponds to the highest density of the polar wind compatible with the observational constraints . our predictions are shown in fig . [ f : d_a_vs_nu ] . it shows the modulation depth ( plotted as @xmath336 ) and the average attenuation as functions of frequency . we see that the maximum modulation depth occurs for @xmath337 ghz , and it is relatively small at higher frequencies . also , the attenuation is relatively moderate , @xmath338 , at all frequencies . we see that we predict much lower modulation depths and attenuation than sz07 . the cause for these differences is that our models yield a much lower wind density in the absorbing region than that assumed by sz07 , which is supported by both the current observational results @xcite , as well as the results of our study of the radio / x - ray correlation in cyg x-1 . our results may explain the lack of observed modulation at 146 ghz and the 2220 ghz spectrum being approximately a single power law @xcite . still , given the uncertainty related to the effect of the focused wind , it would be of high importance to observationally study the orbital modulation of high - frequency radiation . we have studied free - free absorption in an optically - thin x - ray irradiated stellar wind . we have taken into account compton heating and cooling , photoionization heating , bremsstrahlung and line cooling , and advection . we have derived relatively simple formulae for the free - free optical depth , in section [ wind ] . we have also considered constraints from eclipses , counter - jets ( appendix [ eclipses ] ) and compton scattering ( appendix [ scattering ] ) . we have then considered the current constraints on the parameters of the cyg x-1 binary and its stellar wind ( section [ pars ] ) . as found by @xcite , the wind is strongly anisotropic , with the density close to the orbital plane and in the region shadowed by the donor much higher than that in the polar regions , which are responsible for absorption of the radio emission of the jet . in section [ radio ] , we have discussed the radio data , including the available information about the extended character of the emission . it appears that a relatively high fraction of the radio emission is emitted far away from the binary orbit , and thus it is not orbitally modulated by the wind . this effect needs to be taken into account in modelling of the modulation . in section [ cygx1 ] , we have applied the formalism of section [ wind ] to the available radio data from monitoring of cyg x-1 , taking into account the extended character of the emission . in the hard spectral state , the radio light curves show significant modulation at the orbital period , which is due free - free absorption in the wind . we have fitted the folded and averaged light curves at 15 ghz , 8.3 ghz and 2.25 ghz , and obtained the location of the bulk of the emission regions along the jet . in the units of the separation , our best models gave @xmath3391.2 , 1.82.8 , 6.28.8 , respectively . the height at 15 ghz corresponds to @xmath340 , and that at 2.25 ghz is about 10 times higher . these locations , corresponding to the heights at which the jet becomes optically thin to free - free absorption , appear to follow a law relatively close to @xmath341 , expected theoretically ( bk79 ) . on the other hand , @xcite in his estimates of the power of the jet in cyg x-1 based on the model of bk79 used the vlba images of s01 , which imply that 50 per cent of the 8.4 ghz emission comes from distances @xmath342 cm , which is @xmath343 . if the entire jet would follow the emission profile of bk79 , only a tiny fraction of the emission would take place at @xmath245 ( see fig . 1 of @xcite ) , and the resulting orbital modulation would be tiny and probably unobservable , whereas it is at the level @xmath20910 per cent . given that the distance of the emission should scale as @xmath344 ( bk79 ) , it would have been even more difficult to explain the @xmath20930 per cent modulation of the 15 ghz emission . a likely solution to this problem is that the observed resolved emission comes from a secondary dissipation event in the jet whereas the emission of the inner jet , at @xmath245 , is comparable to the resolved one , and it may follow the model of bk79 . this would also solve the problem of that jet power calculated by @xcite being much lower than that inferred from the observed large - scale optical nebula , presumably powered by the jet @xcite . furthermore , this would solve the problem of the jet base inferred from the vlba measurements being an order of magnitude higher than that theoretically expected @xcite . recently , @xcite have calculated the location the jet base in cyg x-1 ( defined by the lowest distance at which significant emission takes place ) at @xmath345 . using the measurement of the turnover frequency of @xmath170 ev of @xcite , and assuming @xmath341 ( bk79 ) , gives a location of the 15 ghz emission in agreement with our results . our best data , at 15 ghz , allow us to study the orbital modulation as a function of the radio flux . we have found a strong dependence of the modulation depth , changing from @xmath346 to @xmath271 when the flux increases by a factor of about 2 . this is naturally interpreted as the height corresponding to the bulk of the emission increasing with the radio flux as a consequence of the corresponding increase of the jet power . in our case , we find @xmath63 approximately proportional to the power of about @xmath347 of the intrinsic radio flux ( i.e. , that before free - free absorption ) , which dependence is somewhat stronger than the relation in the bk79 model . based on the above results , we have estimated the effect of free - free absorption on the form of the hard - state radio / x - ray correlation in cyg x-1 , which correlation was the subject of paper i. we find the absorption has a substantial effect , but , given the uncertainty about the wind density in the polar regions of cyg x-1 , we can not accurately quantify this effect . this effect changes the correlation index , @xmath348 ( defined as @xmath349 ) , from about 1.7 , found in paper i , to about 0.71.3 . thus , cyg x-1 may have a similar correlation to that of the lmxb h1743322 , which has @xmath350 @xcite , or a number of other lmxbs , e.g. , gx 3394 @xcite and v404 cyg @xcite , where the correlation index is @xmath3510.8 . we also discuss predictions for the orbital modulation and attenuation by free - free absorption of radio emission at frequencies higher than 15 ghz . given the relatively low density of the polar wind in cyg x-1 , these effects are likely to be much weaker than estimated before ( in sz07 ) , which may explain the apparent lack of orbital modulation at @xmath352 ghz , and the 2220 ghz radio spectrum of cyg x-1 being represented by a single power law @xcite . still , it would be highly desirable to actually measure the modulation depth at high frequency , which would give us significant constraints on the structure of both the wind and the jet in cyg x-1 . in particular , this would test the possible location of the high - frequency radio emission within the focused wind component . finally , we mention that another high - mass x - ray binary with strong wind , cyg x-3 ( which is likely to contain a black - hole ) does not show any orbital modulation of its very strong radio emission . as discussed by sz07 , this may be due to the binary orbit of that system being @xmath20910 times smaller than that in cyg x-1 . if the jet of cyg x-3 is similar in size to that of cyg x-1 , this would imply values of @xmath66 @xmath20910 times higher , which is likely to reduce the depth of the radio modulation to unobservable values . thanks are due to f. yuan for very valuable suggestions . i also thank j. mikoajewska and j. zikowski for valuable discussions , and the referee for valuable comments . this research has been supported in part by the polish ncn grants n n203 581240 , n n203 404939 , and 362/1/n - integral/2008/09/0 . the ami arrays are operated by the university of cambridge and supported by the stfc . the green bank interferometer is a facility of the national science foundation operated by the nrao in support of nasa high energy astrophysics programs . beloborodov a. m. , illarionov a. f. , 1995 , apj , 450 , 64 blandford r. d. , knigl a. , 1979 , apj , 232 , 34 ( bk79 ) blondin j. m. , 1994 , apj , 435 , 756 brocksopp c. , tarasov a. e. , lyuty v. m. , roche o. , 1999 , a&a , 343 , 861 caballero - nieves s. m. , et al . , 2009 , apj , 701 , 1895 carciofi a. c. , bjorkman , j. e. , 2006 , apj , 639 , 1081 chen x. , abramowicz m. a. , lasota j .- , 1997 , apj , 476 , 61 corbel s. , fender r. p. , tzioumis a. k. , nowak m. , mcintyre v. , durouchoux p. , sood r. , 2000 , a&a , 359 , 251 corbel s. , nowak m. a. , fender r. p. , tzioumis a. k. , markoff s. , 2003 , a&a , 400 , 1007 corbel s. , fender r. p. , tomsick j. a. , tzioumis a. k. , tingay s. , 2004 , apj , 617 , 1272 corbel s. , koerding e. , kaaret p. , 2008 , mnras , 389 , 1697 coriat m. , et al . , 2011 , mnras , 414 , 677 fender , r. p. , pooley , g. g. , durouchoux , p. , tilanus , r. p. j. , brocksopp , c. , 2000 , mnras , 312 , 853 friend d. b. , castor j. i. , 1982 , apj , 261 , 293 gallo e. , fender r. p. , pooley g. g. , 2003 , mnras , 344 , 60 gallo e. , fender r. , kaiser c. , russell d. , morganti r. , oosterloo t. , heinz s. , 2005 , nat , 436 , 819 gies d. r. , bolton c. t. , 1986a , apj , 304 , 371 gies d. r. , bolton c. t. , 1986b , apj , 304 , 389 gies d. r. et al . , 2003 , apj , 583 , 424 gies d. r. , et al . , 2008 , apj , 678 , 1237 guilbert p. w. , 1986 , mnras , 218 , 171 hanke m. , wilms j. , nowak m. a. , pottschmidt k. , schulz n. s. , lee j. c. , 2009 , apj , 690 , 330 heinz s. , 2006 , apj , 636 , 316 heinz s. , sunyaev r. a. , 2003 , mnras , 343 , l59 ibragimov a. , zdziarski a. a. , poutanen j. , 2007 , mnras , 381 , 723 kallman t. r. , mccray r. , 1982 , apjs , 50 , 263 lachowicz p. , zdziarski a. a. , schwarzenberg - czerny a. , pooley g. g. , kitamoto s. , 2006 , mnras , 368 , 1025 lamers h. j. g. l. m. , cerruti - sola m. , perinotto m. , 1987 , apj , 314 , 726 lasala j. , charles p. a. , sith r. a. d. , bauciska - church m. , church m. j. , 1998 , mnras , 301 , 285 malzac j. , belmont r. , fabian a. c. , 2009 , mnras , 400 , 1512 mcconnell m. l. , et al . , 2002 , apj , 572 , 984 merloni a. , heinz s. , di matteo t. , 2003 , mnras , 345 , 1057 miller j. m. , wojdowski p. , schulz n. s. , marshall h. l. , fabian a. c. , remillard r. a. , wijnands r. , lewin w. h. g. , 2005 , apj , 620 , 398 orosz j. a. , et al . , 2007 , nat , 449 , 872 orosz j. a. , mcclintock j. e. , aufdenberg j. p. , remillard r. a. , reid m. j. , narayan r. , gou l. , 2011 , apj , 742 , 84 pooley g. g. , fender r. p. , brocksopp c. , 1999 , mnras , 302 , l1 poutanen j. , zdziarski a. a. , ibragimov a. , 2008 , mnras , 389 , 1427 rahoui f. , lee j. c. , heinz s. , hines d. c. , pottschmidt k. , wilms j. , grinberg v. , 2011 , apj , 736 , 63 reid m. j. , mcclintock j. e. , narayan r. , gou l. , remillard r. a. , orosz j. a. , 2011 , apj , 742 , 83 rushton a. p. , 2009 , phd thesis , university of manchester rushton a. , et al . , 2011 , proceedings of science , 10th evn symposium , 061 russell d. m. , fender r. p. , gallo e. , kaiser c. r. , 2007 , mnras , 376 , 1341 stirling a. m. , spencer r. e. , de la force c. j. , garrett m. a. , fender r. p. , ogley r. n. , 2001 , mnras , 327 , 1273 ( s01 ) szostek a. , zdziarski a. a. , 2007 , mnras , 375 , 793 ( sz07 ) tarter c. b. , tucker w. h. , salpeter e. e. , 1969 , apj , 156 , 943 yuan f. , cui w. , 2005 , apj , 629 , 408 zdziarski a. a. , gierliski m. , mikoajewska j. , wardziski g. , smith d. m. , harmon b. a. , kitamoto s. , 2004 , mnras , 351 , 791 zdziarski a. a. , pooley g. g. , skinner g. k. , 2011a , mnras , 412 , 1985 zdziarski a. a. , skinner g. k. , pooley g. g. , lubiski p. , 2011b , mnras , 416 , 1324 ( paper i ) zdziarski a. a. , lubiski p. , sikora m. , 2012 , mnras , submitted zikowski j. , 2005 , mnras , 358 , 851 the jet emission can be obscured by the donor provided @xmath353 . the lack of such eclipses implies , @xmath354 in case of the lack of x - ray eclipses ( @xmath355 ) , @xmath356 , which implies no eclipses of the jet emission . thus , equation ( [ ecl ] ) is of importance only in systems where x - ray eclipses are observed ( e.g. , m33 x-7 , @xcite ) , in which case observations of eclipses of the radio emission or their lack would impose a constraint on the location of the radio - emitting region . in the case of cyg x-1 , the lack of x - ray eclipses implies @xmath357 for @xmath358 . there will be no eclipse of the counter - jet emission by the donor for either , @xmath359 for example , for @xmath360 and @xmath361 , no counter - jet eclipses correspond to @xmath362 or @xmath363 . in addition , an accretion disc with the outer radius , @xmath364 , will block the jet emission for @xmath365 . equation ( [ tau_ff ] ) also describes the attenuation of the counter - jet emission ( in the constant velocity , constant temperature approximation ) , with the substitution of @xmath366 . note that then there is a singularity in equation ( [ tau_ff ] ) is at @xmath71 and @xmath367 , which is a true singularity , caused by photon paths going through the centre of the donor . note that the counter - jet emission is both relativistically beamed away from the line of sight and more absorbed than the direct jet emission . also , the counter - jet emission can be obscured by the star at some combinations of @xmath63 and @xmath69 , see equation ( [ icju ] ) . all these factors can account for the lack of detection of the radio counter - jet in cyg x-1 ( s01 ) . we now consider scattering or absorption proportional to the density ( which is the case for bound - free absorption ) of photons emitted by the jet and accretion flow . in the latter case , we require that the medium temperature and ionization parameter are constant along the photon length , which is a commonly adopted assumption in calculating bound - free absorption . the optical depth is then , @xmath368 where , in the case of compton scattering , @xmath369 , and @xmath370 is the scattering cross section . we can integrate equation ( [ tau_sc_v ] ) numerically , or , assuming @xmath80 , we obtain , @xmath371 as before , @xmath90 for @xmath88 , @xmath89 , and equation ( [ tau_sc ] ) appears to have a singularity . however , @xmath85 is finite , @xmath372 . the other singularity at @xmath71 , @xmath367 , is a true one . at @xmath355 , equation ( [ tau_sc ] ) yields @xmath373 cyg x-1 shows an orbital modulation at hard x - rays , at which energies the bound - free absorption is negligible , and which is most likely due to compton scattering of photons away from the line of sight by the wind electrons . the source of x - rays in this case is close to the compact object , and thus we may set @xmath355 . however , the assumption of @xmath80 leads to a significant underestimate of the actual @xmath85 . thus , we use the numerical integration of equation ( [ tau_sc_v ] ) . for the parameters of cyg x-1 listed above , we obtain @xmath374 , where the thomson cross section has been used . the results of @xcite and @xcite yield the orbital modulation depth at @xmath20920100 kev hard x - rays ( which photon interactions are dominated by the thomson scattering ) of @xmath3750.04 . we find @xmath376 at @xmath128 , which implies @xmath3772 . this appears to account for the wind density being enhanced near the orbital plane ( see section [ pars ] ) .

we observed the oxygen - rich large magellanic cloud ( lmc ) supernova remnant n132d ( snr052569.6 ) , using all instruments onboard the _ spitzer space telescope _ , irs , irac , and mips ( infrared spectrograph , infrared array camera , multiband imaging photometer for spitzer ) . the 540 irs spectra toward the southeastern shell of the remnant show a steeply rising continuum with ] and ] as well as pah emission . we also present the spectrum of a fast moving ejecta knot , previously detected at optical wavelengths , which is dominated by strong ] and ] emission lines . we interpret the continuum as thermal emission from swept - up , shock - heated dust grains in the expanding shell of n132d , which is clearly visible in the mips 24 image . a emission hump appears superposed on the dust continuum , and we attribute this to pah bending modes . we also detect the well - known 11.3 pah c - h bending feature , and find the integrated strength of the 1520 hump about a factor of seven stronger than the 11.3 band in the shell of the remnant . irac 39 images do not show clear evidence of large - scale , shell - like emission from the remnant , partly due to confusion with the ambient ism material . however , we identified several knots of shocked interstellar gas based on their distinct infrared colors . we discuss the bright infrared continuum and the polycyclic aromatic hydrocarbon features with respect to dust processing in young supernova remnants .

astro-ph0609133

supernovae influence the chemistry and physics in the interstellar medium ( ism ) from galactic scales down to the atomic level . primarily , it is the energy input of supernova explosions that drives important processes in the ism : strong shock waves accelerate , heat , and compress interstellar gas , destroy dust grains , accelerate cosmic rays , and affect interstellar chemistry ( cf . * ? ? ? the role that supernovae play in the formation , destruction , and processing of polycyclic aromatic hydrocarbons ( pahs ) is largely unexplored . pahs are believed to be the carrier of the unidentified infrared ( uir ) bands , well - studied emission features near 3.3 , 6.2 , 7.7 , 8.6 , and 11.3@xmath0 attributed to the c - c and c - h stretching and bending modes of these molecules @xcite . it is established that interstellar grains are readily destroyed by strong shock waves ( e.g. * ? ? ? * ) , and it was suggested that grain destruction through shocks may be a source of pah molecules @xcite . although pah features have been observed in a wide variety of sources ( cf . * ) , no pah emission has yet been detected in snrs as far as we are aware , however , @xcite identified four snrs whose infrared colors suggest emission from pahs . more recently , spectral features at 1520@xmath0 including broad plateaus and distinct , narrower features , most notably near 16.4 and 17.4@xmath0 , were detected in conjunction with the well - known pah features ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * and see @xcite for a summary ) . the 1520@xmath0features are generally attributed to bending modes of large pah molecules ( cf . * ? ? ? * ; * ? ? ? * ; * ? ? ? the relatively narrow observed widths of the discrete 16.4 and 17.4@xmath0 bands suggest that they arise from individual pah molecules in the gas phase @xcite . there is also evidence that the appearance , relative strengths , and profiles of the pah features are coupled to the local physical conditions @xcite . in this paper , we report the detection of weak pah features near 6.2 and 7.78.6@xmath0 , at 11.3@xmath0 , and a prominent , broad hump at 1520@xmath0 from the young supernova remnant n132d ( snr0525@xmath169.6 ) in the large magellanic cloud ( lmc ) . the age of n132d determined from the ejecta kinematics is @xmath22500yr @xcite , and its spatial extent is about 100 , or 25pc assuming a lmc distance of 50kpc . n132d belongs to the small class of young , oxygen - rich snrs that are presumably the product of core - collapse supernovae , with the most famous galactic example being casa . for the progenitor of n132d , @xcite suggested a w / o star of 3035@xmath3 based on relative elemental abundances and comparisons to stellar nucleosynthesis models , which may have exploded as a type ib supernova . n132d shows a radio spectral index of @xmath4 @xcite , consistent with its overall shell - like morphology . it is one of the brightest x - ray sources in the lmc @xcite , and has also been investigated in great detail in the uv / optical @xcite . the large - scale morphology of the gas in n132d shows three types of line - emitting regions in the optical @xcite : ( 1 ) red- and blueshifted oxygen - rich ejecta filaments in the center , with a total velocity range @xmath24400@xmath5 @xcite , ( 2 ) shocked interstellar clouds within the remnant perimeter and in the outer rim , and ( 3 ) a photoionized precursor around the outer edge of the remnant . we observed n132d in the infrared with the _ spitzer space telescope _ @xcite . in section [ sec;observations ] , we describe our observations and the data reduction procedure . we took particular care of the background subtraction in order to avoid confusion with foreground and background emission ( section [ sec;background ] ) . we present our results from the imaging and the spectroscopy in section [ sec;results ] , and give a detailed discussion of the ir emission features and a comparison to pah features observed in other astronomical objects in section [ sec;discussion ] . we observed n132d with all instruments onboard the _ spitzer space telescope _ : irs ( infrared spectrograph , * ? ? ? * ) 2004 december 13 , irac ( infrared array camera , * ? ? ? * ) 2004 november 28 , and mips ( multiband imaging photometer for spitzer , * ? ? ? * ) 2004 november 7 ( _ spitzer _ aorkeys 11019776 , 11018496 , and 11022336 , respectively ) . we used the irs in its low resolution setting , yielding a wavelength coverage of 5.28.7/7.414.5@xmath0 with the short - low ( sl2/sl1 ) and 14.021.3/19.538.0@xmath0 with the long - low ( ll2/ll1 ) modules . the nominal spectral resolution is @xmath6128 as a function of wavelength for each module and order , e.g. @xmath7 at 18@xmath0 . the spatial sampling of the irs is 5.1/1.8 per pixel in the ll / sl modules , and the spatial resolution is diffraction limited by the 85 cm primary mirror . the four irac bands cover the wavelength ranges 3.24.0 , 4.05.0 , 5.06.4 , and 6.49.4@xmath0 , respectively , with a pixel size of 1.2 . the three mips detector bands cover 20.826.1 , 6180 , and 140174@xmath0 , with respective pixel sizes of 2.6 , 5.2 , and 17 in our configuration . the approximate total exposure times are 270s for each irac band , 48 , 38 , and 25s for the mips bands , 280s for each irs sl , and 360s for each ll module . the pipeline processed mips and irac image mosaics were of good quality , and we used them without further processing . the analysis of the irs spectra is complicated by the fact that n132d is a non - uniform , extended source superposed with a complex background adjacent to a molecular cloud ( see fig . [ fig : n132d ] ) . moreover , the _ spitzer _ science pipeline is optimized for point sources , and the extraction of extended sources from the default pipeline products results in biased spectra . our principal data reduction strategy for the irs spectra consists therefore of the following steps : \(1 ) starting from the s13.2.0 pipeline products , we interpolated bad pixels flagged by the pipeline , and removed rogue pixels by subtracting nearby sky positions taken from the _ spitzer _ archive for each sl / ll module ( see sect . [ sec;background ] for details ) . any remaining rogues were removed with the irsclean tool . rogue pixels are single detector pixels that show time - variable and abnormally high flux values . \(2 ) we extracted the spectra with spicev.1.3beta1 using the default extraction widths , and imported the spectra into smart @xcite for postprocessing ; \(3 ) subtraction of a local background spectrum ( see sect . [ sec;background ] ) ; \(4 ) basic aperture and slit loss correction assuming a uniform , extended source , to account for the bias introduced by the point source optimized _ spitzer _ science pipeline . we derived a correction function by modelling the _ spitzer _ point - spread - function ( psf ) for a range of wavelengths with the software stinytim@xmath8 developed by john krist , and then dividing the fraction of the psf admitted by the slit as defined by the slit - width and wavelength dependent extraction aperture by the sum over the whole psf . the resulting flux correction functions have wavelength dependent values of 0.80.6 for sl and 0.850.65 for ll . further division of the corrected flux density by the extracted area given by the slit - width and the extraction aperture yields spectra in surface brightness units , i.e. flux density per steradian , which can be directly compared to the irac and mips fluxes . note that irac fluxes are calibrated for point sources , but established point - to - extended source flux correction factors are available @xcite ; \(5 ) we removed fringes that appear in ll1 at 19.524.0@xmath0 with the smart defringing tool , averaged multiple spectra of the same region , and finally merged sl and ll spectra . note that we extracted the sl and ll spectra of the southeastern rim from two slightly offset positions , because the small overlap region covered by both slits is positioned just outside the brightest regions in the southeastern rim ( cf . [ fig : n132d ] ) . the two extracted regions are separated by about 10 , and the mips24@xmath0 fluxes in both regions differ by only 10 percent , whereas the irac channel 3 and 4 fluxes are comparable . a meaningful background subtraction of the spectra is one of the most critical and difficult tasks , and of major importance for this work . commonly , either a spatially close spectral nod or a dedicated off - source position are subtracted to remove the background . both of these methods are problematic in our case , since we are facing a complex , non - uniform background ( cf . [ fig : n132d ] ) . as a first step , we subtracted archival off - source sky exposures from all spectral modules ( cf . [ sec;datareduction ] , item 1 ) . the selected sky positions are devoid of continuum and line emission , and were observed close in time and position with respect to our observations ( 3hrs , @xmath210 ) . the sky position _ spitzer _ aorkeys are 10964224 , 12935424 , and 12936960 for sl2 , 12936960 for sl1 , and 12935424 for ll2/ll1 . the sky subtraction removes rogue pixels , zodiacal light emission , and flat - field imperfections , which are particularly noticeable in the sl modules . in a subsequent step , we subtracted a rogue - cleaned local background spectrum from the extracted source spectra for each spectral module ( cf . [ sec;datareduction ] , item 3 ) . we selected positions that lie close enough to the source to represent the local background but sufficiently far from the rim to avoid the influence of the radiative precursor that extends noticeably beyond the edge of the remnant ( see sect . [ sec;irac ] ) . the background positions are depicted as blue boxes in fig . [ fig : n132d ] . we averaged two background positions from opposite sides of n132d for sl2 and ll2/ll1 , while ensuring that both positions show similar spectra . for sl1 , we only have slit coverage east of the remnant , and we adopted the same background position as for sl2 . the overall background spectrum extracted from these positions is shown in fig . [ fig : n132dspec+backgr ] . we estimate a 30 percent uncertainty in this background spectrum due to spatial variations by using slightly different extraction positions ( approximately @xmath9 ; blue boxes in fig . [ fig : n132d ] ) and two background positions on opposite sides of the remnant . we measured emission line fluxes , central wavelengths , and full - widths at half - maximum ( fwhm ) from line fitting with the iraf ` spectool ' package . due to the low resolution of the irs spectrograph in our setting , @xmath7 at 18@xmath0 , the lines are usually unresolved or marginally resolved , and represent a convolution of the true line profile with the instrumental profile . we used gaussians to fit the line profiles of atomic lines/@xmath10 and lorentzians for discrete pah features . our quoted fwhm are not deconvolved from the instrumental irs profile . we also extracted spectra from different spatial positions at @xmath11 intervals along the ll spectrograph slit in order to study the spatial variation of line fluxes across the supernova remnant and the molecular cloud to the south ( see sect . [ sec : discussion_spatialvariation ] , fig . [ fig : n132dspec_spatialvariation ] ) . for this particular task , we employed a slightly different data extraction strategy . in order to improve the spatial resolution , we narrowed the default spice extraction aperture by a factor of two , giving extraction widths of about 12 for ll2 and 20 for ll1 near the central wavelengths of each module . this procedure systematically reduces measured fluxes by about a factor of two compared to the calibrated default extractions , which is of no concern as long as we compare only relative fluxes of extended emission extracted in that same fashion . we subtracted the continuum by fitting a low - order legendre polynomial using spectral regions that are free of emission lines , usually below 15 and above 19@xmath0 . for the southeastern rim of n132d , this method gives a similar continuum fit for the ll2 1421@xmath0 region than the modified - blackbody fit shown in fig . [ fig : n132dspec+bbfit ] . we did _ not _ apply a local background subtraction here , which means that we compare emission line fluxes including fore- and background for each spatial position along the ll slit . n132d shows strong emission in the mips24@xmath0 band with a background subtracted peak surface brightness of 50@xmath12 in the southeastern rim . the total , background subtracted mips24@xmath0 flux integrated over the whole remnant is @xmath13jy . the surface brightness distribution follows the x - ray contours very well ( cf . fig . [ fig : n132dmipsxray ] , left panel ) , although note that the mips24@xmath0 emission visibly extends beyond the edge of the remnant as defined by the x - ray contours , particularly along the western , southern , and eastern edges . this is a clear sign of the radiative precursor also seen in the optical @xcite , and is most likely continuum emission from heated dust and/or line emission from ionized oxygen caused by hard photons from the remnant ; it can not solely be explained by the wider mips point - spread - function ( cf . point source in fig . [ fig : n132dmipsxray ] , left panel ) . the remnant is also marginally detected in the mips70@xmath0 band with about 60@xmath12 background subtracted peak surface brightness and a total integrated flux of @xmath14jy , but the background confusion is much stronger ( fig . [ fig : n132dmips70xray ] , right panel ) . there is no conclusive detection in the mips160@xmath0 band , with an upper limit of 5@xmath12 peak surface brightness . [ fig : n132dmips70xray ] the four irac bands do not show clear evidence of emission spatially correlated with the remnant as seen in mips nor with the ejecta seen in the optical ( cf . [ fig : n132dknots ] , left panel ) . this is partly due to confusion with the foreground and background ism material . the irac emission within the remnant perimeter is clumpy and shows voids that may have been created by the passing shockwave of n132d . the irac color ratios measured at different positions within the remnant are consistent with pah emission ( cf . * ? ? ? * fig . 2 ) . we also detected several small emission knots that show irac colors different from the surroundings . in order to remove background stars and confusing pah emission , we subtracted a scaled irac 3.6@xmath0 image from the 4.5@xmath0 image , and a scaled irac 3.6 and 8.0@xmath0 image from the 5.6@xmath0 image ( cf . [ fig : n132dknots ] , right panel ) . we determined the optimal scaling factors empirically to produce a generally flat background . the emission knots have been observed previously in the optical with _ hst _ wfpc2 @xcite , and were identified as clumps of shocked , interstellar / circumstellar material ( cf . [ sec;iracdiscussion ] for a more detailed discussion ) . generally , the strongest emission in all four irac bands is associated with the molecular cloud/-region located south of the remnant ( cf . co contours in fig . [ fig : n132d ] ) . figure [ fig : n132dspec+backgr ] compares the irs spectrum extracted from the southeastern rim of n132d with the local background spectrum . figure [ fig : n132dspec+bbfit ] shows the background subtracted irs spectrum from the southeastern rim of n132d together with a two - component modified - blackbody fit . given the difficult background removal and the slightly offset extractions of sl and ll spectra ( cf . [ sec;background ] ) , it is worthwhile to note that the spectra from all modules seamlessly connect , and the fluxes are consistent with background subtracted flux estimates based on the calibrated irac and mips data . the modified - blackbody fit to the continuum uses an emissivity for 0.1@xmath0 silicate / graphite particles with a silicate - to - graphite ratio of 0.2 , following the ism silicon and carbon elemental abundances derived for n132d by @xcite and silicon abundances derived from x - ray data by @xcite and @xcite . the shape of this simple emissivity function is very similar to the emissivity from the general lmc dust model of @xcite in the 535@xmath0 range . note the shoulder near 10@xmath0 in the modified - blackbody fit due to the silicate part of the emissivity , which reproduces the continuum flux of our spectrum and therefore provides direct evidence that the heated dust in n132d contains at least some fraction of silicates . modelling of the uv to near - ir extinction curve for lmc sightlines @xcite generally seems to indicate a larger silicate content in dust grains , i.e. favouring larger silicate to graphite abundance ratios . for comparison , we also used the emissivity from the lmc dust model by @xcite and alternative silicate - to - graphite abundance ratios of 1 and 4 for our 0.1@xmath0 silicate / graphite dust emissivity . the original emissivity data for silicate and graphite particles are taken from @xcite and @xcitedraine ] . we deliberately excluded the pah component from the emissivity in order to analyze these features later in the continuum subtracted spectrum ( fig . [ fig : n132dspec - bbfit ] ) . the smart package uses a modified - blackbody fitting function of the form @xmath15 where @xmath16 is the normalized emissivity @xmath17 scaled by the factor @xmath18 , and @xmath19 is the planck function at temperature @xmath20 scaled by the factor @xmath21 , which accounts for the size of the emissive region . with the fitted free parameters @xmath18 , @xmath21 , and @xmath20 , all emissivity functions mentioned above produce nearly equally good fits to the spectral data , and the fitted values for @xmath20 and @xmath21 are similar for all models . for a two - component fit with a silicate - to - graphite ratio of 0.2 , we obtain temperatures of @xmath22k and @xmath23k , and a ratio @xmath24 . figure [ fig : n132dspec - bbfit ] shows the irs spectrum of the southern rim . the foreground / background and the fitted dust continuum have been subtracted ( cf . [ sec;background ] and [ sec;irs ] ) . the quoted total error consists of the pipeline - delivered statistical ramp uncertainties , combined with an estimate of the error due to the local background variation ( cf . [ sec;background ] ) . we detect several emission lines from atomic ions , tentative weak pah features near 6.2 and 7.78.6@xmath0 , a weak pah 11.3@xmath0 feature , and a prominent , broad hump at 1520@xmath0 . the ionic emission lines are from ne@xmath25 , o@xmath26 , and s@xmath25 ions . we can not distinguish between ] 25.91@xmath0 and ] 25.99@xmath0 in our low - resolution spectra , but argue that most of the emission is from o@xmath26 ( cf . [ sec;atomiclines ] ) . the pah features at 6.2 , between 7.7 and 8.6@xmath0 , and at 11.3@xmath0 are the well - studied pah c - h in - plane and out - of - plane bending modes ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? the presence of discrete 6.2 and 7.78.6@xmath0 bands is uncertain due to the low quality of our spectra in that region ( cf . total error in fig . [ fig : n132dspec - bbfit ] ) . there is probably an underlying emission plateau at 69@xmath0 , which has been observed in other sources and is probably associated with broad continua due to large pahs , clusters of small pahs , or very small grains ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? the broad hump at 1520@xmath0 is presumably due to pah in- and out - of - plane bending modes @xcite . figure [ fig : n132dspec_gaussfit ] shows the prominent emission hump between 15 and 20@xmath0 in detail . the spectrum without background subtraction clearly shows the 16.4@xmath0 pah feature ( cf . * ? ? ? * ; * ? ? ? * ) , and additional features near 17.1 and 17.9@xmath0 . these distinct bands with widths @xmath20.5@xmath0 may arise from smaller pahs in the gas phase . however , the individual bands are hardly distinguishable after background subtraction , indicating that these features are primarily due to the background ( cf . background spectrum in fig . [ fig : n132dspec+backgr ] ) . table [ tab;gaussfit ] lists the fit parameters for all detected ionic lines and pah features in the rim , the local background , and the molecular cloud/-region near n132d . for the local background spectrum , we assumed a flat continuum of @xmath27 . lcc@c@c@cc i d & & & + & & & + + pah6.2 & & & & & + pah11.3 & 11.28 & ( 0.01 ) & 1.06 & ( 0.12 ) & 0.20 & ( 0.05 ) + [ neiii ] & 15.60 & ( 0.01 ) & 0.81 & ( 0.01 ) & 0.27 & ( 0.03 ) + pah1520 & & & 7.47 & ( 0.67 ) & & + [ siii ] 21 & 18.74 & ( 0.01 ) & 0.49 & ( 0.04 ) & 0.40 & ( 0.03 ) + [ oiv ] & 25.96 & ( 0.01 ) & 0.30 & ( 0.02 ) & 0.36 & ( 0.02 ) + + pah6.2 & & & & & + pah11.3 & 11.26 & ( 0.01 ) & 4.28 & ( 0.15 ) & 0.28 & ( 0.02 ) + [ neiii ] & 15.58 & ( 0.01 ) & 0.30 & ( 0.02 ) & 0.23 & ( 0.01 ) + pah1520 & & & 3.20 & ( 0.68 ) & & + [ siii ] 21 & 18.74 & ( 0.01 ) & 0.24 & ( 0.02 ) & 0.30 & ( 0.03 ) + [ siii ] 10 & 33.52 & ( 0.01 ) & 0.34 & ( 0.01 ) & 0.46 & ( 0.02 ) + [ siii ] & 34.86 & ( 0.01 ) & 0.77 & ( 0.01 ) & 0.43 & ( 0.01 ) + + [ neiii ] ? & 15.63 & ( 0.02 ) & 0.16 & ( 0.04 ) & 0.24 & ( 0.06 ) + pah ? & 15.92 & ( 0.01 ) & 0.25 & ( 0.05 ) & 0.23 & ( 0.05 ) + pah16.4 & 16.46 & ( 0.01 ) & 1.67 & ( 0.08 ) & 0.29 & ( 0.01 ) + pah ? & 16.75 & ( 0.01 ) & 0.83 & ( 0.13 ) & 0.35 & ( 0.04 ) + @xmath10 s(1 ) & 17.06 & ( 0.01 ) & 1.02 & ( 0.01 ) & 0.27 & ( 0.01 ) + pah17.4 & 17.38 & ( 0.01 ) & 1.57 & ( 0.11 ) & 0.41 & ( 0.02 ) + pah17.8 & 17.76 & ( 0.01 ) & 0.48 & ( 0.07 ) & 0.32 & ( 0.04 ) + pah ? & 18.27 & ( 0.03 ) & 0.14 & ( 0.04 ) & 0.30 & ( 0.09 ) + [ siii ] 21 & 18.77 & ( 0.01 ) & 0.68 & ( 0.01 ) & 0.26 & ( 0.01 ) + @xmath10 s(0 ) & 28.25 & ( 0.01 ) & 0.26 & ( 0.01 ) & 0.38 & ( 0.01 ) + [ siii ] 10 & 33.52 & ( 0.01 ) & 0.71 & ( 0.01 ) & 0.41 & ( 0.01 ) + [ siii ] & 34.86 & ( 0.01 ) & 0.97 & ( 0.01 ) & 0.38 & ( 0.01 ) + + [ tab;gaussfit ] the detected pah features in the southeastern rim of n132d are the first evidence of pah molecules surviving a supernova blast wave to our knowledge . pah features also appear in lines - of - sight outside the remnant perimeter , and it is necessary to make certain that the pah emission originates from the blast wave , and not from foreground , background , or ambient ism material . there are three points of evidence that the pah emission is indeed originating from material associated with the blast wave . firstly , we removed the foreground and background emission by subtracting the flux from lines - of - sight just outside the blast wave perimeter ( sect . [ sec;background ] ) . secondly , we analyzed the spatial variation of emission line fluxes along the ll spectrograph slit ( cf . [ sec : discussion_spatialvariation ] , figure [ fig : n132dspec_spatialvariation ] ) , which shows that the pah features between 15 and 20@xmath0 have a well - defined local flux maximum at the position of the rim . this is expected for material that has been swept up by the blast wave and is not yet completely destroyed by the shock . if the majority of the flux would be from unrelated foreground or background emission , there would be no local maximum at the position of the rim . some fraction of the detected pahs may reside in the radiative precursor ( cf . [ sec;irac ] ) , where they would be excited / processed by energetic photons emanating from the supernova remnant . finally , the 1520@xmath0 pah emission changes notably when comparing features from the supernova remnant to the adjacent molecular cloud/-region and the local background ( cf . fig . [ fig : n132dspec_rim+knot+cloud+ngc ] and [ fig : n132dspec_sources ] ) , which we interpret as pah / grain processing in n132d ( cf . [ sec;irsgrainprocessing_discussion ] and [ sec : previousobservation ] ) . [ sec;atomiclines ] [ sec : pahlines ] in figure [ fig : n132dspec_rim+knot+cloud ] , we show the spectrum of the southeastern rim of n132d in comparison to the ejecta knot detected in the ll spectra , the molecular cloud/-region south of n132d , and the average spectrum of the galaxy ngc7331 . the characteristic features of the rim spectrum have been described earlier in this paper . the spectrum of the ejecta knot lacks the steeply rising dust continuum and the 1520@xmath0 pah features , but shows strong ionic emission lines from ne@xmath25 and o@xmath26 , which is consistent with the strong neon and oxygen lines seen in the optical for the same knot ( n132d - p3 , * ? ? ? the line fluxes and fwhm are @xmath28 ( 0.15@xmath0 ) for the neon and @xmath29 ( 0.34@xmath0 ) for the oxygen line . the spectrum of the molecular cloud/-region shows strong pah bands as well as ionic and @xmath10lines . the association of the co emission south of n132d with an -region seen in h@xmath18 was noticed by @xcite . @xcite estimated the mass of the co cloud to be about @xmath30 , and argued that its spatial proximity and matching lsr velocity provide strong evidence that this giant cloud is associated with n132d . this provides an opportunity to relate our observations of dust and pahs in the supernova remnant n132d to the nearby environment its progenitor star probably originated from . such a comparison will also be useful to interpret spectroscopic observations of star - forming galaxies ( e.g. ngc7331 , * ? ? ? * see sect . [ sec : previousobservation ] ) . [ fig : n132dspec_rim+knot+cloud+ngc ] we present a systematic assessment of the pah 1520@xmath0 , ionic , and molecular line fluxes from positions across n132d and the nearby molecular cloud in figure [ fig : n132dspec_spatialvariation ] . we extracted spectra along the ll slit with @xmath31 separation starting at the northern edge of n132d and ending at the southern edge of the molecular cloud about 3.3 south of n132d , resulting in a total number of 15 spectra . we converted the spatial distance along the slit into a parsec scale assuming a lmc distance of 50kpc , and adopted the location of the southeastern rim of n132d as the zero position . each node in figure [ fig : n132dspec_spatialvariation ] represents a line flux derived from continuum and line fitting , with errors given for the positions covering the supernova remnant ( cf . [ sec : analysis_emissionlines ] ) . the ionic and molecular lines detected and examined in our spectra are ] 15.6 , ] 18.7 and 33.5 , ] 25.9 , ] 34.8 , and @xmath1000s(0)28.2 and s(1)17.0@xmath0 . the lines from atomic ions , particularly the strong ] and ] transitions , are mostly unblended and straightforward to fit , which makes them excellent references in terms of flux and central wavelengths . ] and ] lines only appear across the supernova remnant and in the precursor , and are particulary strong at the position of the ejecta knot . generally , ] and ] trace the hot ionized gas of n132d . note that we can not distinguish between ] 25.91@xmath0 and ] 25.99@xmath0 in our low - resolution spectra , but we argue that most of the flux is likely to be from ] because the flux follows ] closely ( cf . fig . [ fig : n132dspec_spatialvariation ] ) ; under the conditions where we observe ne@xmath25 , iron would be mostly present as fe@xmath26 and fe@xmath32 . there is a weak line near 17.9@xmath0 , which may be ] 17.9 , but possible confusion with the pah 1520@xmath0 features renders this line inconclusive . we do not see ] 24.5/35.3@xmath0 , ] /]22.9 , and ] 19.6@xmath0 , which are probably too weak to be detected in our spectra . however , x - ray emission from fe@xmath33 and more highly ionized states was detected in the southern rim @xcite . ] is a common line seen in pdrs and -regions . here , it not only appears across the remnant , but also across the molecular cloud tracing the -region . the ] 1 - 0 transition at 33.5@xmath0 follows the same general trend , with the ratio increasing from 0.08 to about 0.2 from the northern edge of the molecular cloud into the remnant . @xmath10 traces the molecular cloud seen in co ( fig . [ fig : n132d ] ) . we checked for @xmath100 - 0s(0)28.2 , s(2)12.3 , and s(3)9.7@xmath0 , and did not detect these lines across the remnant ; weak s(1)17.0@xmath0 emission is not detectable because of strong blending with pah emission . we conclude that @xmath10 is mostly absent across the remnant , where it was probably destroyed by the fast shock and the strong radiation field , but apparently present in the precursor . the @xmath10s(1)/s(0 ) ratio increases from about 1.4 at the northern edge of the molecular cloud/-region to 11 at the shock precursor of the rim , which implies an increase in the uv radiation field as expected for the presence of a radiative precursor ( cf . sect . [ sec;irac ] ) . ] shows a nearly constant flux , except a noticeable minimum at the position of the rim . this is most likely an effect of the increased radiation field and the elevated temperature . silicon will be much more highly ionized under those conditions , which is confirmed by the detection of x - ray emission from si@xmath34 in the southern rim by @xcite . unfortunately , this conceals a possibly increased amount of silicon in the gas phase due to the destruction of dust grains . there are three well - defined pah components at 16.5 , 17.4 , and 17.8@xmath0 in the region of the molecular cloud ( fig . [ fig : n132dspec_rim+knot+cloud+ngc ] and [ fig : n132dspec_sources ] , and table [ tab;gaussfit ] ) , which have all been previously observed in other sources ( see sect . [ sec : previousobservation ] ) . the fluxes of these features follow the @xmath10 flux very well as expected if the pahs are well - mixed with the gas ( fig . [ fig : n132dspec_spatialvariation ] ) . the feature at 16.8@xmath0 is blended with the stronger pah feature at 16.5@xmath0 and the strong @xmath10s(1 ) line at 17.1@xmath0 , and therefore more uncertain . the situation changes for the supernova remnant , where the discrete 1520@xmath0 pah features all but disappear , and a nearly featureless 1520@xmath0 emission hump remains ( fig . [ fig : n132dspec_gaussfit ] and table [ tab;gaussfit ] ) . we detect shocked interstellar clouds with irac due to their distinct emission in the irac channels ( cf . fig . [ fig : n132dknots ] ) . the irac emission knots seen in the northern central part and the northwestern rim of the remnant coincide with previously detected , bright optical ] and ] emission complexes named `` lasker s bowl '' and `` west complex '' . they were identified as shocked interstellar clouds / circumstellar material based on optical line diagnostics , elemental abundances , and their radial velocities similar to the local ism surrounding n132d @xcite . the small knot visible close to the center of the remnant is probably a smaller analogue of these two complexes . we do not detect the red- and blueshifted , oxygen - rich ejecta of n132d with irac ( cf . the dotted contours in fig . [ fig : n132dknots ] ) . the measured irac band ratios for the shocked clouds are roughly consistent with thermal emission from hot silicate / graphite dust grains . in particular , the west complex shows an irac 4.5/8@xmath0 ratio of 0.18 and a 3.6/5.7@xmath0 ratio of 0.38 . for this region , @xcite presented _ hubble _ fos spectra that show a broad range of ionization stages as expected from shock heating , e.g. o to o@xmath26 and n to n@xmath32 , and elemental abundances consistent with lmc interstellar material . they derive a high electron temperature @xmath35k , a mean electron density of several thousand per @xmath36 , and a preshock cloud density larger than 100@xmath36 . their shock models indicate that a shock speed of at least @xmath2200@xmath5 is needed to produce the observed high - ionization species . comparing this to shocked dust ir emission models calculated by @xcite , we find that the measured irac color ratios are roughly reproduced by models with a preshock density of 100@xmath36 and a shock speed 400@xmath5 , and models with a preshock density of 10@xmath36 and a range of shock speeds between 400 and 1000@xmath5 . this suggests that part of the irac emission in the west complex may be due to a population of shock - heated , small dust grains . however , it is likely that the measured irac colors represent a mix of dust continuum and ionic emission . the green channel 2 excess emission probably includes br@xmath184.05@xmath0 , since the knots appear in the h@xmath18 image from @xcite and the optical spectrum of the west complex shows strong hydrogen recombination lines @xcite . the red channel 3 excess emission could also be due to ] 5.34@xmath0 . if @xmath10 is present in those clouds , the irac 4.5@xmath0 band covers the more highly excited lines , i.e. the @xmath10(0,0 ) s(8)-s(11 ) transitions , whereas the 5.7 and 8@xmath0bands cover @xmath10(0,0 ) s(4)-s(7 ) . a different type of irac emission is mainly characterized by an excess in channel 2 , i.e. it appears as pure green emission in fig . [ fig : n132dknots ] , right panel . the faint green emission aligned along the southwestern rim is spatially coinciding with small clumps of gas seen in the _ hst _ wfpc2 images @xcite . judging solely from the irac band fluxes , it is difficult to determine the nature of the ir emission in these clumps . unfortunately , none of them are covered by our irs spectra . many other isolated patches showing channel 2 excess emission do not have a counterpart in the narrow band _ hst _ wfpc2 ] , ] , nor the h@xmath18 image from @xcite . however , the brightest green patches , one near the center of the ejecta , two blended patches near the northeastern edge of the remnant , and one outside the remnant perimeter near the northeastern corner of the image , have optical counterparts visible in archival , broad band _ hst _ acs data ( proposal 12001 , filters f475w , f550 m , f850lp ; see the related hubble heritage image , stsci-2005 - 30 news release ) . all counterparts are structured , extended sources with color variations . the object outside the remnant perimeter is most likely a background galaxy , whereas the nature of the objects within the perimeter of n132d remains unclear . we interpret the steeply rising continuum in the spectrum of the southeastern rim of n132d as thermal emission from dust grains that have been swept up and shock - heated by the supernova blast wave . most of the mips24@xmath0flux is then due to dust continuum flux , since the ] 25.9@xmath0 line flux is much weaker than the continuum in the wide mips24@xmath0 bandpass ( cf . [ fig : n132dspec+bbfit ] ) . the association of the mips24@xmath0 flux with the blast wave along the remnant perimeter and not with the actual ejecta ( cf . [ fig : n132dmipsxray ] and [ fig : n132dknots ] ) confirms that the dust is part of the swept - up , ambient interstellar / circumstellar material . this is further supported by the appearance of the brightest mips emission along the southeastern rim , where the interstellar gas density is expected to increase due to the presence of a molecular cloud ( cf . co contours in fig . [ fig : n132d ] , and sect . [ sec : discussion_spatialvariation ] ) . generally , the mips24@xmath0flux intensity follows the x - ray emission very well . every peak in the mips24@xmath0 surface brightness is also an x - ray peak ( cf . [ fig : n132dmipsxray ] ) , which demonstrates the connection between the x - ray emitting plasma and the heated dust grains . a two - component modified - blackbody fit to the continuum yields temperatures of 110 and 58k ( cf . sect . [ sec;irs ] ) . these temperatures can be adopted as average dust temperatures if the grains are in thermal equilibrium , i.e. balanced heating and cooling rates , and if the mean energy per grain heating event is much smaller than the thermal energy content of the grain at the equilibrium temperature ( steady vs. stochastic heating , cf . * and references therein ) . under the conditions in the rim of n132d , i.e. an average temperature @xmath37k and a mean preshock density 3@xmath36 ( cf . * ; * ? ? ? * ) , gas - grain collisions probably dominate the grain heating @xcite . if each gas - grain collision transfers an energy of the order of @xmath38j , i.e. similar to the energy of a 0.6kev x - ray photon , an individual heating event changes the temperature of a silicate / graphite grain with radius 0.04@xmath0 by less than 1k . this estimate assumes spherical , homogenous grains with a density of @xmath39 , and a specific heat capacity of @xmath40 for silicate / graphite grains at @xmath41k . small grains 0.04@xmath0 in size can be heated substantially by a single heating event , and thus do not have a well - defined equilibrium temperature ( cf . the fact that we observe two different average grain temperatures may indicate two distinct grain size populations and/or emission originating from separate physical regions with different heating conditions ( see sect . [ sec;irsgrainprocessing_discussion ] for a more detailed discussion ) . we can estimate the dust mass @xmath42 in those two components using @xcite @xmath43,\ ] ] and adopting a density @xmath44 for silicate / graphite grains , a distance @xmath45 for n132d , and an average ratio of grain radius @xmath46 over emissivity @xmath47 of @xmath48 . we evaluated the flux density @xmath49 and the planck function @xmath50 for the fitted dust temperatures @xmath51 at 23.7@xmath0 . [ eqn;dustmass ] is an approximation assuming spherical dust grains of uniform size , composition , and in thermal equilibrium . we derived the dust masses in the entire southern rim ( outlined by the green color in fig . [ fig : n132dmipsxray ] , left panel ) by scaling the result from eq . [ eqn;dustmass ] by the mips24@xmath0 flux ratio of the southern rim and the irs slit aperture , which gives @xmath52 and @xmath53 . if we further scale this result to the total mips24@xmath0 flux of the whole remnant , the total swept - up ism dust mass in n132d becomes about @xmath54 . these numbers have considerable uncertainty , particularly the low temperature component , because of the assumptions made in eq . [ eqn;dustmass ] and the measurement uncertainties . we can compare the derived dust masses with estimates of the swept - up gas mass from x - ray data , which give an approximate upper limit on the total amount of dust . @xcite fitted the bepposax x - ray spectrum of n132d with a two - temperature non - equilibrium ionization model . using the revised best - fit from @xcite , the total mass of the x - ray emitting material is about 380850@xmath3 , assuming a filling factor between 0.1 and 0.5 . @xcite applied sedov models to fit _ spectra of n132d , and derived a swept - up mass of about 650@xmath3 . assuming a total swept - up gas mass of 600@xmath3 for n132d , the amount of swept - up dust is @xmath21.5@xmath3 , if the fraction of dust is 0.25 percent of the gas mass , i.e. about one fourth of the galactic value . this dust mass represents an approximate upper limit for the case of no dust destruction . considering the factor - of - ten lower dust mass estimated directly from eq . [ eqn;dustmass ] , we can state that although some portion of the dust in n132d is surviving , as much as 90 percent might have been destroyed by the blast wave shock ( cf . sect . [ sec;irsgrainprocessing_discussion ] ) . [ sec : dwek_irmodel_spectra ] strong supernova shock waves compress , heat , and accelerate interstellar material , creating physical conditions that destroy interstellar dust grains . the grains are not instantly following the gas acceleration due to their inertia , but are gradually accelerated via gas - grain collisions , which cause a gradual grain erosion through sputtering ( cf . furthermore , the grain population develops velocity differences , and subsequent grain - grain collisions cause shattering and vaporization depending on the relative velocity and grain sizes @xcite . for n132d , @xcite have estimated a mean shock velocity of the x - ray emitting main blast wave of @xmath2800@xmath5 , a mean preshock density of @xmath23@xmath36 , and a postshock temperature of @xmath55k ( cf . also * ? ? ? * ) , which is in good agreement with the sedov model values from @xcite . for such a high - velocity shock , the dominant grain destruction process is thermal sputtering @xcite . the lifetime @xmath56 of silicate / graphite grains against this process was estimated by @xcite , and further refined by @xcite , @xmath57 at a temperature of @xmath55k . thus , if the hydrogen density is @xmath58 , silicate / graphite grains with a radius @xmath59 are completely eroded after roughly 1500yr . likewise , 0.001@xmath0 grains only survive 150yr , whereas 0.1@xmath0 grains would survive 15,000yr . @xcite , @xcite , and @xcite proposed that the supernova of n132d exploded most likely within a low - density cavity in order to reconcile the sedov age with the dynamical age without invoking large explosion energies . this scenario would allow the blast wave to expand unhindered in the beginning until it encountered the denser material at the boundary of the cavity . if we assume an initial , average expansion velocity of @xmath60@xmath5 , the shock reached the cavity wall within about 800 years , supposing that the edge of the cavity lies close to the current mips24@xmath0 rim in the south of n132d and adopting the common center of the oxygen - rich ejecta as the expansion origin ( cf . * ) . since the dynamical age of n132d is @xmath22500yr ( cf . * ) , we can infer that the shock has been interacting with the denser material for maybe 1700yr . hence , we would expect that most of the grains 0.01@xmath0 that were _ initially _ present are destroyed by now through thermal sputtering _ if _ they were continuously residing in the hot plasma ( cf . [ eqn;thermsput ] ) . naturally , as the shock wave keeps expanding into new ism material , the grain population is continually replenished . in addition to thermal sputtering , shattering and vaporization remove some fraction of the large - size grains upon grain - grain collision . this is limited by the probability of a grain - grain encounter before being brought to rest with respect to the gas , which is of the order of the dust - to - gas mass ratio in the absence of betatron acceleration @xcite . therefore , we conclude that thermal sputtering generally dominates the grain destruction in the fast shock of n132d , but shattering may redistribute the grain size distribution appreciably in denser clumps ( cf . * ? ? ? * ) , where the shock speed may be much lower than the estimated mean value of @xmath2800@xmath5 . we compare our results to calculated model spectra from dust grains processed in fast , nonradiative shocks by @xcite . figure [ fig : n132dspec_dwekmodels ] shows the spectrum from the southeastern rim of n132d together with our modified - blackbody fit components and a series of calculated shock model spectra adapted from @xcite . their model assumes silicate / graphite grains in a power - law size distribution after @xcite and grain destruction solely by sputtering without shattering / vaporization . overall , the scaled model spectra reproduce the observed spectrum of n132d remarkably well considering the model parameters are not fitted to the data . the applied scaling factors probably account for the filling factor of the emitting material , and were chosen empirically to match the flux of n132d . the main difference between the observed spectrum and the models is a lack of continuum emission at 69@xmath0 and around 70@xmath0 , although the mips70@xmath0 flux measurement does not provide a very stringent constraint due to the background confusion . the long - wavelength emission may be due to high density material , e.g. a swept - up dense knot , with a shock velocity lower than 400@xmath5 . the lack of 69@xmath0 flux in the models confirms that this emission is due to pahs and not from a population of small , heated silicate / graphite grains . note that the narrow feature near 11.5@xmath0 in the calculated spectra is not the 11.3@xmath0 pah c - h bending mode but an ir active mode of crystalline graphite ( cf . * ? ? ? * ) , which would likely be smoothed out in an imperfect polycrystalline sample @xcite . our high - temperature modified - blackbody fit component is very similar to the 1500@xmath5/10@xmath36 shock model , which justifies that our continuum fit is physically meaningful , particularly for the pah c - c - c emission in the 1520@xmath0 range ( fig . [ fig : n132dspec - bbfit ] ) . generally , however , a single preshock density and/or a single shock velocity are apparently not sufficient to explain the continuum of the observed spectrum . the medium into which the shock wave expands is clumpy ( cf . _ hst _ wfpc2 images of n132d by * ? ? ? * ) , so we expect a combination of different densities and shock speeds . figure [ fig : n132dspec_rim+knot+cloud+ngc ] shows the spectrum of the galaxy ngc7331 ( cf . * ? ? ? * ) in comparison to our spectra for n132d . the spectrum of ngc7331 bears a striking similarity to the spectrum of the molecular cloud/-region south of n132d , and also shows ] and ] / ] lines . since it represents an average galactic ism mixture , the resemblance to our spectra is not surprising . @xcite have fitted the 1520@xmath0 pah features in this spectrum with two drude profile components at 16.40 ( @xmath61 ) and 17.12@xmath0 ( @xmath62 ) , and measured a @xmath63 pah flux of @xmath64@xmath65 . this fit reproduces the overall shape of the emission in this region together with the @xmath10s(1 ) component , but it does not take the blended features near 17.4 and 17.8@xmath0 into account , which are also observed in other sources ( cf . [ fig : n132dspec_sources ] ) . @xcite reported the strong 17.1@xmath0 feature as a possible new pah c - c - c bending mode detection . if we adopt lorentzian components of narrower width , we find the same pah components with similar fluxes and widths as seen in our spectra of the molecular cloud/-region near n132d ( cf . table [ tab;gaussfit ] and fig . [ fig : n132dspec_sources ] ) , and no need for a broad feature near 17.1@xmath0 . the 0.20.4@xmath0 widths of these features are plausible for pah molecules in the gas phase @xcite . our refitted , total pah 1520@xmath0 flux in ngc7331 is @xmath66@xmath65 , which is a factor of 1.6 larger than observed in the southeastern rim of n132d , whereas the pah band at 11.3@xmath0 is about a factor 30 stronger in ngc7331 . * hereafter vhp ) have measured 1520@xmath0 pah plateau strengths as well as the pah bands at 6.2 and 11.3@xmath0 for a variety of galactic sources from _ iso _ sws spectra . table [ tab : pahlineratios ] shows a comparison of different pah band ratios from the _ iso _ results by together with reanalyzed _ spitzer _ archival data and our results for n132d . a comparison of the absolute fluxes is hampered by the larger _ iso _ beam and different source distances , so we restrict ourselves to band ratios using @xmath65units instead . note also that measured the 16.4@xmath0 pah features with a local continuum , i.e. just the excess portion of the feature on top of the 1520@xmath0 emission hump , whereas we generally use the mostly emission - free flux below 15 and above 22@xmath0 to determine the continuum for fitting the 1520@xmath0features . we analyzed _ spitzer _ archival calibration data for two sources in the sample , iras22308 + 5812 and ngc7027 , in order to test whether the results are comparable . we find that the measured pah ratios for these two sources match within 30 percent , which is a reasonable agreement given the uncertainties due to the different beam sizes and continuum choices for the fitting . interpret the observed relative strengths of the broad plateaus and the narrower pah bands in terms of variations in the size distribution , i.e. the larger pahs with a few thousand carbon atoms that give rise to the 1520@xmath0 emission probably originated from dense cloud cores that formed the massive stars in the -regions , whereas smaller pahs with less than a few hundred carbon atoms give rise to strong 6.211.3@xmath0bands . they also tentatively attribute the discrete 16.4@xmath0 band to bending modes of pendant benzene ring substructures in smaller pah molecules ( cf . also * ? ? ? * who propose pah molecules containing pentagonal rings as the carrier of the 16.4@xmath0 band ) . lccccc + source & type & @xmath67 & @xmath68 & @xmath69 & @xmath70 + + lkh@xmath18234 @xmath10-ridge & pdr & 0.2 & 0.08 & 0.1 & 0.04 + ngc7331 ( average ) & galaxy & 0.3 & 0.07 & 0.2 & 0.05 + ngc7023 & rn & 0.3 & 0.14 & 0.1 & 0.06 + ngc7027 ( edge ) & pn & 0.4 & 0.04 & 1.2 & 0.13 + * ngc7027 & pn & 0.5 & 0.04 & 0.7 & 0.06 + * cd-42@xmath7111721 & yso & 0.7 & 0.10 & 0.6 & 0.08 + * iras 03260 + 3111 & haebe & 1.0 & 0.06 & 0.8 & 0.05 + * iras 23133 + 6050 & chii & 2.2 & @xmath1 & 1.5 & @xmath1 + * iras 22308 + 5812 & chii & 2.3 & @xmath1 & 1.8 & @xmath1 + iras 22308 + 5812 & chii & 2.5 & 0.27 & 0.8 & 0.08 + * iras 18317 - 0757 & chii & 2.5 & @xmath1 & 1.2 & @xmath1 + * sh2 - 106 irs4 & hii & 5.1 & @xmath1 & 3.9 & @xmath1 + + n132d se - rim & snr & 7.1@xmath721.0 & @xmath1 & 4 & @xmath1 + n132d local backgr . & & 0.7@xmath720.2 & 0.18 & 0.5 & 0.13 + + [ tab : pahlineratios ] table [ tab : pahlineratios ] is sorted to show the trend of rising pah 1520@xmath0 to 11.3@xmath0 ratio . for a diffuse ism mix of starlight - heated , neutral and ionized pahs , 1520@xmath0 features are efficiently produced by pah particles in the 825 size range ( 300 @xmath73 @xmath74 carbon atoms ) . the 11.3@xmath0 feature can be produced by particles as large as about 20 or @xmath75 , but more efficiently by particles with radius @xmath76 or @xmath77 @xcite . a direct , quantitative comparison of the observed pah 1520@xmath0 to 11.3 and 6.2@xmath0 , or the 16.4@xmath0 to 11.3 and 6.2@xmath0 ratios with model calculations is difficult for a number of reasons . foremost , there is a unique pah size distribution , which differs from source to source . secondly , individual , discrete pah features may be due to specific substructures in pah molecules of a certain size ( cf . interpretation of the 16.4@xmath0 feature above ) . finally , the specific local physical conditions affect the charge state , the amount of dehydrogenation , and the heating process of the pah molecules . nevertheless , it is illustrative to note some general trends when comparing the observed pah emission ratios in table [ tab : pahlineratios ] with calculated pah band radiation efficiencies as function of size for a diffuse ism mix of neutral / ionized pahs ( cf . * ? ? ? we approximate the predicted , total emitted power in the 1520@xmath0 region by scaling the calculated 17.4@xmath0 band power with the total integrated absorption cross section per c atom in the 1520@xmath0 range divided by the cross section for the 17.4@xmath0 band . this gives approximately a scale factor of 4 , using the cross sections per c atom adopted by @xcite . the lowest observed 1520/11.3 , 1520/6.2 , 16.4/11.3 , and 16.4/6.2@xmath0 ratios in table [ tab : pahlineratios ] correspond then to the smallest pahs considered in the calculations by @xcite , which contain about 20 carbon atoms . pahs smaller than about 2050 c atoms are photolytically unstable ( cf . * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? hence , the pah size distribution of the sources near the top of table [ tab : pahlineratios ] is probably dominated by pahs of that size . this is generally consistent with earlier estimates of pah sizes derived from the 3.3 and 3.4@xmath0 c - h stretching modes ( * ? ? ? * table 3 ) , and from profile fitting of the pah 3.3 - 11.3@xmath0 bands with calculated spectra from a distribution of small pahs @xcite . for the supernova remnant n132d , we expect that small grains / pahs are rapidly destroyed by thermal sputtering ( cf . [ sec;irsgrainprocessing_discussion ] ) , i.e. relatively larger grains / pahs tend to dominate the size distribution , which is consistent with the high pah 1520@xmath0 to 11.3@xmath0 ratio and the absence of discrete 1520@xmath0 features . a pah 1520@xmath0 to 11.3@xmath0 ratio of @xmath27 as observed in n132d corresponds to pah sizes of about 20 or 4000 c atoms , if we adopt the calculations by @xcite as a tentative guideline . this number is a rough estimate , since electron collisions rather than starlight dominate grain heating in n132d , changing the amount and frequency of energy deposition into grains compared to the model calculations . in addition , even large pahs may be dehydrogenated in the hot plasma of n132d , which provides another rationale for the weak 11.3@xmath0 feature and the resulting high 1520/11.3@xmath0 ratio . as a consequence , the size estimate from that ratio is probably an upper limit , since pahs in n132d have most likely a higher degree of dehydrogenation than in the diffuse ism . a lower limit estimate of the pah sizes in n132d from the 1520/6.2@xmath0ratio is about 500 c atoms , which is similar to the sizes of the very small grains / large pahs proposed as carriers of the broad plateaus underlying the 614@xmath0 pah features ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * and cf . [ sec : discussion_pahlines ] ) . the lifetime of 5004000 c atom grains , about 0.0010.002@xmath0 in size , is 80160yr according to eq . [ eqn;thermsput ] , which describes thermal sputtering of silicate / graphite grains in the forward shock of n132d . we presume that large pahs / amorphous carbon particles behave similarly to graphite in terms of stability against thermal sputtering , which is a reasonable assumption since both consist mainly of aromatic carbon ring systems . thus , it is plausible for large pahs to survive thermal sputtering in n132d long enough to maintain a detectable population , particularly since both pahs and grains will be replenished as the shock wave keeps expanding into new ism material . we detected strong mips24@xmath0 emission from swept - up , shock - heated dust grains in the young lmc supernova remnant n132d ( sect . [ sec;mipsdiscussion ] ) . in addition , we reported the first detection of pah bands from a snr , with tentative pah features between 6 and 9@xmath0 , a weak 11.3@xmath0 feature , and a prominent 1520@xmath0emission hump . this adds to the growing body of sources showing 1520@xmath0 emission , which is interpreted as bending modes of large pahs ( sect . [ sec : discussion_pahlines ] ) . overall , our observations show that dust grains survive a strong supernova shock , which is consistent with our estimates of the dust destruction rate by thermal sputtering ( sect . [ sec;irsgrainprocessing_discussion ] ) . our results also support the proposed correlation of pah size and the 1520 to 11.3@xmath0 emission ratio @xcite . the ratio of 1520@xmath0 to 11.3@xmath0 emission in n132d is notably higher compared to other types of sources , which we interpret as rapid destruction of small pahs / grains in the supernova blast wave via thermal sputtering , and possible dehydrogenation of the remaining pah population in the high temperature plasma . we conclude that large pahs / amorphous carbon particles with approximately 5004000 c atoms survive the harsh conditions in the forward shock of n132d long enough to be detected in our spectra ( sect . [ sec : previousobservation ] ) . future observations of supernova remnants of various ages located in different environments will probe whether , and what fraction of dust grains and pahs are ultimately able to survive the aftermath of strong supernova shock waves . this work is based on observations made with the _ spitzer _ space telescope , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa . we thank e. dwek , l. allamandola , and a. jones for fruitful discussions , and b. draine for providing a late draft of his manuscript @xcite prior to publication . we thank l. rudnick , t. delaney , and u. hwang for their collaboration on the _ spitzer _ proposal leading to this work and their helpful comments . support for this work was provided by the nasa ltsa program , nra-01 - 01-ltsa-013 , and nasa funding through the _ spitzer _ go program . cartledge , s. i. b. , clayton , g. c. , gordon , k. d. , rachford , b. l. , draine , b. t. , martin , p. g. , mathis , j. s. , misselt , k. a. , sofia , u. j. , whittet , d. c. b. , & wolff , m. j. 2005 , , 630 , 355 fazio , g. g. , et al.2004 , , 154 , 10 higdon , s. j. u. , devost , d. , higdon , j. l. , brandl , b. r. , houck , j. r. , hall , p. , barry , d. , charmandaris , v. , smith , j. d. t. , sloan , g. c. , & green , j. , 2004 , , 116 , 975 van kerckhoven , c. , hony , s. , peeters , e. , tielens , a. g. g. m. , allamandola , l. j. , hudgins , d. m. , cox , p. , roelfsema , p. r. , voors , r. h. m. , waelkens , c. , waters , l. b. f. m. , & wesselius , p. r. 2000 , , 357 , 1013 ( vhp )

the detailed understanding of the antineutrino emission from research reactors is mandatory for any high sensitivity experiments either for fundamental or applied neutrino physics , as well as a good control of the gamma and neutron backgrounds induced by the reactor operation . in this article , the antineutrino emission associated to a thermal research reactor : the osiris reactor located in saclay , france , is computed in a first part . the calculation is performed with the summation method , which sums all the contributions of the beta decay branches of the fission products , coupled for the first time with a complete core model of the osiris reactor core . the mcnp utility for reactor evolution code was used , allowing to take into account the contributions of all beta decayers in - core . this calculation is representative of the isotopic contributions to the antineutrino flux which can be found at research reactors with a standard 19.75% enrichment inu . in addition , the required off - equilibrium corrections to be applied to converted antineutrino energy spectra of uranium and plutonium isotopes are provided . in a second part , the gamma energy spectrum emitted at the core level is provided and could be used as an input in the simulation of any reactor antineutrino detector installed at such research facilities . furthermore , a simulation of the core surrounded by the pool and the concrete shielding of the reactor has been developed in order to propagate the emitted gamma rays and neutrons from the core . the origin of these gamma rays and neutrons is discussed and the associated energy spectrum of the photons transported after the concrete walls is displayed .

1602.07522

this work is motivated by several physics goals , spanning fundamental and applied neutrino physics . reactor antineutrinos are emitted in huge quantities by reactor cores , and this feature has been used by many particle physicists since the late fifties to study the neutrino properties@xcite . in the seventies , was also born the idea that antineutrino detection could be used for reactor monitoring@xcite . in a reactor core , the production of antineutrinos arises from the beta decays of fission products . the distribution of the latter nuclei are inherently linked to the structure of the nucleus undergoing the fission process and to the energy of the impinging neutron . thus the resulting emitted antineutrinos reflect the composition of a reactor core , but also its thermal power . these latter features , combined with progresses in the r&d of small antineutrino detectors and the development of dedicated reactor simulations , could be of interest for the international atomic energy agency ( iaea ) in its task of safeguarding nuclear reactors worldwide@xcite . in parallel , with the advance of the next generation of reactor antineutrino experiments , new calculations of the main uranium and plutonium isotope antineutrino energy spectra have been recently developed@xcite and led to the so - called " reactor anomaly@xcite . several hypotheses have been raised to explain this anomaly , among which the existence of light sterile neutrinos@xcite . since then , short baseline experimental projects have been born at research reactors in order to evidence the existence of such sterile neutrinos , in addition to the ones already under development for applications of neutrino physics@xcite , or other physics goals such as neutrino nucleus coherent scattering experiments @xcite , or the measurement of the neutrino magnetic moment@xcite . originally designed to be a small detector dedicated to reactor monitoring with an optimized detection efficiency , the nucifer experiment@xcite is one of those experiments that could as well bring additional constraints on the phase space associated to a sterile neutrino in the ev@xmath1 range . this experiment has provided the framework of the results presented in this article : a simulation of the osiris reactor core close to which nucifer is installed , a prediction of the associated antineutrino emission , as well as of the gamma and neutron backgrounds generated by the reactor operation . most of worldwide research reactors use fuel enriched up to a maximum of 20% in @xmath0u , which is the upper limit to be qualified as low enriched uranium and is strongly supported by the international atomic energy agency @xcite . this enrichment guarantees that most of the fissions arises from @xmath0u thermal fission , but there may be a small contribution to the fissions coming from the @xmath2pu produced during irradiation from the 80% of @xmath3u composing this kind of fuel . this contribution has to be quantified as the @xmath2pu antineutrino energy spectrum shape differs from the one from @xmath0u , which requires the modelling of the reactor core and of the subsequent antineutrino emission . the reference antineutrino spectra are still the ones based on the integral beta spectra precisely measured at the high flux reactor of the institut laue langevin in the eighties @xcite , and lately the integral beta spectrum produced through fast fission of @xmath3u has been measured in munich at the garching reactor@xcite . currently antineutrino spectra obtained through the conversion of these spectra@xcite are the ones exhibiting the smallest systematic errors . however , it should be mentioned that these systematic errors are the object of numerous discussions , showing that there may be extra sources of systematics to be accounted for due to nuclear physics theoretical and experimental uncertainties , insufficiently considered up to now @xcite . the recent observation of a distorsion of the antineutrino energy spectra@xcite measured by double chooz@xcite , reno@xcite and daya bay@xcite above 4mev with respect to converted spectra reinforces the need for further investigations . reactor neutrino experiments have up to now mostly used converted spectra as predictions to compare their measurements to . as these spectra were measured after a short irradiation time of a target into a research reactor , an extra correction has to be applied to account for the `` off - equilibrium effects '' , due to the build - up of long lived fission products , and to neutron captures on fission products during the core cycle . an alternative to this methodology is to use the summation method@xcite , consisting in summing the contributions of all beta decayers to the antineutrino spectrum . up to now , the published summation method spectra were computed to match the conditions of the individual beta energy spectra of the main uranium and plutonium isotopes , measured at ill@xcite and garching@xcite , i.e. not under the physics conditions of a reactor core under operation . in this article , we have coupled the summation method for the first time with a complete reactor model , allowing us predicting the low energy part of the antineutrino spectrum , below the inverse beta decay process ( ibd ) detection threshold , and evidencing the important contribution of actinides to the antineutrino emission . the obtained antineutrino energy spectra span an energy range beyond the one of the measured ill and garching integral beta spectra , which is the energy range of the converted spectra , making of the summation method spectra the only available predictions in these energy regions ( below 1.8mev and beyond 7 - 8mev ) . in addition , are provided in this paper the corresponding off - equilibrium corrections , deduced for the first time from the simulation of a complete reactor core , that can only be computed with the summation method . experimental conditions at short distance from research reactors are challenging , because the reactor itself produces huge gamma and neutron backgrounds that induce accidental and correlated backgrounds in an antineutrino target . the understanding of these backgrounds is of utmost importance and triggered the second part of the work presented here . we have performed the simulation of the gamma rays and neutrons generated by the osiris research reactor and propagated them through the water and concrete shieldings which separate the core from the antineutrino detector . the background estimation at the core level presented in this article could be used for the optimization of the background rejection of other experimental setups , or to estimate the physics reach of a foreseen experiment in the phase space of the oscillation angle and squared mass difference . in summary , we have thus developed a model of the osiris research reactor core in order to : give orders of magnitude values of fission rates in small thermal research reactors using fuel enriched at about 20% in @xmath0u ; provide the required data to compute the emitted antineutrino energy spectrum , including fine effects such as off - equilibrium corrections ; and eventually give estimates of the reactor - generated neutron and gamma backgrounds and identify their production processes . the paper is organized as follows . a first section describes the osiris research reactor and its fuel elements . a second section describes the simulation tools which were used to perform these studies and the associated validation tests . in a third section , are presented the obtained results for the antineutrino energy spectrum and flux . the last section is dedicated to the neutron and gamma background studies . in appendix a , are provided the antineutrino energy spectrum computed with the complete osiris simulation and the summation method for a reactor at equilibrium and off - equilibrium effects . in appendix b , we provide a table containing the gamma energy spectrum emitted by the core , before its propagation through the specific shielding of the osiris reactor , so that it can be used as an input of a monte carlo simulation of another experimental site . the osiris research reactor is an experimental reactor with a thermal nominal power of 70 mw @xcite . it is a light water reactor with an open core ( 57x57x60@xmath4 ) which main objective is to irradiate , with a high neutron flux , samples such as fuel elements and structure materials from power plants , or to produce radioisotopes for medicine and industry . the reactor delivers a neutron flux of 10@xmath5 neutrons.m@xmath6.s@xmath7 for about 200 days per year as an average , with cycles of 3 to 5 weeks and reactor stop periods of 10 days between cycles . the core vessel with its rectangular cross - section is made of 4 cm thick zircaloy and is surrounded by a 11 m deep , 7.5 m long and 6.5 m wide pool . the core vessel contains a rack of 56 cells used for : fuel elements , reflectors , control rods and experimental devices . its geometry is displayed in fig . [ fig : geo ] . view in the horizontal plane of the geometry of the simulated osiris core vessel with mcnpx 2.5.e@xcite . the colours represent the different materials and universes defined in the core model . one can see the structure of the fuel elements ( f ) , the beryllium walls ( be ) , the control rods ( cr ) and the irradiaton locations ( i ) with one filled in with 4 chouca ovens . ] the core is loaded with 38 standard fuel plate elements ( f ) , 6 control command rods ( cr ) , 5 locations for irradiation experiments ( i ) . two safety rods ensure the reactor stop while 4 rods are compensative and operating rods . an internal wall ( 7 cells ) and an external wall to the vessel ( 5 cm thick ) in beryllium are located on the southern face of the core and act as reflectors . each of the five experimental positions in the core contains a water box which can house four tubes containing for each a device called a `` chouca oven '' used for irradiation experiments . the fuel elements are made with 22 standard fuel plates in a u@xmath8si@xmath9al alloy , which are enriched in @xmath0u up to 19.75% and conform to the iaea requirements . each fuel plate is 0.51 mm thick and encapsulated in an aluminium cladding of 0.38 mm thick . the coolant channel between the plates is about 2.46 mm . two edge plates with boron are added to the fuel elements to comply with the regulatory safety margins and to control the reactivity available at the beginning of irradiation cycle . the upper part of the control command rods is made of an absorbant material ( hafnium ) and the lower part consists in 17 standard fuel plates , with a cooling channel of 2.79 mm between them . the nucifer antineutrino detector is placed at 7 m from the core behind a 2 m thick concrete wall . the detailed description of the design of the detector can be found in@xcite . in order to study either the core physics or the core - generated gamma and neutron backgrounds , the simulation of the reactor was made in several steps : the core alone ( c ) , the core surrounded by the pool ( cp ) and the core surrounded by the pool plus a concrete wall ( cpw ) up to the casemate where the antineutrino detector is located . the core simulation , cf . [ fig : geo ] , has been performed with the mure ( mcnp utility for reactor evolution ) code coupled to the mcnpx 2.5.e monte - carlo code @xcite . all the geometrical parameters were taken from public literature @xcite . in the simulation , fuel elements are discretized into evolving cells and the evolution is temporally discretized . at each time step , an mcnp run is performed in order to compute neutron flux and energy dependence in each evolving cell . the neutron flux is then used , based on a runge - kutta 4@xmath10 order method , to solve the bateman equations which describe the fuel evolution . for each evolving cell , mcnp estimates a neutron flux normalized per neutron source . a scaling of the neutron flux in the cells is then performed to normalize it in order to reach the total thermal power value of the reactor given as an input , taking into account the energies released per fission taken from kopeikin _ et al._@xcite . considering the increase of the cpu time of the simulation due to a large spatial discretization of the geometry , assumptions were made on the discretization of the fuel elements in evolving cells . no axial discretization was considered for all the reactor simulations performed during this work . indeed , such axial dependence of the fission rates is very dependent on the operation of the reactor core with the control rods , which makes studies specific to the considered reactor site mandatory for each concerned neutrino experiment . this would be at variance with the objective of our paper - providing generic predictions and information about the antineutrino and background emission - which could be useful for a wider public than a specific experiment . two core models were studied and compared . in the first model , all the fuel elements are treated as a unique evolving material cell , called a universe . in the monte - carlo code mcnp used in the mure code , the energy is treated continuously for the description of the associated interaction cross sections of the different considered particles . the input nuclear cross sections used in this simulation were preferentially taken from the endfb / vi library and a thermal s(@xmath11,@xmath12 ) treatment was applied to the interaction of neutrons with water @xcite . a multigroup option of the mure code was used to gain cpu time , consisting in discretizing the neutron energy distribution of the nuclear cross sections in 1.79 10@xmath13 groups . this option was tested in the frame of @xcite and showed results in agreement at the level of 0.2% with the results obtained with a continuous energy treatment for the main fission rates of interest ( uranium and plutonium isotopes ) while gaining a factor of 30 in the cpu time of the calculation . in order to ensure a proper convergence of the neutron source in the geometry , a distribution of source neutrons is propagated in the geometry with an mcnp simulation using 50 000 source particles , 1 000 inactive cycles and 300 active cycles . the generated source is then used as an input in the subsequent simulations , with 20 000 source particles , 100 inactive cycles and 250 active cycles . in the second core model , confidential data , especially more detailed informations on the geometry@xcite are used , and each of the 38 fuel assemblies is considered as an independent evolving cell , i.e. 38 fuel universes are considered . in order to validate the ability of these simulations to reproduce correctly the core physics , the neutron energy spectrum in an irradiation cell computed with mure is compared to the results of@xcite . material samples to be irradiated at osiris are placed in experimental devices called `` chouca ovens '' designed to reproduce the operational conditions of a power reactor . these devices ( 32 mm in diameter ) are rods made of an alloy of zircaloy and stainless steel , surrounded by nak and stainless steel layers . they are inserted by groups of 4 in the water boxes . to study the impact of a converter device on the neutron flux in irradiation experiments , f. malouch@xcite computed the neutron energy spectrum in a chouca oven by developing a 2d simulation of the osiris core with the code apollo2 which solves explicitely the neutron transport boltzmann equation on a spatial mesh and using the cea93 cross - sections library consisting in 99 neutron energy groups @xcite . the cea93 cross - sections library is processed from the jef-2.2 evaluations @xcite . this model of the osiris core has been validated with experimental measurements performed at the osiris reactor . for this reason , we consider it as a reference with which we can compare our models . s in fig . [ fig : choucaflux ] , the neutron energy spectrum in a chouca oven computed with apollo2 ( solid black curve ) is compared with the one calculated with mure with the first core model ( solid grey curve ) on the 5@xmath10 day of the power cycle assuming : a core loaded with fresh fuel , a constant nominal power history and all the fuel assemblies grouped in a unique universe , cf fig . [ fig : geo ] , and with the second core model ( dashed black curve ) under the same conditions except that 38 fuel universes are considered . the overall shape of the neutron energy spectrum is well reproduced by the mure calculation . the ratio betwen the neutron fluxes in the thermal and rapid energy ranges is 2.56 with the first mure calculation , and 2.06 with the second one , compared with a ratio of 2.05 with apollo2 . the normalization discrepancy arises from the unknown fuel loading map and the operation history associated to the apollo2 simulation . especially no details are available on the power history or on the different energies released per fission taken in the apollo2 calculation . another origin of differences among the results of the apollo2 and mure simulations arises also from the use of multigroup versus continuous in energy nuclear databases in respectively deterministic and monte - carlo codes , as well as the use of more recent nuclear databases in the case of the mure simulation with respect to the apollo2 one . this feature can be seen in fig . [ fig : choucaflux ] in the anti - resonant behaviour of the neutron energy spectrum between a few hundreds of ev up to a few hundreds of kev . overall , the good agreement obtained on the shape of the neutron spectrum validates our core models . as our objective is to simulate a typical research reactor antineutrino flux , a periodic refueling of the core was implemented . the fuel evolution of the osiris reactor was simulated over 200 days with a refueling of the core every 20 days by seven - th which corresponds to the length of a typical osiris reactor cycle . the cpu cost of such a detailed simulation imposes to simplify it . to this purpose , we have grouped the assemblies in seven universes in order to take into account the refueling of the core by one seven - th every twenty days . the refueling scenario is described on table [ table : refueling ] . the initials cr and i refer respectively to the location of the control rods and irradiation cells . the numbers ( one to seven ) refer to the refueling order of the fuel in the reactor core . at each refueling , the fuel labeled 1 is replaced by the one labeled 2 and so on until 6 . the last fuel block ( 7 ) is replaced by fresh fuel . the cooling time of the fuel blocks between real reactor cycles is neglected in the calculation of the reactor evolution . indeed this cooling time would affect mainly the fission rate of @xmath14pu which contribution to the fissions is very small , considering the enrichment of the fuel . * 4 & * 3 & * 1 & * i & * 1 & * 3 & * 4 * 4 & * 2 & * & * 1 & * & * 2 & * 4 * 4 & * 3 & * 1 & * i & * 1 & * 3 & * 4 * 7 & * 2 & * & * 3 & * & * 2 & * 6 * 7 & * i & * 5 & * i & * 5 & * i & * 6 * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * the distribution of the fission rates for the 4 main fissile nuclei : ( @xmath0u , @xmath3u , @xmath2pu and @xmath14pu ) is presented in fig . [ fig : fissionrates ] for an evolution of 280 days and 13 refuelings by seven - th . the fission rates are constrained by the power maintained at its nominal value along the evolution . by definition , the equilibrium of the reactor is reached when the fission rates show a periodic pattern repeated every 20 days . this is reached for the different contributions after 260 days , i.e the beginning of the third refueling cycle for the different fuel blocks . for a reactor with this fuel composition at equilibrium , the main contributor to the total amount of fissions in - core is @xmath0u . for this isotope , the variation of the fission rate between the beginning and the end of the cycle is about 2 - 3% . antineutrino emission thus mainly arises from the fission of @xmath0u as it will be discussed in the next section . the antineutrino energy spectrum and flux are computed using the summation method@xcite using the mure code coupled to nuclear databases containing beta decay properties of the fission products . the principle is to sum all the contributions of the beta decay branches of the fission products to obtain the total antineutrino spectrum , weighted by the fission product concentrations computed with mure . in this article , the nuclear databases are the same as the ones presented in @xcite , including the latest published total absorption spectroscopy experimental results @xcite and @xcite . in the subsections below , we present antineutrino energy spectra computed in two different ways , using in both cases the summation method . in the first case , the complete core model is used to provide the amount in - core of all beta minus emitters , allowing for the first time to take into account all the contributions , i.e. not only the fission products but also the contribution of the actinides and heavy nuclei produced during core operation . the @xmath12/@xmath15 energy spectrum emitted by the reactor core is thus broken - up into the sum of all @xmath12/@xmath15 spectra of the beta minus emitters ( labelled @xmath16 ) in - core , weighted by their activity @xmath17 : @xmath18 eventually , the @xmath12/@xmath15 spectrum of one beta emitter is the sum over the b branches of all @xmath12 decay spectra ( or associated @xmath15 spectra ) , @xmath19 ( in eq [ sfp ] ) , of the parent nucleus to the daughter nucleus weighted by their respective branching ratios as @xmath20 in the second case , the complete core model is only used to provide the fission rates , which are then weighted with individual uranium and plutonium isotope spectra computed with the summation method , to obtain the total spectrum emitted by the osiris reactor . the latter method is similar to the one used by reactor neutrino experiments i.e. weight individual antineutrino spectra from the main uranium and plutonium isotopes with their fission rates@xcite in the following way : @xmath21 the first term accounting for the number of fissions occuring over the time @xmath22 is the ratio of the thermal power over the average energy released per fission of the four isotopes present in the fuel . @xmath23 stands for the percentage of fissions undergone by the isotope @xmath24 and is computable by reactor evolution codes . its mean energy released per fission , @xmath25 , is stored in nuclear databases . the second term of eq . [ eq_nu_flux ] represents the total @xmath15 spectrum emitted by a reactor per fission . it is defined as the sum over the 4 isotopes ( @xmath26u and @xmath27pu ) of the fraction of fissions undergone by the k@xmath10 isotope times the @xmath15 spectrum per fission of the same isotope @xmath28 . in this paper , the @xmath12/@xmath15 spectrum per fission of a fissible isotope @xmath28 is calculated with the summation method , i.e. it is the sum of all fission product @xmath12/@xmath15 spectra weighted by their activity @xmath29 , @xmath30 where each @xmath31 is computed as in eq.[sfp ] above . the individual spectra @xmath28 are computed after an irradiation time corresponding to or close to the times for which converted spectra were obtained from ill and garching measurements i.e. 12hours for @xmath0u and 36hours for the three other fissible nuclei , and without the contribution of neutron captures on the fission products , as in the ill experimental conditions . the impact of the evolution of the concentration of antineutrino emitters in - core due to reactor physics and nuclide half - lives can be deduced from the discrepancy between the two calculations of the total antineutrino energy spectrum emitted by the reactor presented above . using the summation method in both cases should allow to eliminate a large part of the systematic errors due to the calculation method as all correlated errors vanish in the relative discrepancy between both calculations . in this way , the deduced effects are also directly applicable to correct the converted spectra used for computing the total antineutrino flux by the reactor neutrino experiments . in the next two subsections , are presented the obtained emitted antineutrino flux and energy spectrum and an example of flux and energy spectrum detected using the ibd process with a small liquid scintillator detector doped with gd , placed at 7 m from the osiris reactor core with a detection efficiency of 50% . in fig.[fig : antinusemis ] , the summation is made over the fission product concentrations computed with the osiris core simulation described above , and includes the contributions of the beta decays of actinides which contribute to the energy spectrum at low energy , below the detection threshold of 1.8 mev of the ibd process . it includes also off - equilibrium effects due to neutron capture on fission products or to the evolution of the fission product amounts after core start - up . according to this simulation , the variation of the antineutrino flux emitted along the fuel cycle is of about 2 - 3% . this variation is weak because of the important enrichment of the fuel in @xmath0u , which minimizes the contributions of other nuclei such as @xmath2pu and @xmath3u to the fission rates . on the one hand , this could be an advantage for fundamental physics purposes as the antineutrino energy spectrum associated to the fission of @xmath0u is the most precisely known . it also means that if neutrino detectors with a good energy resolution are used , experiments located close to research reactors could provide new measurements of the @xmath0u antineutrino energy spectrum . on the other hand , it shows that the measurement of the contribution of @xmath2pu to the fissions with an antineutrino detector for non - proliferation purposes is challenging in such a research reactor . the emitted flux computed with the complete core model presented here connected with the summation method is much larger than when using the individual energy spectra of @xmath0u , @xmath2pu , @xmath3u and @xmath14pu , computed with the summation method after respectively 12 , 36 , 36 and 36hours of irradiation , weigthed with the fission rates arising from the same core simulation . this is due to the contribution of @xmath2u and @xmath2np beta decays at low energy , as can be seen on the emitted energy spectra computed with both methods and presented in fig.[fig : emittedenergyspectra]a ) andb ) . these two latter figures correspond to two sets of time steps of the core simulation . fig.[fig : emittedenergyspectra]a ) displays the evolution of the antineutrino energy spectrum at the very beginning after the start - up of a core loaded with fresh fuel . the increase of the antineutrino energy spectra can be seen at 0.5 , 3 , 10 and 19.9days after start - up . the spectra displayed in plain lines show the summation calculation directly issued from the complete core simulation , while the spectra in dashed lines show the calculation using the individual spectra weighted by the fission rates from the same simulation . the power of the reactor core is kept constant in the simulation . one can notice that regarding the spectra computed using the individual uranium and plutonium spectra , the time evolution has no impact ; the four dashed line spectra are superposed . this is awaited as these individual spectra have been computed without off - equilibrium effects or other reactor physics effects , and after a fixed irradiation time . the discrepancy between both calculations in terms of normalization and evolution with time is clearly evidenced , due to the effects mentioned above : contribution of actinides that can be accounted for only in a complete core simulation , and off - equilibrium effects which include the effect of neutron capture on the fission product concentrations and the time evolution of their concentrations in - core with respect to the chosen irradiation time of the individual spectra . these contributions may be different at reactors using a different type of fuel , exhibiting another neutron energy spectrum , or based on other designs such as power stations . indeed the contributions of @xmath2u and @xmath2np arise from successive neutron captures and beta decays starting from @xmath3u at the origin of the creation of @xmath2pu . these contributions will be more pronounced in pwrs in which the creation of @xmath2pu is such that this isotope exhausts nearly half of the total fission rate at the end of a cycle . the discrepancy between emitted antineutrino spectra from both methods is even more pronounced for spectra computed for the core having reached its composition at equilibrium , as illustrated in fig . [ fig : emittedenergyspectra]b ) which is the same as fig.[fig : emittedenergyspectra]a ) but computed after 201 , 203 , 210 and 219.9days of core evolution with a periodic refueling by seventh of the core every 20days . here again , the spectra built with the individual spectra are superposed . the spectra built with the complete core model are much more alike than before , showing that the evolution of the fuel composition is weaker when the reactor has reached equilibrium . in appendix a , a table is provided containing the emitted antineutrino energy spectrum at 210 days after core start - up , computed with the sum of the contribution of all @xmath12-emitters in the complete core model , in order to allow the reader to use it as an input of the simulation of any specific detection setup . this could be useful for experiments chasing sterile neutrinos at research reactors . the energy spectrum can also be used in neutrino nucleus coherent scattering experiments , or any other experiment using another interaction process than the inverse beta decay@xcite . antineutrino flux which would be measured in a detector with a 50% efficiency placed at 7 m from a research reactor with a @xmath0u enriched fuel of 19.75% . the red dashed and black plain lines correspond respectively to the complete core model method and to the individual spectra method . ] in fig . [ fig : detectedantinus ] are displayed the antineutrino fluxes , computed with the two methods exposed above ( in dashed line with the complete core model , in plain line with individual spectra ) , which would be detected in a detector filled in with 0.85 t of liquid scintillator doped with gd at 7 m from the core and with a detection efficiency of 50% . the energy spectra corresponding to fig.[fig : emittedenergyspectra]b ) have been folded by the ibd process cross - section@xcite , weighted by the afore mentioned solid angle and detection efficiency . no energy resolution nor energy dependence of the detection efficiency have been taken into account here , in order to show an example of result rather independent from the used experimental setup . the obtained variation of the detected antineutrino flux along a fuel cycle is only of 0.7% . though very small , this variation is not null , and a misuse of the reactor which would imply a larger @xmath2pu content in - core could result in a visible variation with a high efficiency detector , as shown with a toy model in ref.@xcite . more detailed studies using our reactor model are contemplated , and beyond the scope of this article . a small discrepancy between the detected fluxes computed with the two methods presented above can be seen , which remains well below one percent . indeed , after folding with the ibd cross - section , the discrepancies nearly totally vanish and the two fluxes are very similar . this is well understandable as the effects of the beta decays of actinides , neutron capture and evolution of the fission product concentrations are very much suppressed by the detection threshold of the ibd process . the detected antineutrino energy spectra in the conditions exposed above are displayed in fig.[fig : detectedenergyspectra ] a ) , for a reactor considered at equilibrium ( 201days after core start - up ) , just after refueling , thus at the beginning of a new cycle . the black plain line represents the spectrum built with the distribution of the fission products from the osiris core model , while the red dashed line represents the spectrum built using the individual spectra from the four main uranium and plutonium isotopes . the remaining discrepancy between the two spectra is located between 1.8 and 3.5mev . fig.[fig : detectedenergyspectra ] b ) allows to highlight the relative difference among them , giving the contribution of actinide beta decay , neutron capture and fission product evolution after folding the spectra with the inverse beta decay cross - section , after 201days ( in black ) , 203days ( in grey ) , 210days ( in red ) and 219.9days ( in green ) of core evolution . one can see that the order of magnitude of these corrections with respect to individual spectra taken at the irradiation time of the ill measurements by@xcite reaches 3.5% at the detection threshold and lies between 2.0 and 2.4% at 2.5mev depending on the core irradiation time . these effects fall below 1% after 3mev . the impact of the off - equilibrium effects is not negligible at the ibd detection threshold and is influenced by the short duration of the research reactor cycle ( 20 days in the case of osiris ) which imposes the renewal of one seventh of the fuel . antineutrino experiments installed at research reactors using low enriched uranium fuel will have to cope with such effects . research reactors using high enriched uranium fuel should exhibit similar effects , except at the ibd threshold , where the beta decay of actinides have an impact . indeed , the contribution of the actinide beta decays should be much smaller due to the smaller quantity of @xmath3u in - core . in appendix a , a table is provided containing the relative difference displayed in fig.[fig : detectedenergyspectra ] b ) after 210days of core operation . in a reactor environment , the gamma sources are numerous : prompt gamma rays emitted when the fission process occurs , @xmath32 decays of the fission fragments after their @xmath12@xmath33 decays , ( n,@xmath32 ) reactions in the fuel or in different reactor components , and more generally gamma rays emitted by the excited nuclei created in the different processes . using the attenuation coefficients available in the literature for water , air and concrete , the total attenuation of the gamma rays emitted and transported up to the nucifer detector casemate was estimated to be about 10@xmath34 at 8 mev , 10@xmath35 at 4 mev and up to 10@xmath36 at 1 mev . in order to study in details the production process of gamma rays and their propagation to the detector casemate with enough statistics , the simulated pool and concrete walls were discretized in cells of 50 cm . a variance reduction technique was also used by weighting the importance of the different cells by the number of @xmath32 tracks . the chosen strategy was to perform simulations with the reactor core only ( c ) , core and pool only ( cp ) and core - pool plus concrete walls ( cpw ) with the mcnpx 2.7.d code @xcite using preferentially the endfb / vi cross - section library and the endfb / vii one for photonuclear reactions plus a thermal s(@xmath11,@xmath12 ) treatment of the scattering reactions in the water @xcite . the purpose of these three different simulations was to understand the origin of the gamma rays and neutrons reaching the experiment location . this method also allowed to decrease the cpu cost , as the transportation of the gamma rays from the core through the pool and the concrete walls is challenging . the core model adopted in this study is the first one presented in section iii : a core loaded with fresh fuel and all the fuel assemblies grouped in a unique universe . as a first step in each of the three cases ( c , cp , cpw ) , a simulation of reference using the default physics option of mcnpx 2.7.d for the @xmath32 transport was performed in order to carry out the simplest simulation with the following reactions : compton scattering , pair production , thomson and rayleigh scatterings , photoelectric effect plus k and l shell flurorescences . in a second step , extra simulations were performed by adding only one additional physics phenomenon at a time : delayed gamma rays , bremsstrahlung , photonuclear reactions , in order to estimate their respective contributions in the neutron and gamma production processes . in order to propagate the source correctly in the geometry , a source is generated with a simulation using 50 000 source particles , 1 000 inactive cycles and 700 active cycles . the generated source is then used in the subsequent simulations , with 50 000 source particles , 20 inactive cycles and 470 active cycles . the final gamma spectrum calculated with mure for a reactor fuel composition at equilibrium after 20 days and taking into account all these reactions is presented in fig . [ fig : gammaspectrum ] . the upper part of the figure refers to the gamma rays produced in the reactor core , also tabulated in appendix b , and the lower part to the gamma background expected in the detector casemate . the size of the energy bins is 10 kev . the electron transport is not directly included in these simulations . instead , a thick target bremsstrahlung model is used which assumes that the electrons are generated in the direction of the incident gamma ray and immediately annihilated after generating bremsstrahlung gammas . the 2.2 mev gamma ray , attenuated in the concrete walls , corresponds to the ( n,@xmath32 ) radiative capture by hydrogen nuclei contained in water . one of the most important gamma ray at 7.7 mev is due to the radiative capture on @xmath37al which is included in the fuel cladding . radiative captures on @xmath38be included in the reflector walls were also identified . the neutron and photon transports performed with mcnpx 2.7.d after one time evolution step takes 24 hours using a 3 ghz intel xeon e5450 processor . the contribution of delayed gammas could also be included in our mcnpx 2.7.d calculation using either multigroup data either a multigroup mode coupled to energy lines if individual data are available @xcite . the computing time needed for a multigroup mode coupled to energy line calculation added to the computing time needed for the fuel evolution prevents such a detailed calculation . however , the multigroup data calculation gives the general shape of the gamma energy spectrum and meets our needs with respect to the energy resolution of the nucifer detector . in this case , the number of @xmath32 tracks in the cpw simulation is multiplied by 5 and the computing time increased by 50% . k and l shell flurorescences contribute to 40% of the total number of @xmath32 tracks in the complete simulation . most of the fluorescence reactions ( 85% ) occurs in the core where are located the elements with a high atomic number such as the fuel elements . the respective contributions of the gamma rays created by bremsstrahlung in the core , pool and casemate are 71% , 27% and 2% with typically energies lower that 3 mev and then could be neglected for the background computation in the detector casemate . the neutrons emitted by the core are thermalized in the pool . the influence of the photonuclear reactions and especially the neutron production through the ( @xmath32,n ) reactions is also studied . only photonuclear reactions with a reaction threshold below 8 mev are considered because it corresponds to an important decrease in the gamma flux , as shown in fig . [ fig : gammaspectrum ] . the comparison of c , cp and cpw simulations shows that 85% of the photonuclear reactions occur in the concrete walls , mostly ( @xmath32,n ) reactions on @xmath39k . this secondary neutron production could partly explain the neutron background measured in the detector casemate by the nucifer collaboration@xcite and which increases with the reactor power . further studies are required to know the exact percentage of @xmath39k in the concrete walls and the access to a detailed composition of the different concretes used in the osiris facility , and are beyond the scope of this paper . in fig . [ fig : gammarate ] , the gamma flux rate expected in the detector casemate is calculated with mure for a reactor cycle of 20 days considering a fuel composition at equilibrium obtained with the refueling scenario presented in the section iii . no energy threshold is applied on the gamma detection . the error bars are statistical and computed using the number of @xmath32 tracks . from this calculation , the gamma flux rate in an antineutrinos detector of 10 m@xmath40 surface is estimated to be 2.5 mhz . a more realistic fuel composition could be used in the future to improve the normalisation of the gamma flux rate but this first result is already in good agreement with the order of magnitude of 5 mhz measured in the casemate by the nucifer collaboration@xcite . the simulation of the osiris research reactor ( cea - saclay , france ) , was performed with the mure code , including the core , the pool and a detector casemate located behind concrete walls where the nucifer antineutrino detector has been placed at 7 m from the core . the implementation of the periodic refueling of the core every 20 days by seven - th made it possible to carry out a first study of the emitted antineutrino spectrum which could be measured at such a research reactor . this core simulation was benchmarked with the neutron energy spectrum calculated in an irradiation cell by the deterministic code apollo2 . the antineutrino energy spectrum and flux were computed using the summation method , coupled for the first time to the modelization of a reactor core evolution . the variation of the detected antineutrinos along a 20 days operation cycle was calculated . it was shown that the contribution of @xmath2pu and @xmath14pu to the antineutrinos spectrum is weak , due to the 20% fuel enrichment in @xmath0u , which is a common enrichment at research reactors . the emitted antineutrino energy spectrum associated to the reactor fuel at equilibrium is provided and could be used by experiments studying fundamental properties of neutrinos at such facilities . off - equilibrium effects affecting the antineutrino spectral shape due to neutron capture and time evolution of the fission product concentrations in - core were computed , as well as the impact of actinides created along the operation of the reactor . the obtained corrections could be applied to antineutrino energy spectra computed by conversion of the ill reference beta spectra@xcite like the ones from@xcite in the frame of any sterile neutrino experiment located at a research reactor using a similar fuel enrichment . in a second part , a detailed study of the gamma ray background was performed . this work allowed a better understanding of : the contribution of the fluorescence reactions of fuel , the importance of bremsstrahlung reactions for the gamma rays present in core and their negligible contribution in the casemate , and the origin of the neutron background measured in the casemate . the latter does not come directly from the core but rather from ( @xmath32,n ) reactions on @xmath39k nuclei present in the concrete walls . overall the model succeeded to reproduce the gamma background rate level measured at the location of the nucifer detector . the gamma energy spectra produced by the reactor core , and after propagation up to the detector location are provided , as these results could be used to compute the sensitivity of short baseline reactor neutrino experiments foreseen in the frame of the search of potential sterile neutrinos . the authors thank the needs challenge and the gedepeon research groupment for their financial support . the authors thank the people from the osiris reactor for providing useful information . 9 j. beringer _ et al . _ , particle data group , phys . rev . * d86 * , 010001 ( 2012 ) . v. a. korovkin _ et al . _ , atom . ener . * 65 * , 3 , 712 ( 1988 ) . iaea report sg - eqgnrl - rp-0002 ( 2012 ) . e. christensen , p. huber , p. jaffke and t. shea , phys . lett . * 113 * , 042503 ( 2014 ) . th . et al . _ , phys . rev . * c83 * , 054615 ( 2011 ) . p. huber , phys . rev . * c84 * , 024617 ( 2011 ) . g. mention _ et al . _ , phys . rev . * d83 * , 073006 ( 2011 ) . k. n. abazajian _ et al . _ , http://arxiv.org/abs/1204.5379 . m. fallot , nuclear data sheets * 120 * , 137(2014 ) . a. drukier and l. stodolsky , phys . rev . * d 30 * , 2295 ( 1984 ) . v. i. kopeikin , l. a. mikalyan , v. v. sinev , physics of atomic nuclei * 63 * , issue 6 , 1012 ( 2000 ) . . https://www.iaea.org/ourwork/st/ne/nefw/technical-areas/rrs/conversion.html . k. schreckenbach _ et al . _ , phys . lett . * 99b * , 251 ( 1981 ) . k. schreckenbach _ et al . _ , phys . lett . * 160b * , 325 ( 1985 ) . f. von feilitzsch , a. a. hahn and k. schreckenbach , phys . * 118b * , 162 ( 1982 ) . et al . _ , b218 * ( 1989 ) 365 . et al . _ , . lett . * 112 * , 122501 ( 2014 ) . a. c. hayes _ et al . _ , * 112 * , 202501 ( 2014 ) . double chooz and reno presentations at the xxvi international conference on neutrino physics and astrophysics , boston ( 2014 ) conference http://neutrino2014.bu.edu/ , daya bay presentation at ichep 2014 conference , http://ichep2014.es/. y. abe _ et al . _ , phys . rev . * d 86 * , 052008 ( 2012 ) . j.k . et al . _ , * 108 * , 191802 ( 2012 ) . an _ et al . lett . * 108 * , 171803 ( 2012 ) . m. fallot _ et al . _ , lett * 109 * , 202504 ( 2012 ) . a .- a . et al . _ , lett . * 115 * , 102503 ( 2015 ) . http://www.cad.cea.fr/rjh/add-on/osiris_gb.pdf . mure , mcnp utility for reactor evolution ( 2009 ) , http://www.nea.fr/tools/abstract/detail/nea-1845 . o. meplan , tech . lpsc * 0912 * and ipno-09 - 01 ( 2009 ) . https://mcnpx.lanl.gov/. f. malouch , thse de doctorat , rapport cea * r-6034 * ( 2003 ) . v. kopeikin _ _ , phys . at . nucl . * 67 * , 1892 ( 2004 ) . v. mclane _ , endf / b - vi summary documentation supplement i , endf / he - v1 summary documentation ( 1996 ) . c. jones _ et al . _ , phys . rev . * d86 * , 012001 ( 2012 ) . fission rates calculation for the double chooz experiment , a. onillon _ et al . _ , in preparation . r. sanchez _ et al . . sc . and eng . * 100 * , 352 ( 1988 ) . oecd / nea data bank , 2000 . the jef-2.2 nuclear data library . jeff report 17 . nea / oecd publications , paris . private communication , cea / den / osiris , 2011 . a. algora _ et al . _ , lett . * 105 * , 202501 ( 2010 ) . p. vogel and j. f. beacom , phys d60 * , 053003 ( 1999 ) . m. b. chadwick _ et al . _ , data sheets * 107 * , 2931 ( 2006 ) . j. w. durkee jr . _ et al . _ , prog . in nucl * 51 * , 813 ( 2009 ) . private communication , the nucifer collaboration , 2013 . p4.2 cm p4.2 cm p4.2 cm p4.2 cm * energy ( mev ) & * @xmath15.s@xmath7 & * energy ( mev ) & * @xmath15.s@xmath7 * 0.05 & * 1.740189e+17 & * 6.15 & * 5.893318e+15 * 0.15 & * 4.430448e+17 & * 6.25 & * 5.178195e+15 * 0.25 & * 6.231897e+17 & * 6.35 & * 4.691593e+15 * 0.35 & * 6.231105e+17 & * 6.45 & * 4.252763e+15 * 0.45 & * 5.502131e+17 & * 6.55 & * 3.816455e+15 * 0.55 & * 5.699411e+17 & * 6.65 & * 3.380173e+15 * 0.65 & * 5.622272e+17 & * 6.75 & * 2.979807e+15 * 0.75 & * 5.633882e+17 & * 6.85 & * 2.652472e+15 * 0.85 & * 5.703429e+17 & * 6.95 & * 2.293995e+15 * 0.95 & * 5.550602e+17 & * 7.05 & * 1.995576e+15 * 1.05 & * 5.232802e+17 & * 7.15 & * 1.601314e+15 * 1.15 & * 4.73154e+17 & * 7.25 & * 1.401966e+15 * 1.25 & * 3.965951e+17 & * 7.35 & * 1.163696e+15 * 1.35 & * 3.643601e+17 & * 7.45 & * 9.648392e+14 * 1.45 & * 3.527266e+17 & * 7.55 & * 8.090379e+14 * 1.55 & * 3.385362e+17 & * 7.65 & * 6.895844e+14 * 1.65 & * 3.230806e+17 & * 7.75 & * 5.690034e+14 * 1.75 & * 3.07326e+17 & * 7.85 & * 4.659756e+14 * 1.85 & * 2.887332e+17 & * 7.95 & * 3.686017e+14 * 1.95 & * 2.691265e+17 & * 8.05 & * 2.487741e+14 * 2.05 & * 2.547182e+17 & * 8.15 & * 1.229085e+14 * 2.15 & * 2.350878e+17 & * 8.25 & * 1.057696e+14 * 2.25 & * 2.183044e+17 & * 8.35 & * 8.184408e+13 * 2.35 & * 2.034623e+17 & * 8.45 & * 7.14895e+13 * 2.45 & * 1.89149e+17 & * 8.55 & * 5.35679e+13 * 2.55 & * 1.775137e+17 & * 8.65 & * 4.750343e+13 * 2.65 & * 1.627337e+17 & * 8.75 & * 4.17801e+13 * 2.75 & * 1.534323e+17 & * 8.85 & * 3.666975e+13 * 2.85 & * 1.433899e+17 & * 8.95 & * 3.030192e+13 * 2.95 & * 1.331654e+17 & * 9.05 & * 2.578095e+13 * 3.05 & * 1.258777e+17 & * 9.15 & * 2.236928e+13 * 3.15 & * 1.183501e+17 & * 9.25 & * 1.948864e+13 * 3.25 & * 1.093634e+17 & * 9.35 & * 1.696136e+13 * 3.35 & * 1.012903e+17 & * 9.45 & * 1.457185e+13 * 3.45 & * 9.299688e+16 & * 9.55 & * 1.155702e+13 * 3.55 & * 8.543704e+16 & * 9.65 & * 9.808602e+12 * 3.65 & * 7.804873e+16 & * 9.75 & * 8.031613e+12 * 3.75 & * 7.258166e+16 & * 9.85 & * 6.048608e+12 * 3.85 & * 6.745682e+16 & * 9.95 & * 4.008019e+12 * 3.95 & * 6.190826e+16 & * 10.05 & * 3.61989e+12 * 4.05 & * 5.58247e+16 & * 10.15 & * 3.204932e+12 * 4.15 & * 5.107217e+16 & * 10.25 & * 2.899176e+12 * 4.25 & * 4.615847e+16 & * 10.35 & * 2.667792e+12 * 4.35 & * 4.232445e+16 & * 10.45 & * 2.444314e+12 * 4.45 & * 3.819769e+16 & * 10.55 & * 2.232537e+12 * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * 4.55 & * 3.439567e+16 & * 10.65 & * 2.029009e+12 * 4.65 & * 3.086262e+16 & * 10.75 & * 1.827596e+12 * 4.75 & * 2.807835e+16 & * 10.85 & * 1.638106e+12 * 4.85 & * 2.579924e+16 & * 10.95 & * 1.46054e+12 * 4.95 & * 2.333401e+16 & * 11.05 & * 1.292428e+12 * 5.05 & * 2.134674e+16 & * 11.15 & * 1.130929e+12 * 5.15 & * 1.938639e+16 & * 11.25 & * 9.768183e+11 * 5.25 & * 1.740206e+16 & * 11.35 & * 8.304093e+11 * 5.35 & * 1.524291e+16 & * 11.45 & * 6.587883e+11 * 5.45 & * 1.36163e+16 & * 11.55 & * 5.426998e+11 * 5.55 & * 1.213131e+16 & * 11.65 & * 4.366619e+11 * 5.65 & * 1.089618e+16 & * 11.75 & * 2.559623e+11 * 5.75 & * 9.716239e+15 & * 11.85 & * 2.099621e+11 * 5.85 & * 8.659354e+15 & * 11.95 & * 1.689463e+11 * 5.95 & * 7.50986e+15 & & * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * [ [ relative - difference - between - the - anti - neutrino - energy - spectrum - computed - with - the - full - reactor - simulation - and - the - one - computed - with - individual - uranium - and - plutonium - spectra - weigthed - by - their - fission - rate - for - a - fuel - at - equilibrium - after-210-days - of - operation - after - core - start - up . ] ] relative difference between the anti - neutrino energy spectrum computed with the full reactor simulation and the one computed with individual uranium and plutonium spectra weigthed by their fission rate , for a fuel at equilibrium after 210 days of operation after core start - up . ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ * 9.065 & * 1.674385e+11 & * 9.075 & * 1.707822e+11 * 9.085 & * 1.734625e+11 & * 9.095 & * 1.654341e+11 * 9.105 & * 8.616051e+10 & * 9.115 & * 1.155631e+09 * 9.125 & * 2.009821e+09 & * 9.135 & * 1.208225e+09 * 9.145 & * 1.845366e+09 & * 9.155 & * 1.286536e+09 * 9.165 & * 1.305606e+09 & * 9.175 & * 1.058063e+09 * 9.185 & * 1.061964e+09 & * 9.195 & * 9.791538e+08 * 9.205 & * 1.136450e+09 & * 9.215 & * 9.122502e+08 * 9.225 & * 9.122749e+08 & * 9.235 & * 2.059946e+09 * 9.245 & * 9.669593e+08 & * 9.255 & * 5.374402e+10 * 9.265 & * 9.926257e+10 & * 9.275 & * 1.024625e+11 * 9.285 & * 9.926941e+10 & * 9.295 & * 1.003302e+11 * 9.305 & * 5.550827e+10 & * 9.315 & * 1.460669e+08 * 9.325 & * 1.133726e+09 & * 9.335 & * 1.771312e+08 * 9.345 & * 4.236509e+08 & * 9.355 & * 6.491263e+08 * 9.365 & * 5.813655e+07 & * 9.375 & * 1.647950e+08 * 9.385 & * 2.632036e+08 & * 9.395 & * 2.513725e+08 * 9.405 & * 3.494458e+08 & * 9.415 & * 2.818384e+08 * 9.425 & * 4.144856e+08 & * 9.435 & * 2.633242e+08 * 9.445 & * 3.332542e+08 & * 9.455 & * 1.944464e+07 * 9.465 & * 6.965541e+08 & * 9.475 & * 9.258739e+07 * 9.485 & * 1.055700e+08 & * 9.495 & * 5.605511e+08 * 9.505 & * 2.498472e+08 & * 9.515 & * 8.267858e+07 * 9.525 & * 2.915337e+07 & * 9.535 & * 1.604175e+08 * 9.545 & * 7.775750e+07 & * 9.555 & * 9.802458e+07 * 9.565 & * 1.280754e+08 & * 9.575 & * 5.073984e+08 * 9.585 & * 7.980629e+08 & * 9.595 & * 4.329017e+08 * 9.605 & * 5.345833e+08 & * 9.615 & * 1.665356e+08 * 9.625 & * 3.253000e+08 & * 9.635 & * 1.012343e+07 * 9.645 & * 0.000000e+00 & * 9.655 & * 3.471765e+07 * 9.665 & * 1.990008e+07 & * 9.675 & * 0.000000e+00 * 9.685 & * 1.888757e+08 & * 9.695 & * 8.369541e+07 * 9.705 & * 1.704274e+10 & * 9.715 & * 4.168821e+10 * 9.725 & * 3.882021e+10 & * 9.735 & * 3.733299e+10 * 9.745 & * 3.793519e+10 & * 9.755 & * 4.080612e+10 * 9.765 & * 3.823787e+10 & * 9.775 & * 3.822175e+10 * 9.785 & * 3.958152e+10 & * 9.795 & * 4.000108e+10 * 9.805 & * 2.163756e+10 & * 9.815 & * 6.697928e+08 * 9.825 & * 9.388655e+07 & * 9.835 & * 2.337674e+07 * 9.845 & * 9.830200e+06 & * 9.855 & * 6.407189e+07 * 9.865 & * 4.979948e+07 & * 9.875 & * 1.745777e+07 * 9.885 & * 5.032472e+07 & * 9.895 & * 9.315021e+07 * 9.905 & * 1.029589e+07 & * 9.915 & * 2.917632e+08 * 9.925 & * 9.944524e+07 & * 9.935 & * 0.000000e+00 * 9.945 & * 2.565962e+07 & * 9.955 & * 3.595700e+08 * 9.965 & * 9.820181e+07 & * 9.975 & * 6.477107e+07 * 9.985 & * 4.609961e+08 & * 9.995 & * 1.040360e+08 * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

we present a multiplexed , high - resolution ( r,000 median ) spectroscopic survey designed to detect exoplanet candidates in two southern star clusters ( ngc 2516 and ngc 2422 ) using the michigan / magellan fiber system ( m2fs ) on the magellan / clay telescope at las campanas observatory . with 128 available fibers in our observing mode , we are able to target every star in the core half - degree of each cluster that could plausibly be a solar - analog member . our template - based spectral fits provide precise measurements of fundamental stellar properties ( k ) , [ fe / h ] and [/fe ] ( dex ) , and (.3 km / s)and radial velocities ( rvs ) by using telluric absorption features from 7160 to 7290 as a wavelength reference for 251 mid - f to mid - k stars ( 126 in ngc 2516 and 125 in ngc 2422 ) that comprise our survey . in each cluster we have obtained epochs of our targets . using repeat observations of an rv standard star we show our approach can attain a single - epoch velocity precision of 25 m / s to 60 m / s over a broad range of s / n throughout our observational baseline of 1.1 years . our technique is suitable for non - rapidly rotating stars cooler than mid - f . in this paper we describe our observational sample , analysis methodology , and present a detailed study of the attainable precision and measurement capabilities of our approach . subsequent papers will provide results for stars observed in the target clusters , analyze our dataset of rv time - series for stellar jitter and stellar and sub - stellar companions , and consider the implications of our findings on the clusters themselves .

1603.08673

the realization that planetary systems can harbor hot gas giants led to drastic revisions to both theories of planet formation and dynamical evolution . contemporary formation models typically fall into the broad categories of core accretion @xcite or disk instability models @xcite , neither of which has yet emerged as a dominate driver . it also remains unclear if the orbits of the so - called hot - jupiters evolve inward through coupling to a gaseous disk @xcite , via dynamical scattering off clumps and other bodies ( e.g. , * ? ? ? * ; * ? ? ? * ) , or even secular interactions with a distant stellar companion ( kozai cycles ; e.g. , * ? ? ? though much effort has been spent testing these theories they all remain viable , at least for some systems ( see review by * ? ? ? the core reason is that diagnostics ( e.g. the period - eccentricity distribution ) are muddled by the wide and often unknown ages of host stars in field star exoplanetary systems . a direct solution would be to find young exoplanets with ages @xmath01 myr , however the extreme activity of host stars at t tauri ages has thus far inhibited detecting planets ( e.g. @xcite ; ; @xcite ; @xcite ; @xcite ) ; some very young candidates nevertheless remain ( e.g. * ? ? ? these theories do make various distinct predictions regarding the properties of hot gas giants and their orbits for the first @xmath8 gyr ( c.f . * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * vs. ; * ? ? ? * ; but see also * ? ? ? * ) , and though some constraints require direct measures of properties via transits , which are rare , they are relatively more likely for hot gas giants ( @xmath0 5% for hot gas giants ; * ? ? ? indeed , considerable effort has been expended searching for such systems with known ages ( e.g. , * ? ? ? * ; * ? ? ? the first exoplanets orbiting solar - like stars in a cluster ( f , g , or k and on the main sequence ) were announced by @xcite . to date , gas giant exoplanets orbiting main sequence ( ms ) stars in clusters have been discovered via transits ( ngc 6811 * ? ? ? * ) and rv techniques ( m 67 ; * ? ? ? * ; praesepe , * ? ? ? * ; hyades , * ? ? ? * ) , though none of the latter are also transiting systems . one system discovered in the hyades is noteworthy for its distinctly non - zero eccentricity ( @xmath9 ) , which implies that dynamical scattering has likely played a role in its dynamical evolution , at least in the late stages of the process @xcite . it is already clear that even a few hot gas giant exoplanets in open clusters with precise ages have the potential to strongly constrain planet - migration theories . any such systems that also happen to be transiting will produce even more powerful constraints on gas giant formation models by revealing precise information on the sizes , densities and compositions of exoplanets with well - determined ages . indeed , identifying transiting planets in open clusters is a key science goal of nasa s k2 mission @xcite . the practical problem of finding cluster exoplanets is one of efficiency . over the two decades of exoplanetary searches , more than 600 systems have been studied using rv techniques . in contrast , there are only eight known exoplanets orbiting ms stars in open clusters . the deficiency of known exoplanets in clusters is not a consequence of higher stellar rv jitter . measured values of this in open clusters are @xmath10 m / s at @xmath11 myr @xcite , which is roughly an order of magnitude less than the amplitude induced by a typical hot gas giant . rather , clusters tend to be considerably more distant than the closest individual field stars and are hence fainter and consequently harder to monitor . moreover the same technique used for field stars individual spectra , one star at a time is typical of cluster surveys . lengthy campaigns involving thousands of visits are needed to find the comparatively rare cases with detectable velocity amplitudes : only @xmath12% of all fgk stars in the solar neighborhood harbor hot gas giants @xcite and cluster occurrence rates appear similar @xcite so hundreds of targets must be monitored . the marvels survey @xcite is a notable exception to this with its @xmath011,000 stars and is intended to attain similar rv precision to the work we report here though it is not targeted to clusters . in this paper we introduce a different approach that employs highly - multiplexed rv techniques to detect and study exoplanetary systems containing hot gas giants down to @xmath13 that are orbiting stars in clusters that range in age from about 100 myr to nearly 1 gyr . to our knowledge our survey is the first multi - object spectroscopic survey of open cluster stars to obtain rv precisions at the sub - hundred m / s level . to do this , we have used spectra obtained with the michigan / magellan fiber system ( m2fs ; * ? ? ? * ) which allows us to obtain multiplexed , high - resolution ( r @xmath14 ) optical spectra of solar - analogue stars ( spectra types f5v to k5v ) in nearby ( @xmath151 kpc ) clusters . our approach a variant of the telluric - reference approach first proposed by @xcite and subsequently used or studied by ( * ? ? ? * and references therein ) , @xcite , and @xcite models the observed stellar spectra and telluric absorption features to obtain high - precision velocities and stellar parameters . as part of this process we map out the rv precision attainable using telluric lines across a wide range of wavelengths and quantify the assertions of @xcite from some 43 years ago . we are able to measure rvs to 25 m / s for a slowly rotating , bright rv standard star and 3060 m / s for up to 128 stars simultaneously over a half - degree field - of - view . we expect our precision to remain better than @xmath075 m / s for stars as faint as @xmath16 . though our precision is not particularly notable when compared with e.g. the 0.3 m / s precision of harps @xcite or similar instruments , it is comparable to the expected performance of marvels and our magnitude limit is significantly beyond the survey s v=12 limit . harps is capable of 30 m / s precision at @xmath17 but it requires @xmath01 hour exposure 4000 times less efficient than our multi - object approach . the goal of this paper is to describe the details of our methodology . we will address the stellar properties and rv variability of our target stars in future publications . we describe our target selection and observing procedure in section [ observations ] . section [ reductionsec ] describes our image reduction procedure and extraction to 1d spectra . in section [ analysis ] we present our spectral modelingapproach and describe the procedure in detail . section [ resultssec ] analyzes the quality of the measured rvs and stellar parameters . finally section [ discussionsec ] considers the implications of our results for the future of this survey technique . lrl dec ( deg ) & @xmath18 & visible at magellan + dm & 10.0 & bright enough for prvs + age ( myr ) & @xmath19 & limit stellar activity + @xmath20 ( deg ) & @xmath21 & match to m2fs fov + fe / h & @xmath22 & enhance hj formation + @xmath23 & @xmath24 & match to number of fibers + to select our targets , we first created a list of potential star clusters in the catalog of open cluster data @xcite suitable for this project using a small number of basic selection criteria ( table [ selecttable ] ) chosen to identify systems suitable for multi - object spectroscopy of solar analogs that are close enough to have members sufficiently bright to detect small - amplitude rv variability and old enough to limit stellar jitter . we imposed restrictions on cluster size ( @xmath25 in table [ selecttable ] , the core radius derived by @xcite ) and richness to ensure good multiplexing efficiency , and placed limits on age and metallicity to exclude clusters with stars that exhibit excessive surface activity ( thereby mimicking or masking the doppler rv variations of a companion ) and increase the likelihood of gas giant planet formation . from the list of 124 clusters meeting the first four criteria in table [ selecttable ] we identified @xmath030 matching clusters . ultimately we selected ngc 2516 and ngc 2422 as our targets as they were the richest clusters observable during our first observing run in november 2013 . these @xmath0140 myr and @xmath075 myr old open clusters are within 500 pc , rich in solar analogues , have angular sizes that are well - matched to the multiplexing capabilities of m2fs , and have recent photometric membership catalogs sufficiently deep for selecting solar - analog members ( * ? ? ? * ; * ? ? ? * hereafter j01 and p03 ) . individual targets in ngc 2516 were drawn from the sample of stars studied in j01 . we selected all stars they identified as photometric single ( 79 ) or photometric binary ( 47 ) members having colors and magnitudes consistent with f5v k5v spectral types . this sample of 126 stars was then cross - matched with the ucac4 catalog @xcite and the ucac4 coordinates used to prepare the plug plate . in ngc 2422 we selected all objects with colors and magnitudes consistent with f5v k5v in the membership list of p03 . due to a smaller number of members in our field of view , we expanded our selection out in color from the main sequence defined by p03 members using the ucac4 catalog until we had sufficient targets to fill the available fibers , eventually selecting an additional 25 stars in our adopted pointing in this manner . we then cross - matched the p03 targets with ucac4 for astrometry . with 128 available fibers , we are able to target every star in each half - degree field that could plausibly be a solar - analog member . table [ targtable1 ] provides a summary of the clusters , our adopted pointings , and the stellar targets therein . we selected an additional field ( also within ngc 2516 ) with thirty - two sources with magnitudes ( assuming membership ) and colors consistent with b8-a4 stars ( @xmath26 0.113 ; @xmath27 1.7 ) . these stars possess essentially featureless spectra apart from telluric features in our wavelength range and hence can serve as useful probes to monitor the instrumental point spread function ( psf ) over the full field of view of the spectrograph cameras . finally , we selected six stars with similar ras from the gaia rvs catalog @xcite for use as radial velocity standards . one of these hip 48331 was observed repeatedly and is used as a primary reference to track our rv measurement precision over the duration of the program . although this star hosts an exoplanetary companion , the induced rv semi - amplitude is only 0.8 m / s @xcite , far below our measurement precision and so its variability is irrelevant for our purposes . a summary of the standard stars used for this study is presented in table [ stdtargtable ] . lccc ccc cccc ngc 2516 & ... & 7:58:42 & @xmath2860:46:36 & 141 & 346 & 0.11 & 12 & 126 & 11.6815.09 & 0.461.26 + ngc 2422 & m 47 & 7:36:30 & @xmath2814:29:42 & 72 & 491 & 0.07 & 10 & 125 & 12.2016.10 & 0.451.43 + lllrcccccc hip 48331 & 35 & 7.67 & @xmath29 & 0.9 & k5v & @xmath30 & 4.67 & -0.18 & ... + hip 13388 & 2 & 8.09 & @xmath31 & 2.7 & k1v & @xmath32 & 4.59 & -0.15 & 0.02 + hip 10798 & 5 & 6.33 & @xmath33 & 2.7 & g8v & @xmath34 & 4.63 & -0.44 & 0.17 + hip 22278 & 3 & 8.52 & @xmath35 & 3.6 & g5v & @xmath36 & 4.22 & 0.13 & -0.01 + hip 19589 & 1 & 8.46 & @xmath37 & 3.6 & g0v & @xmath38 & 3.75 & -0.17 & 0.13 + hip 31415 & 1 & 7.70 & @xmath39 & 4.5 & f6v & @xmath40 & 3.94 & -0.31 & 0.12 + the michigan / magellan fiber system ( m2fs ) is a multi - object , optical ( 3700 - 9500 ) , fiber - fed spectrograph that can take simultaneous spectra of up to 256 targets over a half - degree field - of - view at a wide variety of resolutions ( @xmath41 ) . the fibers accept light from the sky at the f/11 nasmyth e focal surface of the magellan / clay 6.5 m telescope at las campanas observatory . each fiber samples the sky through a 1.2 arcsecond aperture ; fibers can be packed to within 14 arcseconds of one another with no restrictions within the m2fs field of view and are held in place by plug plates drilled in advance of an observing run using astrometry for the desired targets . the fibers terminate at the focal surface of the collimator / cameras belonging to a pair of identical quasi - littrow spectrographs . for historical reasons , these are identified as the ` red ' and ` blue ' m2fs arms , each fed by 128 fibers . the spectrographs are equipped with traditional and echellette gratings for low and medium resolution work as well as an r2.0 echelle grating with a prism cross - disperser for high - resolution use . filters located just below the fiber termination surface isolate specific orders , necessary for use in echelle and echellette modes . each arm uses an e2v 4k x 4k anti - fringing ccd with 15 @xmath42 m pixels . additional details can be found in @xcite . figure [ sampleim ] shows an example frame from our program . m2fs incorporates a slit mechanism placed just after the ends of the fibers at the spectrograph collimator / camera focal surface ; details of this mechanism are provided in @xcite . for all of the hires observations used in this paper , we employed the 45-micron wide slit ( the narrowest available ) which projects to approximately three pixels at the detectors . for this configuration we found an effective resolution of in 40 - 60k range for most fibers ( c.f figure [ slitlampimage ] ) , subject to focus variations . because the slits are positioned after the fibers we sacrifice roughly 60% of the incident light . for this study , we employed a two order filter with a design passband of 70507370 ( m2fs echelle orders 49 and 50 ) : the use of a filter is necessary to prevent spectral orders from one fiber overlapping with those of another and a two order filter limits us to every second m2fs fiber for a multiplexing factor of 128 . as few target clusters would offer more than about 100 targets in a single m2fs field this decision does not appreciably affect our multiplexing capability . the passband was selected after careful consideration of the optimal wavelength region to carry out telluric - reference rv measurements of solar analogue stars . we used the formalism of @xcite combined with synthetic spectra from the phoenix grid @xcite to estimate the rv uncertainties for slowly - rotating , main - sequence stars with effective temperatures between 4000 and 7000 k for a range of m2fs orders red - ward of about 6800 (where telluric features become common ) . this uncertainty was then added in quadrature with the wavelength reference uncertainty in each order determined by applying the same formalism to the telluric absorption features present in the empirical telluric spectrum from @xcite . results of this analysis are shown in figure [ precisionfig ] : we show our estimates of attainable velocity precision for sun - like , slowly rotating stars observed at two fiducial resolving powers and s / n levels for each m2fs order with appreciable telluric lines . in practice we only obtain useful data from the first portion of order 49 ( 71607290 ) . our filter was manufactured prior to the discovery of a slight difference in the intended and manufactured echelle grating blaze . this rendered data from order 50 to be of little use , truncated the redder portion of order 49 , and limited system throughput by a factor of two ( see figure [ throughputfig ] ) . this limited bandpass was still able to deliver excellent rv precision ( section [ rvprecisionsec ] ) , and manageable s / n for stars in our target clusters . in early 2016 we started verification tests on a new filter designed to cover the 71607360 region , adding an additional 14 major lines and recovering the lost throughput . we anticipate this will improve our achievable velocity precision by @xmath015% where we are not limited by systematics and do not anticipate any negative impact on our program . [ filterissuesec ] since we were looking for radial velocity variations among stars in clusters that might harbor exoplanetary systems , our observing procedure involved repeat observations of our target fields ( table [ targtable1 ] ) with m2fs . to date , we have observed our pointings in ngc 2516 and ngc 2422 12 and 10 times , respectively . we also obtained 35 observations of our principle rv standard hip 48331 along with a small number of observations of the other comparison standards ( table [ stdtargtable ] ) . the specific dates of the observations will be summarized in a later paper describing our time - series spectroscopy ; for the purposes of the present paper we simply note that the data were obtained during runs in november 2013 , february 2014 and december 2014 . as noted earlier , fibers are positioned at the focal surface of the telescope with m2fs using aluminum plug plates that are manually installed and plugged . for a given field , each assigned fiber is positioned at a specific hole in the plate marked for that fiber . we typically deployed 128 fibers for the ngc 2516 and ngc 2422 fields , though with dead or otherwise inactive fibers excluded . once plugged , a typical observation then consists of acquiring the field using a set of ancillary fibers and imaging optics aligned to reference stars in each field . typical total exposure times for each observation used in this paper were 2 hrs and 2.5 hrs per visit for ngc 2516 and ngc 2422 , respectively . most visits consisted of 3 - 5 individual exposures to aid in cosmic - ray removal and to enable measurement of the photon midpoint for barycentric correction . this yielded a median s / n of @xmath43 ( @xmath44 range 15 - 70 ) per 1d extracted pixel ( @xmath45 per resolution element ) . a detailed table of our individual observations and the targets therein will be presented in our next paper . for each observation we obtain calibration data consisting of a thorium - argon - neon lamp exposure and a quartz lamp exposure either before or following the science frame . on some nights during which our targets were observed we also obtain either evening or morning twilight spectra . rv standard observations are performed by placing a single fiber in a standard hole on the fiber plug plate and offsetting from the field center . remaining fibers are left in their positions and see only sky . typically three exposures of two minutes each are used to obtain a spectrum of s / n@xmath46 per extracted 1d pixel . finally we obtained four epochs ( 1 or 2 per observing run ) of the telluric standard calibration field in ngc 2516 . these spectra have a median s / n of 160 and were obtained in 3 to 5 exposures totaling roughly one half hour . basic data reduction follows a mostly traditional path using a custom set of python tools written for m2fs . the quadrant images produced by the ccd s four amplifiers were bias corrected by subtracting the median overscan column and then row . we then converted counts to electrons and used the python implementation of the l.a . cosmic algorithm @xcite to detect cosmic rays : this algorithm takes the laplacian of the image and identifies cosmic rays using their steep intensity gradient . the quadrants were then packed together and stored with a variance frame consisting of electrons plus the square of each quadrant s read noise and a bad - pixel mask . we created a cleaned , summed image by adding the electrons and variances of each pixel across a sequence of frames . masked pixels in each component frame were repaired with their expectation value based on the other frames . a scaling value was computed for each frame to normalize throughput and exposure time variations by using the total time - normalized counts of all spectra as a proxy for throughput variability . bad pixels in each frame were then repaired using the expectation value determined from the good pixels in other frames and the frame scaling values . the variance of the final , summed pixel was inflated appropriately at every impacted pixel . typically one would flat - field the resulting frames , however quartz trace flats of comparable signal - to - noise take an impractical amount of observing time and m2fs does not presently have a means of obtaining uniformly illuminated ccd frames . engineering work shows m2fs ccds are free from large defects and indicate pixel - to - pixel sensitivity variations of about 1.7% and only 0.25% of pixels are significantly hot or cold . we subtracted a combined dark current and scattered light map , an example of which is shown in figure [ exampleim ] along with the image prior to and post subtraction . this map was computed by first modeling and removing the amplifier glow in each corner by fitting a 2d gaussian surface . all remaining pixels within about a standard deviation of the mean light level in the dark regions between bundles of 16 fibers were then selected as `` scattered light '' pixels and used to fit polynomials across the image . the resulting map was gaussian smoothed using a 32 x 64 pixel rectangle ( @xmath0 1.5 x 3 fiber spacings ) and subtracted from the image . without this step , these components would amount to about 150 @xmath47 per 1d pixel in our extracted spectra , ranging from @xmath48 of our extracted signal for our brightest to faintest targets . each processed frame was then extracted using the pyraf task ` apall ` . we first identified the approximate aperture locations and traces using dome flats taken during the day with all usable fibers plugged . the apertures were then median shifted to the locations of the quartz traces taken with each exposure to account for any temperature drift or repositioning errors in the instrument . we then extracted both the science and variance frames without variance weighting using identical apertures . finally we continuum normalized the spectra by iteratively fitting a polynomial , each time excluding points 1 sigma below or 2 sigma above . an example of order 49 for a @xmath49 f5v , g5v , & k5v star in our sample is shown in figure [ exampleextractions ] . we do not wavelength calibrate our spectra in a traditional sense as wavelengths are determined as part of the modeling process . we measured each target s stellar properties ( e.g. @xmath1 , [ fe / h ] , [ @xmath3/fe ] , @xmath5 ) and line - of - sight radial velocity ( rv ) by fitting a model of each extracted , normalized spectrum to the spectrum in 1d pixel space . this approach is similar to the popular gas cell approach where molecular absorption lines from a well - calibrated gas cell ( e.g. @xmath50 , ammonia ) are used as a simultaneous probe of pixel wavelengths and the instrumental point - spread function ( psf ) . here we make use of the abundant atmospheric @xmath51 lines in the 7230 region as the imprint of a giant gas `` cell . '' a variant of this idea was originally proposed by @xcite more than 40 years ago and has been used with success to measure rvs in both the optical and the infrared @xcite . these studies have demonstrated that telluric features are stable to 10 m / s . this should not come as a surprise as all of the water and the general bulk of our atmosphere is within the first 8 - 16 km where typical bulk motions are below 10 m / s and not along the line - of - sight . we will quantify this source of uncertainty when we discuss our achieved rv precision in section [ rvprecisionsec ] . the model is constructed by combining a template of the telluric absorption spectrum , @xmath52 , one or more synthetic stellar spectra , @xmath53}},{\ \mathrm{[\alpha / fe ] } } , v_r\sin(i ) , { \ \mathrm{rv}})$ ] , a synthetic sky emission spectrum , @xmath54 , and a solar spectrum @xmath55 . @xmath56}},[\alpha/{\ \mathrm{fe } } ] , v_r\sin(i ) , { \ \mathrm{rv}})^\beta+\\\gamma\cdot sun(\lambda+s;rv_{\sun}))+\eta\cdot sky(\lambda ) \end{gathered}\ ] ] this model is then resampled onto pixels , convolved with a model of the 1d projection of the psf , and normalized . @xmath57 in the above equations the scalars @xmath58 and vectors @xmath59 and @xmath60 are model parameters which will be described in the next section and are summarized in table [ modeltabletable ] . ccc l wavelength & @xmath61 & @xmath628 & @xmath63 + psf & @xmath59 & 1 & the psf fwhm + & & 3 & @xmath64 + & & 5 & a 5th order hermite parameterization + & & 22 & @xcite parameterization + nomalization & @xmath60 & 12 & @xmath65 + stellar temperature & @xmath1 & 1 & snaps to 100 k grid + iron abundance & [ fe / h ] & 1 & snaps to 0.1 dex grid + alpha abundance & [ @xmath3/fe ] & 1 & snaps to 0.1 dex grid + stellar rotation & @xmath5 & 1 & + radial velocity & rv & 1 & + airmass & @xmath3 & 1 & scale atmospheric transmission + veiling & @xmath66 & 1 & scales stellar absorption lines in unison . + & & & held at unity when fitting stellar parameters + solar flux & @xmath67 & 1 & fractional contribution of solar flux + solar rv & @xmath68 & 1 & + solar offset & s & 1 & offset between phoenix and kurucz wavelengths + sky emission & @xmath69 & 1 & scale skycalc spectrum + our pipeline uses the phoenix grid as it samples a large stellar parameter space ( far beyond our region of immediate interest ) , is the successor to the grid used in @xcite , and in no small part because it is the grid for which we attained the best rv precision for our rv standard . to verify this point we also checked both the @xcite grid and the ambre grid @xcite and found both to result in larger rv measurement errors for our rv standard . prior to modeling we up - sampled the phoenix grid ( see table [ gridtable ] ) in the parameter space relevant to our target stars using linear interpolation by way of the scipy function ` map_coordinates ` . library spectra are normed by the maximum continuum value in the fitting region and linearly interpolated onto a constant @xmath70 grid , adopting the largest step size present in the raw spectrum , just prior to use in the modeling pipeline . during fitting , the surface gravity is tied to the effective temperature via eqn . [ tempmasslumeqn ] which is derived from the mass - luminosity , temperature - luminosity , and mass - radius relations for lower main - sequence stars as our fits do not appear to be particularly sensitive to variations in @xmath71 among our main sequence targets . this weak dependence is seen in other techniques as well : @xcite reports that even variations as large as a half - dex affect @xmath1 by only a few tens of kelvin . we calibrated the relation using solar values corrected for age per the plot in @xcite , @xmath72 , @xmath73 . @xmath74 this input brings with it the astrophysical parameters @xmath1 , [ fe / h ] , [ @xmath3/fe ] , @xmath5 , and rv along with a feature depth parameter @xmath66 which allows fudging the optical depths of all the stellar lines in unison ( we note that this is a simplification as lines are not expected to scale in unison ) . in the event of a spectroscopic binary we can enable multi - component modeling , using two sets of these parameters and an additional multiplicative parameter for the ratio of flux received from the two stars . ccccc phoenix & 100 & 0.5 & 0.5 & 0.2 + resampled phoenix & 100 & 0.1 & 0.1 & 0.1 + we considered two options for the telluric transmission model : the nso empiric transmission spectrum @xcite and the synthetic tapas model @xcite , nominally tailor made for the atmospheric conditions during each observation . the nso spectrum derives from data obtained on the mcmath - pierce solar telescope using the fourier transform spectrograph ( fts ) in the late 1980s . tapas spectra are computed as described in @xcite for the conditions of each exposure . when using tapas spectra in constructing our models for hip 48331 we measured an rv @xmath75 m / s larger than when using the nso spectrum as the template and also observed a reduction in rv precision ( c.f . section [ rvprecisionsec ] ) . this shift is on the order of the uncertainty of the @xmath76 and @xmath51 line positions in the hitran database @xcite and so is perhaps not unexpected given that is is used as the data source for tapas . it is interesting to note that this is also of order the shift caused by mixing @xcite and @xcite air / vacuum relations ( e.g. converting one way with ciddor and back with edln ) , however we are unable to ascertain a list of conversions applied between the original data and the output tapas spectrum and are unable to offer any firm conclusions regarding the source of the shift . we do measure better @xmath77 values when using tapas spectra and suggest that the tapas pipeline models differences in atmospheric line strengths between kitt peak and las campanas well . given the reduced rv precision we used the nso fts data for our analysis . this input brings with it the parameter @xmath3 to logarithmically scale the absorption features as a proxy for airmass . we used the eso skycalc tool @xcite to obtain night sky emission spectra for our wavelength region . these spectra match the locations of the night sky emission lines well , though they do not always perfectly match their relative strengths . this input adds a multiplicative scaling parameter , @xmath69 , to adjust the predicted count rate . the modeling code also includes inputs for the instrumental dispersion relation , point spread function ( psf ) , and allows for inaccuracies in our continuum normalization . the dispersion relation is a set of legendre polynomial coefficients ( @xmath61 ) which yield the wavelengths at each extracted pixel . we used a second set of legendre polynomial coefficients ( @xmath60 ) to compute a normalization polynomial that accommodates errors in continuum normalization during extraction . the psf is widely understood ( c.f . e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) to have a significant impact on the precision with which line centroids can be recovered and ultimately the rv precision . we investigated this effect by modeling our psf with a gaussian , a gaussian with width quadratically varying along the order , the multi - gaussian parameterization of @xcite , and a 5th order gauss - hermite series @xcite and then determining which prescription minimized the rv standard deviation for our rv standard . to compute the model described in equations [ modeleq1 ] and [ modeleq2 ] the code will first fetch the synthetic spectrum of nearest temperature , surface gravity , iron , and @xmath3-element abundance from our grid ( recall @xmath71 is computed per eqn . [ tempmasslumeqn ] ) along with the telluric absorption , solar , and pointing dependent emission spectra . these spectra are all in excess of @xmath78 . the telluric transmission spectrum is scaled logarithmically by @xmath3 and the sky emission and solar spectra are scaled multiplicatively by @xmath69 and @xmath67 , respectively . the solar spectrum is doppler shifted by multiplying the wavelength grid by the appropriate doppler factor . the stellar spectrum is normalized by the maximum flux in the wavelength region to perform a simple continuum normalization while preserving the slight blackbody effect in our narrow region . in the case of a multi - star fit normalization of each spectrum is still carried out in this manner . the spectrum is scaled logarithmically by @xmath66 to account for any mean discrepancy in line depth and is then rotationally broadened via convolution with a kernel computed based on equation 17.12 in @xcite . finally the spectrum is doppler shifted by multiplying the wavelength grid by the doppler factor . when computing a binary model this process is repeated for each star and a weighted average of the two spectra is taken after they are placed on the sub pixel grid in the following step . in this situation the weight is an additional free parameter ( constrained between 0 and 1 ) . we investigated combining the stellar components in a manner that enforces their relative flux from the synthetic library however this often led to grossly inappropriate minima . when modeling a telluric standard spectrum we model the stellar component as unity . if including scattered solar light the solar spectrum is multiplicatively scaled by @xmath67 , s is added to the solar wavelength grid , and then the grid is multiplied by the appropriate doppler factor . each of the components is linearly interpolated onto a @xmath79 pixel wavelength grid computed from the dispersion relation . we note that cubic spline interpolation significantly decreased our rv stability . these component spectra are then combined as in eqn . [ modeleq1 ] . this model is then convolved with the psf kernel representing the instrument s point spread function as mangled by our simplistic extraction . the kernel size was selected such that the kernel is less than @xmath80 at the window edge and the kernel is always constrained to be positive . to maintain a modicum of speed pixel dependent convolutions are carried out via a fortran subprogram . when using an asymmetric psf we noted that the center of the enclosed power was not generally located at the central sub - pixel nor at some constant offset . we observed a shift for typical , good fits of @xmath81 pixels redward with the hermite parameterization . constraining the centroid to @xmath82 pixels of the center confers substantial improvements to rv precision ( see section [ rvprecisionsec ] ) . in the limit of a linear dispersion relation this effect would correspond to a simple pixel shift and would not be expected to have any impact on results , however our wavelength solutions are not linear . our code constrains the centroid by shifting the psf kernel and hence constrain the center of enclosed power to within @xmath83 m / s of the central sub - pixel . interpolated shifts or a psf parameterization which constrains the enclosed power may be worth future investigation . after convolution , the sub - pixel values are averaged to yield pixel values and the model is divided by the normalization polynomial , yielding a model spectrum . to determine the optimal model , we used the @xmath77 as a merit function . the fitter computes the weighted mean square error of our model with the normalized spectrum , masking pixels based on a wavelength mask ( e.g. for sky lines , if desired ) , an rv dependent wavelength mask for stellar lines , and a pixel mask for detector defects and the littrow ghosts . wavelength masks are additive : for stability , once masked , a change to the wavelength solution will not cause a pixel to unmask . the weights are computed as the ratio of the square of the continuum normalization to the variance spectrum . we noted a significant upward trend in our best - fit @xmath77 with increasing signal ( c.f . figure [ cstrend ] ) , which we attribute to an improved ability to identify finer errors in our computed model . visual inspection shows that phoenix spectra consistently mismatch stellar line depths and with increasing s / n these mismatches become increasingly significant . prior to fitting we visually reviewed all @xmath84 spectra and excluded spectra with average continuum s / n less than 12 ( per pixel ) . we limited all our fits to the extent of order 49 with continuum s / n greater than 12 or the columns between pixels 25 and 2700 ( @xmath85 ) . these column extents were selected such that we have a slight margin at either end with which to estimate the continuum level at the order edges beyond the fitting extent . generally fits were to all 2675 pixels in this region . we masked pixels affected by extraction artifacts , continuum normalization errors , uncorrected cosmic rays , or littrow ghosts . after fitting we inspected the results for any failures ( generally due to a poor initial rv guess ) and either corrected them , flagged them to handle as exceptional cases , or excluded them from analysis . we optimized the model in stages . 1 ) for a subset of spectra we first obtained an initial guess for the wavelength solution by eye . these guesses were used to bootstrap initial relations for the wavelength solution in one frame of each arm on each run . the initial psf width was chosen such that it coincides with m2fs s nominal resolving power in our configuration without asymmetry and an initial spectral type was chosen assuming solar abundance and using the ( b - v)-@xmath1 relation of @xcite with reddening corrected values for b - v . these parameters were then used as the initial values for a round of fits from which we constructed a predictive model of the dispersion parameters as function of m2fs arm , ccd trace position , and night . we found the @xmath86 order wavelength parameters are neither a function of instrument temperature , ( mis)focus , or run and thus adopted a simple polynomial model as a function of aperture position based on the best fit values for all @xmath87 usable spectra . we only adopt the mean values as an initial guess for the @xmath88 order parameters and note that the wavelength zero point parameter is predicted for each exposure separately . 2 ) we then refit all our spectra with the initial wavelength parameters determined in phase one , holding the @xmath86 order wavelength parameters fixed , and adopted the inverse variance weighted means of the best - fit spectral type parameters and @xmath5 values for each star . 3 ) we performed a final round of fits still holding higher order wavelength parameters fixed , now along with the spectral type parameters and @xmath5 . we used the rvs from this final fit as the values we adopt for each star . except in the case of large amplitude binaries the rv was always started from the adopted multi - epoch mean . section [ resultssec ] describes the precision with which we recover these values . as our initial wavelength guesses are crude we investigated the impact priming our wavelength solution parameters with thar calibration fits to verify we were not introducing a fitting bias . for this test we refit all of the rv standard spectra using the iraf identify task solutions to corresponding thar data for the initial wavelength solution . we found no impact . holding the higher order wavelength parameters fixed improves rv precision by approximately @xmath89 m / s at all signal - to - noise levels ( c.f . sec . [ rvprecisionsec ] ) . an example fit is shown in figure [ fitexamplefig ] . optimization was carried out using the ` mpfit ` @xcite package to minimize the weighted errors for each unmasked pixel . in previous iterations of our software we used the amoeba minimizer : the downhill - simplex optimizer @xcite appears more tolerant of poor initial guesses , but minimization takes a greater number of function evaluations , does not yield a parameter covariance matrix , and requires parameter limits be hacked on as they are not inherent to the algorithm . we also investigated using the mcmc core of @xcite , and briefly explored a genetic fitter , the latter of which proved to be of similar quality but highly inefficient . we found these various methods of optimization to all be of comparable end result but with significantly prolonged computation time . we estimated the statistical uncertainty of our stellar parameters from the distribution of best fit values relative to their multi - epoch means . we used the results from the second stage of our fitting pipeline ( where stellar properties are allowed to vary from epoch to epoch , c.f sec . [ modelfittingsec ] ) , exclusive of spectroscopic binaries , to obtain fits to 2283 spectra of 214 targets ( see table [ prectargtable ] ) with which we computed the differences between the single epoch values and the adopted multi - epoch values of the stellar parameter . we then performed kernel density estimation on each parameter and computed confidence intervals . the resulting pdfs are shown in figures [ teffprecision ] , [ fehprecision ] , [ alphaprecision ] , and [ vsiniprecision ] and the @xmath90 confidence intervals given in table [ stellarpropprectable ] . in addition to the overall pdfs , the plots give pdfs for subsamples grouped by spectral type and rotation rate ( @xmath91 km / s ) . the selection of 8 km / s as a grouping is largely arbitrary , though corresponds with the point where @xmath5 becomes the dominate source of line broadening relative to the instrumental psf . though we do not explicitly show pdfs for groupings in s / n , the pdfs for spectral type show this by proxy ; later spectral types are fainter and have lower s / n spectra . we saw no evidence that the psf form ( e.g. gaussian vs. hermite series ) affected the measured values or their uncertainties . we estimated the accuracy of our technique for determining @xmath1 , [ fe / h ] , and [ @xmath3/fe ] by comparing the values we measured with those in the literature . tables [ stdparvals ] gives our values for the six standard stars along with those from the literature . tables [ stderrors ] then reports the differences in these values : we adopt the averages therein as an estimate of our systematic uncertainties . as an additional check we fitted @xmath92 twilight spectra and report the values and differences thereby obtained . though these values are of comparable quality , we excluded them from our average as the large number of spectra would heavily bias the results . we tested our @xmath5 accuracy by comparing our values with those reported in @xcite , with which we have thirty - seven targets in ngc 2516 in common . our values agree to within 5 km / s for all but four stars , which are all spectroscopic binaries . for the remaining 33 stars our adopted , multi - epoch mean values agree with a standard deviation of 2.2 km / s . we recovered the correct @xmath5 to better than 0.1 km / s in fits to our twilight spectra . our code is not able to reliably measure @xmath5 values below @xmath93 km / s ( roughly one third of our velocity resolution ) . based on this analysis , we report our fits yield typical single - epoch precisions of 75 k , 0.05 dex , and 0.75 km / s for @xmath1 , [ fe / h ] and [ @xmath3/fe ] , and @xmath5 . we find this translates to mean multi - epoch precisions of @xmath2 k , @xmath4 dex for both [ fe / h ] and [ @xmath3/fe ] , and @xmath60.3 km / s for @xmath5 . our @xmath1 values are typically cooler than available literature data for our standards by @xmath025 k and we find a similar offset when fitting twilight spectra . iron abundance values appear elevated by a tenth dex but are driven entirely by hip 48331 : excluding hip 48331 @xmath94[fe / h ] becomes @xmath95 dex , consistent with our twilight fits . we do not see any evidence of a systematic offset in [ @xmath3/fe ] or @xmath5 . lccccc ngc 2516 & 11.68 15.09 & 0.46 1.17 & 108 & 18 + ngc 2422 & 12.19 16.05 & 0.45 1.31 & 106 & 19 + total & 11.68 16.05 & 0.45 1.31 & 214 & 37 + lcccccccc @xmath1 ( k ) & @xmath2844 & + 59 & @xmath2862 & + 74 & @xmath2895 & + 121 & @xmath2882 & + 163 + @xmath96 $ ] ( dex ) & @xmath280.06 & + 0.05 & @xmath280.05 & + 0.06 & @xmath280.07 & + 0.06 & @xmath280.07 & + 0.05 + @xmath97 $ ] ( dex ) & @xmath280.04 & + 0.04 & @xmath280.05 & + 0.04 & @xmath280.07 & + 0.04 & @xmath280.07 & + 0.04 + @xmath5 ( km / s ) & @xmath280.5 & + 0.5 & @xmath280.3 & + 0.4 & @xmath280.5 & + 0.7 & @xmath281.0 & + 0.9 + + mean s / n & & & & lrrrrr + this work & @xmath98 & ( 4.7 ) & @xmath99 & @xmath100 + s05 & @xmath101 & @xmath102 & @xmath103 & ... + s08 & @xmath104 & @xmath105 & @xmath106 & ... + n09 & s08 & s08 & s08 & @xmath107 + c11 & @xmath30 & @xmath108 & ... & ... + a12 & s08 & s08 & s08 & @xmath109 + t13 & @xmath110 & @xmath111 & @xmath112 & ... + + + this work & @xmath113 & ( 4.6 ) & @xmath114 & @xmath115 + c11 & @xmath116 & @xmath117 & @xmath118 & @xmath119 + s08 & @xmath120 & @xmath121 & @xmath122 & ... + n09 & s08 & s08 & s08 & @xmath123 + + + this work & @xmath124 & ( 4.6 ) & @xmath125 & @xmath126 + v05 & @xmath127 & @xmath128 & @xmath129 & ... + c11 & @xmath34 & @xmath130 & @xmath131 & @xmath132 + + + this work & @xmath133 & ( 4.5 ) & @xmath134 & @xmath135 + c11 & @xmath136 & @xmath137 & @xmath138 & @xmath139 + + + this work & @xmath140 & ( 4.5 ) & @xmath141 & @xmath142 + c11 & @xmath38 & @xmath143 & @xmath144 & @xmath145 + k13 & @xmath146 & @xmath147 & @xmath148 & @xmath149 + + + this work & @xmath150 & ( 4.4 ) & @xmath151 & @xmath152 + c11 & @xmath40 & @xmath153 & @xmath154 & @xmath155 + + + this work & @xmath156 & ( 4.5 ) & @xmath157 & @xmath158 lcrrrr hip 48331 & k5v & @xmath159 & @xmath160 & @xmath161 + hip 13388 & k1v & @xmath162 & @xmath163 & @xmath164 + hip 10798 & g8v & @xmath165 & @xmath166 & @xmath167 + hip 22278 & g5v & @xmath168 & @xmath169 & @xmath170 + hip 19589 & g0v & @xmath171 & @xmath172 & @xmath173 + hip 31415 & f6v & @xmath174 & @xmath175 & @xmath176 + + average & & @xmath177 & @xmath178 & @xmath179 + + twilights & g2v & @xmath180 & @xmath181 & @xmath182 + radial velocity variations of stars without a companion stem from one of five sources : ( 1 ) an inherent photon noise error ( @xmath183 ) arising from the s / n and the number and shape of the stellar and telluric lines , ( 2 ) an instrumental error ( @xmath184 ) based on the characteristics of m2fs spectra , ( 3 ) an error contribution due to our analysis ( @xmath185 ) , ( 4 ) intrinsic stellar variability ( @xmath186 ) caused by stellar activity ( e.g. stellar flares or star spots ) , and ( 5 ) variability in the bulk atmospheric motion along the line of sight that introduces a doppler shift on our wavelength reference ( @xmath187 ) . we assume that all five sources add in quadrature to produce the observed dispersion ( @xmath188 ) , as follows : @xmath189 under this assumption , the observed velocity dispersion of a star with a known @xmath186 and observed under conditions with a known @xmath187 can be used to estimate the quadrature sum of the first three error terms , which we refer to as an effective measurement error , @xmath190 , @xmath191 here we focus on estimating @xmath190 as a function of s / n , based on observations of the standard star hip 48331 . we observed hip 48331 35 times on 19 different nights ; 9 nights have more than 1 epoch . of these we use the 31 spectra with s / n above 200 and r @xmath192 . eighteen spectra were obtained using the red m2fs arm and 13 using the blue arm . the s / n of these spectra span between 200 and 300 , with a median of 240 . the resolving power of these spectra range from 40,000 to 64,000 , with a median of 55,000 due to variations in the psf width across the m2fs detector and soft instrument focus during early observing runs . the rv measurements of hip 48331 are illustrated in figure [ rvmeasures ] ; these values have a standard deviation ( @xmath188 ) of 23 m / s . to simulate lower s / n spectra that are more representative of the open cluster stars surveyed here , we generated lower s / n versions of these 31 spectra and recomputed the best fit models and rvs from which new @xmath188 can be calculated . we generated the the lower s / n spectra by sampling a poisson process at each pixel with expectation value of the measured electrons multiplied by the desired fractional reduction in mean s / n : e. g. @xmath193 where @xmath194 is the number of electrons measured at the @xmath195 pixel and prime denotes the new values . we also ensured that the simulated variance spectra included in an appropriate amount of gaussian noise to include the effects of detector read noise . the resulting spectra have s / n levels of @xmath0150 , 100 , 80 , 60 , 50 , 40 , and 15 . the spectra were then fit as described in section [ analysis ] , treating each s / n level independently . this resulted in eight rv time - series ( one for each s / n level ) with each standard deviation yielding a measurement of @xmath188 at that s / n level . we also computed @xmath183 for each of the 248 spectra by applying the algorithm described in @xcite to the telluric and stellar components of each best - fit model , adding the results in quadrature . to obtain @xmath190 from the eight @xmath188 values calculated above , we subtracted off a stellar variability of @xmath196 m / s @xcite and an atmospheric variability of @xmath197 m / s ( determined as shown later in this section , see also figure [ noaawindfig ] ) . these values can be compared directly to the the mean @xmath183 values for each of the eight s / n bins . oddly , we found that would result in imaginary errors below a s / n of @xmath198 . in figure [ rvprecisionmeas ] we plot both @xmath190 and @xmath183 , which shows that we measure our rvs with greater precision that anticipated at low s / n . in figure [ rvprecisionmeas ] , we also show the ratio of @xmath190 to the mean of @xmath183 at each s / n bin . as an additional reference we also plot the ratio of each bin @xmath190 to each of the @xmath183 in that bin . this suggests an approximately linear relation between our measurement error and the @xmath183 value we computed for each spectrum . we adopted errors for the ratio from two sources : ( 1 ) the standard deviation of @xmath183 in each s / n bin contributes directly and ( 2 ) an estimate of the error in @xmath188 that was obtained by computing our best - fit models with a small number of slightly perturbed initial rvs for each spectrum in each bin , adopting the standard deviations of the resulting @xmath188 values as an uncertainty on @xmath190 in each s / n bin . we fit the ratio of @xmath190 to the mean of the @xmath183 for each s / n bin and use the result as a scaling relation to convert @xmath183 to @xmath190 provided a s / n . this technique allows us to account for some , if not all , of the increased uncertainty in spectra that are at a lower resolution ( e.g. due to mis - focus ) than the typical rv standard observation , are of more rapidly rotating stars , or otherwise possess a different number or strength of stellar lines . the errors in figure [ rvmeasures ] have been scaled in this manner . from this analysis we find m2fs has a limiting rv precision of about 25 m / s , though the @xmath183 values we computed at high s / n suggest an additional 15 m / s precision gain may be possible at higher s / n ratios . a potential culprit in our modeling process is as yet unclear . in figures [ rvprecisionpsf ] and [ rvprecisionspt ] we use the same process to show the impact various modifications to our analysis have on achievable precision , some of which are discussed in further detail in the following paragraphs . finally , figure [ rvprecision2 ] shows an updated version of figure [ precisionfig ] with the corrections discussed above . [ [ atmospheric - variability ] ] atmospheric variability + + + + + + + + + + + + + + + + + + + + + + + we estimated the impact bulk atmospheric motions have on our wavelength reference by integrating the water vapor weighted wind speed along the line of sight using data from the noaa gfs forecast models @xcite . using the gfs model closest in time to our data the forecast is within 3 hours of the model s initial conditions . these models have an rmse wind vector error of about 3 m / s three days ( ! ) in the future . perturbing the integrals by this error has a maximum impact of about 1 m / s , with typical values less than a tenth of that . the resulting contributions for our data on hip 48331 , ngc 2516 , and ngc 2422 are shown in figure [ noaawindfig ] . while pointing directly into or along the jet - stream would exhibit a clear signature at the @xmath199 m / s level , typical values are not particularly significant to our efforts . as mentioned , we adopt @xmath197 m / s . [ [ sky - emission-1 ] ] sky emission + + + + + + + + + + + + ideally the spectra we obtained of our rv standard would be completely representative of our program stars . our cluster targets are , however , significantly fainter than our rv standard and so many of them exhibit a number of strong sky emission lines ( c.f . figure [ exampleextractions ] and [ fitexamplefig ] ) . to better asses their impact on our rv precision we took a subset of 60 spectra and extracted and fit the spectra from images prior to stacking . this gave us a sample of 4 or 5 rvs from spectra obtained consecutively . these spectra spanned a s / n of about 15 45 for targets of spectral type @xmath0 k3 - f5 . the @xmath188 for these rvs was in agreement with that expected based on our @xmath190 relation . we see some evidence that our better - than - anticipated rv precision stems from our use of the mean rv as a prior . if fit with an initial rv far from the multi - epoch mean we observe an increased @xmath188 at low s / n , though still somewhat below that predicted by ( * ? ? ? figure [ rvprecisionmeas ] ) . [ [ psf - effects ] ] psf effects + + + + + + + + + + + we found very little difference in the results of the multi - gaussian parameterization of @xcite and a gauss - hermite kernel when the center of its enclosed power is not constrained to the central pixel undergoing convolution . the latter is faster with many fewer parameters and once the enclosed power is constrained to the central sub - pixel we find it exhibits enhanced stability ( c.f . figure [ rvprecisionpsf ] ) . both yielded slightly worse performance than a simple gaussian and we did not find an improvement in rv precision by using a variable , but symmetric psf . it may be worth investigating a hybrid approach where the components of the hermite parametrization are allowed to vary with pixel . [ [ model - spectra ] ] model spectra + + + + + + + + + + + + + as an additional test on rv precision and the impact our use of the phoenix grid has we model the twilight spectra both with the phoenix grid and using the empiric solar spectrum of @xcite as the template . we select the @xmath0 600 twilight spectra in images with mean s / n above 100 ( 100 650 , mean of 320 ) . for these spectra we measured a @xmath90 rv scatter within each twilight image of @xmath200 m / s when fitting with the phoenix models and @xmath201 m / s using kurucz s empiric solar spectra . this strongly suggests that the phoenix templates are not limiting our rv precision . we see evidence of a slight quadratic dependence of the measured rv on the spectrum s ccd position . this suggests that the rv zero point and wavelength zero points are slightly affecting our dispersion , though we note that program stars are typically observed in the same fiber . fitting and removing this effect reduced the scatter to @xmath202 m / s and @xmath203 m / s , respectively . we estimated the accuracy of our rvs by looking at the differences between our values and those in the literature for each of our six standard stars . we report these differences in table [ stdrverrors ] and find an offset of @xmath204 m / s from the scale of @xcite , albeit with significant scatter . we also saw a slight indication that rvs measured in our lowest s / n bin are slightly shifted relative to the higher s / n bins by @xmath205 m / s . lccr hip 48331 & k5v & 30 & @xmath206 + hip 13388 & k1v & 2 & @xmath207 + hip 10798 & g8v & 5 & @xmath208 + hip 22278 & g5v & 3 & @xmath209 + hip 19589 & g0v & 1 & @xmath210 + hip 31415 & f6v & 1 & @xmath211 + + average & & & @xmath204 given m2fs s unique ability to efficiently obtain precision optical spectra capable of determining accurate stellar properties and precision rvs , here we briefly summarize the broader scientific impact this instrument could have for both finding exoplanets in open clusters and improving our understanding of the stars in these clusters . though once unexpected , it is now clear that a great many hot - gas giants exist . assuming 1.2% of f5-k5 stars in open clusters harbor hot gas planets ( p @xmath212 days , @xmath213 ) @xcite and given our achieved precision we can predict the limits of our technique . for example , we expect m2fs will be capable of attaining a s / n 25 spectrum of a k5v star at a dm of 9.5 in 4 hours ( effectively 2 minutes per star ) . this would be sufficient for an rv precision of @xmath43 m / s , with brighter members increasingly limited by the systematics in our analysis . in this hypothetical cluster we would then be sensitive to @xmath214% of known hot gas giants . figure [ precisiondm ] shows our anticipated rv measurement precision as a function of distance modulus using the new filter . this implies we could reasonably expect one hot - jupiter per m2fs pointing , provided targets are available for the majority of fibers . table [ futuretargtable ] lists the eight nearby open clusters that matched the cluster selection criteria given in section [ targselecsec ] . from these clusters we can obtain @xmath10 m2fs pointings and would expect about as many exoplanet candidates . we highlight that these clusters span a range of ages and are thus well suited to help build a sample of exoplanets that addresses the formation and migration issues discussed in the introduction . this program would also well characterize the doppler uncertainty commonly referred to as stellar jitter as a function of age . this issue is not yet well constrained ( c.f . * ? ? ? * ; * ? ? ? * ) and the present state of our knowledge is largely summarized in figure [ agejitter ] , which shows the stellar jitter as a function of age for a young association and two open clusters along with the ages of open clusters our expanded survey will target . extant data is far from homogenous a mix of both optical and infrared spectroscopy and suffers from rather small sample sizes . by surveying a large number of stars in clusters over a range of ages we will well constrain this effect with age and determine what truly constitutes `` too young '' to measure precision rvs . we note that m2fs s strength is in identification : though the ability to survey large numbers of stars at this precision is unmatched , we would suggest that followup of promising candidates is better suited to traditional monitoring programs . the success of any large scale rv survey of open clusters for planets depends critically on having a carefully vetted sample to survey . m2fs is poised to do this . first , precision rvs will help confirm membership , especially when combined with gaia proper motions . second , precision rvs can identify spectroscopic binaries that are typically poor targets for exoplanet searches . finally , the high dispersion spectra allow measures of @xmath5 , @xmath1 , and metallicity , with the latter two yielding stellar masses from evolutionary models . when coupled with photometric periods determined by lsst , we could even determine stellar inclinations and identify edge - on systems . the flexible nature of m2fs also means that we could also obtain spectra of the ca ii h and k region for all of our targets with a relatively modest overhead ( @xmath215 min total for targets in ngc 2516 and ngc 2422 ) , further helping calibrate stellar ages . a number of recent papers have also highlighted the importance both stellar multiplicity and metallicity play in star formation and have drawn attention to various gaps in current simulations of cluster evolution @xcite . our stellar properties directly address such gaps by characterizing the cluster chemical environment while our rvs allow robust identification of binaries and brown dwarfs ( constrained by our time baseline ) , characterizing the kinematic environment . such a dataset can contribute to the initial conditions used in dynamical simulations of cluster formation . we also note there is an absence of precision internal kinematics for open clusters . this dataset is useful to study the internal dynamics of open clusters at the 10 m / s level . with an anticipated precision of better than 10 @xmath216as / yr ( @xmath217 20 m / s at 500 pc ) @xcite , once gaia data is available for our targets the combined dataset will offer an unprecedented 3d kinematic picture of stars within open clusters , providing a useful tool to study internal kinematics . we have presented a program to use the michigan / magellan fiber system to obtain multiplexed spectroscopy of solar analog stars in nearby ( @xmath2181 kpc ) open clusters with the intent of identifying exoplanet host stars for subsequent followup . our technique uses telluric lines in the 7230 region as a wavelength reference and is presently capable of measuring rvs with a precision of 25 - 60 m / s , depending on s / n . we also obtain precise and accurate measurements of @xmath1 , [ fe / h ] , [ @xmath3/fe ] , and @xmath5 for all of our target stars thereby enabling characterization of the cluster environment . this paper is the first in a series of papers on our efforts and described our analysis procedure in detail . the next paper in this series will report the rvs and stellar parameters of targets in ngc 2422 and ngc 2516 and carry out an analysis of the companion detectability therein . later papers will use our sample to study binaries in our sample and analyze emergent cluster properties . ic 4725 ( m25 ) & 93 & 560 & 0.5 & -0.3 & 100 + ngc 2546 & 140 & 930 & 0.8 & + 0.12 & 200 + ngc 6475 ( m 7 ) & 180 & 300 & 0.8 & + 0.14 & 150 + ngc 6494 & 330 & 650 & 0.5 & + 0.09 & 200 + ngc 2548 & 520 & 790 & 1.1 & + 0.08 & 100 + ngc 5822 & 900 & 800 & 0.5 & -0.02 & 500 + ngc 2422 ( m 47 ) & 132 & 491 & 0.3 & + 0.02 & 160 + ngc 2516 & 120 & 346 & 0.4 & -0.18 & 330 + the authors gratefully acknowledge valuable discussions with sam quinn , peter gao , justin cantrell , and colin slater . j.b . and m.m . acknowledge support from nsf aag grant 1312997 and the nfs / mri development grant 0923160 . r.w . acknowledges support from nsf aag grant 1009634 and nasa origins of solar systems grant nnx11ac32 g . pyraf is a product of the space telescope science institute , which is operated by aura for nasa . this research made use of astropy , a community - developed core python package for astronomy ( astropy collaboration , 2013 ) . ciardi , d. r. , van eyken , j. c. , barnes , j. w. , et al . 2015 , , 809 , 42 crockett , c. j. , mahmud , n. i. , prato , l. , et al . 2012 , , 761 , 164 huerta , m. , johns - krull , c. m. , prato , l. , hartigan , p. , & jaffe , d. t. 2008 , , 678 , 472 hulamo , n. , figueira , p. , bonfils , x. , et al . 2008 , , 489 , l9 prato , l. , huerta , m. , johns - krull , c. m. , et al . 2008 , , 687 , l103 meynet , g. , mermilliod , j .- c . , & maeder , a. 1993 , , 98 , 477 loktin , a. v. , gerasimenko , t. p. , & malysheva , l. k. 2001 , astronomical and astrophysical transactions , 20 , 607 sung , h. , bessell , m. s. , lee , b .- w . & lee , s .- g . 2002 , aj , 123 , 290 cochran , w. d. 1988 , apj , 334 , 349 crockett , c. j. , mahmud , n. i. , prato , l. , et al . 2011 , , 735 , 78 figueira , p. , pepe , f. , lovis , c. , & mayor , m. 2010 , , 515 , a106 blake , c. h. , charbonneau , d. , white , r. j. , marley , m. s. , & saumon , d. 2007 , , 666 , 1198 blake , c. h. , charbonneau , d. , & white , r. j. 2010 , , 723 , 684 seifahrt , a. , kufl , h. u. , zngl , g. , et al . 2010 , the messenger , 142 , 21 gao , p. , plavchan , p. , gagne , j. , et al . 2015 , american astronomical society meeting abstracts , 225 , 258.22

we first show that a continuous function is nonnegative on a closed set if and only if ( countably many ) moment matrices of some signed measure with , are all positive semidefinite ( if is compact is an arbitrary finite borel measure with ) . in particular , we obtain a convergent explicit hierarchy of semidefinite ( outer ) approximations with _ no _ lifting , of the cone of nonnegative polynomials of degree at most . wen used in polynomial optimization on certain simple closed sets ( like e.g. , the whole space , the positive orthant , a box , a simplex , or the vertices of the hypercube ) , it provides a nonincreasing sequence of upper bounds which converges to the global minimum by solving a hierarchy of semidefinite programs with only one variable ( in fact , a generalized eigenvalue problem ) . in the compact case , this convergent sequence of upper bounds complements the convergent sequence of lower bounds obtained by solving a hierarchy of semidefinite relaxations as in e.g. . [ section ] [ thm]corollary [ thm]lemma [ thm]proposition [ thm]definition [ thm]assumption [ thm]remark c v u !

1009.0125

this paper is concerned with a concrete characterization of continuous functions that are nonnegative on a closed set @xmath1 and its application for optimization purposes . by concrete we mean a systematic procedure , e.g. a numerical test that can be implemented by an algorithm , at least in some interesting cases . for polynomials , stengle s nichtnegativstellensatz @xcite provides a certificate of nonnegativity ( or absence of nonnegativity ) on a semi - algebraic set . moreover , in principle , this certificate can be obtained by solving a single semidefinite program ( although the size of this semidefinite program is far beyond the capabilities of today s computers ) . similarly , for compact basic semi - algebraic sets , schmdgen s and putinar s positivstellenstze @xcite provide certificates of strict positivity that can be obtained by solving finitely many semidefinite programs ( of increasing size ) . extensions of those certificates to some algebras of non - polynomial functions have been recently proposed in lasserre and putinar @xcite and in marshall and netzer @xcite . however , and to the best of our knowledge , there is still no hierarchy of explicit ( outer or inner ) semidefinite approximations ( with or without lifting ) of the cone of polynomials nonnegative on a closed set @xmath4 , except if @xmath4 is compact and basic semi - algebraic ( in which case outer approximations exist ) . another exception is the convex cone of quadratic forms nonnegative on @xmath8 for which inner and outer approximations are available ; see e.g. anstreicher and burer @xcite , and dr @xcite . + * contribution : * in this paper , we present a different approach based on a new ( at least to the best of our knowledge ) and simple characterization of continuous functions that are nonnegative on a closed set @xmath1 . this characterization involves a _ single _ ( but known ) measure @xmath5 with @xmath3 , and sums of squares of polynomials . namely , our contribution is twofold : \(a ) we first show that a continuous function @xmath0 is nonnegative on a closed set @xmath1 if and only if @xmath9 is nonnegative for all polynomials @xmath10 $ ] , where @xmath5 is a finite borel measure on @xmath7 is a nonnegative set function defined on the borel @xmath11-algebra of @xmath7 ( i.e. , the @xmath11-algebra generated by the open sets ) , such that @xmath12 , @xmath13 , and @xmath14 for any collection of disjoint measurable sets @xmath15 . its support ( denoted @xmath16 ) is the smallest closed set @xmath4 such that @xmath17 ; see e.g. royden @xcite . ] with @xmath3 . the measure @xmath5 is arbitrary if @xmath4 is compact . if @xmath4 is not compact then one may choose for @xmath5 the finite borel measure : - @xmath18 if @xmath0 is a polynomial , and - @xmath19 , if @xmath0 is not a polynomial , where @xmath20 is any finite borel measure with support exactly @xmath4 . but many other choices are possible . equivalently , @xmath0 is nonnegative on @xmath4 if and only if every element of the countable family @xmath21 of moment matrices associated with the signed borel measure @xmath22 , is positive semidefinite . the absence of nonnegativity on @xmath4 can be _ certified _ by exhibiting a polynomial @xmath10 $ ] such that @xmath23 , or equivalently , when some moment matrix in the family @xmath21 is not positive semidefinite . and so , interestingly , as for nonnegativity or positivity , our certificate for absence of nonnegativity is also in terms of sums of squares . when @xmath0 is a polynomial , these moment matrices are easily obtained from the moments of @xmath5 and this criterion for absence of nonnegativity complements stengle s nichtnegativstellensatz @xcite ( which provides a certificate of nonnegativity on a semi - algebraic set @xmath4 ) or schmdgen and putinar s positivstellenstze @xcite ( for certificates of strict positivity on compact basic semi - algebraic sets ) . at last but not least , we obtain a convergent _ explicit _ hierarchy of semidefinite ( outer ) approximations with _ no _ lifting , of the cone @xmath24 of nonnegative polynomials of degree at most @xmath25 . that is , we obtain a nested sequence @xmath26 such that each @xmath27 is a spectrahedron defined solely in terms of the vector of coefficients of the polynomial , with no additional variable ( i.e. , no projection is needed ) . similar explicit hierarchies can be obtained for the cone of polynomials nonnegative on a closed set @xmath4 ( neither necessarily basic semi - algebraic nor compact ) , provided that all moments of an appropriate measure @xmath5 ( with support exactly @xmath4 ) can be obtained . to the best of our knowledge , this is first result of this kind . \(b ) as a potential application , we consider the problem of computing the _ global _ minimum @xmath28 of a polynomial @xmath0 on a closed set @xmath4 , a notoriously difficult problem . in nonlinear programming , a sequence of upper bounds on @xmath28 is usually obtained from a sequence of feasible points @xmath29 , e.g. , via some ( local ) minimization algorithm . but it is important to emphasize that for non convex problems , providing a sequence of upper bounds @xmath30 , @xmath31 , that converges to @xmath28 is in general impossible , unless one computes points on a grid whose mesh size tends to zero . we consider the case where @xmath1 is a closed set for which one may compute all moments of a measure @xmath5 with @xmath3 . typical examples of such sets are e.g. @xmath32 or @xmath8 in the non compact case and a box , a ball , an ellipsoid , a simplex , or the vertices of an hypercube ( or hyper rectangle ) in the compact case . we then provide a hierarchy of semidefinite programs ( with only one variable ! ) whose optimal values form a monotone nonincreasing sequence of _ upper _ bounds which converges to the global minimum @xmath28 . in fact , each semidefinite program is a very specific one as it reduces to solving a _ generalized eigenvalue _ problem for a pair of real symmetric matrices .. ( therefore , for efficiency one may use specialized software packages instead of a sdp solver . ) however , the convergence to @xmath28 is in general only asymptotic and not finite ( except when @xmath4 is a discrete set in which case finite convergence takes place ) . this is in contrast with the hierarchy of semidefinite relaxations defined in lasserre @xcite which provide a nondecreasing sequence of _ lower _ bounds that also converges to @xmath28 , and very often in finitely many steps . hence , for compact basic semi - algebraic sets these two convergent hierarchies of upper and lower bounds complement each other and permit to locate the global minimum @xmath28 in smaller and smaller intervals . notice that convergence of the hierarchy of convex relaxations in @xcite is guaranteed only for compact basic semi - algebraic sets , whereas for the new hierarchy of upper bounds , the only requirement on @xmath4 is to know all moments of a measure @xmath5 with @xmath3 . on the other hand , in general computing such moments is possible only for relatively simple ( but not necessarily compact nor semi - algebraic ) sets . at last but not least , the nonincreasing sequence of upper bounds converges to @xmath28 even if @xmath28 is not attained , which when @xmath32 , could provide an alternative and/or a complement to the hierarchy of convex relaxations provided in schweighofer @xcite ( based on gradient tentacles ) and in h and pham @xcite ( based on the truncated tangency variety ) , which both provide again a monotone sequence of lower bounds . finally , we also give a very simple interpretation of the hierarchy of dual semidefinite programs , which provides some information on the location of global minimizers . a borel measure on @xmath7 is understood as a _ positive _ borel measure , i.e. , a nonnegative set function @xmath5 on the borel @xmath11-algebra @xmath33 ( i.e. , the @xmath11-algebra generated by the open sets of @xmath7 ) such that @xmath12 , and with the countably additive property @xmath34 for any collection of disjoint measurable sets @xmath35 ; see e.g. royden @xcite . let @xmath36 $ ] be the ring of polynomials in the variables @xmath37 , and @xmath38\subset\r[\x]$ ] its subset of polynomials that are sums of squares ( s.o.s . ) . denote by @xmath36_d\subset\r[\x]$ ] the vector space of polynomials of degree at most @xmath6 , which forms a vector space of dimension @xmath39 , with e.g. , the usual canonical basis @xmath40 of monomials . also , denote by @xmath38_d\subset\sigma[\x]$ ] the convex cone of s.o.s . polynomials of degree at most @xmath25 . if @xmath41_d$ ] , write @xmath42 in the canonical basis and denote by @xmath43 its vector of coefficients . let @xmath44 denotes the vector space of @xmath45 real symmetric matrices . for a matrix @xmath46 the notation @xmath47 ( resp . @xmath48 ) stands for @xmath49 is positive semidefinite ( resp . definite ) . with @xmath50 being a sequence indexed in the canonical basis @xmath40 of @xmath36 $ ] , let @xmath51\to\r$ ] be the linear functional @xmath52 and let @xmath53 be the symmetric matrix with rows and columns indexed in the canonical basis @xmath40 , and defined by : @xmath54 with @xmath55 . if @xmath56 has a representing measure @xmath5 , i.e. , if @xmath57 for every @xmath58 , then @xmath59_d,\ ] ] and so @xmath60 . a measure @xmath5 is said to be moment determinate if there is no other measure with same moments . in particular , and as an easy consequence of the stone - weierstrass theorem , every measure with compact support is determinate on a compact set @xmath61 are identical if and only if @xmath62 for all continuous functions @xmath0 on @xmath4 , and ( b ) by stone - weierstrass , the polynomials are dense in the space of continuous functions for the sup - norm . ] . not every sequence @xmath56 satisfying @xmath60 , @xmath63 , has a representing measure . however : [ prop - berg ] let @xmath50 be such that @xmath60 , for every @xmath63 . then : \(a ) the sequence @xmath56 has a representing measure whose support is contained in the ball @xmath64^n$ ] if there exists @xmath65 such that @xmath66 for every @xmath58 . \(b ) the sequence @xmath56 has a representing measure @xmath5 on @xmath7 if @xmath67 in addition , in both cases ( a ) and ( b ) the measure @xmath5 is moment determinate . condition ( b ) is an extension to the multivariate case of carleman s condition in the univariate case and is due to nussbaum @xcite . for more details see e.g. berg @xcite and/or maserick and berg @xcite . similarly , with @xmath68 and @xmath41 $ ] written @xmath69 let @xmath70 be the symmetric matrix with rows and columns indexed in the canonical basis @xmath40 , and defined by : @xmath71 if @xmath56 has a representing measure @xmath5 , then @xmath72 , and so if @xmath5 is supported on the set @xmath73 , then @xmath74 for all @xmath75 because @xmath76_d.\ ] ] recall that if @xmath77 is a separable metric space with borel @xmath11-field @xmath33 , the support @xmath78 of a borel measure @xmath5 on @xmath77 is the ( unique ) smallest closed set @xmath79 such that @xmath80 . given a borel measure @xmath5 on @xmath7 and a measurable function @xmath81 , the mapping @xmath82 , @xmath79 , defines a set function on @xmath33 . if @xmath0 is nonnegative then @xmath83 is a borel measure ( which is finite if @xmath0 is @xmath5-integrable ) ; see e.g. royden ( * ? ? ? * and p. 408 ) . if @xmath0 is not nonnegative then setting @xmath84 and @xmath85 , the set function @xmath83 can be written as the difference @xmath86 of the two positive borel measures @xmath87 defined by @xmath88 then @xmath83 is a _ signed _ borel measure provided that either @xmath89 or @xmath90 is finite ; see e.g. royden @xcite . we first provide the following auxiliary result which is also of self - interest . [ aux ] let @xmath77 be a separable metric space , @xmath91 a closed set , and @xmath5 a borel measure on @xmath77 with @xmath3 . a continuous function @xmath92 is nonnegative on @xmath4 if and only if the set function @xmath93 , @xmath79 , is a positive measure . the _ only if part _ is straightforward . for the _ if part _ , if @xmath83 is a positive measure then @xmath94 for @xmath5-almost all @xmath95 . that is , there is a borel set @xmath96 such that @xmath97 and @xmath94 on @xmath98 . indeed , otherwise suppose that there exists a borel set @xmath99 with @xmath100 and @xmath101 on @xmath99 ; then one would get the contradiction that @xmath83 is not positive because @xmath102 . in fact , @xmath0 is called the radon - nikodym derivative of @xmath83 with respect to @xmath5 ; see royden ( * ? ? ? * theorem 23 , p. 276 ) . next , observe that @xmath103 and @xmath104 . therefore @xmath105 because @xmath106 is the unique smallest closed set such that @xmath107 . hence , let @xmath95 be fixed , arbitrary . as @xmath108 , there is a sequence @xmath109 , @xmath31 , with @xmath110 as @xmath111 . but since @xmath0 is continuous and @xmath112 for every @xmath31 , we obtain the desired result @xmath94 . lemma [ aux ] itself ( of which we have not been able to find a trace in the literature ) is a characterization of nonnegativity on @xmath4 for a continuous function @xmath0 on @xmath77 . however , one goal of this paper is to provide a more concrete characterization . to do so we first consider the case of a compact set @xmath61 . let @xmath4 be a compact subset of @xmath7 . for simplicity , and with no loss of generality , we may and will assume that @xmath113^n$ ] . [ thmradon ] let @xmath113^n$ ] be compact and let @xmath5 be an arbitrary , fixed , finite borel measure on @xmath4 with @xmath114 , and with vector of moment @xmath50 , @xmath58 . let @xmath0 be a continuous function on @xmath7 . then : \(a ) @xmath0 is nonnegative on @xmath4 if and only if @xmath115,\ ] ] or , equivalently , if and only if @xmath116 where @xmath117 , @xmath58 , with @xmath118 , and with @xmath119 as in ( [ moment ] ) . if in addition @xmath41 $ ] then ( [ thmradon-2 ] ) reads @xmath74 , @xmath75 , where @xmath70 is the localizing matrix defined in ( [ localizing ] ) . + ( b ) if in addition to be continuous , @xmath0 is also concave on @xmath4 , then @xmath0 is nonnegative on the convex hull @xmath120 of @xmath4 if and only if ( [ thmradon-1 ] ) holds . the _ only if _ part is straightforward . indeed , if @xmath121 on @xmath4 then @xmath122 and so for any finite borel measure @xmath5 on @xmath4 , @xmath123 for every @xmath124 $ ] . next , if @xmath0 is concave and @xmath121 on @xmath120 then @xmath121 on @xmath4 and so the only if " part of ( b ) also follows . _ _ the set function @xmath125 , @xmath79 , can be written as the difference @xmath126 of the two positive finite borel measures @xmath87 described in ( [ decomp])-([hahn ] ) , where @xmath127 and @xmath128 . as @xmath4 is compact and @xmath0 is continuous , both @xmath87 are finite , and so @xmath83 is a finite signed borel measure ; see royden @xcite . in view of lemma [ aux ] it suffices to prove that in fact @xmath83 is a finite and positive borel measure . so let @xmath117 , @xmath58 , be the sequence defined by : @xmath129 every @xmath130 , @xmath58 , is finite because @xmath4 is compact and @xmath0 is continuous . so the condition @xmath131_d,\ ] ] reads @xmath132 for all @xmath133 , that is , @xmath134 , where @xmath119 is the moment matrix defined in ( [ moment ] ) . and so ( [ thmradon-1 ] ) implies @xmath134 for every @xmath63 . moreover , as @xmath113^n$ ] , @xmath135 hence , by proposition [ prop - berg ] , @xmath136 is the moment sequence of a finite ( positive ) borel measure @xmath137 on @xmath138^n$ ] , that is , as @xmath139^n$ ] , @xmath140^n } \x^\alpha\,d\nu(\x)\,=\,\int_{[-1,1]^n } \x^\alpha \,d\psi(\x),\qquad\forall\alpha\in\n^n.\ ] ] but then using ( [ decomp ] ) and ( [ equality ] ) yields @xmath141^n } \x^\alpha\,d\nu_1(\x)\,=\ , \int_{[-1,1]^n } \x^\alpha \,d(\nu_2+\psi)(\x),\qquad\forall\alpha\in\n^n,\ ] ] which in turn implies @xmath142 because measures on a compact set are determinate . next , this implies @xmath143 and so @xmath83 is a positive borel measure on @xmath4 . hence by lemma [ aux ] , @xmath94 on @xmath4 . if in addition @xmath41 $ ] , the sequence @xmath117 is obtained as a linear combination of @xmath144 . indeed if @xmath145 then @xmath146 and so in ( [ thmradon-2 ] ) , @xmath119 is nothing less than the _ localizing _ matrix @xmath70 associated with @xmath50 and @xmath41 $ ] , defined in ( [ localizing ] ) , and ( [ thmradon-2 ] ) reads @xmath74 for all @xmath75 finally , if @xmath0 is concave then @xmath121 on @xmath4 implies @xmath121 on @xmath120 , and so the _ only if _ part of ( b ) also follows . therefore , to check whether a polynomial @xmath41 $ ] is nonnegative on @xmath4 , it suffices to check if every element of the countable family of real symmetric matrices @xmath147 , @xmath63 , is positive semidefinite . an informal alternative proof of theorem [ thmradon ] which does not use lemma [ aux ] is as follows . if @xmath0 is not nonnegative on @xmath4 there exists @xmath148 such that @xmath149 , and so as @xmath4 is compact , there is a continuous function , e.g , @xmath150 close to @xmath151 in some open neighborhood @xmath152 of @xmath153 , and very small in the rest of @xmath4 . by the stone - weierstrass s theorem , one may choose @xmath154 to be a polynomial . next , the complement @xmath155 ) of @xmath152 is closed , and so @xmath156 is a closed set contained in @xmath4 ( hence smaller than @xmath4 ) . therefore @xmath157 because otherwise @xmath158 which would imply that @xmath156 is a support of @xmath5 smaller than @xmath4 , in contradiction with @xmath3 . hence , we would get the contradiction @xmath159 however , in the non compact case described in the next section , this argument is not valid . we now consider the more delicate case where @xmath4 is a closed set of @xmath7 , not necessarily compact . to handle arbitrary non compact sets @xmath4 and arbitrary continuous functions @xmath0 , we need a reference measure @xmath5 with @xmath3 and with nice properties so that integrals such as @xmath160 , @xmath124 $ ] , are well - behaved . so , let @xmath20 be an arbitrary finite borel measure on @xmath7 whose support is exactly @xmath4 , and let @xmath5 be the finite borel measure defined by : @xmath161 observe that @xmath3 and @xmath5 satisfies carleman s condition ( [ carleman ] ) . indeed , let @xmath117 , @xmath58 , be the sequence of moments of @xmath5 . then for every @xmath162 , and every @xmath163 , using @xmath164 , @xmath165 therefore for every @xmath162 , using @xmath166 for every @xmath167 , yields @xmath168 i.e. , ( [ carleman ] ) holds . notice also that all the moments of @xmath5 ( defined in ( [ defmu ] ) ) are finite , and so every polynomial is @xmath5-integrable . [ noncompact ] let @xmath169 be closed and let @xmath20 be an arbitrary finite borel measure whose support is exactly @xmath4 . let @xmath0 be a continuous function on @xmath7 . if @xmath41 $ ] ( i.e. , @xmath0 is a polynomial ) let @xmath5 be as in ( [ defmu ] ) whereas if @xmath0 is not a polynomial let @xmath5 be defined by @xmath170 then ( a ) and ( b ) of theorem [ thmradon ] hold . for a detailed proof see [ appendix ] . it is important to emphasize that in theorem [ thmradon ] and [ noncompact ] , the set @xmath4 is an _ arbitrary _ closed set of @xmath7 , and to the best of our knowledge , the characterization of nonnegativity of @xmath0 in terms of positive definiteness of the moment matrices @xmath119 is new . but of course , this characterization becomes even more interesting when one knows how the compute the moment sequence @xmath117 , @xmath58 , which is possible in a few special cases only . important particular cases of nice such sets @xmath4 include boxes , hyper rectangles , ellipsoids , and simplices in the compact case , and the positive orthant , or the whole space @xmath7 in the non compact case . for instance , for the whole space @xmath32 one may choose for @xmath5 in ( [ defmu ] ) the multivariate gaussian ( or normal ) probability measure @xmath171 which the @xmath172-times product of the one - dimensional normal distribution @xmath173 whose moments are all easily available in closed form . in theorem [ noncompact ] this corresponds to the choice @xmath174 when @xmath4 is the positive orthant @xmath175 one may choose for @xmath5 the exponential probability measure @xmath176 which the @xmath172-times product of the one - dimensional exponential distribution @xmath177 whose moments are also easily available in closed form . in theorem [ noncompact ] this corresponds to the choice @xmath178 the convex cone @xmath179_{2d}$ ] of nonnegative polynomials of degree at most @xmath25 ( a nonnegative polynomial has necessarily even degree ) is much harder to characterize than its subcone @xmath38_d$ ] of sums of squares . indeed , while the latter has a simple semidefinite representation with lifting ( i.e. @xmath38_d$ ] is the projection in @xmath180 of a spectrahedron in a higher dimensional space ) , so far there is no such simple representation for the former . in addition , when @xmath6 is fixed , blekherman @xcite has shown that after proper normalization , the gap " between @xmath181 and @xmath38_d$ ] increases unboundedly with the number of variables . we next provide a convergent hierarchy of ( outer ) semidefinite approximations @xmath182 , @xmath31 , of @xmath181 where each @xmath27 has a semidefinite representation with _ no _ lifting ( i.e. , no projection is needed and @xmath27 is a spectrahedron ) . to the best of our knowledge , this is the first result of this kind . recall that with every @xmath41_d$ ] is associated its vector of coefficients @xmath183 , @xmath184 , in the canonical basis of monomials , and conversely , with every @xmath185 is associated a polynomial @xmath41_d$ ] with vector of coefficients @xmath183 in the canonical basis . recall that for every @xmath186 , @xmath187 as @xmath188 for every @xmath189 . [ th - cone ] let @xmath5 be the probability measure on @xmath7 which is the @xmath172-times product of the normal distribution on @xmath190 , and so with moments @xmath50 , @xmath58 , @xmath191 for every @xmath31 , let @xmath192 , where @xmath193 is the localizing matrix in ( [ localizing ] ) associated with @xmath56 and @xmath41_{2d}$ ] . each @xmath27 is a closed convex cone and a spectrahedron . then @xmath194 and @xmath195 if and only if its vector of coefficients @xmath196 satisfies @xmath197 , for every @xmath163 . following its definition ( [ localizing ] ) , all entries of the localizing matrix @xmath193 are linear in @xmath196 , and so @xmath198 is an lmi . therefore @xmath27 is a spectrahedron and a closed convex cone . next , let @xmath199 and let @xmath5 be as in corollary [ th - cone ] and so of the form ( [ defmu ] ) with @xmath20 as in ( [ normalchoice ] ) . then @xmath5 satisfies carleman s condition ( [ carleman ] ) . hence , by theorem [ noncompact ] with @xmath32 , @xmath0 is nonnegative on @xmath4 if and only if ( [ thmradon-2 ] ) holds , which is equivalent to stating that @xmath198 , @xmath163 , which in turn is equivalent to stating that @xmath200 , @xmath163 so the nested sequence of convex cones @xmath201 defines arbitrary close outer approximations of @xmath24 . in fact @xmath202 is closed and @xmath203 . it is worth emphasizing that each @xmath27 is a spectrahedron with _ no _ lifting , that is , @xmath27 is defined solely in terms of the vector of coefficients @xmath204 with no additional variable ( i.e. , no projection is needed ) . for instance , the first approximation @xmath205 is just the set @xmath206 , which is a half - space of @xmath180 . and with @xmath207 , @xmath208\,\succeq\,0\:\right\},\ ] ] or , equivalently , @xmath209 is the convex basic semi - algebraic set : @xmath210 @xmath211-\ ] ] @xmath212 @xmath213 where we have just expressed the nonnegativity of all principal minors of @xmath214 . a very similar result holds for the convex cone @xmath215 of polynomials of degree at most @xmath6 , nonnegative on a closed set @xmath61 . [ generalk ] let @xmath61 be a closed set and let @xmath5 be defined in ( [ defmu ] ) where @xmath20 is a an arbitrary finite borel measure whose support is exactly @xmath4 . for every @xmath31 , let @xmath216 , where @xmath193 is the localizing matrix in ( [ localizing ] ) associated with @xmath56 and @xmath41_{d}$ ] . each @xmath217 is a closed convex cone and a spectrahedron . then @xmath218 and @xmath219 if and only if its vector of coefficients @xmath185 satisfies @xmath220 , for every @xmath163 . the proof which mimicks that of corollary [ th - cone ] is omitted . of course , for practical computation , one is restricted to sets @xmath4 where one may compute _ effectively _ the moments of the measure @xmath5 . an example of such a set @xmath4 is the positive orthant , in which case one may choose the measure @xmath5 in ( [ exponential ] ) for which all moments are explicitly available . for compact sets @xmath4 let us mention balls , boxes , ellipsoids , and simplices . but again , any compact set where one knows how to compute all moments of some measure with support exactly @xmath4 , is fine . to the best of our knowledge this is the first characterization of an outer approximation of the cone @xmath215 in a relatively general context . indeed , for the basic semi - algebraic set @xmath221 stengle s nichtnegativstellensatz @xcite states that @xmath41 $ ] is nonnegative on @xmath4 if and only if @xmath222 for some integer @xmath223 and polynomials @xmath224 , where @xmath225 is the _ preordering _ associated with the @xmath226 s is the set of polynomials of the form @xmath227 where @xmath228 $ ] for each @xmath229 . ] associated with the @xmath226 s . in addition , there exist bounds on the integer @xmath223 and the degree of the s.o.s . weights in the definition of @xmath224 , so that in principle , when @xmath0 is _ known _ , checking whether @xmath121 on @xmath4 reduces to solving a single sdp to compute @xmath230 in the nonnegativity certificate ( [ stengle1 ] ) . however , the size of this sdp is potentially huge and makes it unpractical . moreover , the representation of @xmath215 in ( [ stengle1 ] ) is not convex in the vector of coefficients of @xmath0 because it involves @xmath231 as well as the product @xmath232 . if in corollary [ generalk ] one replaces the finite - dimensional convex cone @xmath233_d$ ] with the infinite - dimensional convex cone @xmath234 $ ] of _ all _ polynomials nonnegative on @xmath4 , and @xmath235_d$ ] with @xmath236 , then the nested sequence of ( increasing but finite - dimensional ) convex cones @xmath237 , @xmath31 , provides finite - dimensional approximations of @xmath238 . consider the polynomial optimization problem @xmath239 where @xmath1 is closed and @xmath41 $ ] . if @xmath4 is compact let @xmath5 be a finite borel measure with @xmath3 and if @xmath4 is not compact , let @xmath20 be an arbitrary finite borel measure with @xmath240 and let @xmath5 be as in ( [ defmu ] ) . in both cases , the sequence of moments @xmath50 , @xmath58 , is well - defined , and we assume that @xmath241 is available or can be computed , for every @xmath58 . consider the sequence of semidefinite programs : @xmath242 where @xmath243 $ ] is the polynomial @xmath244 . notice that ( [ primal ] ) has only one variable ! [ thm - optim ] consider the hierarchy of semidefinite programs ( [ primal ] ) indexed by @xmath63 . then : \(i ) ( [ primal ] ) has an optimal solution @xmath245 for every @xmath246 . \(ii ) the sequence @xmath247 , @xmath63 , is monotone nonincreasing and @xmath248 as @xmath249 . \(i ) since @xmath250 on @xmath4 , by theorem [ thmradon ] , @xmath251 is a feasible solution of ( [ primal ] ) for every @xmath6 . hence @xmath245 for every @xmath63 . next , let @xmath63 be fixed , and let @xmath252 be an arbitrary feasible solution of ( [ primal ] ) . from the condition @xmath253 , the diagonal entry @xmath254 must be nonnegative , i.e. , @xmath255 , and so , as we maximize and @xmath256 , ( [ primal ] ) must have an optimal solution @xmath257 . \(ii ) obviously @xmath258 whenever @xmath259 , because @xmath253 implies @xmath260 . therefore , the sequence @xmath247 , @xmath63 , is monotone nonincreasing and being bounded below by @xmath28 , converges to @xmath261 . next , suppose that @xmath262 ; fix @xmath31 , arbitrary . the convergence @xmath263 implies @xmath264 . as @xmath167 was arbitrary , we obtain that @xmath265 for every @xmath63 . but then by theorem [ thmradon ] or theorem [ noncompact ] , @xmath266 on @xmath4 , and so @xmath267 , in contradiction with @xmath262 . therefore @xmath268 . for each @xmath63 , the semidefinite program ( [ primal ] ) provides an upper bound on the optimal value @xmath28 only . we next show that the dual contains some information on global minimizers , at least when @xmath6 is sufficiently large . let @xmath269 be the space of real symmetric @xmath270 matrices . one may write the semidefinite program ( [ primal ] ) as @xmath271 which in fact is a generalized eigenvalue problem for the pair of matrices @xmath53 and @xmath70 . its dual is the semidefinite program @xmath272 or , equivalently , @xmath273_d\:\right\}.\ ] ] so the dual problem ( [ dual ] ) is to find a sum of squares polynomial @xmath11 of degree at most @xmath25 ( normalized to satisfy @xmath274 ) that minimizes the integral @xmath275 , and a simple interpretation of ( [ dual ] ) is as follows : with @xmath276 being the space of borel probability measures on @xmath4 , we know that @xmath277 . next , let @xmath278 be the space of probability measures on @xmath4 which have a density @xmath279_d$ ] with respect to @xmath5 . then ( [ dual ] ) reads @xmath280 , which clearly shows why one obtains an upper bound on @xmath28 . indeed , instead of searching in @xmath276 one searches in its subset @xmath281 . what is not obvious at all is whether the obtained upper bound obtained in ( [ dual ] ) converges to @xmath28 when the degree of @xmath279_d$ ] is allowed to increase ! [ thdual ] suppose that @xmath282 and @xmath4 has nonempty interior . then : \(a ) there is no duality gap between ( [ primal ] ) and ( [ dual ] ) and ( [ dual ] ) has an optimal solution @xmath283_d$ ] which satisfies @xmath284 . \(b ) if @xmath4 is convex and @xmath0 is convex , let @xmath285 . then @xmath286 and @xmath287 , so that @xmath288 as @xmath249 . moreover , if the set @xmath289 is compact for some @xmath290 , then any accumulation point @xmath291 of the sequence @xmath292 , @xmath63 , is a minimizer of problem ( [ poly - opt ] ) , that is , @xmath293 . \(a ) any scalar @xmath294 is a feasible solution of ( [ primal ] ) and in addition , @xmath295 because since @xmath4 has nonempty interior and @xmath296 on @xmath4 , @xmath297_d.\ ] ] but this means that slater s condition , slater s condition states that there exists @xmath298 such that @xmath299 for every @xmath300 . ] holds for ( [ primal ] ) which in turn implies that there is no duality gap and ( [ dual ] ) has an optimal solution @xmath283_d$ ] ; see e.g. @xcite . and so , @xmath301 ( b ) let @xmath83 be the borel probability measure on @xmath4 defined by @xmath302 , @xmath79 . as @xmath0 is convex , by jensen s inequality ( see e.g. mcshane @xcite ) , @xmath303 in addition , if @xmath4 is convex then @xmath286 and so , @xmath304 . finally if for some @xmath290 , the set @xmath305 is compact , and since @xmath306 , then @xmath307 for @xmath6 sufficiently large , i.e. , @xmath308 for sufficiently large @xmath6 . by compactness there is a subsequence @xmath309 , @xmath310 , and a point @xmath291 such that @xmath311 as @xmath312 . continuity of @xmath0 combined with the convergence @xmath288 yields @xmath313 as @xmath312 . as the convergent subsequence @xmath314 was arbitrary , the proof is complete . so in case where @xmath0 is a convex polynomial and @xmath4 is a convex set , theorem [ thdual ] provides a means of approximating not only the optimal value @xmath28 , but also a global minimizer @xmath291 . in the more subtle nonconvex case , one still can obtain some information on global minimizers from an optimal solution @xmath283_d$ ] of ( [ dual ] ) . let @xmath315 be fixed , and suppose that @xmath6 is large enough so that @xmath316 . then , by theorem [ dual](a ) , @xmath317 as @xmath250 on @xmath4 , necessarily the measure @xmath318 gives very small weight to regions of @xmath4 where @xmath319 is significantly larger than @xmath28 . for instance , if @xmath320 and @xmath321 , then @xmath322 , i.e. , the set @xmath323 contributes to less than @xmath324 of the total mass of @xmath83 . so if @xmath5 is uniformly distributed on @xmath4 ( which is a reasonable choice if one has to compute all moments of @xmath5 ) then a simple inspection of the values of @xmath325 provides some rough indication on where ( in @xmath4 ) @xmath319 is close to @xmath28 . the interpretation ( [ dual ] ) of the dual shows that in general the monotone convergence is only asymptotic and can not be finite . indeed if @xmath4 has a nonempty interior then the probability measure @xmath326 can not be a dirac measure at any global minimizer @xmath291 . an exception is the discrete case , i.e. , when @xmath4 is a finite number of points , like in e.g. 0/1 programs . indeed we get : [ exception ] let @xmath327 be a discrete set @xmath328 , @xmath329 , and let @xmath5 be the probability measure uniformly distributed in @xmath4 , i.e. , @xmath330 where @xmath331 and @xmath332 denote the dirac measure at the point @xmath333 . then the optimal value @xmath257 of ( [ primal ] ) satisfies @xmath334 for some integer @xmath6 . let @xmath335 ( for some @xmath336 ) be the global minimizer of @xmath337 . for each @xmath338 there exists a polynomial @xmath339 $ ] such that @xmath340 for every @xmath341 ( where @xmath342 denotes the kronecker symbol ) . the polynomials @xmath343 are called _ interpolation _ polynomials . so let @xmath344 $ ] , so that @xmath345 hence as soon as @xmath346 , @xmath283_d$ ] is a feasible solution of ( [ dual ] ) , and so from @xmath347 we deduce that @xmath348 , the desired result . there are several interesting cases where the above described methodology can apply , i.e. , cases where @xmath56 can be obtained either explicitly in closed form or numerically . in particular , when @xmath4 is either : * a box @xmath349\subset\r^n$ ] , with @xmath5 being the normalized lebesgue measure on @xmath350 . the sequence @xmath50 is trivial to obtain in closed form . * the discrete set @xmath351 with @xmath5 being uniformly distributed and normalized . again the sequence @xmath50 is trivial to obtain in closed form . notice that in this case we obtain a new hierarchy of semidefinite relaxations ( with only one variable ) for the celebrated maxcut problem ( and any nonlinear 0/1 program ) . * the unit euclidean ball @xmath352 with @xmath5 uniformly distributed , and similarly the unit sphere @xmath353 , with @xmath5 being the rotation invariant probability measure on @xmath354 . in both cases the moments @xmath50 are obtained easily . * a simplex @xmath355 , in which case if one takes @xmath5 as the lebesgue measure then all moments of @xmath5 can be computed numerically . in particular , with @xmath6 fixed , this computation can be done in time polynomial time . see e.g. the recent work of @xcite . * the whole space @xmath7 in which case @xmath5 may be chosen to be the product measure @xmath356 with each @xmath357 being the normal distribution . observe that one then obtains a new hierarchy of semidefinite approximations ( upper bounds ) for unconstrained global optimization . the corresponding monotone sequence of upper bounds converges to @xmath28 no matter if the problem has a global minimizer or not . this may be an alternative and/or a complement to the recent convex relaxations provided in schweighofer @xcite and h and vui @xcite which also work when @xmath28 is not attained , and provide a convergent sequence of _ lower _ bounds . * the positive orthant @xmath175 , in which case @xmath5 may be chosen to be the product measure @xmath356 with each @xmath357 being the exponential distribution @xmath358 , @xmath79 . in particular if @xmath359 where @xmath360 , then one obtains a hierarchy of numerical tests to check whether @xmath49 is a _ copositive _ matrix . indeed , if @xmath257 is an optimal solution of ( [ eigenvalue ] ) then @xmath49 is copositive if and only if @xmath361 for all @xmath63 . notice that we also obtain a hierarchy of _ outer approximations _ @xmath362 of the cone @xmath363 of @xmath364 copositive matrices . indeed , for every @xmath360 , let @xmath365 be the quadratic form @xmath366 . then , for every @xmath6 , the set @xmath367 is a convex cone defined only in terms of the coefficients of the matrix @xmath49 . it is even a spectrahedron since @xmath368 is a linear matrix inequality in the coefficients of @xmath49 . and in view of theorem [ thmradon](a ) , @xmath369 . in this section we provide three simple examples to illustrate the above methodology . [ ex1 ] consider the global minimization on @xmath370 of the motzkin - like polynomial @xmath371 whose global minimum is @xmath372 , attained at @xmath373 . choose for @xmath5 the probability measure @xmath374 , @xmath375 , for which the sequence of moments @xmath376 , @xmath377 , is easy to obtain . namely @xmath378 for every @xmath377 . then the semidefinite relaxations ( [ primal ] ) yield @xmath379 , @xmath380 , and @xmath381 , showing a significant and rapid decrease in first iterations with a long tail close to @xmath28 , illustrated in figure [ fig1 ] . then after @xmath382 , one encounters some numerical problems and we can not trust the results anymore . if we now minimize the same polynomial @xmath0 on the box @xmath383 ^ 2 $ ] , one choose for @xmath5 the probability uniformly distributed on @xmath383 ^ 2 $ ] , whose moments @xmath376 , @xmath377 , are also easily obtained by @xmath384 . then one obtains @xmath385 , @xmath386 , and @xmath387 , showing again a rapid decrease in first iterations with a long tail close to @xmath28 , illustrated in figure [ fig2 ] . [ fig1],scaledwidth=80.0% ] ^2$][fig2],scaledwidth=80.0% ] [ ex2 ] still on @xmath370 , consider the global minimization of the polynomial @xmath388 whose global minimum @xmath389 is not attained . again , choose for @xmath5 the probability measure @xmath374 , @xmath375 . then the semidefinite relaxations ( [ primal ] ) yield @xmath390 , @xmath391 and @xmath392 , showing again a significant and rapid decrease in first iterations with a long tail close to @xmath28 , illustrated in figure [ fig3 ] ; numerical problems occur after @xmath393 . however , this kind of problems where the global minimum @xmath28 is not attained , is notoriously difficult . even the semidefinite relaxations defined in @xcite ( which provide lower bounds on @xmath28 ) and especially devised for such problems , encounter numerical difficulties ; see ( * ? ? ? * example 4.8 ) . on @xmath394,scaledwidth=80.0% ] [ ex4 ] the following example illustrates the duality results of section [ sec - duality ] . the univariate polynomial @xmath395 displayed in fig [ fig4 ] has two global minima at @xmath396 and @xmath397 , with @xmath398 . in fig [ fig5 ] is plotted the sequence of upper bounds @xmath306 as @xmath399 , with again a rapid decrease in first iterations . one has plotted in fig [ fig6 ] the s.o.s . polynomial @xmath400 , optimal solution of ( [ dual ] ) with @xmath401 , associated with the probability density @xmath402 as explained in [ sec - duality ] . as expected , two peaks appear at the points @xmath403 , @xmath404 . on @xmath383$],scaledwidth=80.0% ] on @xmath383$],scaledwidth=80.0% ] on @xmath383$],scaledwidth=80.0% ] [ maxcut ] we finally consider a discrete optimization problem , namely the celebrated maxcut problem @xmath405 where @xmath406 is a real symmetric matrix whose all diagonal elements vanish . the measure @xmath5 is uniformly distributed on @xmath351 so that its moments are readily available . we first consider the equal weights case , i.e. , @xmath407 for all @xmath408 with @xmath409 in which case @xmath410 . with @xmath411 the successive values for @xmath257 , @xmath412 , are displayed in table [ table1 ] and @xmath413 is relatively close to @xmath28 . next we have generated five random instances of maxcut with @xmath411 but @xmath414 with probability 1/2 , and if @xmath415 it is randomly generated using the matlab rand " function . the successive values of @xmath257 , @xmath416 , are displayed in table [ table2 ] , and again , @xmath413 is quite close to @xmath28 has been computed using the gloptipoly software @xcite dedicated to solving the generalized problem of moments ] . .maxcut : @xmath411 ; @xmath417 for all @xmath409.[table1 ] [ cols="<,^,^,^,^,^,^,^",options="header " , ] the above examples seem to indicate that even though one chooses a measure @xmath5 uniformly distributed on @xmath4 , one obtains a rapid decrease in the first iterations and then a slow convergence close to @xmath28 . if on the one hand the convergence to @xmath28 is likely to be slow , on the other hand , one has to solve semidefinite programs ( [ primal ] ) with only one variable ! in fact solving the semidefinite program ( [ eigenvalue ] ) is computing the smallest _ generalized eigenvalue _ associated with the pair of real symmetric matrices @xmath418 , for which specialized codes are available ( instead of using a solver for semidefinite programs ) . however , one has to remember that the choice is limited to measures @xmath5 with @xmath3 and whose moments are available or easy to compute . hence , the present methodology is so far limited to simple sets @xmath4 as described before . finally , analyzing how the convergence to @xmath28 depends on @xmath5 is beyond the scope of the present paper and is a topic of further research . in nonlinear programming , sequences of upper bounds on the global minimum @xmath28 are usually obtained from feasible points @xmath95 , e.g. , via some ( local ) minimization algorithm . but for non convex problems , providing a sequence of upper bounds that converges to the global minimum @xmath28 is in general impossible unless one computes points on a grid whose mesh size tends to zero . in the above methodology one provides a monotone nonincreasing sequence of upper bounds converging to @xmath28 for polynomial optimization problems on sets @xmath4 , non necessarily compact but such that one may compute all moments of some finite borel measure @xmath5 with @xmath3 . in fact , if there are only finitely many ( say up to order @xmath25 ) moments available then one obtains a finite sequence of upper bounds . in contrast to the hierarchy of semidefinite relaxations in e.g. @xcite which provide lower bounds converging to @xmath28 when @xmath4 is a compact basic semi - algebraic set , the convergence of the upper bounds to @xmath28 is only asymptotic and never finite , except when @xmath4 is a discrete set . however , and even if we expect the convergence to be rather slow when close to @xmath28 , to our knowledge it is the first approach of this kind , and in a few iterations one may obtain upper bounds which ( even if crude ) complements the lower bounds obtained in @xcite ( in the compact case ) . also note that to solve ( [ eigenvalue ] ) several improvements are possible . for instance , we have already mentioned that it could be solved via specialized packages for generalized eigenvalue problems . next , if instead of using the canonical basis of monomial @xmath40 , one now expresses the moment matrix @xmath53 with rows and columns indexed in the basis of polynomials @xmath419 $ ] ( up to degree @xmath6 ) _ orthogonal_$ ] is orthogonal with respect to a finite measure @xmath5 on @xmath190 if @xmath420 . for extensions to the multivariate case see e.g. @xcite . ] with respect to @xmath5 , then @xmath53 becomes the identity matrix . and so problem ( [ eigenvalue ] ) reduces to a standard eigenvalue problem , namely that of computing the smallest eigenvalue of the ( real and symmetric ) localizing matrix @xmath70 ( expressed in the basis of orthogonal polynomials ) ! and it turns out that computing the orthogonal polynomials is easy once the moment matrix @xmath53 is available , since they can be obtained via computing certain determinants , as explained in e.g. @xcite . finally , observe that the above methodology perfectly fits _ inverse problems _ from moments , where precisely some borel measure @xmath5 is known only from its moments ( via some measurement device ) , and one wishes to recover ( or approximately recover ) its support @xmath4 from the known moments ; see e.g. the work of cuyt et al . @xcite and the many references therein . hence if @xmath41 $ ] is fixed then by definition @xmath250 ( on @xmath4 ) provides a strong _ valid _ ( polynomial ) _ inequality _ for the unknown set @xmath4 . so computing an optimal solution @xmath257 of ( [ primal ] ) for @xmath6 sufficiently large , will provide an almost - valid polynomial inequality @xmath421 for @xmath4 . one may even let @xmath41_d$ ] be unknown and search for the best " valid inequality @xmath422 where @xmath0 varies in some family ( e.g. linear or quadratic polynomials ) and minimizes ome appropriate ( linear or convex ) objective function of its vector of coefficents @xmath204 . in this paper we have presented a new characterization of nonnegativity on a closed set @xmath4 which is based on the knowledge of a single finite borel measure @xmath5 with @xmath3 . it permits to obtain a hierarchy of spectrahedra which provides a nested sequence of outer approximations of the convex cone of polynomials of degree at most @xmath6 , nonnegative on @xmath4 . when used in polynomial optimization for certain simple sets " @xmath4 , one obtains a hierarchy of semidefinite approximations ( with only one variable ) which provides a nonincreasing sequence of upper bounds converging to the global optimum , hence a complement to the sequence of upper bounds provided by the hierarchy of semidefinite relaxations defined in e.g. @xcite when @xmath4 is compact and basic semi - algebraic . a topic of further investigation is to analyze the efficiency of such an approach on a sample of optimization problems on simple closed sets like the whole space @xmath7 , the positive orthant @xmath175 , a box , a simplex , or an ellipsoid , as well as for some inverse problems from moments . the _ only if _ part is exactly the same as in the proof of theorem [ thmradon ] . for the _ if _ part , let @xmath117 be the sequence defined in ( [ defnu ] ) . the sequence @xmath136 is well defined because @xmath423 is @xmath5-integrable for all @xmath58 . indeed , if @xmath0 is a polynomial ( so that @xmath5 is defined in ( [ defmu ] ) ) we have seen that all moments of @xmath5 are finite and since @xmath424 is a polynomial the result follows . if @xmath0 is not a polynomial ( so that @xmath5 is defined in ( [ noncompact-1 ] ) ) then @xmath425 where we have used that @xmath426 , and @xmath427 for all @xmath333 . as in the proof of theorem [ thmradon ] , the set function @xmath428 , @xmath79 , is a signed borel measure because again @xmath83 can be written as the difference @xmath429 of the two positive borel measures @xmath87 in ( [ decomp])-([hahn ] ) . with same majorizations as above , both @xmath430 and @xmath431 are finite borel measures and so @xmath83 is a finite signed borel measure . indeed , first consider the case where @xmath0 is a polynomial ( and so @xmath5 is as in ( [ defmu ] ) ) . let @xmath432 be fixed arbitrary , and let @xmath433 . observe that whenever @xmath434 , @xmath435 on the subset @xmath436^n\,:\,\vert x_j\vert=\max_i\vert x_i\vert\}$ ] . and so , @xmath437 for all @xmath438 ( and where @xmath439 ) . hence , @xmath440^n } \vert f(\x)\vert\,x_i^{2k}d\mu(\x)+ % \vert f\vert_1\,\sum_{j=1}^n\int_{\r^n\setminus [ -1,1]^n } \,x_j^{2(k+s)}d\mu(\x)\\ \vert f\vert_1\,\sum_{j=1}^n\int_{\k\cap w_j } \,x_j^{2(k+s)}d\mu(\x)\\ \label{aux - carl1 } & \leq & \vert f\vert_1\mu(\k)+mn\vert f\vert_1\,(2(k+s))\l\,\leq\,2mn\vert f\vert_1\,(2(k+s))\l,\end{aligned}\ ] ] where @xmath441 is as in ( [ bigm ] ) and assuming with no loss of generality that @xmath442 ( otherwise rescale @xmath20 ) . and so we have @xmath443 where @xmath444 is sufficiently large so that @xmath445 . therefore , @xmath446 where the last equality follows from @xmath447 . indeed , @xmath448 and @xmath449 whenever @xmath167 is sufficiently large , say @xmath450 . hence the sequence @xmath136 satisfies carleman s condition ( [ carleman ] ) . if @xmath0 is not a polynomial then @xmath5 is as in ( [ noncompact-1 ] ) and so @xmath451 where we have used that @xmath427 for all @xmath333 , and @xmath452 . and so again , the sequence @xmath136 satisfies carleman s condition ( [ carleman ] ) . next , as @xmath134 for every @xmath63 , by proposition [ prop - berg ] , @xmath136 is the moment sequence of a measure @xmath137 on @xmath7 and @xmath137 is determinate . in addition , from the definition ( [ defnu ] ) of @xmath83 and @xmath136 , we have @xmath453 but then using @xmath454 in ( [ decomp])-([hahn ] ) , ( [ eq1 ] ) reads @xmath455 let @xmath456 , @xmath58 , be the sequence of moments associated with @xmath430 . of course , @xmath457 for all @xmath63 . next , @xmath458 and so , depending on whether @xmath0 is a polynomial or not , we obtain @xmath459 as in ( [ aux - carl1 ] ) or @xmath460 as in ( [ aux - carl2 ] ) . in both cases the sequence @xmath461 satisfies carleman s condition ( [ carleman ] ) and since @xmath457 for all @xmath63 , by proposition [ prop - berg ] , @xmath430 is moment determinate . but then ( [ eq2 ] ) yields @xmath462 , or equivalently , @xmath463 , that is , @xmath83 is a positive measure . the rest of the proof is exactly the same as for proof of theorem [ thmradon ] . v. baldoni , n. berline , j. de loera , m. kppe , and m. vergne . how to integrate a polynomial on a simplex , math . comp . * 80 * ( 2011 ) , pp . 297325 . c. berg . the multidimensional moment problem and semi - groups , proc . appl . math . * 37 * ( 1980 ) , pp . 110124 . g. blekherman . there are significantly more nonnegative polynomials than sums of squares , israel j. math . * 153 * ( 2006 ) , pp . 355380 . a. cuyt , g. golub , p. milanfar and b. verdonk . mutlidimensional integral inversion with application in shape reconstruction , siam j. sci . * 27 * ( 2005 ) , pp . dunkl and y. xu . _ orthogonal polynomials of several variables _ , cambridge univ . press . , cambridge , 2001 . m. dr . copositive programming : a survey , optimization - online , 2009 . h huy vui and t. s. pham . global optimization of polynomials using the truncated tangency variety and sums of squares , siam j. optim . * 19 * ( 2008 ) , pp . 941951 . helton , j.b . lasserre , and m. putinar . measures with zeros in the inverse of their moment matrix , ann . * 36 * ( 2008 ) , pp . 14531471 . d. henrion , j.b . lasserre , y. lfberg . gloptipoly 3 : moments , optimization and semidefinite programming , optim . methods and softwares * 24 * ( 2009 ) , pp . maserick and c. berg . exponentially bounded positive definite functions . ill , j. math . * 28 * ( 1984 ) , pp . global optimization with polynomials and the problem of moments , siam j. optim . * 11 * ( 2001 ) , pp . _ moments , positive polynomials and their applications _ , imperial college press , london , 2009 . lasserre and m. putinar . positivity and optimization for semi - algebraic functions , siam j. optim . * 20 * ( 2010 ) , pp . jensen s inequality , bull . * 43 * ( 1937 ) , pp . . m. marshall and t. netzer . positivstellenstze for real function algebras ( 2010 ) . a.e . nussbaum . quasi - analytic vectors . ark * 6 * ( 1966 ) , pp . 179191 . m. putinar . positive polynomials on compact semi - algebraic sets , indiana univ . * 42 * ( 1993 ) , 969984 . _ real analysis _ , 3rd . , macmillan publishing company , new york , 1988 . k. schmdgen . the @xmath464-moment problem for compact semi - algebraic sets , math . * 289 * ( 1991 ) , 203206 . m. schweighofer . global optimization of polynomials using gradient tentacles and sums of squares , siam j. optim . * 17 * ( 2006 ) , pp . 920942 . g. stengel . a nullstellensatz and a positivstellensatz in semialgebraic geometry , math . 207 * , pp . 8797 . l. vandenberghe and s. boyd . semidefinite programming , siam rev . * 38 * ( 1996 ) , pp .

we use a dense redshift survey in the foreground of the subaru gto2deg weak lensing filed ( centered at = 16; = 43 ) to assess the completeness and efficiency of massive halo identification in the weak lensing map . the redshift survey ( published here ) includes 4541 galaxies ; 4405 are new redshifts measured with the hectospec on the mmt . among the weak lensing peaks with a signal - to - noise greater that 3.7 , only 50% correspond to individual massive systems ; this result is essentially identical to the geller et al . ( 2010 ) test of the deep lens survey field f2 . the subaru map , based on images in substantially better seeing than the dls , enables detection of less massive halos at fixed redshift as expected . we demonstrate that the procedure adopted by miyazaki et al . ( 2007 ) for removing some unreliable peaks from the weak lensing map substantially improves agreement between the lensing map and the redshift survey in the identification of candidate massive systems .

1102.5743

weak lensing maps and redshift surveys are fundamental , complementary tools of modern cosmology . a weak lensing map provides a weighted `` picture '' of projected surface mass density ; a redshift survey provides a three - dimensional map of the galaxy distribution resolving structures along the line of sight within the broad window `` imaged '' by the lensing map . although galaxies are biased tracers of the mass distribution , comparison of a weak lensing map with a foreground redshift survey covering the appropriate range promises progress in resolving some of the issues limiting cosmological applications of weak lensing . here we focus on the use of weak lensing for identifying clusters of galaxies . this paper is the second in a series comparing massive clusters identified in a dense foreground redshift survey with significant weak lensing map convergence peaks . in our first paper ( geller et al . 2010 ) , we compared clusters identified in a hectospec redshift survey ( shels : geller et al . 2005 ) with a deep lens survey ( dls ; wittman et al . 2001 ) weak lensing map derived from deep kitt peak 4-meter observations . here we describe a hectospec redshift survey of the gto2deg@xmath0 field where miyazaki et al . ( 2002 , 2007 ) used subaru telescope observations as the basis for their weak lensing convergence map . these two assessments of the efficacy of weak lensing for massive cluster identification provide insights into both the strengths and limitations of the technique . since the first identifications of rich clusters from weak lensing maps ( wittman et al . 2001 ; miyazaki et al . 2002 ) , the number of detections has risen steeply ( hetterscheidt et al . ( 2005 ) , wittman et al . ( 2006 ) , schirmer et al . ( 2007 ) , gavazzi & soucail ( 2007 ) , miyazaki et al . ( 2007 ) , maturi et al . ( 2007 ) , berg et al . ( 2008 ) , and dietrich et al . ( 2008 ) ) . geller et al . ( 2010 ) discuss the varying approaches and success rates of weak lensing surveys as rich cluster identifiers . their comparison between clusters identified in the shels redshift survey and the dls weak lensing map agrees with the more pessimistic observational and theoretical assessments of the technique ( e.g. hamana et al 2004 ; hennawi & spergel 2005 ; schirmer et al . geller et al . ( 2010 ) conclude that only 4 of the 12 significant dls weak lensing peaks with signal - to- noise greater than 3.5 coincide with clusters massive enough to account for the lensing signal . the shels / dls weak - lensing efficiency for cluster detection is more pessimistic than many previous analyses at least in part because geller et al . ( 2010 ) require that the velocity dispersion determined from the shels redshift survey be large enough to account for the lensing signal . previous studies based on counts or photometric redshifts could not provide a velocity dispersion and often they could not distinguish between a genuine massive cluster and several groups superimposed along the line - of - sight . the underlying causes of the low efficiency of the dls cluster candidate list are hard to identify from the single survey field geller et al . ( 2010 ) studied . to gain further understanding of the relationship between clusters identified in a redshift survey and significant peaks in a weak lensing convergence map , we carried out a redshift survey of a second field . our goal is an independent measurement of the efficiency of weak lensing maps for identifying massive clusters . to explore the efficiency of weak lensing maps further , we chose a weak lensing map and candidate halo list constructed by different observers on a different facility . the subaru weak lensing survey ( miyazaki et al . 2002 ; 2007 ) provides an interesting comparison with the dls because ( 1 ) subaru is an 8.2-meter telescope , ( 2 ) the median seeing for the subaru data was @xmath90.66@xmath8 in contrast with the 0.85@xmath8 seeing for the dls and ( 3 ) the algorithms for constructing the weak lensing map and for identifying significant peaks differ in the two projects . the typical number of sources in the subaru map is 35 arcmin@xmath10 ; in the dls map it is 22.4 arcmin@xmath10 . the _ a priori _ expectation might be that a deeper subaru map with observations in better seeing should correspond more closely to the set of massive systems in the redshift survey . the subaru map should also enable detection of less massive systems at a fixed redshift . we test these expectations here . in section [ data ] we discuss the history of related observations ( section [ history ] ) and we describe the redshift survey ( section [ redshifts ] ) . our table of redshifts in the field includes 4405 new measurements with hectospec ( fabricant et al . 1998 , 2005 ) . next we describe the revised subaru weak lensing map ( section [ newmap ] ) and its sensitivity ( section [ sensitivity ] ) . we sample the redshift survey in section [ probes ] and identify candidate clusters in the weak lensing map and in the redshift survey ( section [ clusters ] ) . we discuss the efficiency of the weak lensing map in identifying systems of galaxies and compare it with our earlier results for the dls field f2 in section [ comparison ] . we conclude in section [ conclusion ] . miyazaki et al . ( 2002 ; 2007 ) used the 8.2 meter subaru telescope to carry out a weak lensing survey of thirteen well - separated fields covering a total of 21.82 deg@xmath0 . they used lensing convergence maps to identify 100 candidate halos with signal - to - noise @xmath11 3.7 . in their gto2deg@xmath0 field ( centered at @xmath1 = 16@xmath204@xmath344@xmath4;@xmath5 = 43@xmath611@xmath724@xmath8 ) , we have completed a dense redshift survey . our gto2deg@xmath0 redshift survey includes redshifts for 4549 galaxies and covers the entire 2.1 deg@xmath0 subaru field . we follow the approach of geller et al . ( 2010 ) and use the redshift survey as a test of the efficiency of the weak lensing map in identifying massive halos . during the commissioning phase of suprime - cam on subaru , miyazaki et al . ( 2002 ) imaged part of a field studied years earlier by gunn et al . ( gho : 1986 ) . here we review previous spectroscopic observations of clusters in this field . in 1998 , postman et al . ( 1998 ) used the keck to make both photometric and spectroscopic observations of the gho cluster cl 1604 + 4304 which lies within the gto2deg@xmath0 field . the cluster is a centrally concentrated , rich system at a redshift @xmath12 with a rest - frame line - of - sight velocity dispersion of 1226 km s@xmath13 ( based on 22 members ) . the mass is @xmath14 m@xmath15 . lubin et al . ( 1998 ) carried out a morphological study of the cluster based on hst observations and later lubin et al . ( 2004 ) used xmm - newton observations to study its x - ray properties . lubin et al . ( 2000 ) studied the clusters cl1604 + 4304 and cl 1604 + 4321 suggesting that they form a supercluster . the cluster cl 1604 + 4321 is at a redshift @xmath16 with a rest - frame line - of - sight velocity dispersion of 935 km s@xmath13 and a mass of @xmath17 m@xmath18 . these clusters are beyond the range of sensitivity of the subaru weak lensing map ( see section [ newmap ] ) . in 1999 , dressler et al . ( 1999 ) included the cluster cl1601 + 42 in their spectroscopic catalog of 10 distant rich clusters of galaxies . girardi & mezzetti ( 2001 ) analyzed the redshift distribution for the 46 cluster members . they obtained a mean redshift of 0.54 and a rest frame line - of - sight velocity dispersion of 646@xmath19 km s@xmath13 . they compute a cluster mass of @xmath20 m@xmath18 . this cluster is well within the sensitivity range of the weak lensing map and we compare these results with our own estimate in section [ clusters ] . in contrast with the observations of small patches of the original gho field geared toward studying individual clusters at reasonably high redshift , the wider field miyazaki et al ( 2002 ) observation provides the basis for a weak lensing convergence map of the region . the weak lensing convergence map is sensitive to systems at lower redshifts ranging from @xmath21 to @xmath22 . in section [ clusters ] , we show that the cluster cl1601 + 42 ( previously studied by dressler ( 1999 ) and girardi & mezzetti ( 2001 ) appears as one of the two most significant peaks in the weak lensing map . hamana et al ( 2009 ) carried out multi - object spectroscopy along four lines - of - sight in the gto2deg@xmath0 field . they observed with a single 6@xmath7 diameter slit mask in each of four patches and measured a total of 92 redshifts . on the basis of 15 of these galaxies coincident in redshift , sl j1602.8 + 4335 is a cluster at a redshift of 0.42 with a rest - frame line - of - sight velocity dispersion of 675 km s@xmath13 ; this cluster is the other of the two most significant weak lensing peaks in the miyazaki et al . ( 2007 ) map . we compare this measurement with our redshift survey results in section [ clusters ] and table [ tbl : vdisp ] . we investigate the association between convergence peaks and clusters ( halos ) in the galaxy distribution based on a densely sampled redshift survey . our goal is identification of the systems of galaxies that should produce a weak lensing signal . we thus seek to identify candidate systems in the redshift survey with rest frame line - of - sight velocity dispersion @xmath23 km s@xmath13 consistent with the sensitivity of the subaru map . we constructed the galaxy catalog from the sloan digital sky survey ( sdss , adelman - mccarthy 2008 ) r - band galaxy list for the gto2deg@xmath0 field . the field covers the region @xmath24 and @xmath25 . because we seek to identify the systems of galaxies corresponding to the weak lensing peaks , we focus the redshift survey on the red galaxy population . these red objects preferentially populate clusters . this procedure also suppresses the blue foreground populations ; the weak lensing map is insensitive to systems at redshift less than @xmath26 ( see figure [ fig : sensitivity.ps ] ) . we acquired spectra for the objects with the hectospec ( fabricant et al . 1998 , 2005 ) , a 300-fiber robotic instrument mounted on the mmt from february 1 , 2009 to april 27 , 2009 . the hectospec observation planning software ( roll et al . 1998 ) enables efficient acquisition of a pre - selected sample of galaxies . the software enables assignment of priorities as a function of galaxy properties . the spectra cover the wavelength range 3500 10,000 with a resolution of @xmath96 . exposure times ranged from 0.75 1.5 hours . we reduced the data with the standard hectospec pipeline ( mink et al . 2007 ) and derived redshifts with rvsao ( kurtz & mink 1998 ) with templates constructed for this purpose ( fabricant et al . 2005 ) . repeat observations yield robust estimates of the median error in @xmath27 where @xmath28 is the redshift and @xmath29 is the speed of light . for emission line objects , the median error ( normalized by @xmath30 ) is 27 km s@xmath13 ; the median for absorption line objects ( again normalized by @xmath31 ) ) is 37 km s@xmath13 . ( geller et al . 2010 ; fabricant et al . 2005 ) . there are 4541 galaxies in the region with a measured redshift ; our hectospec observations provide 4405 of these and the sdss provides 136 . in addition , we measured hectospec redshifts for 54 of the 136 sdss objects . the sdss objects are brighter and at lower redshift than the others we observed with hectospec . table 1 lists all of the objects and their redshifts . the table includes the sdss i d ( column 1 ) , the j2000 right ascension ( column 2 ) , the j2000 declination ( column 3 ) , the sdss petrosian @xmath32 magnitude ( column 4 ) , the sdss fiber color @xmath33 ( column 5 ) , the sdss fiber color @xmath34 ( column 6 ) , the redshift ( column 7 ) , the redshift error ( column 8) , and the redshift source ( column 9 ) . our redshifts agree very well with the 54 objects in common with sdss ; these objects range in redshift from 0.019 to 0.468 . at lower redshifts , the galaxies in common are emission line objects and at higher redshift they are from the sdss red galaxy sample . the overlapping objects are brighter than the typical galaxy in our sample . the mean difference @xmath35 = 3.4 km s@xmath13 ; the median difference is -0.8 km s@xmath13 . the dispersion around the mean is 36.8 km s@xmath13 . our first priority targets in the gto2deg@xmath0 field are galaxies with @xmath36 , @xmath37 and @xmath38 . the bright limit is the sdss completeness limit ( @xmath32 = 17.77 ) ; we observe only fainter galaxies for efficiency . our target selection yields 3992 objects . in this subsample , we obtained spectra for 3021 galaxies and 149 stars . in the original selection , we include everything the sdss classifies as a galaxy ; spectroscopy shows that this selection includes some stars . among the remaining 822 objects without spectroscopy , we expect @xmath39 stars . the integral completeness is thus @xmath40 . figure [ fig : completeness.subaru.histo.ps ] shows that the differential completeness drops to @xmath41 at @xmath42 . in figure [ fig : nozandz.ps ] the open circles show the positions of galaxies within the primary sample with ( upper panel ) and without ( lower panel ) a redshift . the pattern of incompleteness is a natural consequence of the fiber positioning constraints . figure [ fig : nozandz.ps ] also shows the positions of the 6 robust convergence map peaks ( see section [ newmap ] and table [ tbl : vdisp ] ) . note that the numerical designations of the robust peaks are not sequential because there are peaks of intervening rank that are not secure ( section [ newmap ] ) . our secondary targets are ( i ) galaxies with @xmath43 , @xmath37 , @xmath38 ranked by central surface brightness and ( ii ) bluer objects with @xmath44 . we observe the objects with @xmath43 beginning with the highest surface brightness objects . the lower panel of figure [ fig : faintandblue.ps ] shows the distribution on the sky of the 546 faint red galaxies with redshifts . the upper panel shows the distribution on the sky of the 974 bluer objects with a redshift . the faint red objects are reasonably uniform across the field ; the blue objects are concentrated toward the center of the field as a result of the positioning of the hectospec fields . covering the edges of the field is relatively inefficient because the 1 degree diameter hectospec field overlaps the boundary . figure [ fig : rvsz.subaru.ps ] shows the distribution of apparent magnitude , @xmath32 , for all of the galaxies with measured redshifts . different symbols indicate the three subsamples . at every redshift @xmath45 the redshift survey samples more than a one magnitude range in @xmath32 . figure [ fig : redshift.histogram.ps ] shows the redshift distribution for the entire gto2deg@xmath0 field survey and for the primary target sample . the median redshift for the primary target sample of 3025 galaxies is 0.336 ; for the entire sample , the median redshift is 0.349 . the galaxies with @xmath43 populate the high redshift range and the sdss galaxies fill in the survey at low redshift . the sharply defined peaks are the expected signature of the large - scale structure of the universe . figure [ fig : cone.diagram.subaru.ps ] shows cone diagrams for the survey . obvious voids are delineated by walls and filaments . the contrast between the voids and the surrounding structure is enhanced relative to a redshift without color selection for red galaxies . the red galaxies are more clustered and preferentially populate more dense regions . fingers extending along the line - of - sight corresponding to massive clusters of galaxies are evident by eye at @xmath46 and at @xmath47 . in section [ clusters ] we show that these two prominent systems correspond to the two most significant weak lensing peaks . redshifts are crucial for identifying candidate systems that correspond to weak lensing peaks . a dense enough redshift survey discriminates between superpositions along the line - of - sight and systems with large velocity dispersion . it also enables determination of the velocity dispersion . as an example of the inadequacy of photometric redshifts ( photoz hereafter ) for delineating structure in a region and for the identification even of relatively rich systems , figure [ fig : cone.photoz.ps ] shows the appearance the cone diagram for the gto2deg@xmath0 field for simulated perfect 3% photoz s . we construct the photoz plot by randomly sampling a gaussian of width @xmath48 centered on each measured redshift in the cone diagram for the gto2deg@xmath0 field . the accompanying histograms compare the the distribution of measured redshifts with the simulated photoz s . obviously the photoz s wash out all of the structure . we now proceed toward testing an updated version of the miyazaki et al . ( 2007 ) gto2deg@xmath0 convergence map against our dense redshift survey covering the entire field . the redshift survey generally enables separation between massive clusters and superpositions along the line - of - sight . the redshift survey provides an objective test of the fraction of weak lensing peaks that correspond to massive systems and it can , in principle , identify massive systems undetected in the convergence map . miyazaki et al ( 2007 ) first published a convergence map of the gto2deg@xmath0 field . we first reconstruct the weak lensing map to take advantage of improvements in the suprimecam reduction software and to refine the procedures we use to construct the map . we adopt mean stacking of the raw images rather than the median stacking adopted by miyazaki et al . this revised procedure reduces the effect of variations in seeing from one frame to another . otherwise we reduce the raw images as fully described in miyazaki et al . ( 2007 ) . we describe the construction of the revised map in section [ newmap ] . we model the expected sensitivity of the map to massive systems of galaxies in section [ sensitivity ] . the gto2deg@xmath0 field consists of 9 suprime - cam pointings acquired on april 18 - 25 , 2001 . the exposure time varies from subfield to subfield . for each subfield , we select the best seeing images whose summed exposure time exceeds 2000 seconds ( or all the exposures for those fields where the total exposure time is less than 2000 seconds ) . the exposure time per field for the stacked images ranges from 1800 seconds to 2400 seconds . the median seeing for the 9 frames ranges from 0.64@xmath8 to 0.77@xmath8 with an overall median of @xmath49 . we use all galaxies with 23 @xmath50 26 that are detected at significance @xmath51 as sources . on average there are 35 sources arcmin@xmath10 ; a total of 254,010 sources meet our cuts . to derive the ellipticities of the sources , we use the routines _ getshapes _ , _ efit _ , and _ ecorrect _ in the _ imcat _ package ( kaiser , squires & broadhurst 1999 : ksb ) . we measure the object shears using _ imcat _ ( ksb ) . from the shears , we construct lensing maps using a modified version of the ksb algorithm ( hoekstra et al . 1998 ) . first , we measure the size and ellipticity of all the stars identified in the stacked images . we then use a second order polynomial function in pixel @xmath52 and @xmath53 coordinates to fit for the psf size and ellipticity in each subfield image . this procedure yields an estimate of the psf size and shape at each galaxy position , and , therefore , the stellar shear and smear polarizabilities , @xmath54 and @xmath55 , respectively . to calculate the psf shear and smear polarizability appropriate for each galaxy s position and size , we first calculate the polarizability tensor for a set of gaussian kernels with fwhm ranging from 0.8 to 4.0 ( in increments of 0.8 ) . at each position , we calculate the appropriate polarizabilities @xmath56 and @xmath57 by linear interpolation of the polarizability components for the kernels that bracket the galaxy size . we can thus calculate the induced polarizability of the galaxy as ( hoekstra et al . 1998 ; ksb ) @xmath58 because @xmath59 can be affected by uncertainty in the size determination of individual galaxies , our corrected shear measure for each galaxy is @xmath60 where @xmath61 is the measured ellipticity and orientation of the galaxy and @xmath62 is the average polarizability of the 20 nearest neighbors to each galaxy we consider . once we have produced a @xmath63 map from the shear distribution , we estimate a noise map by taking the variance over 100 realizations of the @xmath63 map . in each realization , we randomize the orientation of the galaxies ( leaving their positions fixed ) and regenerate the map . we show the significance map ( the @xmath63 map divided by the variance map ) in figure [ fig : gtokappawopeak.ps ] . we use sextractor to identify peaks in the @xmath63-s the pixels in the map are @xmath64 , and we use a smoothing kernel with a @xmath65 fwhm . a peak must have at least 3 connected pixels significant at the 2@xmath66 level or more to be selected ; @xmath67 is the variance in the @xmath63 -s / n map . the numbers in figure [ fig : gtokappawopeak.ps ] are the significance rank of the 11 peaks we find . table [ tbl : vdisp ] lists the significance of all of these 11 peaks with signal - to - noise greater than 3.7@xmath66 or , equivalently with @xmath68 . as in miyazaki et al . ( 2007 ) we remove peaks within 4@xmath69 of bright stars ( usno b1.0 r@xmath70 or b@xmath70 ) or within 2.7@xmath69 of the edge of the field ; we remove peaks 2 , 4 , 5 , 6 and 8 for these reasons . interestingly , we show below ( section [ clusters ] ) that these unreliable peaks do not correspond to any systems in the redshift survey . to estimate the efficiency and completeness of the weak - lensing selected cluster catalog derived from the @xmath63 map , we estimate the expected sensitivity limits of the convergence map as a function of redshift . for any individual background galaxy in the weak lensing limit , the induced _ measured _ tangential ellipticity ( and hence the s / n for detection ) depends on both the angular diameter distance to the galaxy and on the size of the galaxy relative to the psf . for a collection of galaxies , the effective total weight as a function of redshift is a polarizability - weighted sum over the individual distance ratios . the effective distance ratio for the ensemble of background galaxies is @xmath71 + where the subscripts @xmath72 and @xmath73 refer to the sources and lens , respectively . @xmath74 is the weight of each source galaxy ( which depends on the object size ) , @xmath75 is the lens redshift , @xmath76 is the angular diameter distance between @xmath77 and @xmath78 , ( the ratio is zero if @xmath79 ) , and the sum is over all sources . for a realistic survey an additional suppression results from misidentification of faint foreground galaxies , each of which should have weight zero . one of the advantages of subaru data we use here compared to other deep ground - based surveys comes from the improved seeing . in a typical deep ground - based image where the stellar psf is @xmath80 , up to half the galaxies detected at @xmath81 significance are unresolved ( with size less than @xmath82 ( kubo et al . the images of the gto field were taken in better seeing ( median seeing 0.66 ) . with this image quality , @xmath83% of the galaxies at the object depth cutoff are resolved from stars . for each redshift , we compute the angular diameter distance factors in equation ( 1 ) assuming the concordance " cosmology ( spergel et al . 2007 ) . we derive the @xmath74 terms using data from the cosmos subaru galaxy catalog ( taniguchi et al . 2007 ) and the cosmos acs catalog ( leauthaud et al . 2007 ) . the cosmos combined acs / subaru dataset is currently the best dataset for this calibration because it covers by far the largest area with both ground- and space - based resolution to faint limiting magnitude . in addition , the subaru multi - color observations provide photometric redshift estimates ( mobasher et al . 2007 ) for the majority of galaxies in the survey area . we use the cosmos catalog size and photometric redshift information to estimate the redshift distribution of the galaxies that would be resolved in our weak lensing map . we match galaxies in the cosmos subaru and acs catalogs with the additional requirement that the catalog i magnitudes match to within 0.3 ( this restriction eliminates errors in matching the two catalogs ; it also eliminates most objects that can not be adequately separated from neighboring objects in the subaru imaging ) . because the subaru images of the cosmos field were taken in comparable seeing ( median seeing @xmath90.55 ) , this procedure should not alter the cosmos sample compared with the subaru sample . the final catalog contains approximately 124,000 objects with magnitudes , sizes , and photometric redshift estimates . the galaxy catalog of the gto2deg@xmath0 field has sizes and magnitudes ( r@xmath84 magnitdes ; miyazaki et al . 2007 ) , but not photometric redshifts . to assign redshifts to the galaxies , we assign to every galaxy in the gto2deg@xmath0 field the redshift of a randomly selected galaxy in the cosmos catalog with the same magnitude and ( seeing - deconvolved size ) , where the list of of galaxies is taken from all the galaxies that match to within 0.05 in size and within 0.1 magnitudes . although this procedure does not necessarily assign the correct redshift to any single galaxy , it assigns the correct distribution of galaxy redshifts given the size and magnitude distribution of the galaxies in the gto2deg@xmath0 field , and thus a correct effective total weight via equation ( 1 ) . given this catalog of redshifts ( which by construction has the same number of objects and the same size distribution as the one used in the construction of the lensing map ) , we calculate the effective weight @xmath85 that would be measured as a function of lens redshift . we normalize the redshift dependence of our sensitivity by comparing to simulations . to minimize the calculations , we normalize to the sensitivity at a fixed cluster redshift of 0.3 , and use the calculation of @xmath85 to calculate the redshift - dependence of the sensitivity curve . we follow a modified version of the procedure for generating simulated catalogs in khiabanian & dellantonio ( 2008 ) . we generate simulated galaxies with the same size - magnitude and magnitude - redshift relations as the hdf north and south fields ( williams et al . 1996 ; casertano et al . 2000 ) using the prescription of khiabanian & dellantonio ( 2008 ) . we use these simulated galaxies to populate images ( 0.04 resolution ) representing seven logarithmically - spaced redshift shells . we then distort the images with a lens modeled as a core - softened cutoff isothermal sphere with a given rest frame line - of - sight velocity dispersion , @xmath86 , a 1@xmath8 core radius , and a 100@xmath8 cutoff radius . we resample the distorted images with the subaru pixel scale ( 0.2/pix ) , coadd them , and convolve with a gaussian psf with a 0.66 fwhm and add noise to match the expected noise in the subaru gto observations . we repeat this procedure for different values of @xmath86 ( at a fixed @xmath87 ) to generate maps of the weak lensing signal . we run the kubo et al . ( 2009 ) detection method on these images to determine the value of @xmath86 where the peak s / n reaches our @xmath88 cluster selection limit . in figure [ fig : sensitivity.ps ] , we plot the detection limit as a function of redshift ( @xmath89 ) ; the solid line represents the @xmath88 cutoff in the lensing peak selection sample . our method of calculating the sensitivity fails at low redshift @xmath90 , because the simulation method of khiabanian & dellantonio ( 2008 ) underpredicts the significance of clusters that occupy a large angular region . however , because none of the gto2deg@xmath0 field clusters have @xmath91 , the comparison of sensitivities is valid . there are some fundamental limitations to the accuracy of the calculations . as a result of cosmic variance , the actual number and redshift distribution of the background galaxies behind any cluster may not be well - represented by our model . for clusters near the @xmath88 detection limit , the shear signal is based the shapes of only @xmath92 galaxies . at this level , non - uniformities in the distribution of the background galaxies are important . our detection sensitivities should be viewed as averages of the sensitivity across the field rather than the exact sensitivity to any one cluster at any particular position in the field . although the cosmos / subaru field is extremely valuable , it covers @xmath93 square degree of the sky . therefore , any systematic differences between the properties of the cosmos field and gto2deg@xmath0 can change the normalization . one of the results of the analysis of the dls ( geller et al . ( 2010 ) ) and the gto2deg@xmath0 field ( this work ) is that there are large field - to - field variations in the types and amount of structure present at all measured redshifts even on degree - sized angular scales . thus , although the cosmos field is currently the best field for calibrating weak lensing detections , larger fields will be necessary . the effect of these uncertainties is , however , limited . because the lensing kernel is so broad , and because the angular diameter distances change very slowly with redshift above @xmath94 for the concordance cosmology , problems introduced by cosmic variance are partially mitigated . typical s / n uncertainties due to background galaxy density variations are @xmath95% on the scales we are interested in , except very near bright stars or the cores of rich clusters . in the analysis of the incompleteness of weak lensing cluster detection , we take a 10% uncertainty in the sensitivity limit into account when determining which clusters selected from the redshift survey should be detected in the lensing map . to evaluate the correspondence between systems in the redshift survey and significant peaks in the convergence map , we follow the approach of geller et al . ( 2010 ) for sampling the redshift survey . we examine the redshift distribution in cones with radii of 3@xmath69 and 6@xmath7 . the 3@xmath69 sampling is similar to probes used in previous assessments of the efficiency of weak lensing based on counts ( schirmer et al . 2007 ) , photometric redshifts ( gavazzi & soucail 2007 ) and spectroscopic redshifts ( geller et al . 2010 ) . the combination of 3@xmath7 and 6@xmath7 probes samples the virial radii of massive clusters throughout the redshift range where we might expect the convergence map to detect these systems . the 3@xmath7 radius corresponds to 0.45 mpc for @xmath96 and to 1.1 mpc for @xmath97 . these radii are within r@xmath98 , the radius where the enclosed average mass density , @xmath99 . here @xmath100 is the critical density . the radius @xmath101 , a proxy for the virial radius , ranges from 1.0 to 2.0 mpc for well - sampled clusters with rest frame line - of - sight velocity dispersions in the range 500 km s@xmath13 to 1000 km s@xmath13 ( rines et al . 2003 ; rines & diaferio ( 2006 ) . according to the scaling relations of rines & diaferio ( 2006 ) , the corresponding range of masses , m@xmath98 , within r@xmath98 is 1.3 @xmath102 m@xmath18 to 1.2@xmath103 m@xmath18 . our goal is first to assess the _ efficiency _ of the convergence map in the identification of massive cluster halos . here we define _ efficiency _ as the fraction of convergence map peaks with @xmath68 ( the threshold chosen by miyazaki et al . ( 2007 ) ) that corresponds to an individual massive cluster identified within the redshift survey with a line - of - sight velocity dispersion above the solid @xmath104 threshold curve of figure [ fig : sensitivity.ps ] . in other words , the efficiency is the fraction of weak lensing peaks that correspond to appropriately massive individual systems . we also seek an assessment of the _ completeness _ of the set of massive clusters among the high confidence weak lensing candidate systems . we define _ completeness _ as the fraction of all clusters in the foreground redshift survey with line - of - sight velocity dispersions above the threshold curve that also correspond to weak lensing peaks with @xmath68 . in other words , the completeness is the fraction of appropriately massive individual systems in the redshift survey that are detected as weak lensing peaks . the definitions of _ efficiency _ and _ completeness _ exclude superpositions of small groups from consideration . these superpositions may indeed produce a weak lensing signal , but the weak lensing peak should obviously not be counted as detection of a single massive system in a mass selected cluster catalog . the definitions also exclude coincidences between the abundant low velocity dispersion systems in the redshift survey with convergence peaks . our definitions are similar to those adopted by hamana et al . ( 2004 ) in their simulations of the efficiency and completeness of weak lensing surveys for massive cluster identification . schirmer et al . ( 2007 ) and gavazzi & soucail ( 2007 ) develop uniform procedures for evaluating the coincidence of concentrations of galaxies with convergence map peaks . the approaches of both schirmer et al . ( 2007 ) and gavazzi & soucail ( 2007 ) are based on complete photometric surveys of their fields . they evaluate the significance of cluster candidates internal to their surveys . geller et al . ( 2010 ) develop an analogous approach for evaluating the coincidence of concentrations of galaxies in a foreground redshift survey with convergence map peaks . here we follow the geller et al . ( 2010 ) approach . because the redshift survey is substantially complete we can sample the survey itself to evaluate the significance of galaxy condensations along the line - of - sight to lensing peaks . the sensitivity plot ( figure [ fig : sensitivity.ps ] ) suggests that we need not look deeper than @xmath105 to identify clusters corresponding to the six weak lensing peaks with @xmath68 in the subaru weak lensing map . the most distant systems must be very rich and they should be obvious in the photometric data . we use the r@xmath84-band counts in the image to check for such distant systems . we examine the faint subaru r@xmath84 band counts ( miyazaki et al . 2007 ) in the range @xmath106 by binning the data in 0.3@xmath7 bins and smoothing the map of counts with a 1.5@xmath7 gaussian ( the same smoothing and pixel scale as the subaru weak lensing map ) . the range in apparent magnitude just overlaps the @xmath107 magnitude limit of our main redshift survey sample . the binning and smoothing lengths are essentially the same as those for the weak lensing map . we define @xmath108 as the standard deviation of the pixel value in the map of counts and we note peaks that rise @xmath109 above the median as corroborating evidence for systems of galaxies along the line - of - sight toward weak lensing peaks and/or as evidence for systems beyond the reach of the redshift survey . table [ tbl : vdisp ] indicates the lensing peak directions where there is a count excess . probes through the redshift survey centered on each galaxy in the survey contain many of the selection effects which impact a probe toward a lensing peak . thus comparison of these probes with similar probes toward the peaks should provide a measure of the significance of cluster candidates in the redshift survey . we have demonstrated by experiment that centering of probes on galaxies or on random points in the region has no effect on the results of our analysis . the steps in our candidate cluster identification procedure are : 1 . we sample the redshift survey in _ test cones _ with 3@xmath7 radius centered on each galaxy in the redshift survey . the cones are large enough to detect a galaxy cluster across the redshift range we sample . 2 . in each _ test cone _ we count the number of additional galaxies , n@xmath110 , within @xmath111 of the survey galaxy redshift . this window is comparable with the extent in redshift space of systems we wish to identify . 3 . in each of the redshift bins of item 2 , we evaluate the mean occupation and the variance ( @xmath112 ) across the entire survey . we then identify the set of _ test cones _ at each redshift which are 5@xmath112 above the mean occupation and contain at least 5 galaxies ( n@xmath113 ) . the minimum number of galaxies enables computation of a dispersion ( albeit with large error ) . the 5@xmath112 limit restricts the sample to high peaks which are reasonable candidate clusters especially at the peak sensitivity of the weak lensing map , @xmath114 . both the 5 galaxy and 5@xmath112 limits are generous ; they admit many peaks well below the expected detection threshold for the weak lensing map ( figure [ fig : sensitivity.ps ] ) . we call these probes _ 5@xmath112 probes _ hereafter . 4 . as a final consistency check we examine the subaru images for evidence of a massive system . we consider two indicators : ( 1 ) an @xmath109 excess in the count of galaxies with 21@xmath115 , and ( 2 ) the presence of a visually identifiable red sequence . in figure [ fig : sigmamap.scaled.ps ] the two most significant weak lensing peaks ( labeled 0 and 1 ) are coincident with two well populated clusters . these two clusters appear as prominent `` fingers '' at redshift 0.540 and 0.413 , respectively , in figure [ fig : cone.diagram.subaru.ps ] . there are obviously many well populated 5@xmath112 probes without any associated significant weak lensing peak . many of the well - populated redshift survey probes correspond to groups with small velocity dispersion . from figure [ fig : sensitivity.ps ] we would not expect these systems with rest frame line - of - sight velocity dispersion @xmath116 km s@xmath13 to produce a significant signal in the @xmath63 map . there are also weak lensing peaks with no corresponding peak in the redshift survey . it is interesting to note that none of the 5 unreliable peaks in table [ tbl : vdisp ] ( starred entries ) coincide with well - populated probes in the redshift survey ; removal of these unreliable peaks from consideration evidently increases the efficiency of the lensing map . to further assess the meaning of the weak lensing peaks we examine the redshift distributions along the line - of - sight toward each of the significant weak lensing peaks . we ask which lines - of - sight intersect a cluster with a velocity dispersion large enough to plausibly account for the weak lensing peak ( section [ clusters ] ) . in this section we evaluate candidate systems along the line - of - sight toward the 6 robust gto2deg@xmath0 weak lensing peaks of the revised subaru weak lensing map ( section [ newmap ] ) . figures [ fig : velhisto1.ps ] and [ fig : velhisto2.ps ] show the redshift @xmath28 distribution within 3@xmath69 of the central position of each of the candidate halos . the dark histogram shows the redshift distribution in bins of 0.002(1+@xmath28 ) . the thin histogram shows the redshift distribution in a concentric cone with a @xmath117 radius in each of the candidate halo directions . the probes toward peaks 0 and 1 each contain an impressive peak . these peaks correspond to obvious `` fingers '' in redshift space ( figure [ fig : velhisto1.ps ] ) . the rest frame line - of - sight velocity dispersions of these systems within a 6@xmath69 probe are , respectively 576@xmath11864 km s@xmath13 and 598@xmath11858 km s@xmath13 in agreement with previous measures ( dressler et al 1999 ( peak 0 ) ; hamana et al . 2008 ( peak 1 ) ; see section [ history ] ) . table [ tbl : vdisp ] provides the number of galaxies and the rest frame line - of - sight velocity dispersion for both 3@xmath7 and 6@xmath7 probes . the rest frame line - of - sight velocity dispersions for these two clusters places both of them very near the @xmath88 detection threshold . counts of faint galaxy within these probes show a @xmath119 excess as one might expect ( see table [ tbl : vdisp ] ) . we note that the weak lensing map detects the clusters corresponding to peaks 0 and 1 at high significance @xmath120 . however for @xmath120 , our predicted velocity dispersions for these peaks exceed the measured values by only 20 - 30% . the 1@xmath121 errors in the velocity dispersions are @xmath122% . thus we can not draw any definitive conclusion for this apparent difference between the measured velocity dispersion from the redshift survey and the weak lensing sensitivity analysis . the probe toward weak lensing peak 3 shows a more complex situation . there are two significant peaks in the histogram , one at @xmath123 and the other at @xmath124 . the line - of - sight velocity dispersions are 567 km s@xmath13 and 558 km s@xmath13 , respectively within 3@xmath7 probes ; the rest frame line - of - sight velocity dispersions are systematically smaller in the 6@xmath7 probes , but the errors are large . the faint galaxy counts show an excess ( table [ tbl : vdisp ] ) , probably resulting from the more distant system . both systems may contribute to the lensing signal . probes toward weak lensing peaks 7 and 9 reveal no candidate systems . figure [ fig : sigmamap.scaled.ps ] shows that neither of these peaks overlaps a well - populated probe through the redshift survey . furthermore , there is no excess in the faint galaxy counts . one possibility is that the weak lensing map in these directions detects a cluster at a redshift beyond the limits of the sensitivity of our redshift survey . however , the absence of a faint count excess implies that there is no distant cluster and supports the identification of these peaks as spurious signals ( table [ tbl : vdisp ] ) . along the line - of - sight toward weak lensing peak 10 , there are several narrow peaks corresponding to superposed groups along the line - of - sight . there is a group at @xmath125 with a velocity dispersion of 545 km s@xmath13 within a 6@xmath7 probe . we identify only 6 members of this system . the velocity dispersion of this system places it near the detection threshold albeit with a very large error . the most populated system along this line - of - sight ( as a result of the selection function of the survey ) has a mean redshift @xmath126 and a line - of - sight velocity dispersion of 404 km s@xmath13 . this system corresponds to a maxbcg cluster ( koester et al . 2007 ) at @xmath127 , @xmath128 with a quoted photoz of 0.254 and a bcg spectroscopic redshift of 0.208 . this system may enhance the lensing signal from the more distant system . there is a faint count excess in this direction table ( [ tbl : vdisp ] ) . taken at face value , the system at z = 0.469 alone accounts for the weak lensing signal . occasionally a bright star can be accidentally superimposed near a massive cluster of galaxies or a cluster may be located near the edge of the survey field . we remove 5 weak lensing peaks that we regard as unreliable because they are near a bright star or because they are too close to the survey boundary . neither the redshift survey nor the faint galaxy counts reveal any candidate systems along the line - of - sight toward any of these peaks . figure [ fig : velhisto3.ps ] shows two examples of the redshift distributions toward these peaks ; there are no candidate massive systems . the groups in the probe have rest frame line - of - sight velocity dispersions @xmath129 km s@xmath13 . these redshift survey probes are similar to the probes toward the weak lensing peaks ( peaks 7 and 9 ) and , like these peaks , they do not overlap well - populated probes through the redshift survey ( figure [ fig : sigmamap.scaled.ps ] ) . histograms for the other 3 unreliable lensing peaks are similar . the presence of high significance unreliable peaks demonstrates that identifiable systematic effects can produce spurious high significance peaks in the weak lensing map . identification and removal of these unreliable peaks substantially increases the fidelity of the weak lensing map as a route to identifying massive systems of galaxies . as another test of the nature of the spurious weak lensing peaks , we examine probes through the redshift survey toward the two well - populated positions indicated by heavy squares in figures [ fig : gtokappawopeak.ps ] and [ fig : sigmamap.scaled.ps ] . in the weak lensing map , there are no peaks significant at @xmath130 that are coincident with these redshift survey positions . figure [ fig : velhisto4.ps ] shows probes through the redshift survey in these two directions . by selection , the probes contain more galaxies than the probes toward the spurious weak lensing peaks 7 and 9 . they are populated by a superposition of groups with line - of - sight velocity dispersions @xmath131 km s@xmath13 . according to our sensitivity plot ( figure [ fig : sensitivity.ps ] ) , the systems in these probes should not produce a significant weak lensing signal and we observe that they do not , in fact , produce one . in summary , the 6 weak lensing candidate halos with @xmath68 yield a sample of 2 rich clusters associated with peaks 0 and 1 . a single system along the line - of - sight toward peak 10 may account for the lensing signal , but our error in the determination of the velocity dispersion is very large weak lensing peak 3 , and possibly 10 , might result from boosting of the lensing signal by superpositions of systems along the line - of - sight . thus the efficiency of the weak lensing map for the detection of individual massive systems is at best 50% for @xmath68 . table [ tbl : vdisp ] shows that two of the three peaks with @xmath132 are massive systems suggesting that , as expected , the efficiency may be greater at higher thresholds . geller et al . ( 2010 ) use the larger dls field to show that the efficiency of weak lensing increases with @xmath133 over the range @xmath134 . neither the foreground redshift survey nor the galaxy counts from the subaru images reveal any candidate systems along the line - of - sight toward weak lensing peaks 7 and 9 . the probes through the redshift survey toward these positions are similar to ( or even less well - populated than ) probes toward regions where there should be no significant weak lensing signal and none is observed . the redshift survey provides a route to identifying candidate galaxy systems undetected by the weak lensing map . it thus provides some assessment of the completeness of the cluster candidate list derived from the weak lensing map . there are many approaches to identifying candidate systems in the redshift survey . for consistency we use redshift survey probes to identify systems in the same way that we use them to test for systems along the line - of - sight toward weak lensing peaks . figure [ fig : sigmamap.scaled.ps ] shows the positions of all of the 5@xmath112 probes through the redshift survey ; these probes contain a well - populated redshift bin ( see section [ sampling ] ) . of course the bins are large enough that they may contain a superposition of poor groups rather than a richer system . we compute line - of - sight velocity dispersions for the peaks within each probe . in addition to the clusters corresponding to weak lensing peaks 0 and 1 , we find only one additional concentration in redshift space that corresponds to a massive system . the cluster we identify from the redshift survey at @xmath135 has an impressively dense central region and the subaru image reveals at least one faint strong lensing arc ( figure [ fig : arc.ps ] ) . detection in the weak lensing map is compromised because the cluster center ( @xmath1 = 240.62 ; @xmath5 = 42.57 ) is only @xmath136 arcminutes from the edge of the field . the weak lensing map does have a peak coincident with this cluster at a significance @xmath137 ( we list this cluster as the last entry in table [ tbl : vdisp ] ) . examination of the well - populated redshift survey probes also reveals a dense system with a rest frame line - of - sight velocity dispersion of 419@xmath11824 km s@xmath13 and a mean redshift of 0.185 based on 23 members within 6@xmath69 of the cluster center ( @xmath1 = 241.76 , @xmath5 = 42.80 ) . this system is obvious as a short , well - defined `` finger '' in figure [ fig : cone.diagram.subaru.ps ] . we identify it with the richness class zero abell cluster 2158 ( a2158c : struble & rood 1999 ; see also abell 1958 ) . foreground groups at a redshift @xmath138 have confused the identification of this system in the past . in the lensing map this cluster appears just below our @xmath88 detection threshold ; it is coincident with a weak lensing peak significant at @xmath139 . ( we list this cluster as the next to last entry in table [ tbl : vdisp ] ) . other than the clusters corresponding to weak lensing peaks 0 and 1 , we identify no clusters in the redshift survey that should produce a weak lensing signal @xmath140 . thus , the set of cluster detections appears to be complete . the largest challenge in constructing a weak lensing map is correction for atmospheric seeing ( schneider 2006 ) . because the source galaxies are small compared with typical atmospheric seeing , the signal - to - noise of a weak lensing map is very sensitive to seeing . schneider emphasizes the importance of good seeing for the determination of source ellipticities . he argues that a relatively shallow exposure in good seeing may be a better basis for a weak lensing map than a deeper exposure in poorer seeing ( see his figure 5 ) . the larger number of sources in the deeper exposure is only important if their ellipticities can be measured well . here we begin to examine this conjecture by comparing the gto2deg@xmath0 field ( a relatively short exposure in good seeing ) with the dls f2 field ( a very deep exposure in poorer seeing ) . we discuss some of the consequences for catalogs of weak lensing cluster candidates and for the determination of cluster abundance and the cluster mass function from weak lensing . the gto2deg@xmath0 field consists of subfields with a 30 - 40 minute exposure on the subaru 8.2-meter telescope in 0.66@xmath8 seeing ; the dls f2 field is a 5-hour exposure on the kpno 4-meter in 0.85@xmath8 seeing . figure [ fig : sensitive2.ps ] compares the weak lensing sensitivity of the two fields . the curves show the rest frame line - of sight velocity dispersion for a system which could be detected at a @xmath88 at each redshift . at every redshift , the subaru observations have the potential to yield a detection at a lower line - of - sight velocity dispersion . the maximum sensitivity is shifted toward greater redshift for the subaru data . table [ tbl : maps ] shows the substantial difference in the number of resolved sources for the two weak lensing maps : 22.4 arcminute@xmath10 for the dls and 35 arcminute@xmath10 for the gto2deg@xmath0 field . table [ tbl : maps ] lists other relevant parameters for the two surveys . in constructing table [ tbl : maps ] we have reevaluated the dls f2 results of geller et al . ( 2010 ) using a @xmath88 rather than a @xmath141 threshold for selecting significant peaks in the weak lensing map . we have also included the miyazaki et al . ( 2007 ) procedure for removing unreliable , but apparently high significance weak lensing peaks . figure [ fig : sensitive2.ps ] also shows the rest frame line - of sight velocity dispersions for systems in the foreground redshift surveys corresponding to weak lensing peaks with @xmath11 3.7 in both surveys . the most telling and suggestive difference between the admittedly small samples of clusters in the two surveys is that the demonstration that , as expected from the sensitivity curves , the subaru map cleanly detects clusters with rest frame line - of sight velocity significantly below the limits of the dls sensitivity at the same redshift . for example , the subaru lensing map cleanly detects a cluster at @xmath142 with a rest frame line - of - sight velocity dispersion of @xmath143 km s@xmath13 ; at a similar redshift the dls detects a system with a much larger velocity dispersion of @xmath144 km s@xmath13 at a @xmath145 . this comparison indicates that , in principle , observations with good seeing can be the basis for a weak lensing measurement of the cluster mass function for masses @xmath146 m@xmath18 ( rest frame line - of - sight velocity dispersion @xmath147 km s@xmath13 ) at redshifts @xmath148 . understanding both the completeness and the efficiency of weak lensing map cluster detections are , of course , crucial to their use as cosmological tools for measuring e.g. the cluster mass function . the dls weak lensing cluster sample is @xmath149% complete relative to a sample drawn from the foreground shels redshift survey ( geller et al . the gto2deg@xmath0 cluster catalog is complete ( section [ clusters ] ) , but the sample is so small that we can not draw a general conclusion . these samples certainly emphasize the need for more extensive detailed tests not dominated by small number statistics . another issue in measuring cluster abundance or evaluating the cluster mass function from a weak lensing survey is , of course , the low efficiency of the weak lensing maps . at least @xmath149% of the weak lensing peaks with @xmath68 in both the dls and gto2deg@xmath0 fields do not correspond to individual rich clusters ( massive halos ) ; they are either superpositions in the redshift survey or there is no candidate system at all . in other words , the efficiency of weak lensing maps in identifying massive systems of galaxies is low as predicted by the more pessimistic theoretical models ( hamana et al . 2004 ; hennawi & spergel 2005 ; maturi et al . weak lensing surveys are a potentially powerful tool for identifying massive systems of galaxies . these mass selected cluster catalogs could provide important cosmological constraints . objective tests of the efficiency and completeness are an important foundation for this application of weak lensing . the subaru gto2deg@xmath0 field ( ( centered at @xmath1 = 16@xmath204@xmath344@xmath4;@xmath5 = 43@xmath611@xmath724@xmath8 ) ) is the second where we have carried out a dense foreground redshift survey as a test of the fidelity of the weak lensing map . in contrast with the 0.85@xmath8 seeing of the dls f2 field of our first survey ( geller et al . 2010 ) , the subaru field has median seeing of 0.66@xmath8 . the number of resolved sources in the subaru field is 35 arcmin@xmath10 in contrast with 22 arcmin@xmath10 for the dls field . our tests of these two fields suggest that good seeing is crucial for reliable cosmological tests based on weak lensing cluster samples . to match the sensitivity of the subaru lensing map , the primary component of our redshift survey includes red galaxies ( @xmath150 and @xmath37 ) with sdss petrosian @xmath151 . the total number of galaxies with redshifts in the field is 4959 : 4541 are new redshifts measured with hectospec . this dense survey enables a variety of tests of the weak lensing map . our tests include ( 1 ) searching the redshift survey for massive systems along the line - of - sight toward weak lensing peaks with significance @xmath68 , ( 2 ) searching the redshift survey for all systems massive enough to produce a lensing signal to estimate the completeness of the weak lensing detections and ( 3 ) examining well - populated regions of the redshift survey in directions where there is no significant weak lensing signal to reinforce the redshift survey identification of spurious weak lensing peaks . in the subaru map , at most 50% of the weak lensing peaks with @xmath68 are detections of individual massive systems . the rest either contain superpositions of groups or contain no candidate systems at all . this result is the same as the result of our analysis of the larger dls f2 field , a deeper observation in poorer seeing ( table [ tbl : maps ] ) . in the dls f2 field , the shels foreground redshift survey identifies twice as many massive cluster candidates as the weak lensing map ; in the subaru field all of the cluster candidates are detected in both the redshift survey and the weak lensing map , but the sample is very small . in contrast with the dls f2 field , the subaru map detects less massive systems at the same redshift as one would expect from the greater source density . this comparison suggests that robust measurements of the cluster mass function down to a mass of @xmath152 m@xmath18 for redshifts @xmath153 can be derived from weak lensing maps provided the efficiency and completeness are understood . our results suggest that good seeing is crucial for this measurement and for progress in understanding systematic biases in weak lensing maps . we show that the procedure miyazaki et al . ( 2007 ) used to eliminate unreliable , but apparently high confidence peaks from their weak lensing map improves the fidelity of the weak lensing map in comparison with the foreground redshift survey . we sample the redshift survey in directions where there is no weak lensing signal . the population of redshift survey cones toward significant peaks with no cluster candidate can be even less well populated than these redshift survey cones in selected directions where there is no weak lensing signal . taken together , the subaru gto2deg@xmath0 and dls f2 fields underscore the need for a large - area direct test of weak lensing cluster identification based on high significance peaks derived from weak lensing . it appears from the small sample of gto2deg@xmath0 and dls clusters we have explored so far that these estimates agree within 20 - 30% when the cluster candidate is detected in both the redshift survey and the lensing map . detailed comparison of predicted ( from the weak lensing survey ) and measured line - of - sight velocity dispersions for a large sample might elucidate systematic issues . these surveys suggest that a 50 square degree survey should contain @xmath154 peaks with @xmath155 , a large enough sample of high significance peaks for robust comparison with simulations . we thank p. berlind and m. calkins for their expert operation of the hectospec . d. mink , j. roll , s. tokarz , and w. wyatt constructed and ran the hectospec pipeline . nelson caldwell deftly manages hectospec queue scheduling for optimal scientific results . eduard westra assisted with some of the observations . we thank scott kenyon for insightful comments and discussions . the smithsonian institution generously supported hectospec and this project . nsf grant ast-0708433 supports ian dell antonio s research . rrrrrrccc 587733442122744127 & 240.0929 & 42.7913 & 0.12636 & 0.00009 & 18.97 & 0.94 & 0.52 & hecto + 587733442122744638 & 240.0663 & 42.8125 & 0.48417 & 0.00011 & 20.86 & 1.02 & 0.49 & hecto + 587733442122745038 & 240.0662 & 42.7705 & 0.68305 & 0.00020 & 22.31 & 1.35 & 1.15 & hecto + 587733442122745075 & 240.0767 & 42.7428 & 0.68472 & 0.00022 & 21.03 & 1.55 & 1.20 & hecto + 587733442122809425 & 240.1192 & 42.6976 & 0.59909 & 0.00021 & 20.80 & 1.22 & 0.89 & hecto + 587733442122809463 & 240.2103 & 42.7137 & 0.48322 & 0.00013 & 20.14 & 1.32 & 0.81 & hecto + 587733442122809491 & 240.1748 & 42.6516 & 0.46796 & 0.00012 & 20.46 & 1.87 & 0.87 & hecto + 587733442122809551 & 240.0543 & 42.6959 & 0.09460 & 0.00006 & 17.64 & 0.89 & 0.41 & hecto + [ tbl : redshifts ] lcc field center ( @xmath156)&241.183 , 43.190&139.885 , 30.000 + field area(deg@xmath0)&2.02 & 4.1 + source density ( arcmin@xmath10 ) & 35&22.4 + median seeing ( @xmath8)&0.66 & 0.85 + seeing range ( @xmath8)&0.63 - 0.76 & @xmath157 + pixel scale ( @xmath8)&18&30 + smoothing length ( arcmin)&1.5 & 1.5 + n@xmath158&6 & 8 + clusters detected&3&4 + [ tbl : maps ] lccccccccccc 0&5.784520&240.776778&42.769138&0.540 & 613&20 & 78 & 576&29 & 64&y + 1 & 5.673880 & 240.715968&43.590494&0.413&619 & 15&68 & 598&26&58&y + 2@xmath159&4.451760&241.179641&43.451752 & & & & & & & & + 3 & 4.426710&241.035219&42.658144&0.297 & 595 & 6 & 87&480&18 & 61&y + & & & & 0.673&558&5&76&431&8&77 & + 4@xmath159 & 4.383750&241.483689&43.451752 & & & & & & & & + 5@xmath159&4.352070&241.552100&43.590494 & & & & & & & & + 6@xmath159 & 4.201050&240.396718&42.918980 & & & & & & & & + 7&3.819900&241.977766&42.597097 & & & & & & & & + 8@xmath159&3.802810&240.221891 & 43.630180 & & & & & & & & + 9&3.784870&241.240451&43.645991 & & & & & & & & + 10&3.734410&240.373915&42.769138&0.469&595&5 & 180 & 545 & 6 & 156 & y + 14@xmath160&3.4&241.757&42.791&0.185&463&16&58&419&24&48 & + 37@xmath0&2.9&240.616&42.569&0.602&639&8&97&639&8&97 & + redshift survey completeness . the vertical line separates the primary and faint galaxy samples . the primary sample : @xmath161 , @xmath150 and @xmath37 . for galaxies with @xmath162 , we use sdss redshifts . the faint sample contains galaxies with @xmath43 . , width=672 ] redshift survey . different symbols indicate the three subsets of galaxies plotted in figures [ fig : nozandz.ps ] and [ fig : faintandblue.ps ] . gray dots indicate galaxies in the primary sample . open circles indicate galaxies with @xmath43 and crosses indicate bluer objects . note that the bluer objects concentrate toward lower redshift and the fainter objects concentrate toward greater redshift as expected . [ fig : rvsz.subaru.ps],width=672 ] survey ( 4541 redshifts ) . the bins have width @xmath165.the black histogram shows the redshift distribution for the primary sample of red galaxies . the open histogram shows the additional galaxies in the fainter red galaxy sample ( at large redshift ) and the bluer sample ( lower redshift ) . the narrow peaks are the standard signature of large - scale structure . [ fig : redshift.histogram.ps],width=672 ] redshift survey projected in right ascension . note the fingers corresponding to rich clusters at redshifts 0.413 and 0.540 . these clusters correspond to the two most significant peaks in the subaru weak lensing map ( section [ clusters ] ) . the selection of primarily red galaxies enhances the contrast between the dense structures and the voids . [ fig : cone.diagram.subaru.ps],width=672 ] redshift survey shown in [ fig : cone.diagram.subaru.ps ] . the histograms compare the distribution of measured redshifts ( thin black line ) with the distribution of simulated photoz s ( heavy gray line ) . note that most of the structure disappears ; it is essentially impossible to identifying even the richest clusters in the region . [ fig : cone.photoz.ps],width=672 ] centers of the 3@xmath7 radius 5@xmath112 probes ( section [ sampling ] ; gray dots ) and centers of significant weak lensing peaks ( large numbers are reliable peaks ; small numbers are unreliable ) . the value of the number gives the rank of the convergence map peak . the tick marks on the declination axis are 6@xmath69 apart , the diameter of the probes . note that reassuringly none of the unreliable peaks ( 2 , 4 , 5 , 6 , and 8) overlap the gray dots . the x marks the position of a cluster detected in the redshift survey , but it is too close to the edge of the survey for reliable detection in the lensing map ; open squares are sample directions which are well - populated in the redshift survey but where there is no weak lensing peak significant at @xmath130 ; a denotes the position of the abell cluster 2158 . [ fig : sigmamap.scaled.ps],width=672 ] . the solid histogram shows the redshift distribution for a cone with a 3@xmath69 radius ; the open histogram shows the distribution for a 6@xmath7 radius . bins in redshift are @xmath165 wide . [ fig : velhisto1.ps],width=672 ] . the solid histogram shows the redshift distribution for a cone with a 3@xmath69 radius ; the open histogram shows the distribution for a 6@xmath7 radius . bins in redshift are @xmath165 wide . [ fig : velhisto2.ps],width=672 ] . the solid histogram shows the redshift distribution for a cone with a 3@xmath69 radius ; the open histogram shows the distribution for a 6@xmath7 radius . bins in redshift are @xmath165 wide . these histograms confirm the impression from figure [ fig : cone.photoz.ps ] that there are no systems of galaxies along the lines - of - sight toward these peaks resulting from systematic errors in the map . [ fig : velhisto3.ps],width=672 ] contour of the weak lensing map . the solid histogram shows the redshift distribution for a cone with a 3@xmath69 radius ; the open histogram shows the distribution for a 6@xmath7 radius . bins in redshift are @xmath165 wide . groups with rest frame line - of - sight velocity dispersions @xmath131 km s@xmath13 populate these probes . they are actually more densely populated than probes toward the spurious weak lensing peaks 7 and 9 ( figure [ fig : velhisto2.ps ] . [ fig : velhisto4.ps],width=672 ] and dls f2 weak lensing maps . the curves show the @xmath88 detection threshold for the two maps . the heavy curve refers to the gto2deg@xmath0 field ; the lighter curve refers to the dls f2 field . open points with error bars are the four clusters detected in the dls lensing map and redshift survey . the solid points are groups and clusters in the directions of the gto2deg@xmath0 weak lensing peaks : solid circles are peaks 0 , 1 , and 10 , solid squares are the superposition of peak 3 . the solid triangle indicates the cluster at redshift 0.602 discovered in the redshift survey but too near the edge of the subaru field for reliable weak lensing detection . [ fig : sensitive2.ps],width=643 ]

the extragalactic background light ( ebl ) is of fundamental importance both for understanding the entire process of galaxy evolution and for-ray astronomy , but the overall spectrum of the ebl between 0.1 - 1000 m has never been determined directly from galaxy spectral energy distribution ( sed ) observations over a wide redshift range . the evolving , overall spectrum of the ebl is derived here utilizing a novel method based on observations only . this is achieved from the observed evolution of the rest - frame-band galaxy luminosity function up to redshift 4 ( ) , combined with a determination of galaxy sed - type fractions . these are based on fitting swire templates to a multiwavelength sample of about 6000 galaxies in the redshift range from 0.2 to 1 from the all - wavelength extended groth strip international survey ( aegis ) . the changing fractions of quiescent galaxies , star - forming galaxies , starburst galaxies and active galactic nucleus ( agn ) galaxies in that redshift range are estimated , and two alternative extrapolations of sed - types to higher redshifts are considered . this allows calculation of the evolution of the luminosity densities from the uv to the ir , the evolving star formation rate density of the universe , the evolving contribution to the bolometric ebl from the different galaxy populations including agn galaxies and the buildup of the ebl . our ebl calculations are compared with those from a semi - analytic model , from another observationally - based model and observational data . the ebl uncertainties in our modeling based directly on the data are quantified , and their consequences for attenuation of very high energy-rays due to pair production on the ebl are discussed . it is concluded that the ebl is well constrained from the uv to the mid - ir , but independent efforts from infrared and-ray astronomy are needed in order to reduce the uncertainties in the far - ir . + [ firstpage ] galaxies : evolution galaxies : formation cosmology : observations - diffuse radiation infrared : diffuse background

1007.1459

the formation and evolution of galaxies in the universe is accompanied unavoidably by the emission of radiation . all this radiated energy is still streaming through the universe , although much is now at longer wavelengths due to redshifting and absorption / re - emission by dust . the photons mostly lie in the range @xmath3 0.1 - 1000 @xmath1 m , ultraviolet ( uv ) , optical , and infrared ( ir ) , and produce the second most energetic diffuse background after the cosmic microwave background , thus being essential for understanding the full energy balance of the universe . we will account in this work for the radiation accumulated by star formation processes through most of the life of the universe , plus a contribution from active galactic nuclei ( agns ) to this wavelength range , known as the diffuse extragalactic background light ( ebl ) . the direct measurement of the ebl is a very difficult task subject to high uncertainties . this is mainly due to the contribution of zodiacal light , some orders of magnitude larger than the ebl ( @xcite ; @xcite ) . there are some measurements in the optical ( @xcite ) and in the near - ir ( @xcite ; @xcite ) , but there is not general agreement about the reliability of these data sets ( @xcite ) . in addition , these near - ir data appear to give intensity levels for the ebl in contradiction with the observation of very high energy ( vhe , 30 gev-30 tev ) photons from extragalactic sources ( @xcite ; @xcite ; @xcite ) . little is known about the mid - ir from direct detection due to the higher contamination from zodiacal light at those wavelengths . measurements with the far - infrared absolute spectrometer ( firas ) instrument on board the cosmic background explorer , in the far - ir ( @xcite ; @xcite ) , are thought to be more reliable . other observational approaches set reliable lower limits on the ebl , such as measuring the integrated light from discrete extragalactic sources ( @xcite ; @xcite ) . there are also other authors that focus on studying galaxy properties based on ebl results ( @xcite ) , or on modeling a region of the ebl spectrum ( @xcite ) . on the other hand , there are phenomenological approaches in the literature that predict an overall ebl model ( between 0.1 - 1000 @xmath1 m and for any redshift ) . these are basically of four kinds : 1 . forward evolution , which begins with cosmological initial conditions and follows a forward evolution with time by means of semi - analytical models ( sams ) of galaxy formation , @xcite , @xcite ( hereafter , sgpd10 ) and @xcite ( hereafter , gspd10 ) . backward evolution , which begins with existing galaxy populations and extrapolates them backwards in time , @xcite , @xcite , @xcite ( hereafter , frv08 ) . evolution of the galaxy populations that is inferred over a range of redshifts . the galaxy evolution is inferred here using some quantity derived from observations such as the star formation rate ( sfr ) density of the universe , kneiske et al . ( 2002 ) , @xcite , @xcite . evolution of the galaxy populations that is directly observed over the range of redshifts that contribute significantly to the ebl . the present paper , which we term empirical , belongs in this category . the type ( i ) sgpd10 and gspd10 models discuss the same galaxy formation sam but in different contexts : sgpd10 contains details of the model used in calculating the bolometric luminosity history of the universe and comparison with data , and gspd10 focuses on the derived ebl and @xmath0-ray attenuation . the sgpd10-gspd10 model is based on an updated version of the semi - analytic theoretical approach described in @xcite from the growth of super - massive black holes and their host galaxies within the context of the hierarchical lambda cold dark matter ( @xmath4cdm ) cosmological framework . this is based in part on @xcite , @xcite , and in the simulations summarized by @xcite , and @xcite . we consider that these types of models are complementary to the observational approach taken here . we consider the type ( ii ) frv08 model the most complete observationally - based work of those mentioned above . they base their ebl modeling on galaxy luminosity functions ( lfs ) , quantities which are directly observed and well understood . frv08 exploit a variety of data to build evolutionary schemes according to galaxy morphology . they account for the contribution from early , late - type galaxies and a starburst population to the ebl . they use observed near - ir lfs from the local universe to @xmath5 for describing the early and late - type galaxies . for the starburst population they use an optical and only local lf . different prescriptions are used to extrapolate the evolution of the different morphological types to higher redshifts , and corrections to fit their results to other observational data are applied . type ( iii ) models are not directly based on galaxy data . instead , they are built from some parametrization of the history of the sfr density . this is a quantity derived using several different methods , each of which have different and signitficant uncertainties and biases . the sfr density is combined with uncertain assumptions about the emitted galaxy spectral energy distribution ( sed ) evolution as well . our type ( iv ) ebl estimates ( the first approach in this category ) will be compared in detail with the type ( i ) forward evolution semi - analytical galaxy formation model by sgpd10 and gspd10 , and the type ( ii ) oservationally motivated model by frv08 . the other works mentioned are briefly compared with our ebl calculations in sec . [ discussion ] . our aim in this paper is to develop an ebl model that is as observationally - based and realistic as possible , yet fully reproducible , including a quantitative study of the main uncertainties in the modeling that are directly due to the data . this constrains the range of the background intensity and its implications to @xmath0-ray astronomy . one important application of the ebl for @xmath0-ray astronomy is to recover the unattenuated spectra of extragalactic sources . our goal is to measure the ebl with enough precision that the uncertainties due to the ebl modeling , in these recovered unattenuated spectra , are small compared with other effects such as systematic uncertainties in the @xmath0-ray observations . examples of this are discussed in sec . [ attenuation ] . our model is based on the rest - frame @xmath2-band galaxy lf by @xcite ( hereafter , c10 ) and on multiwavelength galaxy data from the all - wavelength extended groth strip international survey ( aegis , @xcite ; @xcite ) of about 6000 galaxies in the redshift range 0.2 - 1 . these data sets are put together in a very transparent and consistent framework . the c10 lf is used to count galaxies ( and therefore to normalize the total ebl intensity ) at each redshift . the lf as well as our galaxy sample are divided in three magnitude bins according to the absolute rest - frame @xmath2-band magnitude faint , middle , and bright ( defined quantitatively later ) . within every magnitude bin a sed - type is statistically attached to each galaxy in the lf assuming sed - type fractions that are function of redshift within those magnitude bins . this is estimated from fitting our aegis galaxy sample to the 25 galaxy sed templates from the swire library . then , luminosity densities are calculated from these magnitude bins from every galaxy population at all wavelengths , and finally all the light at all redshifts is added up to get the overall ebl spectrum . the results are linked with @xmath0-ray astronomy and with the current understanding on galaxy evolution . the paper is organized as follows : section [ data ] describes the lf , our multiwavelength galaxy catalogue and the galaxy templates . section [ method ] explains our methodology . the results for galaxy sed - type fractions , luminosity densities , sfr densities , ebl buildup , and ebl intensities are given in section [ results ] . section [ attenuation ] shows the attenuation computed from our ebl model for some vhe sources taken from the literature . in section [ discussion ] our results are discussed including a detailed study on the uncertainties from the modeling , a comparison between our observationally - based ebl and that given by theoretical sams of galaxy formation . finally , in section [ summary ] a summary with our main results and conclusions is presented . throughout this paper , a standard @xmath4cdm cosmology is assumed , with matter density @xmath6 , vacuum energy density @xmath7 , and hubble constant @xmath8 km s@xmath9mpc@xmath9 . the evolving galaxy lf in rest - frame @xmath2-band provided by c10 from @xmath10 to 4 is used . this evolving lf is the most accurate measurement to date of cosmological galaxy evolution in the near - ir , where dust absorption is less severe than in optical bands . the @xmath11-corrections in this band are less severe than in the optical as well . the choice of the c10 lf to normalize the model is also based on the smooth and well - studied shape of the galaxy seds in the near - ir , unlike others in uv or mid - ir wavelengths . the resulting evolving lf is based on the ukidss ultra deep survey ( @xcite ) , which has a large area and depth , and hence reduces the uncertainties due to cosmic variance and survey incompleteness . we refer the reader interested in details to that work . it is important to note that they give a parametrization of the evolution of the lf corrected from incompleteness and fitted by a schechter function ( @xcite ) over redshift , @xmath12 , where @xmath13 is rest - frame @xmath2-band absolute magnitude at redshift @xmath14 . the strongest assumption that they make is to keep constant the faint - end slope @xmath15 in their parametrization . .the photometric bands in our galaxy sample . for each we show the effective wavelength , the data source , the requirement for that band to be included for a given galaxy in our sample ( det : a detection in this band is required ; obs : observation in this band is required , but not necessarily a detection ; ext : this band is considered extra information when available ) , and the 5@xmath16 upper limit in that band in cases where there is no detection . [ cols="<,^,^,^,^",options="header " , ] fig . [ fig12 ] shows the history of the sfr density of the universe computed from our modeling using eq . [ star ] . it is also plotted the prediction using the same equation , from the luminosity densities provided by sgpd10 , and a compilation of observational works from @xcite using different estimators , assuming a salpeter stellar imf . we are aware that this imf is not as good description of the observations as other imfs such as @xcite , but we are concerned here on showing a comparison with the compilation of sfr data , which is given by a salpeter imf . the data from @xmath17 are roughly reproduced . our results are in agreement within errors with the upper data envelope from @xmath18 . we systematically predict a factor @xmath19 higher sfr than the observational data between @xmath20 . for the _ high - starburst _ assumption a considerably higher sfr density is estimated . this _ high - starburst _ case is motivated by the increasing star formation rate density to @xmath21 in fig . [ fig12 ] , and the increasing specific star formation rate to @xmath21 ( @xcite ; @xcite ) . but fig . [ fig12 ] also indicates that our _ high - starburst _ is an extreme assumption . we want to call attention to the large uncertainties on the observational data estimates for the sfr for all redshifts . these uncertainties are especially important for the higher redshifts , mainly because local calibrations are used in the estimations , and also the uncertainties of the corrections due to dust absorption . the same is true for eq . [ star ] which is calibrated using observed local galaxy properties and these might indeed evolve in redshift . the local ebl ( @xmath10 ) estimated using our method is shown in fig . [ fig13 ] . the solid - black line is the ebl calculated by our _ fiducial _ model using eq . [ ebl_spectrum ] . observational data from direct measurements ( empty symbols ) and from galaxy counts ( filled symbols ) are plotted . it is usual to consider data from galaxy counts as lower limits . we find a very low background from uv to mid - ir , along the lower limits from galaxy counts . in the uv our model is lower than the @xcite data , but we consider these data suspect , due to very poor statistics on their number counts at the faintest magnitudes and the fact that they are systematically higher than the uv data from galex ( @xcite ) , an experiment with higher sensitivity and better statistics . in the mid - ir region between 7 - 15 @xmath1 m our results are a factor @xmath22 higher than other models . a lower background than frv08 is estimated from 15 - 50 @xmath1 m by a factor as large as @xmath23 . our results are still compatible with the limits from galaxy counts . on the contrary , we predict about the same far - ir light than frv08 and a factor @xmath24 larger than gspd10 , higher than the galaxy counts and in very good agreement with most of the direct measurements . the high flux we predicted in the far - ir ( in comparison gspd10 ) is a characteristic of the swire galaxy seds we use , given by the grasil code which is used to calculate the far - ir emission , and the relation between the near - ir and the far - ir in the templates . in the same figure , we also plot upper limits using solid - colour lines from @xmath0-ray attenuation studies . the cyan and yellow solid lines by @xcite were computed for the so - called _ realistic _ and _ extreme _ cases , where the authors considered different upper limits for the spectral slopes of vhe emission from blazars of @xmath25 ( @xcite ; @xcite ) and @xmath26 , respectively . our calculation is compatible with the upper limits from the extreme case , but marginally disagrees with the realistic case for the largest wavelengths . we will discuss these issues further in sec . [ attenuation ] . . for example , according to the _ fiducial _ model @xmath27% of the local ebl at 2.2 @xmath1 m comes from @xmath28 , but only @xmath29% of the local ebl at 240 @xmath1 m . ] [ fig13 ] also shows the uncertainties in our modeling due to the uncertainties on the schechter parameters of the lf given by c10 , the errors in the photometric catalogue , as explained at the beginning of this section , the uncertainties on the @xmath30 cut applied , and uncertainties due to the extrapolations for the galaxy - sed types for @xmath31 . all the possibilities are calculated and the extreme cases are plotted . the uncertainties from the uv up to the mid - ir are dominated by the errors in the photometry and the cuts . the directions from both effects are different : the uncertainties from the photometry are below the _ fiducial _ model , the uncertainties from the cuts are above it . in the far - ir the uncertainties in the extrapolations to @xmath31 dominate with a factor @xmath23 . these effects will be thoroughly discussed in sec . [ sec6.1 ] . the evolution of the ebl is important to account for the history of the galaxy emission and the processes involved , as well as to properly calculate the attenuation for vhe @xmath0-rays for the high - redshift universe . we show , in fig . [ fig14 ] , the co - moving intensity level of the ebl for different redshifts , the contribution to the ebl at those redshifts from the four main sed groups to our _ fiducial _ extrapolation , and the predictions for the ebl by other models . in table [ tab4 ] we quantify this evolution , where the bolometric intensity is defined according to eq . [ bolometric ] , @xmath32 we should note that the starburst population contributes 54% to the co - moving bolometric ebl at @xmath33 , but only 30% for the local universe . we note as well that the far - ir peak in the sed is higher relative to the near - ir peak at these redshifts ; this is due to the fact that a large fraction of the energy radiated from starburst systems is at far - ir wavelengths . we also note that the total bolometric intensity peaks at @xmath34 , because the far - ir peaks at higher energetic wavelengths there as seen in fig . [ fig14 ] . m obtained from different phenomenological models , normalized to @xmath35 , compared with the spitzer / mips data by @xcite . for example , according to our _ fiducial _ extrapolation ( see sec . [ sec4.1 ] ) , about 75% of the local ebl at 24 @xmath1 m was already in place at @xmath361.5 . uncertainties in the modeling are shown with a shadow region ( see fig . [ fig13 ] ) . the curve from franceschini et al . ( 2008 ) has been calculated by us from their published ebl densities . ] another important observable is the buildup of the local intensities for different wavelengths . this is the fraction of the local ebl at a given wavelength that was already in place at a given redshift . this is shown in fig . [ fig15 ] for several wavelengths . as an example , we see that @xmath370% of the local ebl at @xmath37 m and 2.2 @xmath1 m comes from @xmath28 , 50% of the ebl below @xmath3180 @xmath1 m was already in place at @xmath38 , but it is only @xmath29% at 240 @xmath1 m . it is significant that the ebl at shorter wavelengths mostly come from sources at much lower redshifts than the larger ones ( see @xcite ) . [ fig16 ] shows a comparison between the ebl buildup for our model , frv08 , gspd10 , and the observational work by @xcite based on data from mips at 24 @xmath1 m up to @xmath36 1.5 in the cosmos field . the main contribution to the ebl at 24 @xmath1 m comes from star - forming and starburst - type galaxies . this region of the seds is highly dependent on the non - smooth pah features . we observe a general agreement , but reaching a factor 40% difference at @xmath39 for the _ fiducial _ extrapolation . the uncertainties here are large ( see sec . [ sec6.1 ] ) . the ebl has important implications for the interpretation of data taken using recent vhe experiments ( the fermi satellite , @xcite ; and iacts , such as magic , veritas and hess ; @xcite ; @xcite ; @xcite , respectively ) , due to the photon - photon pair production between @xmath0-ray photons traveling across cosmological distances and ebl photons ( see @xcite ; @xcite ) . blazars are an important source of extragalactic @xmath0-ray emission and have become a relevant tool for indirectly measuring the ebl . these objects are believed to be an extreme category of agns . their emission , which occurs at all wavelengths of the electromagnetic spectrum , comes from supermassive black holes ( with masses @xmath40 10@xmath41 m@xmath42 ) swallowing matter accreted from their surroundings . in general , agns are characterized by a beamed emission perpendicular to the accretion disc known as jets , which are pointing toward us in the case of blazars . the current theoretical models for the emission by this class of objects are of two kinds : leptonic or hadronic . both models predict a spectrum with two peaks , the first one localized from radio to @xmath43-rays due to synchrotron radiation from relativistic electrons ( leptonic model ) , or protons ( hadronic model ) . however , the second peak has a different nature . while in the leptonic model it is due to inverse compton ( ic ) scattering of the same population of electrons that produce the synchrotron peak ( @xcite ) , in the hadronic model , nuclear photo - disintegration is advocated to explain the second peak ( @xcite ) . both models face serious difficulties in explaining intrinsic ( ebl - corrected ) vhe power law indices harder than 1.5 , and fail to explain slopes harder than 2/3 . the intrinsic spectrum is the spectrum that we would observe if there were no effect from the ebl . -ray photons for sources at different redshifts ( from bottom to top @xmath35 , 0.3 , 0.6 and 1 ) , due to the extragalactic background light computed for our model in solid - black line , for franceschini et al . ( 2008 ) in dashed - magenta line , and for @xcite in orange - dot - dashed line . _ lower panel _ : flux attenuation versus observed energy of @xmath0-ray photons for fictitious sources at different redshifts ( from right to left @xmath35 , 0.3 , 0.6 and 1 ) . we have calculated attenuation for the franceschini et al . ( 2008 ) and @xcite models using the ebl data provided by the authors . the ebl uncertainties in fig . [ fig13 ] are propagated to the optical depth and flux attenuation . they are shown here with a shadow region.,height=377 ] the egret satellite observed agns in the local universe ( hence not very attenuated ) , claiming that all of them have spectral indices @xmath44 with @xmath45 in the high energy ( 30 mev-30 gev ) regime ( @xcite ) . this result has been confirmed by the fermi collaboration ( within uncertainties ) , which has published a catalogue of agns detected by the fermi large area telescope ( lat ) all - sky survey during its first year in operation ( @xcite ) . from this observational fact , and the theoretical issues above , it is usually conservatively consider that no agn could have an intrinsic vhe spectrum fitted by a power law with an index harder than 1.5 . some authors such as @xcite , @xcite , @xcite and @xcite provide some mechanisms within standard physics to reach slopes harder than 1.5 , but never harder than @xmath46 . the ebl may be constrained using vhe observations of extragalactic sources if their intrinsic emitted spectra are known . as mentioned in section [ intro ] , @xmath0-ray photons coming from cosmological distances are attenuated by photon - photon pair production by ebl photons . the cross section of this reaction depends of the product shown in the left side of eq . [ cross ] , @xmath47 @xmath48 where , in the rest - frame at redshift @xmath14 , @xmath49 is the energy of the @xmath0 photon , @xmath50 is the energy of the ebl photon , and @xmath51 is the angle of the interaction , which defines an energy threshold @xmath52 for the ebl - photon energy given in eq . [ threshold ] with @xmath53 the electron mass . the cross section peaks at about twice @xmath52 , which produces a peak in the interaction at @xmath54 [ @xmath1 m ] = 1.24@xmath49 [ tev ] . from this property , a @xmath0-ray with energy 1 tev interacts mainly with a photon of the ebl with wavelength @xmath3 1 @xmath1 m . the details may be found for example in @xcite . for a given observed spectrum of a source at redshift @xmath14 we can find the intrinsic spectrum by assuming a particular ebl model and multiplying by the attenuation factor to _ de - absorb _ the spectrum using eq . [ correct ] , @xmath55}\ ] ] where the subscript @xmath56 means observed , @xmath57 is intrinsic , and @xmath58 is the optical depth dependent on the observed energy @xmath49 of the @xmath0 photon for a given ebl photon density and redshift , @xmath59 @xmath60 with @xmath61 given by eq . [ peebles ] , @xmath62 , @xmath63 the photon - photon pair production cross section , @xmath64 is given by eq . [ beta ] and @xmath65 is the proper number density per unit energy of ebl photons . we show in fig . [ fig17 ] the optical depth and attenuation for sources at @xmath35 , 0.3 , 0.6 and 1 . since the ebl produces an attenuation of the vhe spectra , a mere detection of vhe photons ( using some constraint on the intrinsic blazar power spectrum ) places an upper limit on the ebl density . some upper limits have been derived by different authors , fitting ebl models to the density level where the condition @xmath66 is satisfied , building _ ad - hoc _ ebl models . we plotted those limits in fig . [ fig13 ] ( @xcite ; @xcite ; @xcite ) . each of those upper limits comes from the study of different blazars with a different measured energy spectrum . due to the peak of the interaction previously mentioned , each of the studies constrains different ranges on the ebl . @xcite used the vhe spectrum of the blazar 1es 1101 - 232 at @xmath67 observed from 0.2 - 3 tev , scaling the model by @xcite multiplying the total ebl intensity by a constant to satisfy the @xmath66 condition . @xcite used the spectrum of 3c 279 at @xmath68 observed from 0.08 - 0.5 tev , scaling a slightly modified model by kneiske et al . @xcite used a compilation of blazars at different redshifts and observed at different energies , and splines from a grid as ebl densities . they make two different assumptions about the maximum @xmath69 leading to two different upper limits ( called by the authors _ realistic _ and _ extreme _ ) . we see in fig . [ fig13 ] that the _ fiducial _ ebl model ( hereafter all the results in this section are discussed for this , unless otherwise stated ) is below the upper limits at all wavelengths , except at the largest wavelengths , where slightly exceeds the limits from the _ realistic _ case by @xcite . this fact is discussed in sec . [ mrk501 ] and it is explained why we do not consider this a major problem . another limit not plotted comes from the blazar 1es 0229 + 200 at @xmath70 ( @xcite ) . its study set a lower limit in the slope of the local ebl spectrum between 2 - 10 @xmath1 m , @xmath15 @xmath40 1.10 @xmath71 0.25 , to satisfy the limit on agn s spectra @xmath69 @xmath40 1.5 . we remark that they set the limit only on the slope , not on the intensity level . we have fitted our model in that wavelength range , to a power law @xmath72 getting @xmath73 @xmath71 0.07 . our model is thus compatible with this constraint . it is also possible to set upper limits on the unknown redshift of blazars assuming an ebl model and finding the redshift by which the ebl - corrected spectrum satisfies @xmath66 ( @xcite ; @xcite ) . we apply that method to the pg 1553 + 133 spectrum observed by magic ( @xcite ) , assume an ebl - corrected spectrum given by a power law and find an upper limit at @xmath74 in agreement with the lower limit ( @xmath75 ) found by @xcite using absorption features in the ly-@xmath15 forest of the blazar . as shown in fig . [ fig17 ] , our ebl model implies about the same attenuation as other recent models we compare to over all the energy range observed by the current generation of iacts . larger transparency than the observationally - based model by frv08 is found ( roughly a factor @xmath76 in flux , but still within the uncertainties ) for @xmath0-ray photons with energies between @xmath77 tev for @xmath78 , but a factor @xmath79 in flux less transparent than the gspd10 theoretical approach around @xmath80 tev . for the large - redshift case , our model predicts about the same attenuation as frv08 , but a factor @xmath23 more transparency than gspd10 for sub - tev energies . note that a small difference in the optical depth has large effects on the spectra due to the exponential in eq . [ correct ] , a factor 1.5 in optical depth leads to a factor @xmath81 in attenuation . we now proceed to test whether the observed spectra of the three most constraining agns known in the vhe range due to their hard spectrum , or to their large redshift , satisfy the condition that the intrinsic spectrum corrected by the attenuation derived with our model has @xmath69 @xmath40 1.5 . we consider the blazars : mrk 501 at @xmath82 detected by the hegra system of cherenkov telescopes in 1997 ( @xcite , with a reanalysis by @xcite ) , fsrq 3c 279 at @xmath68 observed by magic ( @xcite ) , and the blazar 3c 66a observed by veritas at @xmath83 ( @xcite ) , all of them seen in a flaring state . all these blazars are plotted in fig . [ fig18 ] showing in the legends that the condition @xmath84 is satisfied . the highest energy bins in this measurement , where it is observed a significant deviation from a power law ( see fig . [ fig18 ] upper - left panel ) , are affected by the far - ir ebl at @xmath85 m . this is the region of the ebl spectrum where it was found a disagreement with the _ realistic _ ( but not _ extreme _ ) upper limits of @xcite . the problem comes from the very low statistics and high systematic uncertainties at such high energies ( @xcite ) . a later reanalysis of the same observation done in @xcite accounts for larger systematic uncertainties as shown in fig . [ fig18 ] upper right panel . this exponential behaviour for the highest energy bin was already observed from the first ebl models ( @xcite ; @xcite ; kneiske et al . 2002 ) , whose ebl levels were higher than the more recent ones . this fact was discussed thoroughly in @xcite , and even some exotic explanations such as lorentz invariance violation ( @xcite ) were proposed . more recent ebl models with a more transparent universe ( such as our model , frv08 and gspd10 ) relax such predictions . the solutions to exponential spectra and photon pileup could involve widespread problems with the photon statistics and systematic uncertainties in the observations ( as suggests the results from the later reanalysis ) , or new mechanisms extending the normal ssc model , using external regions close to the @xmath0-ray source with target photons . the ebl uncertainties in the far - ir leading to the attenuation uncertainties at these high energies as shown in fig . [ fig18 ] , might contribute to the solution as well . another observed flare with better statistics with the current generation of iacts up to such high energies as @xmath86 tev would be very helpful in constraining these possibilities . [ fig18 ] shows in the lower - left panel the ebl - corrected vhe spectrum for this source . an external photon field providing target photons for ic ( such as that provided by a broad - line emission region ) might be necessary to explain the flat behavior at the largest energy bins , as discussed in @xcite . instrumental systematic uncertainties might explain this behavior as well . we note here that our model is already matching the lower limits from galaxy counts at the wavelengths where @xmath0-ray attenuation with the observed energies occurs , and a much lower ebl density than the one calculated in this work does not seem realistic . the attenuation uncertainties from the ebl modeling are too low at this redshift and these energies to explain that spectral behaviour . [ fig18 ] shows in the lower - right panel the ebl - corrected vhe spectrum for this source , whose ebl - corrected slope is well within the 1@xmath16 limit of the @xmath45 according to the calculated intrinsic index . it is important to note that the redshift considered for this object is calculated using just one emission line and is thus not very secure ( see discussion in @xcite ) , and its attenuation might be indeed overestimated if the redshift is lower than assumed . it is concluded from the study of these extreme blazars that our ebl is generally compatible with the hardness of the ebl - corrected slopes expected from theoretical arguments . however , it is clear that a simple ssc model can not explain any flatness at the highest energies of the ebl - corrected spectra of either mrk 501 or 3c 279 , which suggests that some extension to the model may be necessary such as an external photon region , a better understanding of the iact systematic uncertainties , or even a revision of the propagation mechanisms mainly through the intergalactic medium ( see @xcite ) . as shown in fig . [ fig17 ] the uncertainties in the attenuation are dependent on the observed @xmath0-ray energy as well as the redshift : the higher the energy or the redshift , the higher the uncertainties in the attenuation . the attenuation uncertainties shown were calculated from the uncertainties in our ebl modeling , which were shown in fig . [ fig13 ] , explained in sec . [ results ] and will be thoroughly discussed in sec . [ sec6.1 ] . for sub - tev energies up to around 1 tev the uncertainties in the flux attenuation are never higher than a factor @xmath79 and generally lower . the uncertainties in the ebl - corrected spectra in this case are dominated by other effects ( see indices in fig . [ fig18 ] ) . for energies larger than 10 tev the uncertainties are around a factor of several . the uncertainties in the ebl - corrected spectra up to such high energies due to the ebl modeling are considerable . these high uncertainties are derived from the ebl in the far - ir region due to the very fast increment of the ebl photon density ( @xmath65 , see eq . [ attenu ] ) with longer wavelengths . observations of sources at low redshift but energies larger than @xmath80 tev will set constraints on these uncertainties . as explained in sec . [ method ] , we adopt a lower limit to the photometric errors higher than those in the aegis catalogue . different lower limits are set from 1 - 10% of the photometric measurements . that is , if the error in any band is lower than our limit then we set it to the limit . the results are sensitive to the limit choice in @xmath87 and @xmath88 ( where the errors in the catalogue are the lowest ) , but not for the other bands . the galaxy sed - type fractions change for lower limits 1 - 6% , but there is little change if the level is set higher than 6% . the change is mostly for the quiescent and star - forming galaxy fractions . if we use the errors in the catalogue without any change , we find 10% more quiescent galaxies at @xmath89 than for a lower limit of 6% , which decreases to @xmath90 more quiescent galaxies at @xmath91 than for a lower limit of 6% , as shown in fig . the change is mostly in ell2-type galaxies , according to the swire classification . we have investigated those quiescent galaxies that change their best fit to star - forming galaxies upon raising the lower limit on the errors , and find that they are often fitted much better by a star - forming sed . in many cases they even have detection in mips 24 , clearly indicating ongoing star formation . on the other hand , based on the comparison of our photometric measurements to those of other catalogues we estimate that any error in the photometry lower than @xmath92 should not be considered very reliable . for those reasons we set the lower limit at 6% for model . the uncertainties due to this are below the _ fiducial _ model in fig . [ fig13 ] , and for the reasons stated an ebl in this region should not be considered very likely ( and therefore , neither is their derived attenuation in fig . [ fig17 ] ) . another source of uncertainty accounted for in fig . [ fig8 ] and fig . [ fig13 ] is the @xmath30 cut that separates good and bad sed fits . the main change occurs for agns , where for a relaxation in the cut ( from @xmath93 to 20 ) , the fraction can increase by as much as 10% at @xmath91 , and by a smaller fraction at @xmath89 . these changes affect the ebl in the following way : higher agn fractions increase the uv as well as the mid - ir , while higher quiescent fractions decrease the flux at those wavelengths . this effect affects the uncertainties above the _ fiducial _ model and an ebl intensity in this shadow region is considered more reliable than the region below the _ fiducial _ model ( the same for their derived attenuation in fig . [ fig17 ] ) . while @xmath30 cuts do not have an appreciable effect on the far - ir flux , there is a substantial change arising from the choice of extrapolation in sed - types above @xmath94 , as we find in the _ high - starburst _ assumption . [ fig13 ] also accounts for the uncertainties in the schechter parameters of the lf given by c10 but these are small . two major potential problems for our modeling might be a colour - dependent selection effect and the extrapolation of the galaxy sed - type fractions for @xmath31 . it was already shown in sec . [ sec3.2 ] that the colour - selection effect is rather small . from the fact that most of the light in the ebl comes from the knee of the lf around @xmath95 , where our sample does not suffer any color - dependent selection effect , we do not consider this to be a significant problem for our ebl calculation . our estimated galaxy sed - type fractions appear to be consistent with works by others as well . for example , our results agree with @xcite and @xcite , who find roughly no evolution for late - type ( blue ) galaxies from @xmath96 within a 10% range , and an increment of the early - type ( red ) population in the same redshift range by at least a factor 2 . we also highlight that the galaxy sed - type fractions that we calculated for the local universe smoothly link with our independently - derived results at @xmath97 . regarding the galaxy sed - type fraction extrapolations , we have considered two rather different approaches which basically lead to the same evolving luminosity densities and ebl for the optical / near - ir range where @xmath0-ray attenuation occurs , as shown in fig . [ fig11 ] and fig . [ fig14 ] , respectively . this fact is due to the shape of the stellar emission , because the contribution to the optical / near - ir peak is very similar for quiescent , star - forming and starburst galaxies for a given @xmath98 . we recall here that the normalization to our model is fixed by the rest - frame @xmath2-band lf by c10 . the only difference between our extrapolations is at the far - ir peak , where our results are considered to be less robust for this reason , as well as for the reasons stated below regarding the swire templates and the lack of photometric data . deeper observations by future galaxy surveys will help in characterizing the galaxy sed - fractions up to higher redshifts . it was also checked how the fractions change if the detection limit is relaxed from @xmath99 to @xmath100 for the bands where there are observations , but no detections . many more quiescent galaxies than in the @xmath99 case were found , even a factor larger than 2 , due to the low detection limit on the mips 24 instrument , but generally they are not good fits . in our work we have not differentiated between the spectroscopic and secure photometric redshifts . this is an approximation , and it is necessary to check that this is consistent with our results . we find that the galaxy sed - type fractions derived from both sub - catalogues are clearly compatible and the trends are the same . galaxies fitted to a starburst sed may instead be very late - type star - forming galaxies ( or viceversa ) , because both sed templates are rather similar in the regions where we have data . this may be called mis - typing and its effect is expected to be larger for faint galaxies , because the major fraction of faint galaxies are star - forming or starburst and not massive quiescent or agn galaxies . such small galaxies are probably rather metal poor and thus lacking dust . hence , their seds are probably more like star - forming galaxies rather than starbursts ( in agreement with our results in fig . [ fig9 ] ) . it is a source of uncertainty in the mid and far - ir ( underestimating or overestimating light ) and might explain the excess found in fig . [ fig11 ] and [ fig12 ] compared with the data . far - ir data would help in resolving this issue , but the number of galaxies with detection in mips 70 is rather low to make statistical estimations . herschel data will be very useful thanks to its good spatial resolution and deep photometry in the far - ir . another source of uncertainties in our model that quantitatively we have not accounted for arises from the use of local sed templates to fit galaxies at @xmath101 . this comes from the fact that the swire templates are based on observations of local galaxies , and we expect that they become worse fits when the redshift is increasing . this problem will be addressed by new data from wfc3 on hubble space telescope and the next generation of ground and space optical / near - ir telescopes such as the james webb space telescope . the lowest ebl flux in fig . [ fig13 ] is given by the case with the highest number of quiescent galaxies and lowest number of agn galaxies , which corresponds to the case of using the low errors in the catalogue and our @xmath30 cuts . the highest ebl flux occurs with fewer quiescent galaxies and the highest fraction of agn galaxies . this is the case with the 6% lower limit for the errors in the photometry and without a @xmath30 cut . using the lower limits from galaxy counts in the uv and in the optical we may rule out at @xmath102 the mixing of galaxy sed - type fractions predicting the highest fractions of quiescent galaxies and fewest agns in fig . we do not consider the vhe observations to exclude the models with higher far - ir , because the discrepancy is for wavelengths longer than 60 @xmath1 m where those limits may not be reliable for the reasons stated in sec . [ attenuation ] . further vhe observations might indeed constrain our galaxy sed - type fractions . thus to recap , the ebl uncertainties from the uv up to the mid - ir are low enough to recover the spectra of @xmath0-ray sources with energies lower than @xmath80 tev , but the ebl uncertainties have to be reduced in the far - ir ( for neglecting uncertainties due to the ebl modeling ) to correct higher energy sources . additional photometry is needed there to clearly distinguish between star - forming and starburst galaxies , therefore to reduce the mis - typing , as well as a better understanding on the far - ir region of the galaxy seds at @xmath101 . characterizing the galaxy sed - type fractions at @xmath31 will reduce these uncertainties in the far - ir as well . the local luminosity density from galaxies is observationally well constrained over all wavelengths from 0.1 - 1000 @xmath1 m . as shown in fig . [ fig10 ] , the prediction of the local luminosity densities is in very good agreement with observational results . fig . [ fig11 ] showed the evolving rest - frame uv luminosity density as well . these results agree well with the observational data by @xcite and @xcite within uncertainties , but they are a factor @xmath103 below the data for @xmath104 and 2.2 . it was also compared in fig . [ fig11 ] the evolution of the rest - frame luminosity in the @xmath2-band from our calculations to independent observational works . some disagreement was found that in the case of @xcite might be explained by the fact that they do not correct their sample for incompleteness and only consider the brightest sources , unlike the lf by c10 . therefore their results should be considered as lower limits . the direct comparison in fig . [ fig11 ] with the rest - frame luminosity density in @xmath105-band showed that our luminosity in that band is not in contradiction with other independent works . we are doing really well reproducing the data from @xcite , @xcite and @xcite . we might be indeed overestimating the light in this band 15 - 20% for @xmath28 according to the data by @xcite and a factor @xmath76 for @xmath31 according to the data by @xcite , but this latter does not significantly affect our results for the local ebl because as we already showed in fig . [ fig15 ] , most of the optical / near - ir light comes from @xmath28 . the comparison with the bolometric ir luminosity density with the observational works by frv08 and @xcite is very good , even though we are a factor @xmath79 higher around @xmath94 . a good agreement was found with the upper envelope of the data cloud on the calculations for the sfr history from @xmath106 down to the local universe using eq . [ star ] ( see fig . [ fig12 ] ) , using our _ fiducial _ extrapolation . according to @xcite at around @xmath94 the main contributor to the star formation is the obscured ir contribution , instead of the uv . we may be overpredicting some of this obscured ir light around @xmath94 due to the lack of far - ir photometric data in our galaxy catalogue that allow a clear classification between late star - forming and starburst galaxies as discussed in sec . [ sec6.1 ] . with the _ high - starburst _ extrapolation of the galaxy - sed - type fractions was checked that increasing the starburst - like population a factor @xmath107 from @xmath108 , we may get a flatter sfr density history up to @xmath21 , but even higher than the observational data . this _ high - starburst _ assumption does not change our general picture of the local ebl , but increases the far - ir peak a factor @xmath23 ( as it was considered in fig . [ fig13 ] and was shown explicitly in fig . [ fig14 ] for some other redshifts ) . the ebl calculated in this work is matching the data from galaxy counts from the uv up to the mid - ir ( see fig . [ fig13 ] ) , except the data found in @xcite . higher intensities than the data from galaxy counts were calculated in the far - ir but in agreement with direct detections . the ebl evolution shown in fig . [ fig14 ] is in good agreement with frv08 up to @xmath38 . at higher redshift our results are different in the uv and optical / near - ir . this may be due to the fact that frv08 extrapolate the galaxy evolution , while in our model this evolution is entirely based on the observed lf by c10 up to @xmath109 . see sec . [ sams ] for a comparison with the results by gspd10 . there are some works in the literature where the contribution from agn galaxies to the ebl is studied . according to recent works that focus in the mid - ir ( @xcite ; @xcite ) this contribution should not be larger than 10 - 20% . this is in agreement with our results : we find that the agn - galaxy contribution to the bolometric ebl is 6% for the _ fiducial _ extrapolation ( table [ tab4 ] ) and 13% for the case with the largest agn fraction in fig . [ fig8 ] . for the wavelength range between 1 - 20 @xmath1 m the agn contribution from our model is also between 8 - 16% . we estimate that this contribution to the co - moving bolometric ebl slightly increases with redshift . the ebl buildup was studied in fig . [ fig15 ] and [ fig16 ] . it was found that most of the local uv / optical / near - ir ebl was built up at @xmath28 , while the far - ir ebl was mostly built up at @xmath31 . this result for the far - ir light agrees well with the observational work by @xcite , but disagrees with @xcite . in any case , our uncertainties in the far - ir are very large . differences up to 40% were found in the buildup of the local ebl at 24 @xmath1 m . these differences are due to the fact that a very small change in the mid - ir region leads to a very strong difference in this buildup plot , and to the mid - ir peak that we get at larger redshifts ( see fig . [ fig14 ] ) due to the shape of the galaxy seds . we point out here that the ebl buildup is on how the light is being built up , and not about the absolute intensity value . we already discussed our results on @xmath0-ray attenuation in sec . [ attenuation ] in this section we compare our ebl estimation against the ebl model described in sgpd10 and gspd10 , which is based on sams of galaxy formation . the comparison for @xmath0-ray attenuation has been thoroughly discussed in sec . [ attenuation ] . we notice that slightly different cosmological parameters were used for our model and that by sgpd10 . the latter uses the latest values from wmap5 , which slightly affect all the results in the local universe as well as their evolution . we already saw in fig . [ fig11 ] the comparison between our observational luminosity densities and the theoretical prediction by sgpd10 for the co - moving luminosity density versus redshift in the uv , in the near - ir ( @xmath2-band ) , in the optical ( @xmath105-band ) and for the bolometric ir luminosity . we note that our @xmath2-band luminosity density evolution is given exclusively by the c10 lf , because at that band our choice of galaxy sed fractions does not affect our results . this quantity is above the prediction by sams by a factor around 20% from @xmath21 down to the present universe . the uv from this sam is above our results for all redshift , except at @xmath110 . at @xmath111 is a factor 4 higher . for the @xmath105-band luminosity density the agreement is excellent from @xmath94 down to the local universe . for @xmath31 sams predict a factor of several more light than our observationally - based approach . we may see the consequences of this for the ebl evolution in fig . [ fig14 ] for high redshifts where the excess of light has not been diluted by the expansion of the universe . for the bolometric ir luminosity sams seem to systematically predict at least a factor @xmath79 less light than our calculations and the one by frv08 . this difference is maximized around @xmath110 up to a factor @xmath112 . [ fig12 ] showed a comparison between our sfr density estimation and that predicted by sgpd10 as calculated by using the eq . [ star ] . from @xmath113 our sfr densities have different a behaviour : for our observational model increases up to @xmath38 and for sgpd10 keeps constant down to @xmath114 . for lower redshifts both models decline down to the local universe . in general a very good agreement between the local ebl from the uv up to the mid - ir predicted by our method and the sam of sgpd10-gspd10 was seen in fig . [ fig13 ] . a factor @xmath23 higher intensity is found in the local uv from sams , and around the same factor lower intensity around 15 @xmath1 m . for the far - ir peak , the difference comes from the different templates used for the dust component in the far - ir , which is given by the grasil code in the case of the swire templates ( which we use ) , and by a interpolation between the observed 70 and 160 @xmath1 m by mips in the case of the templates used by these sams ( @xcite ) . the agreement on the evolution is very good as well as seen in fig . [ fig14 ] , even though at high redshift ( @xmath115 ) , gspd10 predicts a factor of several more light in the uv . this is due to galaxies within the faint end of the theoretical lf at @xmath115 . we recall that our observational model seems to already be overproducing light in the @xmath105-band for @xmath116 according to data in @xcite ( see fig . [ fig11 ] ) . the evolutionary path that we have in mind to interpret the evolution of the galaxy sed - type fractions is the following : agns are mostly formed by mergers between galaxies during hierarchical growth of dark matter halos ( @xcite ) . when the merging galaxies are gas rich ( also known as wet mergers ) , inflows of gas are produced leading to starburst galaxies and to the mass growth of the central black hole . the central black hole activity begins to expel the gas . eventually , the gas is exhausted , switching off the agn . the galaxy continues forming stars as a star - forming galaxy until the gas is fully depleted , then becomes a quiescent galaxy ( @xcite ; @xcite ) . it is now a well known observational fact that galaxies are bimodal in some properties such as colours ( @xcite ; @xcite ) . they group in two different regions in colour - magnitude diagrams defining the red sequence and the blue cloud . galaxies forming stars are in the blue cloud . some galaxies have their star formation quenched when they become satellite galaxies in a larger halo , they cease to accrete gas , and they join the red sequence . central galaxies form in the blue cloud , but they join the red sequence when they form a supermassive black hole and/or their halo mass exceeds approximately 10@xmath117m@xmath42 and/or they become satellite galaxies in a cluster . the most massive red galaxies can not have simply be quenched central blue galaxies , since the latter are not massive enough ; thus they must have been created by mergers without much star formation ( also known as dry mergers ) . this effect is shown in fig . [ fig4 ] , taking into account that the @xmath2-band absolute magnitude @xmath98 is a good tracer of the galaxy stellar mass , as shown in @xcite . massive galaxies today ( very bright @xmath98 ) form their stars first , which is known as _ downsizing _ ( @xcite ; @xcite ) . this initially seemed at odds with the hierarchical nature of the @xmath4cdm paradigm , in which small halos form first and agglomerate into larger ones . but the idea that star formation is efficient only in dark matter halos with a narrow range of masses naturally explains how the phenomenon of downsizing arises : halos that are massive today passed through the star - forming mass band between 10@xmath118 - 10@xmath117m@xmath119 earlier and thus formed their stars earlier than halos that are less massive today ( @xcite ; @xcite ) . a careful examination of fig . [ fig4 ] reveals some interesting trends . we find that all the oldest galaxies ( @xmath120 gyr old , ell13 ) are in the red sequence . however , the younger quiescent galaxies ( @xmath81 gyr old , ell5 ) can be found in the red sequence as well as in the green valley ( the region between the red sequence and the blue cloud ) . for the youngest quiescent galaxies ( @xmath79 gyr old , ell2 ) we find that for @xmath121 they populate the green valley , while for @xmath122 they belong to the blue cloud . all the early - type star - forming galaxies ( s0 , sa ) are in the red sequence . later - type star - forming galaxies such as sb and sc start to populate the green valley as well as the red sequence . most of the very late - type star - forming galaxies ( sd , sdm , spi4 ) populate the blue cloud . starburst galaxies are mainly in the green valley , but some of them are in the bluer region of the red sequence and in the redder region of the blue cloud . the same happens to agns , but they tend to be in the blue cloud more than in the red sequence . we note that the increasing rate of quiescent galaxies as @xmath14 declines is roughly the same as the decreasing rate of starburst - type galaxies from @xmath123 . one possible explanation would be the direct transformation of starbursts ( either merger or huge - cold - gas reservoir triggered ) directly to quiescent galaxies , without an intervening stage of significant star formation . another explanation is that the characteristic time in which starburst - like galaxies consume their cold gas is the same that in which star - forming galaxies consume their lower cold - gas reservoir . thus the specific sfrs of these populations are different , but the rate at which starbursts enter the star - forming sequence is the same as the rate at which star - forming galaxies become quiescent . from @xmath124 the fraction of starbursts is very low , so the constant increase of the red sequence is modeled as due to agns preventing gas from cooling and forming stars . a novel , robust and powerful method based on observations to derive the evolving spectrum of the extragalactic background light ( ebl ) between 0.1 - 1000 @xmath1 m was presented . this model is based on the observed rest - frame @xmath2-band galaxy luminosity function ( lf ) over redshift by @xcite ( c10 ) , combined with an estimation of galaxy sed - type fractions based on a multiwavelength sample of @xmath125 galaxies from aegis . this model has the following main advantages over other existing ebl models : transparent methodology , reproducibility , and utilizing direct galaxy data . the best available data sets are used ( c10 s lf and the aegis galaxy catalogue ) observed over a wide redshift range . the galaxy evolution is directly observed in the rest - frame @xmath2-band up to @xmath109 . observed galaxies up to @xmath38 from the uv up to 24 @xmath1 m with spectral energy distributions ( seds ) of 25 different types ( from quiescent to rapidly star - forming galaxies , and including agn galaxies ) are taken into account in the same observational framework . a study of the uncertainties to the model directly from the data ( such as uncertainties in the schechter parameters of the c10 lf and the errors in the photometric catalogue ) was done , and their propagated uncertainties to the @xmath0-ray attenuation were studied . a brief comparison with results from other recent ebl models is made here : stecker et al . ( 2006 ) estimate a local ebl in the uv a factor of several higher than us , and in contradiction with recent @xmath0-ray observations ( @xcite ) . a comparison with the frv08 s results was throughly presented through sec . [ results ] . in general , our results are in good agreement , despite the fact that our modelings are different . finke et al . ( 2010 ) have five different models based on different parametrizations of the sfr density of the universe and imfs . the local ebl from the uv to the near - ir are similar to ours for their models _ e_. @xcite claim to model a strict lower limit for the ebl . however our results for the local ebl in the uv are lower than the calculation by @xcite , but are in agreement with the robust lower limits from galaxy counts in the uv by @xcite and @xcite . in the near - ir , the model by @xcite is not compatible with the lower limits by @xcite . our observationally - based approach was also throughly compared with the results from the semi - analytic model ( sam ) of galaxy formation by sgpd10 and gspd10 in sec . our ebl results are in general in good agreement at least for @xmath126 , even though this sam predicts more light ( by up to a factor of several ) than our observational approach in the uv , and a factor @xmath127 less light in the far - ir . our methodology provides a tool for calculating the ebl more accurately at the longest wavelengths when a better understanding on the far - ir galaxy seds , new photometry , and deeper lfs at those wavelengths are available from the herschel space observatory . two extrapolations of the galaxy sed - type fractions to @xmath31 were considered , showing that these assumptions only affect the far - ir . it was calculated that the population with sed features of quiescent local galaxies increases a factor @xmath79 since @xmath94 . the star - forming population remains roughly constant , while the starburst - like population decreases very quickly from around @xmath128 at @xmath94 . the agn - like population decreases slower than the starburst - like population from almost 20% at @xmath94 to just around 2% at @xmath97 . data from the future james webb space telescope will help to determine the galaxy sed - type fractions at @xmath31 . a low intensity local ebl ( @xmath10 ) was found , matching the lower limits from galaxy counts up to @xmath129 @xmath1 m . for longer wavelengths , our model predicts higher intensities than the data from galaxy counts , in agreement with direct measurements . our results are also compatible with all the upper limits from @xmath0-ray astronomy according to the standard framework for the propagation of vhe photons through the universe , even though to account for the highest energies detected by @xcite for mrk 501 we have to assume a @xmath130 , appeal to statistical and systematic uncertainties on this vhe spectrum , or attenuation uncertainties due to uncertainties in the ebl for such high energies as discussed in sec.[attenuation ] . the ebl uncertainties in far - ir leading to attenuation uncertainties of a factor of several for energies larger than @xmath80 tev needs to be addressed by the current and next generation of ir telescopes providing new photometric data and a better understanding of the galaxy ir emission . @xmath0-ray astronomy may constrain these uncertainties from a better understanding of the emission mechanisms at those high energies ( helped with simultaneous multiwavelength observations ) and of the instrumental systematic uncertainties . observations aimed to measure photons with energies higher than @xmath80 tev at @xmath131 are encouraged . it is worth mentioning that high energy ( 30 mev-30 gev ) @xmath0-rays are detected by fermi for @xmath132 from agns ( @xcite ) and for @xmath133 ( @xcite ) from grbs . the reasons for these high redshift detections include a larger @xmath0-ray flux at lower energies and a lower density of ebl target photons that can interact with these @xmath0-rays . understanding the evolution of the ebl at uv wavelengths is essential to interpreting observations of these high - redshift sources . new observations of agns as well as the first grb detection in the vhe range would help to make new and stronger constraints on the ebl ; see @xcite . the universe , according to our observationally - based model , is more transparent than the estimation from frv08 ( a factor @xmath76 in flux ) for vhe photons coming from low - redshift sources ( @xmath134 ) for energies between @xmath135 tev , but still the uncertainties here from the ebl modeling are large ( a factor of several ) . the same attenuation than frv08 is estimated for other energies . for vhe photons coming from larger - redshift sources ( @xmath94 ) , roughly the same attenuation as frv08 is estimated . here the attenuation uncertainties ( for energies available to @xmath0-ray telescopes ) due to the uncertainties on the ebl modeling are low in comparison with other effects . at these redshifts the uncertainties on the ebl - corrected spectra are dominated by instrumental systematic uncertainties . we may conclude that it is not expected to observe any such high redshift ( @xmath94 ) multi - tev @xmath0-ray photons from blazars with the current or even next telescope generation such as the cherenkov telescope array ( cta , @xcite ) or the advanced gamma - ray imaging system ( agis , @xcite ) , but we indeed expect a promising future for sub - tev detections at these high redshifts . we are grateful to ranga - ram chary , darren croton , amy furniss , valentino gonzlez , patrik jonsson , nepomuk otte , miguel ngel snchez - conde , david williams and li yan - rong for fruitful discussions , and to the anonymous referee for useful comments . we also thank valentino gonzlez for his help running fast . a.d . warmly thanks for their hospitality the staff at scipp and the astronomy department - ucsc , where most of this work was done . a.d.s work has been supported by the spanish ministerio de educacin y ciencia and the european regional development fund ( feder ) under projects fis2008 - 04189 and cpan - ingenio ( csd2007 - 00042 ) , and by the junta de andaluca ( p07-fqm-02894 ) . a.d . also acknowledges the spanish mec for a fpi grant . j.p.s and r.g.s research is supported by fermi theory grants nnx08aw37 g , nnx09at98 g and nnx10ap54 g . d.r . acknowledges the support of the national science foundation through grants ast-0507483 and ast-0808133 . thanks the support of the spanish micinn s consolider - ingenio 2010 programme under grant multidark csd2009 - 00064 . r.g . was also supported by a research fellowship from the sissa astrophysics sector . this study makes use of data from aegis , a multiwavelength sky survey conducted with the chandra , galex , hubble , keck , cfht , mmt , subaru , palomar , spitzer , vla , and other telescopes and supported in part by the nsf , nasa , and the stfc .

we present the results of the first x - ray study of a sample of 16 young radio sources classified as compact symmetric objects ( csos ) . we observed six of them for the first time in x - rays using _ chandra _ , re - observed four with the previous _ xmm - newton _ or _ beppo - sax _ data , and included six other with the archival data . all the sources are nearby , with the age of their radio structures ( years ) derived from the hotspots advance velocity . our results show heterogeneous nature of the csos indicating a complex environment associated with young radio sources . the sample covers a range in x - ray luminosity ,ergs , and intrinsic absorbing column density of . in particular , we detected extended x - ray emission in 1718 ; a hard photon index of in 2021 and 1511 consistent with either a compton thick absorber or non - thermal emission from compact radio lobes , and in 0710 an ionized iron emission line atkev and ew.4kev , and a decrease by an order of magnitude in the 2 - 10 kev flux since the 2008 _ xmm - newton _ observation in 1607 . we conclude that our pilot study of csos provides a variety of exceptional diagnostics and highlights the importance of deep x - ray observations of large samples of young sources . this is necessary in order to constrain theoretical models for the earliest stage of radio source evolution and study the interactions of young radio sources with the interstellar environment of their host galaxies . [ firstpage ]

1603.00947

theory and simulations predict that radio sources are the most luminous when they start growing within the very central regions of their host galaxies @xcite . during this initial expansion a radio source is expected to interact strongly with the interstellar medium ( ism ) , inducing shocks , accelerating particles , heating the ism and accelerating gas forming an outflow @xcite . empirical evidence for such interactions comes from observations of galactic scale x - ray structures in nearby radio galaxies ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , as well as from correlations between velocities of optical line - emitting gas and the morphology of radio outflows in parsec scale radio sources @xcite . such interactions , leading to energy exchange between a radio source and the ism , are thought to contribute to feedback that governs the evolution of galaxies over cosmological time scales ( e.g. * * ; * ? ? ? * ) . at present , though , we have only been able to study such interactions in relatively old , @xmath13yrs , large radio sources , while studies of young compact systems have been limited . although the youngest sources would only impact the very central regions of their galaxy , they can strongly influence the sites directly responsible for the agn fuel supply , e.g. , by heating and accelerating material away from the central regions ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? thus , these youngest sources could be critical to our understanding of the feedback processes impacting galaxy evolution . rlcrrrll # & source & @xmath14 & size & velocity & age@xmath15 & refs.@xmath16 & refs.@xmath17 + & name & & [ pc ] & [ @xmath18 & [ yr ] & ( age ) & ( x - ray observations ) + + + 1 . & 1718@xmath8649@xmath19 & 0.014 & 2.0 & 0.07 & 91 & ( 1 ) & this work / c + 2 . & 1843@xmath10356 & 0.763 & 22.3 & 0.39 & 180 & ( 2 ) & this work / c + 3 . & 2021@xmath10614 & 0.227 & 16.1 & 0.14 & 368 & ( 2 ) & this work / c + 4 . & 0035@xmath10227 & 0.096 & 21.8 & 0.15 & 450 & ( 1 ) & this work / c + 5 . & 0116@xmath10319@xmath20 & 0.059 & 70.1 & 0.45 & 501 & ( 1 ) & this work / c + 6 . & 0710@xmath10439 & 0.518 & 87.7 & 0.30 & 932 & ( 2 ) & this work / c ; ( 6/n ) + 7 . & 1946@xmath10708 & 0.101 & 39.4 & 0.10 & 1261 & ( 1 ) & this work / c ; ( 7/b ) + 8 . & 1943@xmath10546 & 0.263 & 107.1 & 0.26 & 1308 & ( 1 ) & this work / c + 9 . & 1934@xmath8638 & 0.183 & 85.1 & 0.17 & 1603 & ( 1 ) & this work / c ; ( 7/b ) + 10 . & 1607@xmath1026@xmath21 & 0.473 & 240 & 0.34 & 2200 & ( 3 ) & this work / c ; ( 8/n ) + 11 . & 1511@xmath100518 & 0.084 & 7.3 & 0.15 & 300 & ( 4 ) & ( 9/c ) + 12 . & 1245@xmath10676 & 0.107 & 9.6 & 0.16 & 188 & ( 1 ) & ( 10/n ) + 13 . & oq@xmath10208@xmath22 & 0.077 & 7.0 & 0.10 & 219 & ( 5 ) & ( 11/n ) + 14 . & 0108@xmath10388 & 0.669 & 22.7 & 0.18 & 404 & ( 1 ) & ( 6/n ) + 15 . & 1031@xmath10567 & 0.460 & 109.0 & 0.19 & 1836 & ( 2 ) & ( 6/n ) + 16 . & 2352@xmath10495 & 0.238 & 117.3 & 0.12 & 3003 & ( 2 ) & ( 6/n ) + + @xmath16 ( 1 ) giroletti & polatidis ( 2009 ) ; ( 2 ) polatidis & conway ( 2003 ) ; ( 3 ) nagai et al . ( 2006 ) ; ( 4 ) an et al . ( 2012 ) ; ( 5 ) luo et al . + @xmath17 c : _ chandra _ ; n : _ xmm - newton _ ; b : _ bepposax _ ; ( 6 ) vink et al . ( 2006 ) ; ( 7 ) risaliti et al . ( 2003 ) ; ( 8) tengstrand et al . ( 2009 ) ; ( 9 ) kuraszkiewicz et al . ( 2009 ) ; ( 10 ) watson et al . ( 2009 ) ; ( 11 ) guainazzi et al . + @xmath19 ngc 6328 ; @xmath20 4c@xmath1031.04 ; @xmath21 ctd 093 ; @xmath22 mkn 668 . + @xmath23 many different age estimates are given in the literature for each source , as the measurements depend on radio frequency , radio components and the number of epochs . thus , the systematic errors could be 20%-40% and therefore we only use the age in general discussion and do not perform any statistical correlations with age . the existing velocity measurements of radio hot spots and spectral aging studies indicate that compact symmetric objects ( csos ) , a subclass of gigahertz - peaked spectrum ( gps ) sources , are among the _ youngest _ extragalactic radio sources , with ages between @xmath24100 and 3000yrs @xcite . interestingly , there seems to be an excess of csos younger than @xmath24500yrs @xcite . this excess together with a large number of compact radio sources ( @xmath25 kpc ) in comparison to large size radio galaxies @xcite may suggest that the radio jet activity is intermittent @xcite or that the compact sources are short - lived @xcite . in either case the youngest csos expand within the central ( @xmath26pc ) regions of the galaxies and contribute to the feedback . the young age of the cso sources has been challenged by a _ frustrated jet _ scenario in which the presence of a dense medium forces the jets to decelerate rapidly and prevents them from propagating beyond the central parsec scale regions @xcite . this would introduce a systematic bias to the kinematic age measurements . however , until now there has been no evidence for a dense enough medium to halt the jets of gps sources , and this lack of evidence gives further support to the _ youth _ scenario ( e.g. , @xcite ; see however , @xcite ) . at the same time , recent broad - band low frequency radio observations reveal inhomogeneous absorber surrounding compact radio lobes in a few csos ( e.g. * ? ? ? * ; * ? ? ? * ) implying that complex interactions between expanding jets and the surrounding medium do take place . x - ray observations of csos can potentially resolve this issue by providing evidence in favor or against the presence of dense obscuring matter able to frustrate the jets in the youngest sources . in general the gps sources are faint in x - rays and studying them in this waveband became possible only during the last decade , thanks to the _ chandra _ x - ray observatory ( e.g. , * ? ? ? * ; * ? ? ? * ) and _ xmm - newton _ ( e.g. , * ? ? ? * ; * ? ? ? the early x - ray results indicate that the gps x - ray emission could be intrinsically absorbed , although the absorption properties of the gps sources are similar to those found in general source populations . the intrinsic gps x - ray radiation could be related to the accretion , to expanding radio lobes , to relativistic jets , or in some cases to the thermal hot medium found in the central regions of the host galaxy @xcite . however , only a few cso sources with known redshift and measured age has been observed in x - rays ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? thus , we initiated our _ chandra _ program to study the cso sample defined by requiring the availability of kinematic age measurements and @xmath0 ( section [ sec : sample ] ) . our main goal was to establish the x - ray properties of these csos , determine the origin of their x - ray emission , and study the properties of the environment into which they are expanding . we describe our methods of x - ray data analysis in section [ sec : xdata ] , and present the x - ray analysis results in section [ sec : results ] . we discuss the implications of the cso x - ray properties on the models of the high - energy processes in compact radio sources in section [ sec : discussion ] and conclude our findings in section [ sec : conclusions ] . we use the most recent constraints on the cosmological parameters to convert the observed fluxes into luminosities ( hinshaw et al . 2013 ; @xmath27 km@xmath28mpc@xmath4 , @xmath29 implemented as wmap9 in the astropy.cosmology package , @xcite ) . we compiled from the literature a sample of 16 cso sources with redshift and expansion velocity of the radio hot spots available at the time of our proposed _ chandra _ project . table [ tab : cso ] lists these sources and their general radio properties ( the linear size , @xmath30120kpc ; the expansion velocity , @xmath310.4@xmath32 ; and the age inferred through the kinematic argument , @xmath331003000yrs ) . ) is marked in light grey . the dark grey indicates additional csos from @xcite ( except for pks b1413 + 135 classified as core - dominated ) without x - ray data , with the ages derived after our _ chandra _ program has been completed . , width=283 ] we obtained new _ _ observations of 10 csos . we observed the six csos with no prior x - ray information ( 5ksec exposures ) . additionally , we imaged with _ chandra _ the two sources observed previously with _ bepposax _ ( 1946@xmath10708 , 5ksec ; 1934@xmath863 , 20ksec ) , or with _ xmm - newton _ ( 0710@xmath10439 , 38ksec ; 1607@xmath1026 , 38ksec ) . the goal of the long , 2040ksec , exposures was to study in detail the cso environments through the high - resolution spatial and spectral analysis allowed by the _ chandra _ data . table [ tab : chandra ] contains the log of our _ chandra _ observations . it includes also 1511@xmath100518 with archival 2ksec _ chandra _ data . figure [ fig : ages ] shows the distribution of csos kinematic ages for 16 sources in our pilot study and 7 additional sources from @xcite with their ages measured after our _ chandra _ program has been completed . the uncertainties on the age measurements are typically within @xmath34 and related to the quality of the radio monitoring data . note that our x - ray sample contains all the youngest csos with the ages @xmath35 years . in summary , table [ tab : cso ] contains a complete list of cso sources with derived redshift and kinematic age that have been observed in x - rays to date , while table [ tab : chandra ] contains the sources analyzed in this paper . with our _ chandra _ program , we have increased the size of the cso x - ray sample by @xmath2468% . rlrrrrrrc # & source & ra ( j2000 ) & dec ( j(2000 ) & date & obsid & exposure & total@xmath16 & net@xmath16 + & name & & & & & [ s ] & counts & counts + + + 1 . & 1718@xmath8649 & 17 23 41.0 & @xmath865 00 36.6 & 2010 - 11 - 09 & 12849 & 4783 & 231 & 224.9@xmath36 + 2 . & 1843@xmath10356 & 18 45 35.1 & @xmath1035 41 16.7 & 2010 - 10 - 26 & 12850 & 4783 & 11 & 10.8@xmath37 + 3 . & 2021@xmath10614 & 20 22 06.7 & @xmath1061 36 58.8 & 2011 - 04 - 04 & 12853 & 4784 & 54 & 53.8@xmath38 + 4 . & 0035@xmath10227 & 00 38 08.1 & @xmath1023 03 28.4 & 2010 - 10 - 13 & 12847 & 4783 & 10 & 9.9@xmath39 + 5 . & 0116@xmath10319 & 01 19 35.0 & @xmath1032 10 50.0 & 2010 - 11 - 05 & 12848 & 4742 & 3 & @xmath40 + 6 . & 0710@xmath10439 & 07 13 38.2 & @xmath1043 49 17.2 & 2011 - 01 - 18 & 12845 & 37845 & 1679 & 1676.8@xmath41 + 7 . & 1946@xmath10708 & 19 45 53.5 & @xmath1070 55 48.7 & 2011 - 02 - 07 & 12852 & 4742 & 110 & 109.8@xmath42 + 8 . & 1943@xmath10546 & 19 44 31.5 & @xmath1054 48 07.0 & 2011 - 05 - 04 & 12851 & 4783 & 12 & 11.9@xmath43 + 9 . & 1934@xmath863 & 19 39 25.0 & @xmath863 42 45.6 & 2010 - 07 - 08 & 11504 & 19793 & 362 & 361@xmath44 + 10 . & 1607@xmath1026 & 16 09 13.3 & @xmath1026 41 29.0 & 2010 - 12 - 04 & 12846 & 37845 & 213 & 212.5@xmath45 + & secondary@xmath17 & 16 09 12.7 & @xmath1026 41 17.5 & 2010 - 12 - 04 & 12846 & 37845 & 32 & 30.8@xmath46 + 11 . & 1511@xmath100518@xmath19 & 15 11 41.2 & @xmath1005 18 09.2 & 2003 - 05 - 18 & 4047 & 1994 & 49 & 48.8@xmath47 + + notes : @xmath16 total and background subtracted counts in @xmath48 circle with energies between 0.5kev and 7kev ; @xmath17 the secondary x - ray source resolved in our _ chandra _ image . @xmath19 _ chandra _ archival data . the _ chandra _ acis - s data were collected during the 20102011 epoch ( see table [ tab : chandra ] for details ) . all targets were placed at the aim point on the back - illuminated acis ccd ( s3 ) . the observations were made in vfaint mode with 1/8 ccd readout to avoid pileup if sources were to be bright . all 10 targets were detected by _ chandra _ with a number of counts between 3 and 1,677 . the x - ray data analysis was performed with the ciao version 4.6 software @xcite using caldb version 4.4 . we processed the data by running ciao tool acis - process - events and applied the newest calibration files , filtered vfaint background events , and ran a sub - pixel event - repositioning algorithm ( and set pix_adj = edser ) . this final step provides the highest angular resolution x - ray image data for the most up - to - date acis - s calibration . all spectral modeling was done in _ sherpa _ @xcite . we used the cash and cstat fitting statistics @xcite and the nelder - mead optimization method @xcite . we inspected the _ chandra _ data using ds9 and visually confirmed the location of each source , and also defined the source and background regions for further spectral extractions . figure [ fig : images ] displays the @xmath49 acis - s images of all the targets ( denoted in the panels with green circles ) , except for @xmath50 which is detected with 3 counts only ( @xmath51 ) is the probability that the observed counts are due to the poisson background fluctuations . @xmath52 indicates a detection . it is derived via simulations . ] . the target is the brightest source in the image . we stress that due to the different redshifts of our cso sources , the _ chandra _ observations probe physical scales that differ by @xmath241.5 orders of magnitude ( 0.411.2kpc ; figure [ fig : images ] ) . hints of extended x - ray emission in the three sources with the highest number of counts ( 0710@xmath100439 , 1718@xmath8649 , 1934@xmath8638 ) , or a secondary x - ray source ( in 1607@xmath10268 ) could be seen in the images . we investigated the significance of the extended x - ray emission using surface brightness analysis . we compared the observed surface brightness profiles to the chart simulated psf profile expected for a point source . rlccccccr + # & source & @xmath53 & @xmath54 & @xmath55 & norm & flux ( soft ) & flux ( hard ) & cstat / d.o.f + & name & [ @xmath56@xmath7 ] & [ @xmath58@xmath7 ] & & [ 10@xmath59ph@xmath7s@xmath4 ] & [ @xmath61erg@xmath7s@xmath4 ] & [ @xmath61erg@xmath7s@xmath4 ] + + + 1 . & 1718@xmath8649 & 7.15 & 0.08@xmath62 & 1.6@xmath63 & 80.2@xmath64 & 18.6@xmath65 & 35.3@xmath66 & 334.9/443 + 2 . & 1843@xmath10356 & 6.75 & 0.8@xmath67 & ( 1.7 ) & 4.64@xmath68 & 1.1@xmath69 & 2.2@xmath70 & 68.0/444 + 3 . & 2021@xmath10614 & 14.01 & @xmath71 & 0.8@xmath72 & 10.6@xmath73 & 2.7@xmath74 & 19.3@xmath75 & 212.2/443 + 4 . & 0035@xmath10227 & 3.37 & 1.4@xmath76 & ( 1.7 ) & 8.2@xmath77 & 1.8@xmath78 & 3.6@xmath79 & 72.8/444 + 5 . & 0116@xmath10319@xmath16 & 5.67 & @xmath40 & @xmath40 & @xmath40 & @xmath80 & @xmath40 & @xmath40 + 6 . & 0710@xmath10439@xmath17 & 8 . & @xmath40 & @xmath40 & @xmath40 & @xmath40 & @xmath40 & @xmath81 + 7 . & 1946@xmath10708 & 8.57 & 1.7@xmath82 & 1.7@xmath83 & 105.3@xmath84 & 24.4@xmath85 & 36.6@xmath86 & 296.3/443 + 8 . & 1943@xmath10546 & 13.15 & 1.1@xmath87 & ( 1.7 ) & 7.8@xmath88 & 1.8@xmath74 & 3.7@xmath89 & 76.5/444 + 9 . & 1934@xmath863 & 6.16 & 0.08@xmath90 & 1.67@xmath91 & 29.6@xmath92 & 5.0@xmath93 & 12.3@xmath94 & 330.2/443 + 10 . & 1607@xmath1026 & 4.1 & @xmath95 & 1.4@xmath96 & 7.03@xmath97 & 1.6@xmath98 & 4.8@xmath99 & 347.1/443 + & secondary@xmath19 & 4.1 & @xmath100 & 1.4@xmath69 & 1.06@xmath101 & 0.2@xmath96 & 0.6@xmath102 & 143.6/443 + 11 . & 1511@xmath100518 & 3.29 & @xmath103 & 1.0@xmath63 & 22.8@xmath104 & 5.6@xmath105 & 31@xmath106 & 203.8/443 + + notes : unabsorbed flux in the observed energy range ( soft : 0.52kev ; hard : 210kev , extrapolated from the model fitted over the 0.57kev range ) ; all uncertainties are given as @xmath107 for one interesting parameter ; upper limits are 3@xmath108 limits ; @xmath16 3 counts were detected in the source region and no model fitting was performed for this source ; we list the 3@xmath108 flux limit estimated using srcflux tool which performs simulations @xcite . @xmath17 see table [ tab:0710 ] ; @xmath19 the secondary x - ray source resolved in our _ chandra _ image with @xmath109 limit at z=0 . the x - ray spectra and corresponding calibration files ( arf and rmf ) were extracted with the ciao script specextract for all the sources . we assumed a circular region with 1.5@xmath110 radius centered on the position of each source for the spectral extraction . a local background was extracted assuming an annulus with inner and outer radii equal to @xmath111 and @xmath112 , respectively . we compared the total number of counts and net counts for each source ( table [ tab : chandra ] ) . we assessed that the background was negligible in all cases ( @xmath113 count in the source extraction region for all targets , except for 0710@xmath10439 with 2.1 counts , e.g. @xmath114 of the source counts ) , and therefore we ignored it during the spectral analysis ( i.e. we modeled only the source spectra ) . we applied an absorbed power law model to the spectral data : @xmath115\}$ ] ph@xmath7s@xmath4kev@xmath4 , where @xmath116 is the normalization at 1kev , @xmath55 is the photon index of the power law , and @xmath117 and @xmath118 are the equivalent column densities for the two absorption components , galactic and intrinsic to the source . in the _ sherpa _ nomenclature , this baseline spectral model was defined as xsphabs*xszphabs*powlaw1d and fit to the data in the 0.57kev energy range . the galactic absorption was always fixed at the respective value , @xmath53 listed for each source in table [ tab : chandra - fit ] , while the intrinsic absorption column was allowed to vary . the photon index @xmath55 was fixed to 1.7 ( a typical value for unabsorbed @xmath119@xmath7 agn , burlon et al . 2011 ) while modeling the sources with a low number of counts , @xmath120 , in order to obtain a limiting value or detection of the intrinsic absorption . the range of the photon index @xmath121 results in the intrinsic column density consistent to within 1@xmath108 with the values listed in table [ tab : chandra - fit ] . the photon index was a free parameter in sources with a larger number of counts . we detected x - ray emission associated with all the csos in our _ chandra _ sample . however , because the majority of our _ chandra _ observations were short , designed for x - ray detections , the detailed analysis were only possible for about half of the sources in the sample . in the short @xmath245ksec observations the detected source counts ranged from 3 to 231 counts ( see table [ tab : chandra ] ) . the background in such a short _ chandra _ observation of a point source is very low and even 3 counts in 0116@xmath10319 constitute a detection at high significance level ( p-@xmath122 ) . however , we can only provide a limiting flux in this case ( see table [ tab : chandra - fit ] ) , while for the other sources we were able to extract _ chandra _ spectrum and apply parametric models . the three longer exposures provided data of high enough sensitivity to explore the csos x - ray environment on arcsecond scales . we extracted the x - ray surface brightness profiles assuming the annuli regions centered on the radio source position to quantify the presence of an x - ray emission outside the central point source . in two cases , 0710@xmath100439 and 1934@xmath8638 , the profiles are consistent with the background level outside a @xmath246@xmath110 radius . the residuals in the innermost circular regions ( @xmath123 ) , slightly deviate from zero suggesting the possibility of an additional emission component . however , the formal significance test gives a p - value of 0.33 , which does not allow us to reject the null hypothesis that the source is point - like . we concluded that in these two sources the observed x - ray emission originates in an unresolved region within @xmath124 ( corresponding to @xmath125kpc and @xmath126kpc for each source , respectively ) . there is no extended x - ray emission on large scales detected above the background level of 0.0048cts arcsec@xmath127 . the surface brightness profile for the third object , 1718@xmath8649 , is shown in figure [ fig : profile ] . an excess of the observed emission above the psf profile is visible at the distances within 26 ( @xmath128kpc ) from the radio source position indicating an extended x - ray emission present in this source . we add another model component ( the standard beta model , beta1d in _ sherpa _ ; king 1962 ) to account for this emission . we obtained a 1@xmath108 lower limit of @xmath129 in our best fit model . chandra _ observations are needed to better constrain the properties of this emission . in general , our baseline absorbed power law model provided a good description of the x - ray continua of the cso sources . the modeling results are given in table [ tab : chandra - fit ] which displays the galactic and intrinsic absorption columns , photon index , and the unabsorbed flux in the soft ( 0.52kev ) and hard ( 210kev ; extrapolated from the model fitted over the 0.57kev band ) x - ray bands . the uncertainties are given as 1@xmath108 for one interesting parameter . in seven csos the photon index was a free parameter and we obtained @xmath1301.7 , with the exception of 2021@xmath10614 and 1511@xmath100518 where @xmath131 . the intrinsic absorption component was required at @xmath132 confidence level in three csos ( 0710@xmath10439 , 1946@xmath10708 , and 0035@xmath10227 ; note though that in 0035@xmath10227 the photon index was fixed at 1.7 ) with @xmath133@xmath134@xmath7 . in three other cases we were able to derive only upper limits on the intrinsic column densities ( but see section [ sec:2021 ] ) . the x - ray spectra of the remaining five sources are suggestive of the presence of an intrinsic absorbing neutral hydrogen column with @xmath13510@xmath6@xmath7 , although at low statistical significance . even though we detected an extended x - ray emission in 1718@xmath8649 , the lowest redshift cso in our sample , we find that the x - ray spectrum of this source is quite well modeled with an absorbed power law . however , our _ chandra _ spectrum has only 224@xmath136 counts ( see table [ tab : chandra - fit ] ) and any detailed analysis of potential contributions from emission associated with the extended medium would require higher quality x - ray data . below we provide a detailed description of the x - ray results for three particular csos in our sample : 0710@xmath10439 has a hard band narrow emission line ( section [ sec:0710 ] ) ; 2021@xmath10614 and 1511@xmath100518 have an extremely hard x - ray photon index compared to typical agn ( section [ sec:2021 ] ) ; and 1607@xmath1026 is accompanied by a secondary x - ray source resolved within the prior _ xmm - newton _ extraction region ( section [ sec:1607 ] ) . we also comment on 1934@xmath863 and 1946@xmath10708 observed previously with beppo - sax and revisited with our _ chandra _ program ( section [ sec : beppo ] ) . llccccccccc + # & model & @xmath137 & @xmath55 & kt & @xmath138 & @xmath139 & ew & flux & flux & cstat / d.o.f . + & description@xmath16 & [ @xmath58@xmath7 ] & & [ kev ] & & [ kev ] & [ ev ] & [ soft ] & [ hard ] & + + + a & pl & 0.58@xmath140 & 1.59@xmath62 & @xmath40 & @xmath40 & @xmath40 & @xmath40 & @xmath141 & @xmath142 & 490.3/443 + b & pl and g & 0.61@xmath143 & 1.64@xmath144 & @xmath40 & @xmath40 & 6.62@xmath145 & 154@xmath146 & @xmath147 & @xmath148 & 480.3/441 + c & tb and g & 0.49@xmath90 & @xmath40 & 13.6@xmath149 & @xmath40 & 6.63@xmath145 & 164@xmath150 & @xmath151 & @xmath152 & 475.8/441 + + + d & pl and apec & 0.58@xmath140 & 1.42@xmath153 & 5.4@xmath154 & @xmath40 & @xmath40 & @xmath40 & @xmath155 & @xmath156 & 485.0/441 + e & pl and mekal & 0.56@xmath140 & 1.39@xmath157 & 5.2@xmath158 & @xmath40 & @xmath40 & @xmath40 & @xmath159 & @xmath160 & 484.4/441 + + + f@xmath17 & scattered pl and g & 1.02@xmath161 & 1.75@xmath162 & @xmath40 & 0.13@xmath163 & 6.62@xmath145 & 1.40@xmath164 & @xmath165 & @xmath166 & 477.2/440 + + notes : + @xmath16 pl power law ; g gaussian line ; tb thermal bremsstrahlung ; @xmath138 is the fraction of the intrinsic power law scattered into our line of sight by a photoionized medium that is also responsible for the fe xxv iron line emission ; unabsorbed flux in the soft ( 0.52kev ) and hard ( 210kev ) bands extrapolated from the model fitted over the 0.57kev range , in units of [ @xmath61erg@xmath7s@xmath4 ] . phabs * zphabs * powerlaw + b phabs * zphabs * ( powerlaw + zgauss ) + c phabs * zphabs * ( zbremss + zgauss ) + d phabs * zphabs * ( apec + powerlaw ) + e phabs * zphabs * ( mekal + powerlaw ) + f phabs * ( zphabs * powerlaw + fsc * powerlaw + zgauss ) + all models modified by galactic absorption with @xmath167@xmath7 ; abundances in apec and mekal models fixed at 0.5 solar value ; mekal density fixed at 1@xmath168 ; width of the fe line fixed at @xmath169kev . + @xmath17 ew calculated with respect to the unabsorbed scattered power law , fsc * powerlaw . the photon index and the amount of intrinsic absorption in our _ chandra _ spectrum of 0710@xmath10439 agree very well with the earlier _ xmm - newton _ results of vink et al . the source did not vary between the _ xmm - newton _ and _ chandra _ observations showing consistent 210kev flux of about @xmath170erg @xmath7 s@xmath4 . however , we find clear residuals in the hard x - ray band suggestive of the presence of an emission line ( figure [ fig:0710 ] , top / left ) . to account for this feature we added a gaussian component to the absorbed power - law spectral model ( model b in table [ tab:0710 ] ; figure [ fig:0710 ] , top / right ) . the line is unresolved ( @xmath171ev ; fixed ) with the best fit line energy of @xmath172kev ( rest frame ) and the equivalent width @xmath173ev . visual inspection of the _ xmm - newton _ modeling presented by vink et al . ( 2006 ) tentatively hints at the presence of an emission line at similar energy also in the _ xmm - newton _ dataset . emission lines at the energies within @xmath246.46.9kev due to iron are common in x - ray spectra of accreting black holes @xcite , including radio - quiet and non - blazar radio - loud agn ( e.g. , @xcite , winter et al . 2010 ; fukazawa et al . 2011 ; zhou , zhao & soria 2011 ) and galactic black hole binaries ( e.g. , done et al . in particular , a narrow fluorescent neutral fe k@xmath174 line at 6.4kev ( rest frame ) seems to be omnipresent in the spectra of radiatively efficient agn ( bianchi et al . 2004 ; nandra et al . 2007 ; iwasawa et al . 2012 ; and references therein ) , and is believed to originate from a cold matter , e.g. a molecular torus , illuminated by hard x - rays . narrow @xmath246.66.9kev emission lines from an ionized iron have been observed e.g. in luminous pg quasars , including the radio - loud source mrk 1383 ( porquet et al . 2004 ) , and in a substantial fraction of seyfert 1 galaxies ( patrick et al . these lines could be emitted by a hot ionized region of an accretion disk illuminated by an external x - ray source ( e.g. , * ? ? ? * ; * ? ? ? * ) ; a distant gas , photoionized by the nuclear illumination ( e.g. , bianchi & matt 2002 ) ; or a diffuse thermal plasma at @xmath175kev @xcite the best fit parameters of the line detected in 0710@xmath10439 assuming model b ( see table [ tab:0710 ] ) are consistent with the first scenario and with the he - like fe xxv line emission in a system with an inclination angle @xmath1760.6 and an accretion rate @xmath177@xmath178 ( matt et al . however , in order to study the origin of this emission line in more details we also tested four other spectral models of increased complexity that are typically considered for agn x - ray emission(table [ tab:0710 ] ) . given the quality of the current data all the models are statistically equivalent , but considering the range of the best fit parameters we can potentially rule out some of them . first , we parametrized the x - ray continuum with a thermal bremsstrahlung model ( zbremss ) and kept the gaussian profile to fit the emission feature ( model c in table [ tab:0710 ] ) . we found a statistically good fit , but with a relatively high temperature , @xmath179kev . next , we replaced the bremsstrahlung continuum with a single temperature plasma ( mekal ) model ( metal abundances fixed at 0.5 the solar value ) and found that at this high temperature the strength of the ionized fe lines was overpredicted , especially at the energy of the fully ionized fe xxvi , while the data did not provide evidence for the presence of the fe xxvi line . similar result was obtained when we used the apec plasma model instead of the mekal model . hence we concluded that single component thermal models are not likely given the physical properties required by the best fit parameters . therefore , we tried a two - component model , consisting of a power law and either mekal or apec ( models d and e in table [ tab:0710 ] ) . we obtained the best fit plasma temperature @xmath180kev , and a hard photon index , @xmath1811.5 , which are consistent with the prediction of jet / lobe emission . finally , we tested a model in which the spectrum is composed of the intrinsically absorbed primary emission ( intrinsic emission modeled with a power law ) and an unabsorbed fraction of the same emission scattered by a warm medium . the warm medium is also producing the iron emission line ( model f in table [ tab:0710 ] ) . in this scenario the photon index is @xmath182 , the primary continuum is absorbed with a column of @xmath183@xmath7 , and the scattering fraction is @xmath2413% . the ew of the line ( computed with respect to the scattered continuum ) is @xmath241.4kev , in agreement with the values found by bianchi & matt ( 2002 ) for column densities of the order of a few @xmath184@xmath7 ( see their figure 4 ) . the best fit parameters of the models a , b , e and f , and the ratio of the data to the models are presented in figure [ fig:0710 ] . we conclude that with the present data we are not able to state if the soft x - rays and the he - like iron line in 0710@xmath10439 are due to scattered nuclear emission ( models b or f ) or thermal emission of the ism heated through interactions with the expanding jet ( models d and e ) . the neutral iron k@xmath174 line has been previously reported in cso sources observed with _ xmm - newton _ ( oq208 , guainazzi et al . 2004 ; and 1607@xmath1026 , @xcite ) , while risaliti et al . ( 2003 ) detected neutral and highly ionized ( @xmath185kev ) iron emission lines in 1934@xmath8638 and 1946@xmath10708 ( respectively ) using _ beppo - sax_. these two emission lines are not present in the low counts _ chandra _ spectra63 and 0.4 for 1946@xmath186708 when testing for the line with the null hypothesis of the power law model , and using the simulations as described in @xcite ( plot_pvalue in _ sherpa _ ) . ] of 1607@xmath1026 ( see sec . [ sec:1607 ] ) and of the two _ beppo - sax _ sources ( see sec . [ sec : beppo ] ) . @xcite detected a narrow emission line in a cso candidate ( pmnj1603@xmath84904 ) , and derived the redshift of the source assuming that the most likely rest frame energy of the line is 6.4kev . the presence of a mildly ionized fe line in the spectrum of 0710@xmath10439 indicates that there might be a systematic error in the redshift derivation of pmnj1603@xmath84904 because of a range of ionization states of iron producing the emission lines in the cso sources . lastly , we note that 0710@xmath10439 has the highest number of counts ( 1,679 ) among the sources in our sample . the remaining csos were detected with 3362 counts . a higher number of counts in _ chandra _ spectra are needed for detecting narrow spectral features . the absorbed power - law model applied to two csos , 2021@xmath10614 and 1511@xmath100518 , resulted in an unusually hard photon index of @xmath187 and @xmath188 respectively . typically , such a low value indicates a strong nuclear obscuration able to cover completely the source of the primary x - ray radiation . the observed spectrum is then dominated by a compton reflection hump ( and an accompanying neutral iron line ) from a distant material , presumably a molecular torus , resulting in an artificially low photon index . agn with these properties are called ` compton thick ' , and the amount of the hydrogen absorbing column required to distort the x - ray spectrum in this manner is @xmath189 ( e.g. , * ? ? ? the first radio - loud compton thick agn was found in oq208 , a cso source observed with _ xmm - newton _ by @xcite . @xcite speculated that 1607@xmath1026 is also a compton thick cso . they based their argument on the photon index @xmath190 obtained when an absorbed power - law model was applied to the _ xmm - newton _ data of this source . by analogy with oq208 and 1607@xmath1026 , it is likely that 2021@xmath10614 and 1511@xmath100518 are compton thick cso candidates . our result on the intrinsic absorption in these sources would then be significantly underestimated because in our fit we do not account for the column density that would be needed to absorb the primary continuum . indeed , when we fit the data of 2021@xmath10614 with the sum of an intrinsically absorbed power law and unabsorbed reflection ( pexrav ) component we obtain cstat / d.o.f . = 209/442 , @xmath191@xmath7 . the resulting photon index of the underlying power law would be , however , very soft , @xmath192 , and the unabsorbed intrinsic 210kev luminosity would increase by an order of magnitude making 2021@xmath10614 the second cso in our sample ( in addition to 0710@xmath10439 ) to exceed @xmath193ergs@xmath4 ( comparable to the eddington luminosity of a 10@xmath194m@xmath195 black hole ) . in 1511@xmath100518 the compton thick model results in cstat / d.o.f . = 198.2/443 , @xmath196@xmath7 , @xmath197 and @xmath198ergs@xmath4 . on the other hand , we do not see any evidence for the presence of a neutral iron line in the spectra of 2021@xmath10614 and 1511@xmath100518 . the line is unconstrained in 1511@xmath100518 . in 2021@xmath10614 we can only derive an upper limit on the ew of this line , @xmath199kev , inconsistent with the expected range derived by matt et al . ( 1996 ; @xmath2002.7kev ) . thus , it is interesting to note that the low x - ray photon index and the low intrinsic absorption can be understood in the context of a non - thermal x - ray lobe emission model for the gps / cso sources @xcite . unfortunately , the quality of our present _ chandra _ data is not sufficient to discriminate between the compton thick absorption and x - ray lobe emission scenarios for x - ray emission in 2021@xmath10614 and 1511@xmath100518 ( only 54 and 49 counts collected in 5ksec and 2ksec _ chandra _ observations , respectively ) . deeper x - ray observations with _ chandra _ or _ xmm - newton _ accompanied with the hard @xmath201kev band exposures ( _ nustar _ ) would be needed in order to resolve this issue . owing to the superior spatial resolution of _ chandra _ we resolve the source region of 1607@xmath1026 observed with _ xmm - newton _ by @xcite into two point - like x - ray sources ( see figure [ fig : images ] ) . the location of 1607@xmath1026 matches the position of the upper left source visible in figure [ fig : images ] with the detected 213 counts . the secondary source separated by @xmath2413.5@xmath110 ( lower right corner in figure [ fig : images ] ) has 32 counts and its spectrum could be modeled with an absorbed power law ( assuming only galactic absorption since the redshift to the source is unknown ) with a photon index @xmath202 ( see table [ tab : chandra - fit ] ) . this means that the _ xmm - newton _ spectrum obtained by @xcite represents a combination of emission from 1607@xmath1026 and from the secondary source . we detect an order of magnitude decrease between the hard x - ray fluxes of _ xmm - newton _ ( @xmath203ergs@xmath4@xmath7 ) and _ chandra _ ( @xmath204ergs@xmath4@xmath7 ) . this change is larger than the sum of the fluxes from the two sources , and it is larger than the absolute calibration uncertainties between the _ chandra _ and _ xmm - newton _ satellites @xcite . therefore , we concluded that the x - ray flux of one or both of the sources decreased during the @xmath205years interval between the two observations . we searched the sdss , 2mass and first catalogs to check the identification of the secondary x - ray source , but did not find any corresponding emission . if the source were associated with 1607@xmath1026 then it would be located @xmath2480kpc away along the axis of the cso radio structure , on the extension of its radio jet ( see @xcite for a discussion of the cso radio morphology ) . it is interesting to speculate that the observed secondary x - ray source might be a hot spot due to the past radio activity in the framework of the intermittent scenario for the gps sources . however , assuming the expansion velocity of @xmath206 ( table [ tab : cso ] ) the age of this feature would be @xmath241myr , while the terminal hot spots are expected to die very quickly , on timescales @xmath207myr once the jet switches off @xcite . therefore , the nature and the association of this secondary source with 1607@xmath1026 remains unclear . contrary to @xcite who reported @xmath208 in 1607@xmath1026 , we find @xmath209 , an x - ray photon index that is fairly common in a population of nearby radio - quiet and radio - loud agn ( e.g. winter et al . 2009 , fukazawa et al . we do not detect any evidence of the neutral iron k@xmath174 line in the spectrum . this line was , however , required by the _ xmm - newton _ data of @xcite as part of their compton reflection dominated model for this source . one explanation of this result could be that 1607@xmath1026 is a transitional compton thick / thin agn . transitions between the compton thick and thin regimes have been previously reported in the so - called ` changing look ' agn ( see , e.g. , @xcite ; risaliti et al . 2005 ; 2010 ; 2011 ; and references therein ) . the compton thick / thin switches are thought to be due to an x - ray absorber made of clouds with @xmath21010@xmath211@xmath7 crossing our line of sight on timescales between hours and weeks . if the flux from the secondary source were constant between the two exposures , then 1607@xmath1026 would have been @xmath212 times brighter in the reflection dominated state recorded with _ xmm - newton _ than in our observation . this clearly contradicts the ` changing look ' agn scenario . to reconcile the hard x - ray flux variability with the changing look hypothesis , 1607@xmath1026 could not have contributed more than @xmath24200cts to the _ xmm - newton _ data , meaning that , the secondary x - ray source would have to be @xmath212 times brighter than 1607@xmath1026 in the _ xmm - newton _ observation . this would imply that the x - rays detected and modeled by tengstrand et al . ( 2009 ; including the iron line ) originated from the secondary source rather than from 1607@xmath1026 , eliminating any spectral information about the state of 1607@xmath1026 during the _ xmm - newton _ observation . it would also indicate that the photon index of the secondary source changed from @xmath213 to @xmath214 , its iron line disappeared , and the hard flux increased by a factor of @xmath215 between the two exposures . these spectral properties would be very difficult to explain , especially with no radio , optical and @xmath216-ray identification available at present for the secondary source . thus , we reject the possibility that x - ray variability observed in 1607@xmath1026 is due to the compton thick / thin switches . the alternative explanations include fading of the nuclear and reflected 1607@xmath1026 emission , or a heavy obscuration with a column density @xmath217@xmath7 , as a result of which _ chandra _ would observed thermal emission of the ism . finally , in the non - thermal scenario of @xcite and @xcite any potential intrinsic variability of the x - ray continuum is expected to take place on timescales much too long to account for the results of the _ chandra _ observations of 1607@xmath1026 . clearly , future x - ray monitoring of this young radio source is needed in order to determine its nature and x - ray variability properties . @xcite reported that 1934@xmath863 and 1946@xmath10708 showed _ beppo - sax _ spectra typical of the compton - thick sources with @xmath218@xmath7 , strong fe k@xmath174 line , and the reflection hump . they also considered compton - thin absorbers as a secondary explanation of the data . we have revisited both sources with _ chandra_. we could only obtain a 3@xmath108 upper limit on the line equivalent width of @xmath219kev in 1934@xmath863 and of @xmath220kev in 1946@xmath186708 , assuming that the lines are unresolved ( @xmath221ev ) and located at the rest energies , 6.3 kev and 6.9 kev , respectively ( as in @xcite ) . future more sensitive x - ray observations are needed to constrain the parameters of these lines . we measure a relatively low intrinsic absorption column density in both sources ( @xmath222@xmath7 ) and do not confirm their compton - thick nature ( see table [ tab : chandra - fit ] ) . we conclude that either the _ bepposax _ spectra have been contaminated by other x - ray sources in the field , given the 2 arcmin source extraction regions in the _ bepposax _ observations , or the compton - thin model is more appropriate explanation of the data , or that these sources belong to the agn class that switches between the compton thick and thin states . we have investigated properties of the x - ray emission in a sample of 16 young radio sources with known redshift and kinematic age measurements of their radio structures . all sources in our sample have symmetric radio morphology , have been classified as csos and have been associated with radio galaxies . they do not display strong agn emission in the optical band because they either do not contain an optical - uv bump characteristic of the agn , or because the optical emission has been absorbed by a dense medium in the central region of the host galaxy . we detected six of the sources in our sample for the first time in x - rays . the angular sizes of the cso radio emission are of the order of milliarcseconds ( e.g. * ? ? ? * ) , and as such are too small to be resolved with _ chandra_. as a result , x - ray emission associated with any compact radio jets , lobes , or bow shocks potentially present around the lobes is integrated within the 1.5@xmath110 radius of the circular source extraction region , corresponding to 0.411kpc in our sources , depending on redshift . consequently , the csos x - ray emission contains information on both the circumnuclear region ( on sub - pc and pc scale ) , and the surrounding ism ( kpc scale ) . different physical processes may , thus , contribute to the observed x - ray emission ( see , e.g. * ? ? ? * for a review ) . the csos x - ray emission could arise from the agn nucleus ( x - ray corona ) indicating the nature of the nucleus and its accretion state @xcite . however , if the absorption column is larger than @xmath24 10@xmath211@xmath7 , the nucleus could be hidden and the observed x - rays would be dominated by the thermal emission of the hot ism from shocks driven by the expanding jet @xcite , or by a reprocessed nuclear emission due to the reflection and/or scattering effects ( e.g. * ? ? ? * ) . on the other hand x - rays can also be emitted by compact radio lobes through inverse - comptonization of local radiation fields @xcite . these models result in different predictions on the x - ray photon index , amount of intrinsic x - ray absorption , and the luminosity and could be tested with deep observations . the csos in our sample are rapidly expanding , as demonstrated by their measured hot spot expansion velocities spanning the 0.070.4@xmath32 range ( table [ tab : cso ] , and references therein ) . based on these velocities , the csos kinematic age estimates range between @xmath24100 and 3000yrs . these are , thus , the youngest radio sources which can provide information about the accretion state of supermassive black holes shortly after the onset of the jet formation , as well as about the properties of the environment into which the newly born jets are evolving , and with which they are interacting . jets and lobes in young radio sources , confined to the central parts ( @xmath223kpc ) of their host galaxies , should interact with a denser and less homogeneous medium than their evolved ( @xmath224kpc - long ) analogs . indeed , our results support this idea . we detected multi - phase galactic environments consisting of diverse gaseous components co - spatial with the young expanding cso radio structures : ( 1 ) the extended thermal component in 1718@xmath8649 ( section [ sec : images ] , figure [ fig : profile ] ) ; ( 2 ) possibly an ionized diffuse plasma giving rise to the he - like fe emission line in 0710@xmath10439 ( section [ sec:0710 ] , figure [ fig:0710 ] ) ; ( 3 ) neutral hydrogen with moderate column density , @xmath22510@xmath6@xmath7 , in the majority of the sources ( table [ tab : chandra - fit ] , figure [ fig : ages ] , middle ) ; ( 4 ) possibly a dense compton - thick absorber in 2021@xmath10614 and 1511@xmath100518 ( section [ sec:2021 ] ) . ; filled and open symbols indicate fits with @xmath55 free and fixed , respectively ; vertical lines connect multiple measurements ( or measurements obtained through alternative models as in section [ sec:2021 ] ) available for the same source . * bottom : * intrinsic x - ray luminosity ( 210kev ) vs. age based on our _ chandra _ results ( squares , k - corrected ) and _ xmm - newton _ results ( triangles ; sources 6 , 10 , 1316 in table [ tab : cso ] ) . symbols as in the top panel ; note the large scatter within the youngest sources . , title="fig : " ] ; filled and open symbols indicate fits with @xmath55 free and fixed , respectively ; vertical lines connect multiple measurements ( or measurements obtained through alternative models as in section [ sec:2021 ] ) available for the same source . * bottom : * intrinsic x - ray luminosity ( 210kev ) vs. age based on our _ chandra _ results ( squares , k - corrected ) and _ xmm - newton _ results ( triangles ; sources 6 , 10 , 1316 in table [ tab : cso ] ) . symbols as in the top panel ; note the large scatter within the youngest sources . , title="fig : " ] recent multi - wavelength studies also provide evidence that cso radio sources are embedded in complex gaseous environments . in the ir band the spitzer data show a variety of the ir line properties indicating an enhanced star formation rates in host galaxies of compact radio sources @xcite . one explanation could be a jet - induced star formation in csos @xcite , suggestive of the interactions between the jet and dense environment . @xcite points to a possible selection bias in flux limited radio samples that impacts general conclusions about the epoch of star formation and triggering the radio source . however , these ir results are consistent with the cso source being embedded in a dense environment . in the radio band , @xcite show that the properties of the hi absorption line in the spectra of gps and compact steep spectrum sources ( css ; a more evolved stage compared to the gps sources ) differ substantially from those of the extended radio sources . based on the correlation between the amount of hi absorption and the type of a radio source ( @xmath2455% hi detection rate , and optical depths higher in the compact sources than in the extended sources ) the authors argued that the compact radio sources are embedded in a medium that is rich in atomic gas . moreover , the hi kinematics in the gps / css sources ( broader and more asymmetric , blue - shifted profiles ) suggested that the gas is unsettled and may form outflows . they concluded that the evidence of interactions between the radio sources and rich ambient environment is more frequent in the young / compact sources than in the extended sources . finally , studies of the relationship between the x - ray and radio absorbing columns reveal a correlation between the @xmath226 and @xmath227 absorption in gps / cso sources @xcite , suggesting that the x - ray and radio absorbers are co - spatial . this is an important constraint , given the latest radio results from the low frequency radio observations @xcite . they seem to rule out the synchrotron self absorption process and favor the free - free absorption as the origin of the radio obscuration in the gps source , pksb0008@xmath842 . this again points to a rich , multi - phase environment impacting the evolution of csos . it is , thus , interesting to interpret our findings on the csos intrinsic neutral hydrogen column density in the context of the absorption properties of an unbiased population of agn . we compare our results with those obtained by @xcite for their subsample of sy1.82 galaxies from the hard x - ray swift / bat survey ( note that csos are observed at large inclinations to the jet axis , and as such should be compared with type 2 agn ) . in the sample of csos presented in table [ tab : cso ] ( see figure [ fig : agedependence ] top ) we observe that 2 of 15 ( 13% ) csos show evidence for compton - thick absorption with @xmath228@xmath7 ( 33% if we include 2021@xmath10614 , 1511@xmath100518 and 1607@xmath1026 proposed to be compton - thick candidates in this work and in @xcite ) , respectively , as compared to @xmath2420% in sy1.8 - 2 galaxies in @xcite . conversely , 9 of 15 ( 60% ) sources have @xmath229@xmath7 ( 40% if we exclude 2021@xmath10614 , 1511@xmath100518 and 1607@xmath1026 ) . instead , only 7% of the sy1.82 galaxies have column densities in this range . while the fraction of compton - thick csos seems to be in a relatively good agreement with that found in a sy1.82 population of @xcite , especially given the small size of our csos sample , the overabundance of sources with only moderate intrinsic obscuration seems puzzling . @xcite pointed out that the narrow line radio galaxies ( nlrg ) , considered to be the more evolved analogues of cso , have @xmath230@xmath7 associated with the absorption of the nuclear emission by a molecular torus . consequently , the large fraction of csos with hydrogen column densities lower than those typically found in nlrgs , may indicate echoing the conclusions of @xcite that the cso x - ray obscuration is caused by the gas located on much larger scales than molecular tori . the 210kev x - ray luminosities of our sources cover @xmath2313 orders of magnitude ( figure [ fig : agedependence ] bottom ) , @xmath232@xmath233ergs@xmath4 , typical for the @xmath0 agn ( e.g * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? interestingly , we observed that the oldest csos ( @xmath234years ) tend to be relatively bright in x - rays , with luminosities @xmath235ergs@xmath4 , while the youngest csos can be found at any x - ray luminosity covered by the sample . we found hard x - ray spectra with a photon index @xmath2361.7 in five ( out of seven ) sources , where we were able to model the photon index . among non - blazar agn , this range of photon indices is often seen in radio - quiet agn , although it appears to be the most compatible with the average photon index of the non - blazar radio - loud agn ( @xcite but see @xcite ) . the difference between the average x - ray spectral slopes in radio - loud and radio - quiet agn populations may result from either a dilution of the x - ray coronal emission , or a non - negligible contribution from the non - thermal emission of jets @xcite . it is , thus , likely that in the particular case of csos the x - ray jets and/or lobes contribute to the observed x - ray radiation , as suggested by @xcite and @xcite . the photon indices we observed would then be consistent with a non - thermal scenario in which low - energy photons are inverse compton upscattered to x - ray energies by electrons with a power - law spectral energy distribution with a rather standard spectral index , @xmath237 . thus , the shape of the absorbed power law x - ray continuum that we derived for the csos in our sample is consistent with either the nuclear ( x - ray corona ) emission absorbed on the torus scale ( @xmath238pc ) , or an extended ( x - ray jets / lobes ) emission absorbed on the host galaxy ( @xmath24kpc ) scale . we note that the results described in section [ sec : environment ] , along with that of @xcite , favor the latter possibility . instead , high intrinsic luminosities , exceeding 10@xmath239ergs@xmath4 , derived for several csos ( figure [ fig : agedependence ] , bottom panel ) , together with the likely detection of the reprocessed / scattered nuclear light in 2021@xmath10614 , 1511@xmath100518 ( section [ sec:2021 ] ) and 0710@xmath10439 ( section [ sec:0710 ] ) may indicate the dominant agn contribution to the observed x - ray emission of young radio sources . our results do not appear to favor thermal emission of the diffuse ism plasma as the origin of the cso x - ray continuum , unless the soft x - rays and the he - like iron emission in 0710@xmath10439 are interpreted as a sole example in which the ism is heated by the expanding radio structure ( section [ sec:0710 ] , figure [ fig:0710 ] , model e ) . we examined the x - ray properties of 16 young radio sources with _ chandra _ and observed variety of interesting characteristics , which we summarize below . * we found that our cso sample covers a wide range in x - ray luminosity , @xmath2@xmath3ergs@xmath4 and that the majority of sources contains a modest amount of the intrinsic absorption ( @xmath24010@xmath6@xmath7 ) . * we obtained the highest quality x - ray spectrum of the cso source to date in our long _ chandra _ observation of 0710@xmath10439 and found an ionized iron emission line , @xmath11kev . however , we were not able to determine the origin of the line , as it could be related to the emission from the hot ionized accretion disk illuminated by an external source , from a distant photoionized medium , or from a diffuse hot thermal plasma within the nuclear region . depending on the model , we derived ew of the line ranging from @xmath241ev ( model b ) to @xmath242kev ( model f ) . * we found a hard photon index of @xmath243 in 2021@xmath10614 and @xmath188 in 1511@xmath100518 consistent with either a compton thick absorber or non - thermal emission from compact radio lobes . * we discovered that the 210kev x - ray flux decreased by an order of magnitude since the 2008 _ xmm - newton _ observation of 1607@xmath1026 . future monitoring of this source is needed in order to characterize and understand the nature of the variability . * we examined the highest angular resolution x - ray images of the csos and detected extended x - ray emission in 1718@xmath8649 which needs to be studied with the deeper x - ray observations in the future . we found that short @xmath245ksec x - ray observations with @xmath244 have proven to be effective in detecting the csos sources and deriving their basic x - ray properties . with our _ chandra _ program , we observed six csos for the first time in x - rays , and we increased the size of the known csos sample with measured redshift , kinematic age , and x - ray information by @xmath2468% . the cso x - ray population can be further extended through snapshot _ chandra _ and/or _ xmm - newton _ observations of seven sources whose kinematic ages have been recently derived ( an & baan 2012 , figure [ fig : ages ] ) . however , it is apparent that deeper exposures are needed in order to study details of the csos environment ( e.g. , the amount of intrinsic absorption , the gaseous media co - spatial with the radio sources ) and the origin of the csos intrinsic x - ray emission ( diffused thermal vs. nuclear emission , evidence for reprocessing of the nuclear x - rays by a circumnuclear matter , thermal vs. non - thermal electron population generating the high - energy csos spectra , long term x - ray variability ) . deep x - ray observations are also necessary in order to place critical constraints on current theoretical models for the earliest stage of radio source evolution . we thank the anonymous referee for detailed comments . this research is funded in part by national aeronautics and space administration contract nas8 - 03060 . partial support for this work was provided by the nasa grants go1 - 12145x , go4 - 15099x , nnx10ao60 g . g.m . acknowledges the financial support from the univearths labex program of sorbonne paris cit ( anr10labx0023 and anr11idex000502 ) l.o . acknowledges support from the infn grant indark , the grant prin 2012 `` fisica astroparticellare teorica '' of the italian ministry of university and research , and the `` strategic research grant : origin and detection of galactic and extragalactic cosmic rays '' funded by the university of torino and compagnia di san paolo . .s . was supported by polish nsc grant dec-2012/04/a / st9/00083 . this research has made use of data obtained by the _ chandra _ x - ray observatory , and software provided by_chandra _ x - ray center ( cxc ) in the application packages ciao , chips , and _ sherpa_. refsdal , b. et al . 2009 , in proceedings of the 8th python in science conference ( scipy 2009 ) , g varoquaux , s van der walt , j millman ( eds . ) , pp . 51 - 57 http://conference.scipy.org/proceedings/scipy2009/paper_8

the interpretation of experimental measurements at the lhc requires accurate theoretical predictions for exclusive observables , and in particular the summation of soft and collinear radiation to all orders in perturbation theory . we report on recent progress towards the automated calculation of multi - parton lhc cross sections at next - to - leading order in qcd , including the summation of next - to - leading logarithmic corrections through the combination with parton showers . higher - order qcd corrections , parton shower

1502.01521

theoretical calculations within fixed - order perturbation theory allow for accurate predictions of inclusive observables like total cross sections . the analysis and interpretation of experimental signatures at the lhc , however , require theoretical predictions for exclusive final states , i.e. predictions for differential distributions or cross sections with cuts on kinematic variables . higher - order calculations for such exclusive final states involve in general large corrections from soft or collinear parton emission , which need to be summed to all orders . an efficient way to achieve such a summation is through a parton shower . parton showers typically include the leading logarithmic contributions from soft and collinear gluon or quark emission to all orders in perturbation theory . they form a central part of monte carlo event generators and are thus essential to connect theoretical models with realistic experimental signatures . on the other hand , standard parton shower event generators often rely on leading - order expressions for the hard scattering processes and can therefore not predict inclusive cross sections accurately . a central goal of recent and current theoretical work in lhc phenomenology is thus the automated calculation of next - to - leading ( nlo ) lhc cross sections including the summation of large corrections from multiple quark and gluon emission through parton showers . such a calculation should combine the accuracy of nlo predictions for inclusive cross sections with the power of parton shower monte carlo programs to reliably describe differential distributions and cross sections with cuts on kinematic variables . a naive combination of nlo calculations with parton showers would , however , lead to double counting of higher - order contributions that are included in both the nlo cross section and the parton shower . these contributions have to be identified and subtracted from the calculation by means of a matching procedure . much work in recent years has been devoted to the formulation of matching schemes that allow for a consistent combination of parton showers and nlo calculations including loop - corrections @xcite . so far , however , only few publicly available computer codes exist that implement such schemes , including in particular mc@nlo @xcite , powheg @xcite , sherpa @xcite and madgraph5_amc@nlo @xcite . the problem of double counting contributions that are included in both the nlo calculation and the parton shower has been solved in all approaches in a similar way : the parton shower evolution is generically described by a sudakov factor of the form @xmath0 shown here for gluon emission off a quark with virtuality @xmath1 . in the collinear limit @xmath2 the function @xmath3 reduces to the altarelli - parisi splitting function . the sudakov factor includes short - distance contributions which are also part of the nlo cross section and which can be identified by expanding the exponential in powers of @xmath4 . when combining parton showers with nlo calculations these contributions need to be subtracted to avoid double counting . early work by many authors has focused on implementing specific processes into the mc@nlo program and within the powheg framework . while the original mc@nlo code is very successful and now includes a large number of processes , it is tied strongly to the parton shower of herwig ( both herwig6 and herwig++ ) . this restriction has been lifted recently within madgraph5_amc@nlo , which is not only fully automatic , but also parton shower independent and may be used with various versions of herwig and pythia . we note that sherpa contains an implementation of mc@nlo as well . powheg , on the other hand , allowed for matching with a generic parton shower from the beginning and could thus be interfaced with various monte carlo programs . in practice , the implementation of specific processes has been greatly simplified thanks to the powheg - box tool . in each of the frameworks , multiple parton emission is summed at the leading - logarithmic ( ll ) level only ( although leading - color subleading logarithmic effects might be included in parton showers based on nlo subtraction schemes , such as the catani - seymour subtraction ) , and the results can thus not compete with the accuracy of dedicated resummation calculations which are routinely performed at next - to - leading logarithmic accuracy ( nll ) . the formulation and improvement of parton showers , which are in general based on various assumptions and approximations , has also been addressed by various authors recently @xcite . notably , refs.@xcite have proposed a parton shower that includes quantum interference , spin correlations and sub - leading color effects . interference is treated in standard parton showers only approximately by means of angular ordering , while spin information is generally ignored completely . spin correlations , on the other hand , are crucial for example to explore new physics models in cascade decays at the lhc . while refs.@xcite describe the theoretical formulation of parton showers with quantum interference , the first implementation @xcite so far is based on the standard spin - averaged and leading - color treatment of parton splitting only . the set - up of the parton shower also has implications for the matching with nlo calculations . if the splitting functions of the parton shower , @xmath5 , and the momentum mapping follow closely the definition of the subtraction terms used to regularize the soft and collinear divergences in the nlo calculation , then the matching is simplified considerably . with that in mind , various authors have formulated parton showers that are based on commonly used subtraction schemes @xcite . to improve on existing nlo plus parton shower implementations , quantum interference contributions , spin correlations and sub - leading color effects in the parton shower should be included systematically . ultimately , one would like to develop a fast and numerically robust code for the automated calculation of multi - parton lhc cross sections at next - to - leading order , including the summation of next - to - leading logarithmic corrections through combination with parton showers . in this contribution we shall describe recent progress towards this goal . the article is structured as follows . in section [ sec : subtraction ] we shall describe the construction of a nlo subtraction scheme as derived from a nll parton shower , and its implementation into the program package helac - nlo . section [ sec : bbbb ] reports on the application of such a scheme for the automated calculation of nlo - qcd corrections to the production of four bottom quarks at the lhc . section [ sec : nll_shower ] finally discusses progress towards developing and implementing a nll parton shower and the matching with helac - nlo . we conclude in section [ sec : conclusions ] . in the following , a new subtraction scheme based on a parton shower introduced by nagy and soper @xcite is described . this new subtraction has been implemented in the publicly available helac - dipoles software @xcite and has already been tested in the calculation of the nlo qcd corrections to @xmath6 production at the lhc ( see the next section for more details ) . our main motivation , however , was to provide a framework for a simple matching between a fixed - order calculation and the new parton shower . however , before addressing this issue , the problem of the integration of subtraction terms over the unresolved phase space needed to be solved . this is the non - trivial part of any subtraction scheme . contrary to the usual practice at nlo , we did not perform involved analytic integrations , but rather used a semi - numerical approach . after a suitable parameterization , numerical integrations , inspired by recent nnlo methods @xcite , have been performed . this has allowed us to cover both massless and massive cases with comparable effort . let us emphasize that our semi - numerical approach distinguishes our work from earlier publications on the subject presented e.g. in refs.@xcite . let us start with the inclusive nlo qcd cross section for a generic process involving @xmath7 final - state qcd partons with momenta @xmath8 , which can be written as follows @xmath9 where @xmath10 @xmath11\,,\end{aligned}\ ] ] and @xmath12 , @xmath13 , @xmath14 describe the born , one - loop and real - emission matrix elements , respectively . the integration measure for the @xmath15 and the @xmath16parton phase space is denoted by @xmath17 and @xmath18 , whereas @xmath19 and @xmath20 are the jet functions . for @xmath21 well separated hard jets , the born contribution is finite , whereas the virtual and the real - emission terms are individually divergent due to the presence of soft and collinear singularities . for an infrared - safe definition of partonic jets , all soft and collinear divergencies that affect the virtual and real corrections should cancel for the inclusive cross section , except for the singularities arising from the emission of nearly - collinear partons off the initial state , which are absorbed into a re - definition of the parton distribution functions ( pdfs ) . this absorption is achieved by introducing suitable collinear counterterms , @xmath22 . however , the individual pieces , @xmath23 and @xmath24 , still suffer from soft and collinear divergencies and can not be integrated numerically in four dimensions . to solve this problem , local counterterms , @xmath25 , that are designed to match the singular structure of the integrand in the soft and collinear limits , can be introduced : @xmath26 @xmath27\ ] ] @xmath28 @xmath29 \ , f_{m } \,.\ ] ] they are defined on the @xmath30-parton phase space , denoted @xmath31 and are subtracted from @xmath32 and added back to @xmath33 after integration over the phase space of the unresolved parton . this procedure , called subtraction method , makes the integrals individually convergent and thus well suited for a monte carlo integration . the construction of the local counterterms is inspired by the well known property of the universal factorization of qcd amplitudes in the soft and collinear limits . the singular structure of an @xmath30-parton squared amplitude for two partons @xmath34 and @xmath35 that become collinear can be expressed as follows : @xmath36 @xmath37 where @xmath38 is an amplitude for @xmath21 on - shell external partons , @xmath39 is an operator acting on the spin part of the amplitude and @xmath40 describes the reduced @xmath21-parton kinematics in the limit where partons @xmath34 and @xmath35 become collinear . the structure of the real - emission contribution can therefore be reduced to the product of a finite born amplitude squared times a divergent , collinear splitting kernel @xmath41 associated with the splitting @xmath42 : @xmath43 where @xmath44 denotes spin correlations . when a parton @xmath35 becomes soft , on the other hand , factorization can be written in the following form : @xmath45 @xmath46 where @xmath47 , @xmath48 are operators acting on the color part of the amplitude and @xmath49 is the soft limit of the kinematical configuration @xmath31 . in the limit where @xmath50 , one can write the real - emission contribution as @xmath51 where the factorization is expressed in terms of @xmath21 soft splitting kernels @xmath52 , one for each external parton , and the symbol @xmath44 denotes color correlations . factorization properties described so far dictate general rules for constructing local counterterms for a given subtraction method . in the first step , a complete set of transformations , which map the original @xmath7 partons phase space , @xmath31 , into a new one , @xmath53 , that describes @xmath21 on - shell partons , needs to be defined . subsequently , a set of splitting functions @xmath54 , matching the behavior of the soft and collinear kernels in the singular limits , needs to be worked out . consequently , the local counterterms take the general form : @xmath55 there is a freedom in defining both mappings and splitting functions away from the singular limits , each choice leading to a different subtraction scheme . the most widespread version is the so - called catani - seymour scheme ( cs ) @xcite , where the local counterterms take the following form : @xmath56 here , each mapping @xmath57 is labeled by three parton indices . for a large number , @xmath21 , of external partons , the number of mappings and matrix elements required by the calculation scales cubically : @xmath58 . for our new nagy - soper subtraction scheme ( ns ) , on the other hand , we can write @xmath59 @xmath60 therefore , each mapping is characterized by two labels @xmath61 only , which implies that @xmath62 , i.e. the number of mappings and subsequent matrix element evaluations is reduced by a factor @xmath21 compared to the catani - seymour case . our particular choice of the mapping and a form of splitting functions for the nagy - soper subtraction scheme can be found in ref.@xcite . let us only mention that the latter are based on the original matrix elements for @xmath63 , @xmath64 and @xmath65 . let us also add here that the general structure of the real - emission contribution to @xmath66 in this scheme takes the following form : @xmath67 @xmath68\ ] ] @xmath69 @xmath70 @xmath71 f_{m } \,,\ ] ] where @xmath72 and @xmath73/@xmath74 correspond to the integrated subtraction terms . more precisely @xmath72 encodes the full soft / collinear structure of the matrix element in the form of single and double poles in @xmath75 , with @xmath76 the number of space - time dimensions , together with a finite part . the @xmath73/@xmath74 operator consists of purely finite pieces coming from the initial - state splitting ( @xmath73 ) as well as from the collinear counterterms ( @xmath74 ) and involves an additional integration over the momentum fraction @xmath77 of an incoming parton after splitting . an important feature of the nagy - soper mapping , that we would like to emphasize , is a change of all spectator momenta at the same time . this fact impacts the way the factorization of the phase space is performed . while in the case of initial - state emission there are no substantial differences with respect to the catani - seymour scheme , the factorization is derived in a slightly different way when the splitting occurs in the final state . the lorentz invariant phase space for a generic final state with @xmath7 partons is organized in terms of recursive splittings @xmath78 @xmath79 @xmath80 @xmath81 @xmath82 where @xmath83 and k is the so - called collective spectator momentum that is built by all spectator momenta @xmath84 moreover @xmath85 the masses of the on - shell final - state partons are denoted by @xmath86 , while @xmath87 is the invariant mass of the collective spectator . one can observe that @xmath88 which follows from the fact that the mapping @xmath89 is a lorentz transformation , and the phase space is lorentz invariant . in the frame where the total momentum @xmath90 is at rest , the two - body phase space can be parameterized in terms of angular variables , @xmath91 where @xmath92 represents the solid angle in @xmath76 dimensions and @xmath93 is the standard kllen function , @xmath94 when the total momentum @xmath90 is at rest , the integral @xmath95 for the two phase space elements @xmath96 and @xmath97 is the same . this implies that the jacobian related to the mapping @xmath98 is given by @xmath99 the phase space for the final - state emission can therefore be written in the fully factorized form @xmath78 @xmath100 where @xmath101 is the measure of the splitting phase space in @xmath76 dimensions . the parameters @xmath102 and @xmath103 are uniquely fixed by setting @xmath104 and are given by @xmath105 therefore , in the singular limit one would simply have @xmath106 , @xmath107 , @xmath108 and @xmath109 . . here @xmath110 is the spectator parton selected to define the azimuthal variable @xmath111.__,scaledwidth=50.0% ] the phase space of the splitting has three degrees of freedom in @xmath112 dimensions . one possible way of parameterization is to use lorentz - covariant scalar products and splitting variables as proposed in ref.@xcite . this simple choice results in compact formulae of the integrated dipoles for massless partons @xcite . applying the same strategy to the fully massive case , however , the kinematical bounds of the splitting become much more complicated and the resulting expressions turn out to be very cumbersome . to keep the final expressions reasonably compact , we have adopted an alternative parameterization . for example , in the case of the final - state emission , for a set of momenta @xmath113 the splitting @xmath114 and the set of momenta @xmath31 had to be constructed out of three parameters that we have called collinear , soft and azimuthal variables . in the singular limit , where @xmath115 , the collinear and the soft variables have to correspond to the relative angle between the nearly - collinear partons , @xmath116 , and the energy of the unresolved parton , @xmath117 . on the other hand , the azimuthal variable , @xmath111 , which is the second angular parameter uniquely fixes the kinematics of the splitting . our reference frame consists of an orthogonal set of axes @xmath118 , where the @xmath119-axis is identified with the spatial direction of the vector @xmath120 , as shown in figure[fig : angles_fin ] . the azimuthal variable is then the angle , which separates the unresolved parton @xmath35 from the the @xmath77-@xmath119 plane . since there is complete freedom in selecting the direction of the @xmath77-axis , we have decided to place it in the plane @xmath120-@xmath110 , where @xmath110 is the momentum of the spectator . in the frame , where the total momentum @xmath90 is at rest , the definition of the soft variable is as follows : @xmath121 it is convenient to divide this soft variable by its kinematically allowed maximum value to get a normalized soft variable @xmath122 : @xmath123 the integration of the splitting phase space in case of the final - state emission runs over the variables @xmath124 \;,\ ] ] @xmath125 \,,\ ] ] @xmath126 \;.\ ] ] the soft and collinear limits correspond to @xmath127 and @xmath128 , respectively . for the calculation of the integrated subtraction terms in the nagy - soper scheme , we have adopted the spin - averaged version of the splitting functions as described in ref.@xcite . our goal is the integration in @xmath129 dimensions over the whole phase space of the splitting . however , as a consequence of the increased complexity of the mapping , a fully analytic evaluation of the integrals turns out to be demanding . thus , as an alternative we have used numerical approaches to integrate over the splitting phase space . more precisely , we have decided to adopt a semi - numerical approach to consider analytic integration when possible , and monte carlo integration otherwise . since the general dependence of the integrands on the azimuthal variable @xmath130 was simple and all the azimuthal integrals could be classified into three groups , we carried out this part of the integration analytically . the dependence on the soft and collinear variables was not as simple and led to more complicated expressions that we treated numerically . further details can be found in the original publication @xcite . we have incorporated the new subtraction method based on the nagy - soper formalism into the helac - dipoles package , preserving at the same time all the optimizations already available in the code . for a detailed description of the package functionalities , we refer to the existing literature @xcite . all elements of the calculation that do not dependent on a specific subtraction scheme , like the born matrix elements and the color correlators , were already provided by the framework of helac - dipoles . this fact has dramatically simplified our implementation . the construction of the nagy - soper subtraction terms is dictated by the form of the splitting functions . they contain generic spinors and polarization vectors , which enables them to treat simultaneously fixed helicities as well as random polarization states . we have provided random polarization sampling as a further option available for the nagy - soper scheme . this is an alternative to the existing random helicity sampling optimization , which uses stratified sampling over the different ( incoherent ) helicity assignments of partons @xcite . the option for the spin sum treatment can be controlled by the user in the configuration file ` dipoles.conf ` as described in the appendix of ref.@xcite . besides random polarization sampling , which is an important speedup in every calculation , we have also provided random sampling over color , or color monte carlo , for the subtracted real radiation part . this functionality has provided an important speedup for matrix elements with a large number of colored external states . the general ideas from @xcite have been adopted , which are also an essential ingredient of the helac-1loop package @xcite . our implementation of the nagy - soper subtraction scheme ( ns ) has been tested and compared to the catani - seymour subtraction scheme ( cs ) for some specific processes . more precisely proton - proton collisions at the lhc with a center - of - mass energy of 8 tev have been considered and the following partonic subprocesses @xmath131 , @xmath132 , @xmath133 , @xmath134 have been studied . they give dominant contributions to the subtracted real emissions at @xmath135 for the corresponding processes @xmath136 , @xmath137 , @xmath138 and @xmath139 . moreover , they represent a high level of complexity and test almost all aspects of the software , as they involve both massive and massless states . we have imposed basic selection cuts on jets @xmath140 which have been defined through the anti-@xmath141 jet algorithm @xcite with radius parameter @xmath142 . the mass of the top quark was set to @xmath143 gev and the bottom quark was considered to be massless . results have been presented for the nlo ct10 parton distribution functions @xcite with five active flavors and the corresponding two - loop @xmath4 . the renormalization and factorization scales were set to the scalar sum of the jet transverse masses @xmath144 where for the top quark @xmath145 and for light jets ( also tagged bottom - jets ) @xmath146 a factor of @xmath147 has been included in the scales for all but the @xmath133 process where @xmath148 has been chosen instead . in the following , a few examples from the comparison of both schemes are given . if not specified otherwise , full summation over all color configurations has been assumed together with random helicity sampling . ._[tab : real - emission - dipoles ] number of catani - seymour ( cs ) and nagy - soper ( ns ) subtraction terms for dominant partonic subprocesses contributing to the subtracted real emission contributions at @xmath135 for the @xmath149 , @xmath150 , @xmath151 and @xmath139 processes at the lhc . the number of feynman diagrams ( fd ) corresponding to the subprocesses is given as well . _ [ cols="<,^,^,^ " , ] differential cross section for @xmath152 at the lhc ( @xmath153 = 14 tev ) in the 5fs as a function of the transverse momentum of the hardest bottom jet . the dash - dotted ( blue ) curve corresponds to the lo and the solid ( red ) curve to the nlo result . the scale choice is @xmath154 . the hashed area represents the scale uncertainty , and the lower panels display the differential k factor . the cross sections are evaluated with the mstw2008 pdf sets._,scaledwidth=55.0% ] differential cross section for @xmath152 at the lhc ( @xmath153 = 14 tev ) in the 4fs and 5fs as a function of the transverse momentum of the hardest bottom jet , normalised to the corresponding inclusive cross section . the scale choice is @xmath155 , the cross sections are evaluated with the 5fs and 4fs mstw2008 pdf sets , respectively._,scaledwidth=55.0% ] within the four - flavor scheme , bottom quarks are treated massive and are not included in the parton distribution functions of the proton . we define the bottom quark mass in the on - shell scheme and use @xmath156gev , consistent with the choice made in the mstw2008 four - flavor pdf @xcite . the central cross section predictions in lo and nlo for @xmath157 using the 4fs mstw2008 @xcite pdf are shown in table[tab:5fs4fs ] , in comparison with the 5fs results . we observe that the bottom mass effects decrease the cross section prediction by @xmath158 at lo and @xmath159 at nlo . the residual scale dependence at nlo is approximately 30% , similar to the 5fs calculation . the difference between the massless 5fs and the massive 4fs calculations has two origins . first , there are genuine bottom mass effects , the size of which depends sensitively on the transverse momentum cut . for @xmath160gev we find a 10% difference between the 5fs and 4fs from non - singular bottom - mass dependent terms . this difference decreases to about @xmath161 for @xmath162gev . second , the two calculations involve different pdf sets and different corresponding @xmath163 . while a 4fs pdf has , in general , a larger gluon flux than a 5fs pdf , as there is no @xmath164 splitting , the corresponding four - flavor @xmath165 is smaller than for five active flavors . for @xmath152 the difference in @xmath163 is prevailing and results in a further reduction of the 4fs cross section prediction by about @xmath166 . this latter difference should be viewed as a scheme dependence rather than a bottom mass effect . in figure[fig : nlo2:4f ] we present the differential distribution in the transverse momentum of the hardest bottom jet , as calculated in the 5fs with massless bottom quarks and in the 4fs with @xmath167gev , normalised to the corresponding inclusive cross section . we find that the difference in the shape of the distributions in the 5fs and the 4fs is very small . to conclude this section , we have presented selected results for the differential cross - sections for @xmath168 at the lhc at the centre - of - mass energy of @xmath169tev@xcite . we find that the higher - order corrections significantly reduce the scale dependence , with a residual theoretical uncertainty of about 30% at nlo . the impact of the bottom quark mass is moderate for the cross section normalisation and negligible for the shape of distributions . the fully differential nlo cross section calculation for the process @xmath170 presented in ref.@xcite provides an important input for the experimental analyses and the interpretation of new physics searches at the lhc . in this section , we will discuss the nagy - soper shower in more detail , including its particular implementation in the c++ library deductor @xcite . among other topics , we will elaborate on the theoretical framework necessary to include quantum interference effects . furthermore , we will point out the inherent ambiguities of the approach . ultimately , we will discuss the matching to fixed order calculations at the next - to - leading order in qcd , and present some results for a non - trivial process : the production of a top - anti - top - quark pair in association with a jet in hadronic collisions . this section is based on @xcite . we start from a generic @xmath171 process , which is defined by two initial state partons @xmath172 and @xmath173 and @xmath174 final state particles . each particle is described by a set of quantum numbers to define the flavor @xmath175 , spin @xmath176 and color @xmath177 of the particle and its momentum @xmath34 . the initial state parton kinematics is described by momentum fractions , @xmath178 and @xmath179 , with respect to the original colliding hadrons , instead of their momenta . thus , a complete parton ensemble can be described by @xmath180 , [ \eta_b ,- f_b , s_b , c_b],\\ [ p_1,f_1 , s_1,c_1], ... ,[p_m , f_m , s_m , c_m ] \ } \ ; .\end{gathered}\ ] ] the state of the parton shower evolution is described by a quantum density matrix @xmath181 , which gives the probability is related to color ordered amplitudes , it may become negative for subleading color configurations . thus , one can not naively interpret @xmath181 as a probability . nevertheless , we use the terminology of statistical mechanics . ] to find a certain parton ensemble @xmath182 . the expectation value of an observable @xmath183 for any final state multiplicity is given by @xmath184 = & \sum_m \frac{1}{m!}\int [ d\{p , f\}_m ] \frac{f_a(\eta_a,\mu_f^2)f_b(\eta_b,\mu_f^2 ) } { 4n_c(a ) n_c(b ) \times \text{flux } } \\ & \quad \quad \quad \times \bra{m(\{p , f\}_m)}f(\{p , f\}_m)\ket{m(\{p , f\}_m ) } \\ = & \sum_m \frac{1}{m!}\int [ d\{p , f\}_m ] tr [ \rho(\{p , f\}_m ) f(\{p , f\}_m ) ] \ ; . \end{split}\ ] ] here the @xmath185 is the symmetry factor for identical particles in the final state and @xmath186 $ ] is the @xmath21 particle phase space measure . @xmath187 are the parton density functions evaluated at the momentum fraction @xmath188 and factorization scale @xmath189 . the factor @xmath190 in the denominator is the spin averaging factor . @xmath191 represents the averaging color factor for the initial state partons , where @xmath192 and @xmath193 . the expectation value of f and the trace in the second line are meant to be a summation over indices in color @xmath44 spin space . the quantum density @xmath181 introduced in the second line , is thus given by @xmath194 it is useful to define basis vectors @xmath195 , such that @xmath196 one can then write the expectation value of the observable @xmath183 as @xmath197 = ( f|\rho ) \ ; .\ ] ] a particularly important observable for the defintion of the parton shower is the total cross section measurement function . it is defined as @xmath198 the shower evolution equation describes the propagation of the quantum density matrix from some initial shower `` time '' , @xmath199 , which represents the hard interaction , to the final `` time '' @xmath200 in the low energy regime . the final shower time @xmath200 characterizes the physical scale at which parton emissions can not be described perturbatively anymore . the definition of shower time @xmath201 is not unique and is explained later . the parton shower evolution will transform a few partons at the matrix element level , to a realistic final state with jets typically made of many partons . after this evolution , a phenomenological hadronization model must be applied . the perturbative evolution itself is described by a unitary operator @xmath202 . the observable @xmath183 , after showering , has the expectation value @xmath197 = ( f|\rho(t_f ) ) = ( f|u(t_f , t_0)|\rho(t_0 ) ) \ ; .\ ] ] the unitarity of the evolution operator is a consequence of the requirement that it should not change the total cross section . thus @xmath203 . the evolution operator can be obtained from a differential equation involving two operators @xmath204 and @xmath205 , corresponding to the concepts of real and virtual corrections respectively : @xmath206u(t , t_0 ) \ ; . \label{diffeq}\ ] ] here @xmath204 describes the emission of a resolved particle , i.e. the momenta , flavor , spins and color configuration will change after its application . @xmath205 describes the unresolved emission and therefore does not alter momentum or flavor configurations . nevertheless it can change color configurations , which will affect further emissions . for convenience of calculations , the virtual operator can be further decomposed into @xmath207 , where @xmath208 is diagonal in color space , while @xmath209 is not . interestingly , the evolution equation takes the same form as the time evolution of a statistical ensemble in liouville space @xmath210 = l\rho(t ) \ ; , \ ] ] where the liouville operator can be identified as @xmath211 $ ] . traditional parton showers are constructed using the large @xmath212 limit . thus , the state is always color diagonal implying @xmath213 . in this case , eq . yields @xmath214 with the sudakov form factor @xmath215 since @xmath205 is diagonal in the traditional approach , @xmath216 is a number and not a matrix in color space . in the general case with non - diagonal @xmath205 , it is not practical to exponentiate a matrix in color space . the idea is , therefore , to exponentiate only the diagonal color part @xmath208 , and treat @xmath209 iteratively as a perturbation . this can be justified by noting that the off diagonal color contributions are always suppressed by a relative factor of @xmath217 compared to the leading color contributions . therefore , the solution of eq . with full color evolution , using the decomposition of @xmath207 , is given by @xmath218 n(\tau , t_0 ) \ ; . \ ] ] with @xmath219 the real evolution operator , @xmath220 , describes the transition from an @xmath21-particle ensemble to an @xmath30-particle ensemble . this is achieved by splitting a chosen parton into two , which would physically correspond to a decay of a slightly off - shell parton . the splitting is constrained by flavor conservation , @xmath221 , and momentum conservation , @xmath222 . the description of the transition is ambiguous , as only the singular limits of amplitudes are uniquely determined in qcd . after emitting a particle , it is necessary to correct the momenta in the event in order to ensure momentum conservation and preserve the on - shellness of all particles . this is done by certain momentum mapping operators , which define momenta and flavors of the new ensemble @xmath223 where @xmath224 . in deductor , a global momentum mapping has been chosen . whenever a particle is emitted , the momentum of all final state particles is affected , see the discussion in section [ sec : subtraction ] . in contrast , e.g. the sherpa @xcite parton shower is based on catani - seymour dipoles @xcite , which have local momentum mappings . in this case a single parton momentum is modified . an explicit description of the original momentum mapping used in deductor can be found either in ref.@xcite or in ref.@xcite . recently , the initial state momentum mapping has been slightly modified . a study @xcite showed , that the generated @xmath225 spectrum in @xmath226 strongly depends on the momentum mapping for initial state parton splittings . for this reason , deductor uses a momentum mapping , which allows for an improved resummation of higher - order corrections @xcite . besides momentum and flavor mapping operators , @xmath204 contains splitting functions which correspond to the factorisation of amplitudes in the soft , collinear and soft - collinear limits . in these limits the amplitude can be written as @xmath227 the operator @xmath228 acts in color and spin space . the behaviour of amplitudes in singular limits translates into a similar behaviour of the density matrix , which can be written more explicitly as @xmath229 where @xmath230 is an operator in color space , while @xmath231 is an splitting operator in spin space . the general prescription to obtain deductor s splitting functions @xmath232 has been summarised in @xcite , whereas the complete set of splitting functions can be found in @xcite . the approximation in eq . can be cast into an operator equation , @xmath233 . the operator @xmath234 describes all possible splittings of the emitter parton @xmath235 . @xmath204 is then defined by the splitting operators @xmath234 at a fixed shower time @xmath236 @xmath237 where the sum runs over all possible emitters @xmath235 . the shower time @xmath238 corresponds to an infrared sensitive scale , discussed in section [ subsec : time ] . the virtual evolution operator , @xmath205 , represents the unresolved virtual corrections . nevertheless its content is fixed due to the unitarity condition of the parton shower . by applying @xmath239 from the left and @xmath240 from the right to eq . one obtains @xmath241 which should be valid for any @xmath242 . parton shower unitarity corresponds to simplified virtual corrections , whose main function is to cancel the divergences of the real corrections . in order to obtain an expression for @xmath205 , we write @xmath243 this equation has an ambiguous solution in color space . we shall not reproduce here the explicit form for @xmath205 used in deductor . it can be found in @xcite . in the previous section , we discussed the real and virtual evolution operators . while they allow for an exact treatment of color , the current implementation of deductor is based on the so - called lc+ approximation @xcite , which amounts to only allowing for color evolution without non - local modifications of the color state . in this case , we identify @xmath244 and @xmath245 . the color diagonal part of the virtual operator will be exponentiated , which yields a parton shower formulation similar to the traditional one . the evolution equation is given by @xmath246 where @xmath247 the color off - diagonal splittings are included perturbatively @xmath248u^{\rm lc+}(\tau , t_0 ) \ ; , \end{gathered}\ ] ] where @xmath249 and @xmath250 . let us investigate the logarithmic accuracy of observables . we suppose that the parton shower is used to calculate an observable @xmath251 which contains large logarithms l of some invariant . for definiteness , we could image this invariant to be the transverse momentum of a gauge boson . then @xmath252 has the form @xmath253 we further suppose that a shower with full color , generated by @xmath254 reproduces all coefficients @xmath255 and @xmath256 correctly . then @xmath257 will reproduce @xmath255 exactly , because the lc+ approximation is exact with respect to the soft - collinear singularities . one insertion of @xmath258 $ ] generates a contribution @xmath259 because it contains a correction for soft wide - angle gluon emission . this term multiplies contributions of order @xmath260 and , therefore , corrects the coefficient @xmath256 . a second insertion of @xmath258 $ ] would only affect the coefficients @xmath261 . therefore , one insertion is sufficient to obtain nll accuracy @xcite . ( an illustration of the logarithm counting is given in figure[fig : log_counting ] . ) this implies that nll accuracy is in fact obtained with the evolution equation @xmath262u^{\rm lc+}(\tau , t_0 ) \ ; .\end{gathered}\ ] ] and a diagonal step is given by an insertion of @xmath263 . one can see , that two insertions of @xmath264 only contribute to the coefficient of @xmath265.,scaledwidth=50.0% ] emissions generated by a parton shower are strongly ordered in some kinematic variable in order to correctly resum the leading logarithms . it turns out that the choice of the ordering variable , which we call shower time , is ambiguous . the essential approximation made in the parton shower description is that , in each step of the evolution , all partons are on - shell . thus , the parton shower time should allow us to neglect the virtuality of a splitting parton . here , we summarize the discussion from ref.@xcite for final state radiation . the discussion for initial state radiation is analogous . consider a mother parton @xmath266 which splits in two partons , @xmath267 . the momentum @xmath268 is given in light - cone and @xmath269 variables @xmath270 , where @xmath271 denotes the transverse momentum . we denote the virtuality of the parton as @xmath272 . the momentum takes the form @xmath273 the daughter momenta are given in the sudakov parameterisation as @xmath274 from momentum conservation we obtain @xmath275 in order to be allowed to neglect @xmath276 and @xmath277 at each step of the evolution , we must require @xmath278 inserting the momentum fraction @xmath119 @xmath279 where the last approximation is valid in the singular limit @xmath280 . here , @xmath281 is the total final state momentum . we finally arrive at the conditions @xmath282 they should always be fulfilled . therefore , we define @xmath283 and enforce the emissions to be ordered in @xmath284 . this can be achieved by defining the dimensionless shower time as @xmath285 generally , one can choose other ordering variables . e.g.pythia 8 @xcite uses the transverse momentum , @xmath225 , in the parton splittings to order the emissions , herwig @xcite uses angualar ordering , and pythia 6 @xcite uses a virtuality ordering . a consequence of the @xmath286 ordering is an enlarged phase space for initial state splittings as compared to @xmath225 ordering @xcite . as we have already pointed out at various places , the construction of the parton shower is not uniquely defined . in the following we list the main ambiguities . they should be kept in mind , since future findings might require modifications of their solutions . momentum mappings : : : the way the parton shower distributes momentum among the particles after splitting may influence the resummation accuracy . in case of drell - yan z - boson production , a study @xcite showed that a different momentum mapping @xcite generates a different @xmath225 spectrum of the z - boson . splitting functions : : : splitting functions are only required to reproduce the singular limits of qcd matrix elements , but can have an arbitrary finite remainder . soft partition function : : : soft emissions are spread on different collinear emitters by means of a partition function . the latter is , however , arbitrary . in @xcite it was argued , that the partition function should not depend on the emitter s energy . a choice , which has not been explored , would be to make the partition function spin dependent . color treatment : : : in deductor , the color diagonal part of the evolution is exponentiated , whereas the off - diagonal part is treated perturbatively . the separation depends on the representation of the color algebra , if the perturbative insertion of the off - diagonal color operator is truncated . in ref.@xcite a different approach for a full color treatment in a parton shower has been proposed . spin treatment : : : the spin basis ( not discussed in this proceedings ) is arbitrary . a different choice could modify the spin weights in the spin evolution of the shower . shower time : : : the parton shower emissions are strongly ordered . as found in ref.@xcite , @xmath286 ordering provides a wider phase space for initial state splittings than ordering in @xmath225 . in many cases , nevertheless , different ordering variables give the same results . a counter example would be , e.g. angular ordering vs. @xmath225 ordering . angular ordering preserves color coherence of soft gluon emissions , whereas @xmath225 ordering can not account for this effect @xcite . pdf evolution : : : traditional parton showers interface lhapdf @xcite to obtain the ratio of pdfs occurring in the backward evolution for initial state radiation . therefore , those parton showers explicitly depend on the pdf kernels they use . deductor tries to minimize this dependence by evolving the pdfs according to the shower splitting functions . since the splitting functions are not uniquely defined , the pdf evolution is also not unique . additionally , if massive quarks are assumed in the initial state , then the mass dependent terms of the splitting kernel depend on the definition of the shower time . matching nlo calculations with parton showers is a widely explored subject and there already exist several matching schemes @xcite . the most popular ones are the powheg method @xcite and the mc@nlo formalism @xcite . a general comparison between those two major schemes can be found in @xcite . we choose to work in analogy to mc@nlo instead of powheg , because we are looking for a general solution , which can be easily automated . before we discuss the problems of the mc@nlo formalism and their solutions , we want to give a brief overview of the general objectives of parton shower matching . this section presents original results obtained in @xcite . independently of the accuracy of the matching we obtain the following benefits : connection to low energy physics : : : inclusive distributions are not affected by showering . nevertheless , the evolution of partons down to a scale @xmath200 allows to include decays of unstable particles , as well as the non - perturbative hadronization and multiple interactions models . logarithmic accuracy : : : infrared sensitive observables , which are ill defined at fixed order , are replaced by finite predictions due to resummation of large logarithms generated by collinear , soft and soft - collinear splittings . matching at the next - to - leading order gives us furthermore : cross section normalization at nlo : : : when considering an inclusive observable @xmath183 we want to keep the fixed order normalization of the cross section . thus , the parton shower must not modify the total cross section @xmath287 \ ; .\ ] ] high-@xmath288 emission according to matrix elements : : : the parton shower is valid in the soft and collinear regime . thus , a parton shower description of high @xmath225 emissions is not reliable . since nlo calculations are indeed valid in this region , one wants to recover the nlo predictions for high @xmath225 emissions after showering . meaningful events : : : matching to parton shower is the only way to define events at nlo . without matching , the weights of the real matrix element and the subtraction terms belong to different kinematics and diverge separately . due to the matching scheme , they are combined and one obtains real emission phase space configurations with a finite , but not necessarily positive , weight . we start our discussion of matching from the quantum density matrix . for a generic @xmath289 process at nlo , one can write it in a perturbative expansion @xmath290 note that we normalize the leading order contribution in the counting of the coupling to be of order @xmath291 . @xmath292 and @xmath293 correspond to tree level matrix elements , whereas @xmath294 to the one - loop amplitude . the definitions of these densities are analogous to the definition given in eq . . based on this quantum density matrix , the observable @xmath183 after showering naively reads @xmath295^{ps } = ( f|u(t_f , t_0)|\rho ) = \\ \sum_{\lambda = m}^{\infty } \frac{1}{\lambda ! } \int [ d\phi_\lambda ] ( f|\phi_\lambda)(\phi_\lambda|u(t_f , t_0)|\rho ) \ ; , \ ] ] where we use the shorthand @xmath296 . the quantum density @xmath240 accounts for the hard matrix elements for @xmath297 . finally @xmath202 describes the parton evolution during showering , as defined in eq . . this naive description of the cross section suffers from double counting , which is demonstrated as follows . the evolution equation of the parton shower has an iterative solution . since , at first , we are only interested in corrections up to @xmath298 , it is sufficient to expand the evolution equation linearly . evolving the density state @xmath240 from @xmath199 to @xmath200 yields @xmath299|\rho_m^{(0)})+ \mathcal{o}(\alpha_s^2 ) \ ; . \ ] ] as we can see from the unitarity condition @xmath300 = 0 $ ] , the total cross section @xmath301 is conserved . on the other hand , for inclusive observables @xmath183 one does not recover the nlo prediction , because in general @xmath302 \neq 0 $ ] . even if we would not insists on recovering the nlo prediction , the result would have to be considered wrong , since it would contain real emission contributions twice : from the real emission quantum density @xmath303 , and from its parton shower approximation @xmath304 . this problem is solved by matching . the solution is slightly simpler for processes , which can be defined without any cuts at the born level , e.g. @xmath305 or @xmath306 . at the end , we will obtain a color and spin correlated version of the original mc@nlo formalism . the additional parton shower contribution can be cancelled by including the same term with an opposite sign in the quantum density state @xmath240 . thus , we can avoid double counting by introducing a modified quantum density state @xmath307|\rho_m^{(0)})+ \mathcal{o}(\alpha_s^2 ) \ ; . \label{rhobar}\ ] ] first notice that @xmath308 is unchanged , due to the unitarity condition . on the other hand , considering @xmath309 and expanding the evolution equation again shows that the undesired parton shower contributions are cancelled up to @xmath298 . notice that this cancellation is non trivial , since the modified quantum density @xmath310 , depends now explicitly on the parton shower splitting kernels and the choice of @xmath199 . let us investigate the expectation value for an infrared safe observable @xmath183 given by the density state in eq . @xmath311 = & \frac{1}{m!}\int [ d\phi_m ] ( f|u(t_f , t_0 ) |\phi_m ) \\ & \times \left [ ( \phi_m|\rho_m^{(0 ) } ) + ( \phi_m|\rho_m^{(1 ) } ) + \int_{t_0}^{t_f } d\tau ( \phi_m|v(\tau)|\rho_m^{(0 ) } ) \right ] \\ + & \frac{1}{(m+1 ) ! } \int [ d\phi_{m+1 } ] ( f|u(t_f , t_0 ) |\phi_{m+1 } ) \\ & \times \left [ ( \phi_{m+1}|\rho_{m+1}^{(0 ) } ) - \int_{t_0}^{t_f } d\tau ( \phi_{m+1}|h_i(\tau)|\rho_m^{(0 ) } ) \right ] \ ; . \end{split}\ ] ] written in this way , the matched cross section suffers from infrared divergences in the virtual @xmath312 and real contributions @xmath303 , which appear in two separate integrals . in the mc@nlo approach one uses the parton shower splitting kernels as subtraction terms , thus one drops the infrared cutoff , which is imposed in the parton shower , and takes the limit @xmath313 . in the subtracted real cross section we can make use of the definition of the real splitting operator in eq . and write @xmath314 here the sum runs over all external legs and @xmath315 is the total splitting kernel for a given external leg . we want to emphasize that @xmath315 also contains completely finite contributions like the massive @xmath316 splitting . @xmath317 is the shower time defined in section [ subsec : time ] . hence @xmath318 represents the ordering of the emissions . the @xmath199 dependence provides a dynamical restriction of the subtraction phase space . the real subtracted cross section is now finite in @xmath112 dimensions , since @xmath317 is allowed to approach infinity and therefore the subtractions terms can resemble the singular limits of the qcd matrix element . integrating the virtual operator @xmath319 without an infrared cutoff is more complex , since there is an explicit integration over the splitting variables . thus , we have to integrate this part in @xmath129 dimensions to extract the @xmath320 and @xmath321 poles analytically . @xmath319 takes the form @xmath322 where @xmath323 is the phase space of the additional parton . the decomposition of the integrated @xmath319 into @xmath324 , which contains all integrated final - state splittings , and @xmath325 which contains the initial - state splittings is arbitrary . we emphasize this structure only to show that the parton shower naturally incorporates a subtraction scheme as in the catani - seymour framework @xcite . the only part which can not be generated by the parton shower are the collinear counterterms , denoted by @xmath74 , needed for the pdf renormalization . thus , the matched cross section reads @xmath311 & = \int \frac{[d\phi_m]}{m ! } ( f|u(t_f , t_0 ) |\phi_m ) \\ \times & \left [ ( \phi_m|\rho_m^{(0 ) } ) + ( \phi_m|\rho_{m}^{(1 ) } ) + ( \phi_m|[\mathbf{i}(t_0 ) + \mathbf{k}(t_0 ) + \mathbf{p}]|\rho_m^{(0 ) } \right ] \\[0.2 cm ] + \int & \frac{[d\phi_{m+1}]}{(m+1 ) ! } ( f|u(t_f , t_0 ) |\phi_{m+1 } ) \\ \times & \left [ ( \phi_{m+1}|\rho_{m+1}^{(0 ) } ) - \sum_l ( \phi_{m+1}|\mathbf{s}_l|\rho_m^{(0 ) } ) \theta(t_l - t_0)\right ] \ ; . \label{masterformula } \end{split}\ ] ] in practice eq . is not solved in a single step . let us define the shorthands ( as in @xcite ) @xmath326|\rho_m^{(0 ) } ) \ ; , \\[0.2 cm ] ( \phi_{m+1}|h ) \equiv\ ; & ( \phi_{m+1}|\rho_{m+1}^{(0 ) } ) - \sum_l ( \phi_{m+1}|\mathbf{s}_l|\rho_m^{(0 ) } ) \theta(t_l - t_0 ) \ ; .\end{aligned}\ ] ] one can use the total cross section @xmath327 = \frac{1}{m!}\int [ d\phi_m ] ( 1|\phi_m ) ( \phi_m|s ) \\ + \frac{1}{(m+1 ) ! } \int [ d\phi_{m+1 } ] ( 1|\phi_{m+1 } ) ( \phi_{m+1}|h ) \ ; , \ ] ] to generate the hard events according to @xmath328 and @xmath329 . the generated events are subsequently interfaced to the parton shower and after the evolution one can investigate the desired observables . thus , after showering one has performed the following integrals @xmath330^{ps } = \frac{1}{m!}\int [ d\phi_m ] ( f|u(t_f , t_0)|\phi_m ) ( \phi_m|s ) \\ + \frac{1}{(m+1 ) ! } \int [ d\phi_{m+1 } ] ( f|u(t_f , t_0)|\phi_{m+1 } ) ( \phi_{m+1}|h ) \ ; .\end{gathered}\ ] ] in the case of processes , which require cuts already at born level to yield a finite cross section , one has to modify the matching prescription eq . . naively , one would simply introduce a generation cut function given by a state @xmath331 as follows @xmath332 applying the parton shower to these ensembles , shows that double counting is still present @xcite . it turns out that it is also necessary to modify @xmath329 to be defined with account of the generation cut in the subtraction phase space @xmath333 where we have introduced the momentum mapping operator @xmath334 with @xmath335 as the inverse transformation to that given by the operator @xmath336 in eq . . it can be shown that this matching prescription is correct as long as the generation cuts are looser than those implied by the final observable , and the latter also only amounts to cuts . the matching scheme of the previous subsection has been implemented in the framework of helac - nlo . in order to test the implementation , we have chosen to study the process of top - quark pair production in association with an additional jet at the large hadron collider . the results reported in this subsection are taken from @xcite . we point out that the nlo qcd corrections to @xmath337 production have been previously obtained in refs.@xcite . furthermore , nlo + parton shower predictions have been studied in ref.@xcite . results for @xmath337 production are given for @xmath338 collisions at the lhc with a center - of - mass energy of 8 tev . the top quark is assumed to be stable and its mass is set to @xmath143 gev , while the bottom quark is considered as massless . we use the nlo mstw2008 pdf set @xcite with five active flavors and the corresponding two - loop running of the strong coupling . we set the renormalization and factorization scale to the top quark mass , @xmath339 , and the starting shower time @xmath199 to @xmath340 , with @xmath341 where @xmath342 is the partonic center - of - mass energy . since the @xmath343 process is divergent already at leading order , we have to impose cuts on the hard jet in the event generation . these cuts have to be as minimal as possible to ensure the inclusiveness of the events before they are passed to the parton shower . we require the reconstructed jets to have @xmath344 in the event generation , and @xmath345 in the final analysis of several observables . jets are clustered using the anti-@xmath141 jet algorithm @xcite , with @xmath346 used at both the generation and analysis levels . only particles with pseudo - rapidity @xmath347 are passed to the jet algorithm . in the parton shower , the top quark is kept as a stable particle ( i.e. no decay allowed ) , hadronization and multiple interactions are not included . in order to address the theory uncertainties we investigated the scale dependence on the unphysical scales @xmath348 and @xmath340 . here @xmath348 is varied between @xmath349 and @xmath350 . whereas the parton shower starting time @xmath199 is varied between @xmath351 and @xmath352 . first we check the total cross section for @xmath337 production . including scale variation of @xmath353 we obtain our final prediction as @xmath354 these results are fully consistent , which proves that the initial cuts during the generation phase were chosen appropriately . additionally we observe that the parton shower does not improve the theory uncertainty of the total cross section at all . this is expected because we removed all parton shower effects for the total cross section by construction . nevertheless , we expect some reduction of the scale dependence for differential distributions , because of the summation of the leading logarithms . several distributions showing a comparison between the fixed order nlo calculation and several nlo+ps predictions ( amc@nlo+pythia8 @xcite and powheg+pythia8 @xcite ) are shown in figures[fig : dist1 ] and [ fig : dist2 ] . @xmath355 in figure[fig : dist2 ] is the scalar sum of the relative transverse momenta of the particles in the first jet , defined with respect to the jet axis in the frame where the first jet has zero rapidity @xmath356 where @xmath357 is the momentum of the @xmath358 particle in @xmath359 . comparison between nlo and several nlo+ps predictions.,title="fig:",width=264 ] + production . comparison between nlo and several nlo+ps predictions.,title="fig:",width=264 ] + production . comparison between nlo and several nlo+ps predictions.,title="fig:",width=264 ] production . comparison between nlo and several nlo+ps predictions.,title="fig:",width=264 ] + production . comparison between nlo and several nlo+ps predictions.,title="fig:",width=264 ] + production . comparison between nlo and several nlo+ps predictions.,title="fig:",width=264 ] we observe that distributions shown in figure[fig : dist1 ] , which are not expected to be affected by showering effects have , to a good approximation , the same shape as at fixed order . the resummation effect is visible first and foremost in the transverse momentum distribution of the @xmath360 system and other distributions shown in figure[fig : dist2 ] . a more detailed analysis is presented in @xcite . we have discussed recent progress towards matching next - to - leading qcd calculations for lhc processes with parton showers at next - to - leading logarithmic accuracy . matched nlo+nll calculations will provide accurate predictions for differential distributions and exclusive observables with experimental cuts , and are thus essential to fully exploit the potential of the upcoming lhc run . to facilitate the matching of nlo calculations with parton showers , the subtraction terms needed to combine virtual and real corrections should by constructed from the splitting functions that define the parton shower . we have presented a subtraction scheme based on a parton shower with quantum interference and its implementation into the helac - nlo software . the new subtraction scheme has been applied to a number of challenging processes , including the production of up to four heavy quarks at the lhc . the new scheme performs well compared to established methods . it not only provides an important internal check of multi - parton nlo calculations , but also forms the basis of current work on parton shower matching at nll accuracy . first results of such a new nlo+nll calculation have been presented . we find that the resummation is important for a wide range of phenomenologically relevant distributions . work is in progress to further improve on the accuracy of the calculation by adding sub - leading color effects and spin correlations . this work was supported by the deutsche forschungsgemeinschaft through the collaborative research centre sfb - tr9 computational particle physics " , and by the u.s . department of energy under contract de - ac02 - 76sf00515 . we would like to thank our collaborators giuseppe bevilacqua , heribertus bayu hartanto , manfred kraus and michael kubocz . mk is grateful to slac and stanford university for their hospitality . y. kurihara , j. fujimoto , t. ishikawa , k. kato , s. kawabata , t. munehisa and h. tanaka , `` qcd event generators with next - to - leading order matrix elements and parton showers , '' nucl . b * 654 * ( 2003 ) 301 . j. alwall , r. frederix , s. frixione , v. hirschi , f. maltoni , o. mattelaer , h .- s . shao and t. stelzer _ et al . _ , `` the automated computation of tree - level and next - to - leading order differential cross sections , and their matching to parton shower simulations , '' jhep * 1407 * , 079 ( 2014 ) . g. corcella , i. g. knowles , g. marchesini , s. moretti , k. odagiri , p. richardson , m. h. seymour and b. r. webber , `` herwig 6 : an event generator for hadron emission reactions with interfering gluons ( including supersymmetric processes ) , '' jhep * 0101 * ( 2001 ) 010 . s. dittmaier , p. uwer and s. weinzierl , `` hadronic top - quark pair production in association with a hard jet at next - to - leading order qcd : phenomenological studies for the tevatron and the lhc , '' eur . j. c * 59 * ( 2009 ) 625 .

it is common , in both theoretical and experimental studies , to separately discuss quark and gluon jets . however , even at parton level , widely - used jet algorithms fail to provide an infrared safe way of making this distinction . we examine the origin of the problem , and propose a solution in terms of a new ` flavour- ' algorithm . as well as being of conceptual interest this can be a powerful tool when combining fixed - order calculations with multi - jet resummations and parton showers . it also has applications to studies of heavy - quark jets . hep - ph/0601139 + bicocca - ft-05 - 28 + cavendish - hep-05/25 + cern - ph - th-06/02 + damtp-2005 - 134 + fermilab - pub-06 - 003-t + lpthe-05 - 34 + january 2006 + * * + andrea banfi , gavin p. salam and giulia zanderighi + _ damtp , centre for mathematical sciences , cambridge cb3 0wa , uk_. + _ cavendish laboratory , university of cambridge , cambridge cb3 0he , uk_. + _ university of milano bicocca and infn sezione di milano , 20126 milan , italy_. + _ lpthe : universit pierre et marie curie paris 6 ; + universit denis diderot paris 7 ; cnrs ; 75252 paris 75005 , france . _ + _ theory group , fermilab , p.o . box 500 , batavia , il , us . _ + _ theory division , physics department , cern , 1211 geneva 23 , switzerland . _ +

hep-ph0601139

a search through the spires database reveals over 350 articles whose titles contain the expressions ` quark jet(s ) ' or ` gluon jet(s ) ' @xcite . the idea of quark and gluon jets appears so intuitive that it hardly seems necessary to examine the question of what it means . yet , when going beyond leading order perturbative qcd , the concept of quark and gluon jets is only meaningful once a procedure has been defined to classify an ensemble of partons into a set of jets , each with a well - defined flavour a flavour that is insensitive to the addition of extra soft or collinear branchings . to our knowledge the question of how to do this in general has not been addressed in the literature . as well as being of intrinsic interest , the question of how to define the flavour of a partonic jet is becoming of increasing practical importance as the study of qcd is extended to multi - jet ensembles ( by jets we mean both incoming and outgoing ones ) : in studies of @xmath9 jets one knows that the basic @xmath10-jet born configuration consists of quark jets ; but for jet production at hadron colliders , the born configuration involves @xmath10 incoming and @xmath10 outgoing jets and many flavour channels are possible , @xmath11 , @xmath12 , @xmath13 , etc . the ability to assign flavours to the jets is especially useful when combining fixed - order predictions with all - order calculations ( be it for parton showers as in @xcite or for analytical resummation @xcite ) . this is because all - order calculations are carried out for a fixed born configuration , with a single flavour channel at a time , while fixed - order calculations implicitly sum over all flavour channels and can at best be split up a posteriori to match onto the individual flavour channels of the all - order calculation . as a concrete example , consider the calculation of higher - order corrections to the process @xmath14 , fig . [ fig:2to3]a . an all - order calculation treats the addition of any number of soft / collinear gluons and extra @xmath15 pairs implicitly , leaving the underlying @xmath16 flavours unchanged . when trying to supplement this with results of a fixed order calculation one encounters the problem that higher - order contributions can not be uniquely assigned to any given @xmath16 flavour channel the @xmath17 corrections to @xmath14 include e.g. a @xmath18 piece , but a fixed order calculation gives only the squared sum of all @xmath19 diagrams , among them @xmath20 and @xmath21 , illustrated in fig . [ fig:2to3]b and [ fig:2to3]c respectively . there can exist no unambiguous procedure for separating the @xmath22 contribution into its different underlying channels , both because the different channels are not individually gauge invariant and because they interfere when squaring the amplitude . flavour channel for a @xmath23 parton scattering process ; ( b ) higher - order diagram that can be seen as a correction to ( a ) ; ( c ) higher - order diagram that can be seen as a correction to the process @xmath24 , but with the same final - state partons as ( b).,scaledwidth=95.0% ] one therefore needs a prescription to assign @xmath22 either to the @xmath25 or the @xmath26 underlying born @xmath23 process ( or else to declare it irreducibly @xmath27 like ) , it only being in the @xmath25 case that one needs to put it together with the @xmath14 all - order calculation . this reclassification of a @xmath28 event as a @xmath23 event is similar conceptually to what is done in a normal jet algorithm , except that not only should the momenta of the resulting @xmath23 configuration be infrared and collinear safe , but so should the flavours . accordingly we call it a jet - flavour algorithm . an obvious approach to defining jet flavours at the perturbative level would be to start with an existing jet algorithm , such as the @xmath0-clustering @xcite or cone @xcite algorithm , that defines jets such that each particle belongs to at most one jet . one can then determine the net flavour content of each of the jets , as the total number of quarks minus antiquarks for each quark flavour . jets with no net flavour are identified as gluon jets , those with ( minus ) one unit of net flavour are ( anti ) quark jets , while those with more than one unit of flavour ( or both a flavour and a different antiflavour ) can not be identified with a single qcd parton . pair ( @xmath29 , @xmath30 ) with the @xmath31 and @xmath32 then clustered into different jets ( @xmath33 , @xmath34 ) . ] applied to the @xmath0 or cone algorithms , this procedure yields a jet flavour that is infrared ( ir ) safe at ( relative ) order @xmath35 discussed in our example above . however at ( relative ) order @xmath36 a large - angle soft gluon can split into a widely separated soft @xmath37 pair and the @xmath31 and @xmath32 may end up being clustered into different jets , ` polluting ' the flavour of those jets , see fig . [ fig : badjets ] . because this happens for arbitrarily soft gluons branching to quarks , the resulting jet flavours are infrared unsafe from order @xmath36 onwards . we are not aware of this problem having been discussed previously in the literature , though there do exist statements that are suggestive of ir safety issues when discussing flavour @xcite . in section [ sec : k_t - flav - algor ] we shall discuss ir flavour unsafety with respect to the @xmath0 ( or ` durham ' ) algorithm in @xmath38 @xcite . there we shall recall that the @xmath0 closeness measure is specifically related to the divergences of qcd matrix elements when producing soft and collinear gluons . however there are no divergences for the production of soft quarks and , as we shall see , it is the use for quarks of a distance measure designed for gluons that leads to the infrared unsafety of jet flavour in the @xmath0 algorithm . by taking into account the absence of a soft - quark divergence when designing the jet - clustering distance measure , one can eliminate the infrared divergence of the jet flavour . the essence of the modification to the @xmath0 distance is that instead of the @xmath39 factor that appears usually , one needs to use @xmath40 when the softer of @xmath41 is a quark . in section [ sec : flav - algor - hadr ] we will examine how this can be extended to processes with incoming hadrons . there the added difficulty is the need for a particle - beam distance measure . traditionally this involves only one dimensionful scale , related to the squared transverse - momentum @xmath42 of the particle . there is a sense in which this can be understood as @xmath43 , where @xmath44 is some transverse scale associated with the beam that is larger than all @xmath42 and so could up to now be ignored . in order to obtain a sensible jet - flavour algorithm we shall however need to consider also @xmath45 and therefore in section [ sec : flav - algor - hadr ] we shall investigate how to construct sensible ` beam scales ' . as well as explaining how to build jet algorithms that provide an infrared safe jet flavour , we shall also examine how they fare in practice . in @xmath38 it will be possible to carry out tests both with an nlo code ( which explicitly reveals the ir unsafety of flavour in traditional jet algorithms ) and with parton - shower monte carlo codes . for hadron - hadron collisions only parton - shower monte carlo tests will be possible because none of the currently available nlo codes provides access to the final - state parton flavour information . the aim of clustering algorithms is to recombine particles into jets in a manner that approximates the inverse of the nearly probabilistic picture of ordered qcd branching . since , however , the branching itself is a quantum mechanical process , there is no unique way of inverting it for a given final ensemble of particles . what can at most be done is to design it to work correctly in limits in which the qcd branching behaves classically , e.g. when a given particle is emitted as if from a single identifiable parent . the design of good jet algorithms is therefore more a craft than a deductive science . nevertheless certain general principles will help us identify how to extend existing jet algorithms to deal properly with flavour . let us start by considering the most widespread clustering algorithm , the standard @xmath38 durham ( or @xmath0 ) algorithm @xcite : 1 . introduce a distance measure @xmath48 between every pair of partons @xmath49 , @xmath50 : @xmath51 where @xmath52 is the energy of particle @xmath49 , @xmath53 is the angle between particles @xmath49 and @xmath50 and @xmath54 is the centre of mass energy . 2 . find the specific @xmath49 and @xmath50 that correspond to the smallest @xmath48 and recombine them according to some recombination scheme ( we shall here use the @xmath55 scheme , which sums the four - momenta ) . 3 . repeat the procedure until all @xmath56 ( or , alternatively , until one reaches a predetermined number of jets ) . the defining characteristic of such clustering algorithms is the distance measure , because it determines the order in which emissions are recombined . it is closely related to the divergences of the qcd matrix elements for a gluon @xmath50 that is soft and collinear to a gluon @xmath49 the product of phase - space and matrix element for a parent gluon to branch to @xmath49 and @xmath50 is @xmath57 |m^2_{g\to g_i g_j}(k_j)| \simeq \frac{{\alpha_s}{c_a}}{\pi } \frac{de_j}{e_j } \frac{d\theta_{ij}^2}{\theta_{ij}^2}\ , , \qquad ( e_j \ll e_i\ , , \;\ , \theta_{ij } \ll 1)\,.\ ] ] thus with the distance measure eq . ( [ eq : yij ] ) , two particles are deemed to be close when either of the parameters in which the matrix element has a divergence , @xmath58 or @xmath53 , is small . this is a key characteristic of a good distance measure because where there is a strong divergence there will be many splittings that are independent of the ` hard ' properties of the event such splittings should be undone ( recombined ) at the early stages of the clustering to leave at the end only well - separated hard pseudo - jets . a second key characteristic of a distance measure can be understood by examining the jade algorithm @xcite , which is identical to ( and predates ) the @xmath0 algorithm , except that its distance measure is @xmath59 again , for @xmath60 , @xmath61 , the distance @xmath62 becomes smaller when either @xmath63 or @xmath53 is reduced , i.e. whenever the matrix - element divergence is made stronger . however it also becomes smaller when @xmath52 is reduced , even though a modification of @xmath52 has no effect on the divergence of the matrix element in eq . ( [ eq : soft - col - mat ] ) . the undesirable consequence of this is that the jade algorithm strongly ` prefers ' to recombine pairs of soft particles at large relative angle , instead of combining the individual soft particles with any collinear but harder neighbours , and so ` pulls ' particles out of their natural jet . from this brief discussion , one can see that the distance measure should satisfy two main characteristics : ( a ) two particles should be considered close when there is a corresponding divergence in their matrix elements ; and ( b ) the measure should not inadvertently introduce ` spurious ' extra closeness for a variation of the momenta that does not lead to any extra divergence ( see however discussion below eq . ) . for generic hadron - level jet studies the durham measure eq . ( [ eq : yij ] ) is a good choice because the majority of emissions are gluons the correct matrix element to consider in the design of the measure is that for soft gluon emission ( be it from a quark or a gluon ) and it always has both a soft ( energy ) and collinear ( angular ) divergence . for flavour algorithms one should remember that the matrix elements for @xmath64 or @xmath65 ( with a soft quark ) have no soft divergence , but just the collinear divergence , @xmath66 |m^2_{g\to q_i\bar q_j}(k_j)| \simeq \frac{{\alpha_s}{t_r}}{2\pi } \frac{de_j}{e_i } \frac{d\theta_{ij}^2}{\theta_{ij}^2}\,,\qquad\quad ( e_j \ll e_i\ , , \;\ , \theta_{ij } \ll 1)\,,\ ] ] ( note the index @xmath49 in the energy denominator ) and analogously for @xmath67 . with the @xmath48 measure , eq . ( [ eq : yij ] ) , a branching that produces a soft quark , @xmath60 , has the same closeness as in the case of the gluon however this closeness is now spurious because , in contrast to the gluon - emission case , there is no divergence for @xmath68 . the replacement of the @xmath63 denominator in the gluon - emission case , eq . ( [ eq : soft - col - mat ] ) , with @xmath52 in the ` soft - quark ' emission case , eq . ( [ eq : soft - col - mat - g2qq ] ) , suggests that the closeness measure for soft @xmath69 branching should become @xmath70 . a similar argument holds in the case of @xmath71 with @xmath60 . thus we should use a distance measure that depends on the flavours of the particles being considered : @xmath72{ll } \max(e_i^2 , e_j^2)\ , , & \quad\mbox{softer of $ i , j$ is flavoured,}\\ \min(e_i^2 , e_j^2)\ , , & \quad\mbox{softer of $ i , j$ is flavourless , } \end{array } \right.\ ] ] where the softer of @xmath41 is the one with the smaller energy and where we use the terms flavoured and flavourless rather than quark - like and gluon - like so as to allow also for situations with diquarks or other multi - flavoured objects . with eq . ( [ eq : yij - flavour ] ) soft - quark ` emission ' leads to no smaller a distance measure than non - soft quark emission , in accord with the absence of a soft divergence for quark emission . furthermore if a quark is to recombine with a harder particle it will favour one that is not too hard , in accord with the presence of @xmath73 in the denominator of eq . ( [ eq : soft - col - mat - g2qq ] ) , which implies that the harder the parent , the less likely it is that it will produce a quark of a given softness . with such a distance measure , for configurations as in figure [ fig : badjets ] the soft @xmath31 and @xmath32 will have similar energies , @xmath74 . thus @xmath75 , whereas @xmath76 . so independently of the precise ( large ) angles of the soft @xmath77 pair , @xmath78 and @xmath79 , it is that soft pair that will recombine first to give a gluon - like pseudo - jet @xmath80 . this will have @xmath81 and now the soft gluon pseudo - jet will recombine with either @xmath82 or @xmath10 ( which one depends on the angles ) and the net flavour of the hard particles will remain unchanged . therefore , at order @xmath36 , our new measure correctly eliminates the soft flavour - changing divergence that exists for the plain durham algorithm . sometimes in the above algorithm a quark can be recombined with another quark or with an antiquark of a different flavour . this can happen for example if there are two large - angle @xmath77 pairs . as long as the resulting ` doubly - flavoured ' object is treated in the same way as a quark in the definition of @xmath83 , the algorithm will remain infrared safe , because in the subsequent clustering steps there will be a strong preference for recombining the multiply - flavoured object with other objects of similar softness , until all soft large - angle multiply - flavoured objects combine between themselves to produce gluon - like objects ( these then recombine normally with the hard partons ) . one may wish to avoid the appearance of multiply - flavour pseudo - jets altogether , since they can not be associated with qcd partons . this can be achieved by vetoing any recombination that would lead to a multiply - flavoured object , i.e. by replacing step @xmath10 with 1 . find the specific @xmath49 and @xmath50 that correspond to the smallest @xmath83 among those combinations of @xmath49 and @xmath50 whose net flavour corresponds either to an ( anti)quark or a gluon , and recombine them . we call this a ` bland ' variant of the jet - flavour algorithm , since ` excessively flavoured ' clusterings are forbidden . we note that a blandness requirement on clusterings has been discussed also in @xcite ( though a simple ` bland ' durham algorithm with the standard @xmath84 remains infrared unsafe ) . an interesting question is that of how much freedom exists in the definition of the distance measure for a flavour algorithm . returning to the analysis of fig . [ fig : badjets ] the main requirement for infrared flavour safety is that the soft fermions @xmath78 and @xmath79 should recombine between themselves before recombining with harder particles . this property is maintained for the following _ class _ of distance measures , @xmath85{ll } [ \min(e_i , e_j)]^{2-\alpha } [ \max(e_i , e_j)]^{\alpha}\ , , & \quad\mbox{softer of $ i , j$ is flavoured,}\\ \min(e_i^2 , e_j^2)\ , , & \quad\mbox{softer of $ i , j$ is flavourless , } \end{array } \right.\ ] ] where @xmath86 is a continuous parameter in the range @xmath87 ( so far we have implicitly discussed @xmath88 ) . above , we stated the requirement that the distance measure should not introduce ` spurious ' extra closeness for a variation of the momenta that does not lead to any extra divergence . here though , for @xmath89 such a spurious extra closeness is present . infrared flavour safety is nevertheless preserved , because the extra closeness is _ weaker _ than that that arises in the case of a divergence , i.e. for a soft gluon @xmath50 , @xmath90 vanishes as @xmath91 , whereas for a soft quark @xmath50 it only vanishes as @xmath92 . naively it would seem that @xmath88 should give the best identification of flavour . however there are situations where a hard quark loses energy through multiple collinear gluon emission and thus becomes a relatively soft quark . in principle there are no large ratios between the quark energy and the softest of the harder gluons it has emitted . however if that gluon is a bit harder than the quark , a value of @xmath93 can make it easier for them to recombine . accordingly below we shall examine both @xmath94 and @xmath95 . events that after jet clustering have their flavour badly identified , i.e. identified as consisting of two gluon jets ( that is , each of zero net flavour ) or two jets each of net flavour larger than 1 ; the coefficient of @xmath96 , as generated with event2 @xcite , is plotted as a function of the durham @xmath97 three - jet resolution threshold ; results are shown for the durham and flavour algorithms ( for two values of @xmath86).,scaledwidth=65.0% ] a rigorous test of jet flavour algorithms can be obtained by numerically investigating the infrared safety of the jet flavour in fixed - order calculations . for example , one generates events @xmath98 together with higher orders and clusters them to two jets . with a jet algorithm that provides a good reconstruction of the flavour , one expects that each of the two jets should have net flavour corresponding to an ( anti)quark . sometimes this does not happen for example each of the two jets may have no net flavour , i.e. be gluon - like . this is legitimate in events in which there has been a hard branching ( there is not a unique clustering to two jets ) , but for an infrared safe flavour jet algorithm , the probability of this happening should vanish in the limit in which there are only soft and collinear emissions . to measure the hardness of a given event we use @xmath99 , the threshold value of the durham jet - resolution below which the event is clustered to three jets of more . figure [ fig : event2 ] shows the differential cross section at next - to - leading order ( nlo , order @xmath36 ) for producing events in which the flavour of the two jets is badly identified . it has been obtained with event2 @xcite , to our knowledge the only nlo code that provides information on the flavour of the final - state partons . one sees that for the durham algorithm the differential cross section for events whose jet flavour does not corresponds to @xmath15 goes to a constant as @xmath100 goes to @xmath101 . this is the sign of the infrared unsafety of flavour identification in the durham jet algorithm . in contrast , in our flavour algorithms ( for both values of @xmath86 ) the corresponding cross section vanishes for @xmath102 . detailed examination of the events with badly identified flavour at small @xmath99 reveals that one of the ( anti)quarks has lost nearly all of its energy to a hard splitting and goes into the same hemisphere as the other quark , i.e. identification of the event as consisting of two gluon jets is actually legitimate . such configurations appear at order @xmath35 where their cross section is @xmath103 . at nlo , sudakov suppression of an extra soft gluon leads to a contribution @xmath104 which is found to be consistent with the observed numerical results , thus confirming the interpretation given above for the origin of the small fraction of @xmath105-like events at small @xmath99 . given that one of the possible applications of jet flavour algorithms is in the merging of matrix - element and parton - shower calculations , we also wish to examine how flavour algorithms behave for monte - carlo generated parton - level ensembles of quarks and gluons . this is interesting for various other reasons too : monte carlo generators produce multiple soft and collinear gluon emissions and @xmath69 splittings , so they are more likely to ` stress - test ' a jet flavour algorithm ; also we can study a much wider variety of processes with them for example one can simulate a fake @xmath106 to examine jet flavour algorithms in a simple gluonic context ; one can also easily use them for studies of hadron - hadron events ( next section ) where currently none of the nlo programs gives direct access to information on the flavour of the outgoing partons . while monte carlo event generators provide considerable flexibility , it can be difficult to interpret their results . for example infrared unsafety of the flavour in fixed - order programs manifests itself as a non - vanishing probability of misidentification of the flavour as @xmath107 . with an event generator one is instead likely to see this probability vanishing with an anomalous dimension , e.g. @xmath108 where @xmath109 is some coefficient ( assuming , for the purpose of the discussion , fixed coupling ) . for an infrared safe jet flavour one expects that for the clustered jets to have a different flavour from the born channel there should have been a hard branching , as in the discussion above for the nlo @xmath38 calculation . this would lead to flavour misidentification vanishing as @xmath110 where @xmath111 is some pure number ( above , @xmath112 ) . this too may however be modified by an anomalous dimension , becoming for example @xmath113 where @xmath114 is some further pure number . quark production in parton distribution functions , known to be enhanced by an all - order double logarithmic series @xcite . such a series might also appear in the jet - flavour case , leading to a more complex modification of the naive @xmath110 behaviour than stated in the main text . ] in the presence of anomalous dimensions it is difficult to establish from monte carlo events exactly which functional form one is seeing . yet another complication is that monte carlo event generators often do not contain the full structure of soft large - angle divergences , so that in any case the anomalous dimensions observed may not correspond to the true ones . jet resolution threshold ; a large value of @xmath54 has been chosen for illustrational purposes , so as to provide a correspondingly large range in @xmath99 ; the left - hand plot shows results for @xmath115 , while the right - hand plot shows fake `` @xmath106 '' process as generated by herwig ( code=107).,title="fig:",scaledwidth=48.0% ] jet resolution threshold ; a large value of @xmath54 has been chosen for illustrational purposes , so as to provide a correspondingly large range in @xmath99 ; the left - hand plot shows results for @xmath115 , while the right - hand plot shows fake `` @xmath106 '' process as generated by herwig ( code=107).,title="fig:",scaledwidth=48.0% ] despite these complications , for an infrared safe jet flavour algorithm one expects flavour misidentification to vanish visibly faster as @xmath107 than for the infrared unsafe case . this signal can be made clearer by going to large @xmath54 so as to have access to a large range in @xmath99 ( note though that a large value of @xmath54 also ` stresses ' the jet flavour algorithm , since it increases the phase space for extra soft @xmath15 production ) . figure [ fig : herwig - ee2xx ] ( left ) shows the fraction of events , for each @xmath99 value , where the flavour has been misidentified in various jet algorithms . it has been generated for @xmath116 gev , using herwig @xcite ( chosen because it provides default access also to a fake @xmath106 reaction , code @xmath117 ) . one sees clearly different @xmath99 dependences for the durham versus the flavour jet algorithms , with the flavour jet algorithm misidentification vanishing considerably more rapidly ( actually as @xmath118 ) . here all the flavour algorithms behave similarly . note also that the bland durham algorithm works considerably better than the plain durham algorithm and only at very small @xmath99 values does one see it doing worse than the flavour algorithms : for the bland algorithm to generate a wrong - flavour event there must be a soft @xmath15 pair of the same flavour as the hard @xmath15 , and additionally the directions of the soft @xmath119 must be such as to lead to jets with net gluon flavour rather than diquark flavour . this situation changes in the right - hand plot of figure [ fig : herwig - ee2xx ] , where we consider fake @xmath106 events . here the bland durham algorithm behaves almost identically to the normal durham algorithm . this is expected , since a soft @xmath119 pair encounters no blandness problems when contaminating the flavour of gluon jets . the flavour algorithms all work systematically better than the durham - based algorithms , clearly vanishing faster with @xmath99 . one sees differences in normalisation between the different flavour algorithms and the blandness requirement provides a non - negligible advantage , especially for @xmath88 . this implies that the flavour misidentification involves more than one @xmath120 pair . nevertheless , the algorithm remains infrared safe even for multiple soft or collinear @xmath119 pairs , as discussed above pairs , spread densely in rapidity from large - angles all the way to the hard - fragmentation region can lead to a systematic worsening of the flavour identification . ] ( see also the appendix for a more general outline of the discussion of ir safety ) . for hadron - hadron collisions ( and dis ) the @xmath0 jet algorithm is similar to that described in section [ sec : k_t - flav - algor ] , with a few modifications in the definition of the distances @xcite . given that there is no unique hard scale @xmath54 , instead of examining dimensionless @xmath90 s one looks at dimensionful @xmath121 s . these need to be invariant under longitudinal boosts and the most widespread convention is to take @xmath122 where @xmath123 , @xmath124 and @xmath125 , @xmath126 and @xmath127 are respectively the transverse momentum , rapidity and azimuth of particle @xmath49 , with respect to the beam . a particle @xmath49 can also recombine with the beam and here too one needs a distance measure , usually taken to be @xmath128 it is the smallest of the @xmath129 and the @xmath121 that determines which recombination takes place . if it is @xmath129 that is smallest at a given step , then @xmath49 recombines with the beam ( or else gets called a jet , in the `` inclusive '' version of the algorithm ) . the modification of the @xmath121 needed to obtain a flavour - safe jet algorithm is directly analogous to that used for the @xmath38 algorithm : @xmath130{ll } \max(k_{ti}^2 , k_{tj}^2)\ , , & \quad\mbox{softer of $ i , j$ is flavoured,}\\ \min(k_{ti}^2 , k_{tj}^2)\ , , & \quad\mbox{softer of $ i , j$ is flavourless , } \end{array } \right.\ ] ] where by ` softer ' we now mean that having lower @xmath0 and where temporarily , for simplicity , we consider only the case @xmath88 . it is less obvious how to modify the beam distance . the problem is that @xmath129 involves just a single scale , @xmath42 , and so there is no `` minimum '' that one can replace with a `` maximum '' . however one could imagine that @xmath129 is actually the minimum of @xmath42 and some transverse scale associated with the beam , @xmath44 , which has never been explicitly needed so far because it was always larger than any of the @xmath42 . the analogue of eq . ( [ eq : dij - flavour ] ) would then be to take @xmath131{ll } \max(k_{ti}^2 , k_{tb}^2)\ , , & \quad\mbox{$i$ is flavoured,}\\ \min(k_{ti}^2 , k_{tb}^2)\ , , & \quad\mbox{$i$ is flavourless . } \end{array } \right.\ ] ] the question that remains is how to define @xmath132 . a first issue is that we will want to identify the flavour of each of the incoming beams . so whereas for the normal @xmath0 algorithm one recombines particles with ` the beams ' , here we will need to specify _ which _ of the two beams a particle recombines with . therefore we will need to define @xmath132 for the beam moving towards positive rapidities ( right ) and @xmath133 for the other beam . in line with the dglap idea @xcite of logarithmic ordering , such that harder emissions are at successively larger angles with respect to the beam that produced them , it makes sense for the beam hardness to be a function of rapidity , @xmath134 . in the definition of @xmath129 , eq . ( [ eq : dib - flavour ] ) , one would then use @xmath135 . for the right - moving ( positive rapidity ) beam , one scale that appears naturally is ( with @xmath136 ) , @xmath137 i.e. the beam scale should be at least as hard as all emissions that have already occurred from that beam ( i.e. all emissions that are at larger rapidity ) . another scale that arises is @xmath138 when one performs a sudakov decomposition of all momenta @xmath139 ( @xmath140 and @xmath141 ) , in the massless approximation , this scale is just the sum of the @xmath142 components of all particles that are still to be emitted by this beam ( i.e. are at smaller rapidity ) . it is equivalent to the light - cone momentum still left in the beam . this scale depends on the reference frame , but can be transformed into a boost invariant , local ` transverse ' hardness by multiplying it by @xmath143 , giving algorithm for dis and hadron - hadron collisions @xcite , the beam distance is @xmath144 . replacing @xmath145 with the effective beam energy @xmath146 ( i.e. taking the larger of @xmath52 and the effective beam energy ) and taking the small - angle limit gives precisely @xmath147 . ] @xmath148 by adding the two measures , @xmath149 and @xmath150 for the beam scale , one obtains an overall beam hardness measure , @xmath151 that takes into account both emissions that have already occurred at a certain rapidity ( in the picture of ordering of emissions ) and those that will occur further on . similarly one defines a scale for the other beam @xmath152 in the same way that one updates the @xmath121 and @xmath129 after each clustering , one should update also the @xmath132 and @xmath133 . and @xmath133 for a multi - jet parton - level lhc event , generated by herwig ; also shown is the histogram of the rapidity distribution of transverse momenta.,scaledwidth=60.0% ] to illustrate the properties of @xmath132 and @xmath133 , fig . [ fig : ktb - event ] shows these two quantities for a typical multi - jet lhc event ( represented as a histogram of total transverse momentum per bin of rapidity ) . towards positive rapidities , @xmath134 decreases as @xmath143 , while @xmath153 approaches a constant , so that as is natural , positive - rapidity particles combine with @xmath154 , while negative rapidity particles combine with @xmath155 . at the point where @xmath132 and @xmath133 cross , they are of the same order of magnitude as the total transverse momentum in the event , i.e. its overall hardness . note also that @xmath134 and @xmath153 are always at least as hard as the hardest emission at rapidity @xmath156 . let us now summarise the jet flavour algorithm for hadron - hadron collisions : 1 . introduce a distance measure @xmath157 between every pair of partons @xmath49 , @xmath50 : @xmath158{ll } \max(k_{ti } , k_{tj})^\alpha \min(k_{ti } , k_{tj})^{2-\alpha}\ , , & \quad\mbox{softer of $ i , j$ is flavoured,}\\ \min(k_{ti}^2 , k_{tj}^2)\ , , & \quad \mbox{softer of $ i , j$ is flavourless , } \end{array } \right.\ ] ] as well as distances to the two beams , @xmath159{ll } \max(k_{ti } , k_{tb}(\eta_i))^\alpha \min(k_{ti } , k_{tb}(\eta_i))^{2-\alpha } \ , , & \quad\mbox{$i$ is flavoured,}\\ \min(k_{ti}^2 , k_{tb}^2(\eta_i))\ , , & \quad\mbox{$i$ is flavourless , } \end{array } \right.\ ] ] and an analogous definition of @xmath160 involving @xmath161 instead of @xmath135 ( both defined as in eqs . ( [ eq : ktb ] ) and ( [ eq : ktbbar ] ) ) . ) and ( [ eq : ktbbar ] ) have been constructed by considering situations with just massless partons . however , their definition can be extended to cases with massive particles in the final state by replacing @xmath125 with @xmath162 . notice that any heavy non - qcd particles should also be included in the sums ( [ eq : ktb ] ) and ( [ eq : ktbbar ] ) , even if they do not enter the clustering . in dis , in the breit frame , @xmath134 should include an additional contribution related to the virtual photon , given by @xmath163 , while @xmath164 should have an additional contribution @xmath165 , where @xmath54 is the photon virtuality . ] as in section [ sec : k_t - flav - algor ] we have introduced a class of measures , parametrised by @xmath166 . 2 . identify the smallest of the distance measures . if it is a @xmath167 , recombine @xmath49 and @xmath50 ; if it is a @xmath168 ( @xmath160 ) declare @xmath49 to be part of beam @xmath154 ( @xmath155 ) and eliminate @xmath49 ; in the case where the @xmath168 and @xmath160 are equal ( which will occur if @xmath49 is a gluon ) , recombine with the beam that has the smaller @xmath135 , @xmath169 . 3 . repeat the procedure until all the distances are larger than some @xmath170 , or , alternatively , until one reaches a predetermined number of jets.@xmath171 algorithm to a geometrical nearest neighbour problem @xcite to reduce its computational complexity to @xmath172 , it is worth commenting that the simultaneous use here of both @xmath173 and @xmath174 invalidates the lemma of @xcite that was central in making the connection with a nearest neighbour problem . it is therefore not clear whether it would be possible to write the flavour algorithm such that its complexity goes as @xmath175 . the implementation that we use has a complexity that scales roughly as @xmath176 . ] in the ` bland ' variant of the algorithm one considers only those @xmath121 for which the product of the recombination would have at most one flavour . similarly one considers only a subset of the @xmath129 in this case the blandness requirement is imposed on the flavour of the parton entering the hard interaction , or equivalently on the difference between the flavour of the incoming hadron and the flavour contained in the outgoing beam jet . the infrared safety of this algorithm follows from the same arguments that were used in the @xmath38 context . the beam scales simply ensure that @xmath15 pairs that are soft but separated by @xmath177 recombine with each other before recombining with the beam . this eliminates the potentially dangerous situation that would otherwise occur , in which first the @xmath31 recombines with one beam and then the @xmath32 recombines with the other beam . therefore it is not just the flavours of the outgoing jets that are infrared and collinear safe , but also those of the incoming beam jets ( the determination the beam - jet flavours of course also requires knowledge of the incoming parton flavours ) . a concrete demonstration of the infrared safety of the hadron - hadron algorithms , analogous to figure [ fig : event2 ] for @xmath38 , is not possible with currently available tools , because none of the higher - order nlo jet codes @xcite provide direct access to information about final - state flavour . even if they did , there would be an additional complication compared to @xmath38 . in @xmath38 at born level , there is only one flavour channel , i.e. @xmath178 . therefore one could identify flavour infrared unsafety by examining , for example , the 3-jet nlo cross section for jets classified as @xmath105 . in hadron - hadron collisions all flavour channels are present at born order , therefore to verify the infrared safety of , say , the @xmath179 channel one must supplement the nlo 2 + 3 jet calculation with the @xmath180 born contribution and its two loop corrections , i.e. one must carry out a nnlo 2 + 2 jet calculation , which is beyond today s technology . fortunately an alternative method exists for verifying the ir safety of flavour identification using just a nlo 2 + 3 jet calculation , namely by examining the cross section for doubly - flavoured jets , since these do not appear at born level , but are infrared unsafe in the plain @xmath0 algorithm . we hope that flavour information will soon become available in @xmath181 jet nlo codes , making it possible to demonstrate this explicitly . in the absence of any way of obtaining a fixed - order illustration of the infrared safety of the flavour algorithms , we resort to investigations of reconstruction of the flavour in parton - level monte carlo events . this is achieved by comparing , event - by - event , the flavours in the hard @xmath16 partonic scattering with those of the beam and outgoing jets after clustering of the event to @xmath182 jets . since the normal @xmath0 algorithm does not usually distinguish between the two beams , we extend it ( both normal and bland variants ) such that a particle destined to recombine with ` the beams ' is assigned to that with the smaller of @xmath183 and @xmath184 . for three channels ( in the case of @xmath185 for both herwig @xcite and a recently developed parton shower algorithm in pythia @xcite ) ; lhc kinematics are used and the events selected are those where the hardest @xmath0-algorithm jet has a transverse momentum greater than @xmath186 and where the two hardest jets have @xmath187 . the range of most common values of @xmath188 depends on the subprocess but is typically roughly @xmath189.,title="fig:",scaledwidth=47.0% ] for three channels ( in the case of @xmath185 for both herwig @xcite and a recently developed parton shower algorithm in pythia @xcite ) ; lhc kinematics are used and the events selected are those where the hardest @xmath0-algorithm jet has a transverse momentum greater than @xmath186 and where the two hardest jets have @xmath187 . the range of most common values of @xmath188 depends on the subprocess but is typically roughly @xmath189.,title="fig:",scaledwidth=47.0% ] + for three channels ( in the case of @xmath185 for both herwig @xcite and a recently developed parton shower algorithm in pythia @xcite ) ; lhc kinematics are used and the events selected are those where the hardest @xmath0-algorithm jet has a transverse momentum greater than @xmath186 and where the two hardest jets have @xmath187 . the range of most common values of @xmath188 depends on the subprocess but is typically roughly @xmath189.,title="fig:",scaledwidth=47.0% ] for three channels ( in the case of @xmath185 for both herwig @xcite and a recently developed parton shower algorithm in pythia @xcite ) ; lhc kinematics are used and the events selected are those where the hardest @xmath0-algorithm jet has a transverse momentum greater than @xmath186 and where the two hardest jets have @xmath187 . the range of most common values of @xmath188 depends on the subprocess but is typically roughly @xmath189.,title="fig:",scaledwidth=47.0% ] + the proportion of events where the original and reconstructed @xmath23 flavours do not match is shown in fig . [ fig : hhxx2yy ] , as a function of @xmath190 . here @xmath191 is the threshold value of @xmath170 below which the event is clustered to 3 or more jets in the standard exclusive longitudinally - invariant @xmath0 algorithm @xcite ; @xmath192 and @xmath193 are the transverse energies of the two last jets to be recombined with the beam if there is no @xmath170 @xcite ( equivalently the two hardest jets when running the inclusive @xmath0 algorithm @xcite ) . we consider simulated lhc events and require the hardest jet to have a transverse energy larger than @xmath186 and the two hardest jets to have @xmath194 . three representative channels , @xmath195 ( including @xmath196 ) , @xmath24 and @xmath185 are shown in fig . [ fig : hhxx2yy ] , as obtained with herwig @xcite . the standard parton showering in pythia @xcite gives similar results ( with a slightly higher normalisation ) . we also illustrate the @xmath185 channel using the recently developed transverse momentum ordered shower in pythia @xcite . in all cases one sees that the rate of flavour misidentification falls significantly more rapidly towards small @xmath188 for the flavour algorithms than for the normal @xmath0 algorithm or its bland variant . algorithm sometimes behaves worse than the normal @xmath0 algorithm ( e.g. for @xmath24 ) . to see why this happens , consider a beam corresponding to incoming @xmath197 flavour , together with a soft collinear @xmath198 pair . in the normal @xmath0 algorithm , the @xmath197 and @xmath199 can recombine with the beam in any order . in the bland variant the @xmath199 is prevented from recombining first ( because the parton entering the reaction would then implicitly have @xmath200 flavour ) and if it has the lower @xmath201 it will instead try to recombine with the other ( wrong ) beam . therefore the bland algorithm actually has an extra source of infrared - collinear flavour unsafety relative to the plain @xmath0 algorithm . ] this is similar to what was observed for @xmath38 in section [ sec : k_t - flav - algor ] , and is a sign of the infrared safety of the flavour algorithms . one notes that for all algorithms the fall - off is less rapid in the hadron - hadron case than in @xmath38 . this is natural given the increased number of jets and therefore of sources of radiation which can lead to extra flavour in the final state . another difference compared to @xmath38 is that now the @xmath94 flavour algorithms sometimes fare better than the @xmath88 case . this is not systematic and also depends on the monte carlo program used to generate events ( compare figs . [ fig : hhxx2yy]c and [ fig : hhxx2yy]d ) . the overall normalisation of the curves also depends on the monte carlo program used and one sees that pythia with transverse - momentum ordered showers produces parton - level final states in which it is systematically harder to cluster back to the original flavour . we have shown in this article that it is possible to define parton - level jets in a manner that ensures that their flavour is infrared safe . the key ingredient in doing this was a modification of the @xmath0 distance measure , inspired by the different structures of divergences that appear in quark production and gluon production . in the case of hadron - hadron collisions it was also necessary to introduce the concept of a hardness associated with the beam at any given point in rapidity . where possible , explicit nlo verifications confirm the infrared safety of the new ` flavour ' jet algorithms . parton - level monte carlo studies also indicate a significant improvement in the identification of flavour relative to the @xmath0 algorithm . to make use of our new algorithms to accurately study jet flavour , it is necessary to have access to information about the flavour of final - state partons in nlo jet codes . currently however , most nlo jet codes have been designed assuming that the user has no need for information about final - state parton flavour ( an exception is event2 @xcite ) . in light of the developments presented here , we look forward to flavour information being made available in the future ( e.g. @xcite ) . our original motivation for studying the problem of jet flavour was the need to accurately combine resummed predictions for hadron - collider dijet event shapes @xcite with corresponding fixed - order predictions @xcite . another simple flavour - related study would be the investigation of how the relative fractions of quark and gluon jets at hadron colliders are modified by nlo corrections and how they vary with jet transverse momentum . apart from its intrinsic interest , such information could be of relevance also to the tuning of monte carlo event generators and studies of hadron multiplicities in jets , both of which are sensitive to the proportions of quark and gluon jets . one drawback of the algorithms presented here is that , when considering light flavours , they can only be applied to partonic and not hadronic events . this is because at each recombination they require knowledge of which objects are flavoured ( quark - like ) rather than flavourless ( gluon - like ) and that information is not present in hadronic final states .. it would be interesting to find a jet algorithm based purely on particle momenta , that nevertheless provides a good infrared - safe determination of the flavour at parton level . it is not clear to what extent this is possible . at @xmath202 . however it has numerous other drawbacks , which we suspect are part of the reason why in monte carlo studies we find that its flavour identification properties are no better than those of the ( flavour ) infrared unsafe durham algorithm . ] there is nevertheless one hadron - level context in which this article s flavour algorithms could be used directly , that is for heavy - quark jets @xcite . currently a heavy - quark jet is defined as a jet containing one or more heavy quarks ( or heavy - quark hadrons ) . the fraction of jets of transverse energy @xmath203 containing a heavy quark of mass @xmath204 is enhanced by terms @xmath205 for @xmath206 , due to the large multiplicity @xmath207 of gluons above scale @xmath204 , combined with the possibility that they split collinearly @xmath208 , responsible for a further factor @xmath209 @xcite . therefore , at high @xmath203 the majority of so - called heavy - quark jets are not jets induced by a heavy quark , but rather jets in which a heavy quark has appeared from the internal branching in the jet . this implies that the current definition of heavy - quark jet will lead to large qcd backgrounds in searches for new particles which aim to tag an ` intrinsic ' heavy quark jet among the decay products of the new particle . an alternative approach to the study of heavy quark jets would be to consider the _ net _ heavy flavour of jets , pairs @xcite . ] i.e. the number of heavy quark hadrons minus heavy antiquark hadrons in a given jet . with the cone or @xmath0 algorithms such a definition would eliminate nearly all the final - state logarithmically enhanced terms , leaving just @xmath210 contributions ( involving a final - state bfkl - type resummation @xcite ) . these remaining terms come from the same diagrams that led to the infrared unsafety of light flavour of a jet . they can therefore be eliminated altogether by applying our flavour jet algorithm with the minor modification that every occurrence of `` flavour '' is to be replaced with `` heavy flavour '' . in this way it becomes possible to give meaning to a concept of intrinsic heavy flavour , i.e. heavy flavour that originates exclusively from the heavy - flavour component of parton distribution functions , from hard qcd flavour `` creation '' ( e.g. @xmath211 ) and from the decay of other heavier particles . we look forward to future phenomenological investigation of this concept . all three of us are grateful to the milano bicocca theoretical physics group for hospitality during the initial stages of this work . gps also wishes to thank fermilab , the cavendish laboratory and cern for hospitality . we thank yuri dokshitzer for several discussions and suggestions , matteo cacciari and giuseppe marchesini for comments , bryan webber for bringing to our attention the original use of the ` bland durham ' algorithm in ckkw matching , michael seymour for assistance with event2 , peter skands for help with pythia , and william kilgore and zoltan nagy for information about the accessibility of flavour information in hadron - collider nlo jet codes . this work was supported in part by grant anr-05-jcjc-0046 - 01 from the french agence nationale de la recherche . the arguments for the infrared safety of our jet flavour algorithms , as discussed in section [ sec : k_t - flav - algor ] , applied only to the case of one or two extra soft @xmath119 pairs . here we give an outline of a general all - order discussion of infrared and collinear safety of the flavour . it will be framed in the context of @xmath38 collisions , and then in closing we will briefly mention hadronic collisions . for a general discussion of the infrared and collinear safety of flavour one needs to examine all divergent cases in which flavour is either produced or moved from one part of the event to another . production of flavour arises from gluon splitting . this has just a collinear divergence ; additionally the gluon itself has soft and collinear divergences with respect to other quarks and gluons . flavour can ` move ' during the branching process when a quark recoils due to emission of a gluon of similar hardness to the quark . this has no divergences , but there may be divergences associated with the original production of the quark itself . flavour can also move during the jet - clustering procedure whenever a quark recombines with a parton that is not collinear to it and whose momentum is of the same order of magnitude as ( or larger than ) the quark . let us first consider flavour production by collinear splitting of a gluon . the durham algorithm always recombines collinear particles into the same jet . since in @xmath69 splitting there is no soft divergence , the @xmath31 and @xmath32 have commensurate hardnesses . therefore the ` flavour ' distance measure eq . ( [ eq : yij - flavour ] ) is of the same order of magnitude as the durham distance measure and so the @xmath15 from a collinear splitting of a gluon will end up in the same jet also in the flavour algorithm , leaving the jet flavour unchanged as is required for irc safety of the flavour . next we consider non - collinear splitting of a gluon into @xmath119 . this has divergences when the original gluon is collinear to some other parton and/or soft . if the gluon itself is collinear to some other parton @xmath212 , angle @xmath213 , then the gluon splitting to @xmath119 is strongly suppressed unless @xmath214 , i.e. non - collinear splitting is not possible from a gluon that is collinear to some other parton . this is the basis of the widely used angular ordering approximation . therefore a @xmath119 produced from a collinear ( and optionally soft ) gluon will always recombine , in the flavour algorithm as in the durham algorithm , ensuring the safety of the flavour of any resulting jet . this leaves the case discussed already in the main text , in which a large - angle @xmath119 pair is produced from a large - angle soft gluon . we have already presented the arguments that explain the ir unsafety of the durham algorithm in this case and the ir safety of the flavour algorithms . in generalising the analysis to higher orders one needs also to examine potential ` motion ' of the soft large - angle @xmath31 and @xmath32 . it will be useful to introduce the compact notation @xmath215 for the set of distance measures @xmath216 . firstly the quark ( or anti - quark ) can itself emit a large - angle gluon of similar softness ( @xmath217 ) , fig . [ fig : badjets - app ] ( left ) . this will change the direction of the quark ( @xmath30 ) . in the durham algorithm , each of the @xmath218 is of the order of the soft gluon @xmath219 , and the recombination sequence depends significantly on the angles . in particular the emission of @xmath217 from the quark may have moved it further away from the antiquark making it more likely that the soft @xmath15 end up in different jets . in contrast , in the flavour algorithm @xmath220 are of order @xmath82 , whereas @xmath221 and @xmath222 are of order of the soft gluon @xmath219 . therefore @xmath78 , @xmath79 and @xmath223 will all recombine together first , or @xmath223 will recombine with the hard jets and then @xmath78 and @xmath79 will recombine together . in both cases the flavour of the soft quarks is neutralised . the analysis of the right - hand diagram of figure [ fig : badjets - app ] is largely similar as long as @xmath217 is at large angles and of the same hardness as @xmath29 and @xmath30 . the additional issue is that now @xmath217 has a collinear divergence with respect to @xmath34 . one might generally worry that semi - hard radiation collinear to @xmath34 might pull @xmath29 far away from its original direction . this could happen if @xmath217 is collinear to @xmath34 and if @xmath29 and @xmath217 recombine , with @xmath224 such that the recombination product ends up collinear to @xmath34 . however if @xmath225 then @xmath226 and the @xmath29@xmath30 will recombine first , neutralising the flavour . note that if @xmath227 and @xmath217 is collinear to @xmath34 then the @xmath228 recombination will occur first , leaving the usual ( safe ) configuration consisting of a soft @xmath15 pair . one can straightforwardly extend this analysis to multiple @xmath15 pairs and multiple gluons . the originally soft large - angle quark can be dragged further and further towards the hard jet in ensembles with multiple gluons of similar @xmath0 s but successively larger ( but not strongly ordered ) energies . however , given any fixed number of recombinations , in the soft limit the resulting quark - like object always has energy @xmath229 and will recombine with the soft antiquark rather than with the hard particles . a final comment concerns hadron - hadron collisions . there , the beam jets have a hardness @xmath134 , which is of the same order of magnitude as any hard final - state jets that might have been emitted at the rapidity @xmath156 . therefore there is no difference from the point of view of irc safety between recombination into final - state jets and into beam jets and all the arguments given here apply equally well in the hadron - hadron context . it would be excessive to cite them all here . instead the reader may examine them by submitting the spires search `` ` find t jet # and ( t quark or t gluon ) ` '' and then extracting from the results only those articles whose title contains `` quark jet(s ) '' or `` gluon jet(s ) '' . s. catani , f. krauss , r. kuhn and b. r. webber , jhep * 0111 * ( 2001 ) 063 [ hep - ph/0109231 ] . n. kidonakis , g. oderda and g. sterman , nucl . b * 531 * , 365 ( 1998 ) [ hep - ph/9803241 ] . r. bonciani , s. catani , m. l. mangano and p. nason , phys . b * 575 * ( 2003 ) 268 [ hep - ph/0307035 ] . a. banfi , g. p. salam and g. zanderighi , jhep * 0503 * ( 2005 ) 073 [ hep - ph/0407286 ] . s. catani , y. l. dokshitzer , m. olsson , g. turnock and b. r. webber , phys . b * 269 * , 432 ( 1991 ) . s. catani , y. l. dokshitzer , m. h. seymour and b. r. webber , nucl . b * 406 * , 187 ( 1993 ) . s. d. ellis and d. e. soper , phys . d * 48 * , 3160 ( 1993 ) [ hep - ph/9305266 ] . see e.g. f. abe _ et al . _ [ cdf collaboration ] , phys . d * 45 * , 1448 ( 1992 ) ; g. c. blazey _ et al . _ , hep - ex/0005012 . z. nagy and d. e. soper , jhep * 0510 * ( 2005 ) 024 [ hep - ph/0503053 ] . y. l. dokshitzer , g. d. leder , s. moretti and b. r. webber , jhep * 9708 * , 001 ( 1997 ) [ hep - ph/9707323 ] ; m. wobisch , ph.d . thesis , rwth aachen , 1999 , desy - thesis-2000 - 049 . w. bartel _ et al . _ [ jade collaboration ] , z. phys . c * 33 * ( 1986 ) 23 ; s. bethke _ et al . _ [ jade collaboration ] , phys . b * 213 * ( 1988 ) 235 . s. catani and m. h. seymour , nucl . b * 485 * ( 1997 ) 291 [ erratum - ibid . b * 510 * ( 1997 ) 503 ] [ hep - ph/9605323 ] ; phys . b * 378 * ( 1996 ) 287 [ hep - ph/9602277 ] . g. marchesini , b. r. webber , g. abbiendi , i. g. knowles , m. h. seymour and l. stanco , comput . commun . * 67 * ( 1992 ) 465 . g. corcella _ et al . _ , jhep * 0101 * ( 2001 ) 010 [ hep - ph/0011363 ] . gribov and l.n . lipatov , sov . . phys . * 15 * ( 1972 ) 438 ; + g. altarelli and g. parisi , nucl . b * 126 * ( 1977 ) 298 ; + yu.l . dokshitzer , sov . jetp * 46 * ( 1977 ) 641 . s. catani , y. l. dokshitzer and b. r. webber , phys . b * 285 * ( 1992 ) 291 . m. cacciari and g. p. salam , hep - ph/0512210 . w. b. kilgore and w. t. giele , phys . d * 55 * ( 1997 ) 7183 [ hep - ph/9610433 ] . z. nagy , phys . * 88 * , 122003 ( 2002 ) [ hep - ph/0110315 ] ; phys . d * 68 * , 094002 ( 2003 ) [ hep - ph/0307268 ] . t. sjostrand and p. z. skands , eur . j. c * 39 * ( 2005 ) 129 [ hep - ph/0408302 ] . g. grunberg , y. j. ng and s. h. h. tye , phys . d * 21 * ( 1980 ) 62 ; for recent work see s. frixione and m. l. mangano , nucl . b * 483 * , 321 ( 1997 ) [ hep - ph/9605270 ] and b. abbott _ et al . _ [ d0 collaboration ] , phys . rev . * 85 * ( 2000 ) 5068 [ hep - ex/0008021 ] .

in this article we study the geodesic motion of test particles and light in the five - dimensional ( rotating ) black string spacetime . if a compact dimension is added to the four - dimensional schwarzschild or kerr spacetime , the new five - dimensional metric describes a ( rotating ) black string . the geodesics in the schwarzschild and kerr spacetime have been studied in great detail , however , when a compact dimension is added new behaviour occurs . we present the analytical solutions of the geodesic equations and discuss the possible orbits . the motion in the ordinary four - dimensional schwarzschild and kerr spacetime is compared to the motion in the ( rotating ) black string spacetime .

1303.6863

in 1916 schwarzschild introduced the first exact solution of the vacuum einstein equations in four dimensions @xcite . this unique solution describes a spherically symmetric black hole . about fifty years later , in 1963 , the solution of a rotating black hole was discovered by kerr @xcite . if a compact dimension is added to the schwarzschild metric , it describes a five - dimensional black string . analogously the kerr spacetime turns into the spacetime of a rotating black string if a compact dimension is added . the idea of compact dimensions came up in the 1920s after kaluza suggested to add a fifth dimension to general relativity @xcite . this was one of the first attempts to unify gravity and the electromagnetic forces . in 1926 klein proposed a physical interpretation @xcite : the added dimension is compact , i.e. , it is curled up in itself and has a certain length , which is too small to be observed . but still the theory of kaluza and klein could not explain the weakness of gravity in comparison to the electromagnetic forces . in the 60s and 70s the string theory was developed . here the existence of higher dimensions is essential for the internal consistency of the theory , mostly they are also assumed to be compact . + the equations of motion for test particles in the schwarzschild spacetime were solved in 1931 by hagihara @xcite in terms of the elliptic weierstra @xmath0- , @xmath1- and @xmath2-functions . in the taub - nut @xcite , reissner - nordstrm @xcite and myers - perry @xcite spacetime the equations of motion were also solved in terms of the elliptic weierstra functions . the integration of geodesics was advanced in the papers of hackmann and lmmerzahl @xcite . they integrated the geodesics in the four - dimensional schwarzschild - de sitter space - time analytically in terms of the hyperelliptic @xmath3- and @xmath1-functions . the mathematical problem is based on the jacobi inversion problem which can be solved if restriced to the @xmath3-divisor . this method was also applied to find the analytical solution of the equations of motion in higher dimensional schwarzschild , schwarzschild-(anti)de sitter , reissner - nordstrm and reissner - nordstrm -(anti ) de sitter spactime @xcite as well as in kerr-(anti ) de sitter spacetimes @xcite and in higher dimensional myers - perry spacetime @xcite . recently , the geodesics equations in special cases were solved analytically in the singly spinning black ring spacetime @xcite and in the ( charged ) doubly spinning black ring spacetime @xcite . + the test particle motion in different spacetimes containing black strings was studied in @xcite . in @xcite and @xcite the analytical solutions of the equations of motion in the schwarzschild and kerr spacetime pierced by black string were presented . furthermore , the geodesic motion of test particles in field theoretical cosmic string spacetimes were investigated , namely abelian - higgs strings @xcite , two interacting abelian - higgs strings @xcite and cosmic superstrings @xcite . + in this paper we will consider static and rotating black strings , which can be obtained by adding an extra compact dimension to the schwarzschild and kerr metric . the geodesic motion in the ordinary schwarzschild spacetime and the kerr spacetime has already been analyzed in detail , however , when a compact dimension is added new behaviour occurs which is studied in this article . we present the analytic solutions of the geodesic equations in terms of the elliptic weierstra functions and discuss the corresponding orbits . in the first part the non - rotating black string is studied and compared to the schwarzschild black hole . in the second part the rotating black string is analysed and compared to the kerr black hole . in this section we will discuss the geodesics in the static black string spacetime and present analytical solutions of the equations of motion . the well known schwarzschild metric is the unique static spherically symmetric solution to the vacuum einstein equations in four dimensions . if an extra compact spatial dimension @xmath4 is added , the metric takes the form @xmath5 where @xmath6 is proportional to the mass of the black hole . this solution describes a neutral uniform black string . as usual the singularity is located at @xmath7 . the horizon is located at @xmath8 and covers the extra dimension . the gravitational constant and the velocity of light are set to 1 . the hamilton - jacobi equation @xmath9 has a solution of the form @xmath10 here @xmath11 is an affine parameter along the geodesic . the parameter @xmath12 is equal to @xmath13 for particles and equal to @xmath14 for light . @xmath15 , @xmath16 and @xmath17 are the conserved momenta ; @xmath15 is the energy , @xmath16 denotes the angular momentum and @xmath17 is a new constant of motion according to the compact dimension @xmath4 . we set @xmath18 since the orbits lie in a plane if plotted in cartesian @xmath19-@xmath20-@xmath21 coordinates ( @xmath22 , @xmath23 , @xmath24 ) due to the spherical symmetry of the original schwarzschild metric . for convenience , we introduce dimensionless quantities ( @xmath25 ) @xmath26 in these dimensionless quantities the horizon is at @xmath27 . the hamilton - jacobi equation separates and yields a differential equation for each coordinate @xmath28 with the polynomial @xmath29 we also used the mino time @xmath30 as @xmath31 @xcite . the new constant of motion @xmath17 due to the compact dimension is only present in the @xmath32-equation and of course in the @xmath33-equation . the other equations are the same as in the original schwarzschild spacetime without the compact dimension . before discussing the @xmath32-motion in detail , we introduce a list of all possible orbits : 1 . _ terminating orbit _ ( to ) with ranges @xmath34 or @xmath35 $ ] with @xmath36 . the tos end in the singularity at @xmath37 . 2 . _ escape orbit _ ( eo ) with range @xmath38 with @xmath39 . 3 . _ bound orbit _ ( bo ) with range @xmath40 $ ] with @xmath41 and @xmath42 . + the right hand side of the differential equation has the form @xmath43 with the coefficients @xmath44 @xmath45 is required in order to obtain real values for @xmath32 . the regions for which @xmath45 are bounded by the zeros of @xmath46 . the number of zeros can be determined with the help of the rule of descartes . one can distinguish four cases : 1 . if @xmath47 and @xmath48 , then two or no zeros are possible . 2 . if @xmath47 and @xmath49 , no zeros are possible . 3 . if @xmath50 and @xmath48 , then one or three zeros are possible . 4 . if @xmath50 and @xmath49 , one zero is possible . since @xmath15 is real and therefore @xmath51 , @xmath52 is required . so if @xmath53 ( where the equations of motion reduce to the original schwarzschild case without the compact dimension ) the cases 3 . and 4 . are not possible for light ( @xmath54 ) . but if @xmath55 there are 3 possible zeros of @xmath46 for @xmath54 , that means in the schwarzschild spacetime with an extra compact dimension we have the possibility of stable bound orbits for light , which are not possible in the original schwarzschild spacetime . + we define an effective potential @xmath56 from the equation by @xmath57 thus @xmath58 in the limit @xmath59 the effective potential @xmath56 converges to @xmath60 . if @xmath61 , we have @xmath62 . @xmath63 determines the turning points of an orbit . depending on the number of turning points and the shape of the effective potential four different types of orbits are possible : * type a : no zeros . only tos are possible . * type b : one zero . only tos are possible . the turning point can coincide with the event horizon . * type c : two zeros . tos and eos are possible . if the energy @xmath15 coincides with the maximum of the effective potential then there is a single zero , which corresponds to an unstable circular bound orbit . * type d : three zeros . tos and bos are possible . + if the energy @xmath15 coincides with the minimum of the effective potential then there are two zeros . tos and circular bos are possible . table [ tab : type - orbits ] shows an overview of the types of orbits . some examples for energies corresponding to certain orbits in the effective potential can be seen in figure [ pic : potential ] . a & 0 & ( -2.5,-0.2)(3,0.2)(-2.5,0)(3,0 ) ( -1,-0.2)(-1,0.2 ) ( -2.5,0)(3,0 ) & to + b & 1 & ( -2.5,-0.2)(3,0.2)(-2.5,0)(3,0 ) ( -1,-0.2)(-1,0.2 ) ( -2.5,0)(0,0 ) & to + b@xmath64 & & ( -2.5,-0.2)(3,0.2)(-2.5,0)(3,0 ) ( -1,-0.2)(-1,0.2 ) ( -2.5,0)(-1,0 ) & to + c & 2 & ( -2.5,-0.2)(3,0.2)(-2.5,0)(3,0 ) ( -1,-0.2)(-1,0.2 ) ( -2.5,0)(0,0 ) ( 1.0,0)(3,0 ) & to , eo + d & 3 & ( -2.5,-0.2)(3,0.2)(-2.5,0)(3,0 ) ( -1,-0.2)(-1,0.2 ) ( -2.5,0)(0,0 ) ( 1,0)(2,0 ) & to , bo + in this section we present the analytical solution of the equations of motion - . the polynomial @xmath43 can be reduced to cubic order by the substitution @xmath65 : @xmath66 . a further substitution @xmath67 transforms @xmath68 into the weierstra form so that equation turns into @xmath69 where @xmath70 the differential equation is of elliptic type and is solved by the weierstra @xmath0-function @xcite @xmath71 where @xmath72 with @xmath73 . then the solution of acquires the form @xmath74 equation has the trivial solution @xmath76 using equation and the subsitution @xmath77 the @xmath33-equation becomes @xmath78 the right hand side is an elliptic differential of the third kind . after the substitution @xmath79 with @xmath80 and integration of equation one gets @xmath81 where @xmath82,@xmath83 and @xmath84 . the solution of this integral is given in terms of the elliptic @xmath0- , @xmath1- and @xmath2-function @xcite : @xmath85 using equation the @xmath86-equation becomes @xmath87 now we substitute @xmath77 and apply a partial fractions decomposition . then we get @xmath88 where @xmath89 , @xmath90 , @xmath91 and @xmath92 . the right hand side consists of elliptic differentials of the third kind . after the substitution @xmath79 with @xmath80 and integration of equation one gets @xmath93 where @xmath82,@xmath83 and @xmath94 . the solution of this integral is given in terms of the elliptic @xmath0- , @xmath1- and @xmath2-function @xcite : @xmath95 with these analytical results we have found the complete set of orbits for light and test particles in the black string spacetime . depending on the parameters @xmath12 , @xmath96 , @xmath17 and @xmath15 , bos , eos , and tos are possible . bos can be seen in figure [ pic : bo - light ] and [ pic : bo - particles ] . an eo is depicted in figure [ pic : eo - light ] , and figure [ pic : to - particles ] shows a to . in cartesian @xmath19-@xmath20-@xmath21 coordinates the orbits lie in a plane . however , this is not the case if the orbit is plotted in @xmath19-@xmath20-@xmath4 coordinates . in the black string spacetime bos for light are possible ( see figure [ pic : bo - light ] ) . such orbits do not exist in the ordinary four - dimensional schwarzschild spacetime . in this section we will discuss the geodesics in the rotating black string spacetime and present analytical solutions of the equations of motion . the rotating black string metric is derived from the kerr metric by adding an extra compact dimension . a detailed analysis of the geodesics in the original kerr spacetime without a compact dimension can be found in e.g. @xcite . the kerr metric is the rotating axially symmetric solution of the einstein equations in four dimensions . if an extra compact spatial dimension @xmath4 is added , the metric takes the form @xmath97 where @xmath98 and @xmath99 . this solution describes a rotating uniform black string . @xmath6 is proportional to the mass of the black string and @xmath100 is proportional to the angular momentum . the gravitational constant and the velocity of light are set to 1 . the ( ring ) singularity is located at @xmath101 , i.e. at @xmath7 and @xmath18 . hence geodesics with @xmath7 and @xmath102 do not end in the singularity . so in the rotating black string spacetime negative @xmath103-values are allowed as in the kerr spacetime @xcite . there are two horizons defined by @xmath104 and for @xmath105 they are given by @xmath106 note that the metric is given in boyer - lindquist coordinates @xmath103 , @xmath107 , @xmath75 . cartesian coordinates can be obtained by the transformation @xmath108 the hamilton - jacobi equation @xmath109 has a solution of the form @xmath110 here @xmath11 is an affine parameter along the geodesic . the parameter @xmath12 is equal to @xmath13 for particles and equal to @xmath14 for light . @xmath15 , @xmath16 and @xmath17 are the conserved momenta ; @xmath15 is the energy , @xmath16 denotes the angular momentum and @xmath17 is a new constant of motion according to the compact dimension @xmath4 . for convenience , we introduce dimensionless quantities ( @xmath25 ) @xmath111 @xmath112 is the famous carter constant resulting from the separation of the hamilton - jacobi equation , see @xcite . in these dimensionless quantities the horizons are located at @xmath113 . the hamilton - jacobi equation separates and yields a differential equation for each coordinate @xmath114 - \frac{1}{\sin^2\vartheta}({\tilde{a}}e \sin^2\vartheta -{\tilde{l } } ) \label{eqn : kerr - phi - equation}\\ \frac{{{\rm d}}{\tilde{w}}}{{{\rm d}}\gamma } & = & j{\tilde{\rho}}^2 \label{eqn : kerr - w - equation } \\ \frac{{{\rm d}}{\tilde{t}}}{{{\rm d}}\gamma } & = & \frac{{\tilde{r}}^2+{\tilde{a}}^2}{{\tilde{\delta}_r}}[({\tilde{r}}^2+{\tilde{a}}^2)e-{\tilde{a}}{\tilde{l } } ] - { \tilde{a}}({\tilde{a}}e \sin^2\vartheta -{\tilde{l } } ) \label{eqn : kerr - t - equation}\end{aligned}\ ] ] with the polynomial @xmath46 and the function @xmath115 : @xmath116 ^ 2 - { \tilde{\delta}_r}({\tilde{k}}+(\delta+j^2){\tilde{r}}^2 ) \\ \theta & = { \tilde{k}}- ( \delta + j^2){\tilde{a}}^2\cos^2\vartheta - \frac{1}{\sin^2\vartheta}({\tilde{a}}e \sin^2\vartheta -{\tilde{l}})^2 \ , . \end{split}\ ] ] we also used the mino time @xmath30 as @xmath117 @xcite . the new constant of motion @xmath17 due to the compact dimension is present in the @xmath32-equation , the @xmath107-equation and of course in the @xmath33-equation . the other equations are the same as in the original kerr spacetime without the compact dimension . to obtain the geodesic equations of the original kerr spacetime , one can choose @xmath53 . the function @xmath115 and the polynomial @xmath46 define the properties of the orbits depending on the parameters of the metric and the test particle . in this section we will analyze the function @xmath115 and the polynomial @xmath46 to determine the possible types of orbits . a similar analysis of the kerr spacetime can be found in @xcite , for the kerr-(anti- ) de sitter spacetime see @xcite . + to obtain real values of @xmath107 and @xmath32 the conditions @xmath118 and @xmath119 have to be fullfilled , otherwise no motion is possible . from @xmath118 we can immediately conclude that @xmath120 . if @xmath121 , the motion takes place in the equatorial plane @xcite . furthermore we see that a geodesic hits the singularity at @xmath122 if @xmath123 : a solution of @xmath124 is @xmath37 and simultaneously @xmath18 . since @xmath125 \geq0 \ , , \end{split } \label{eqn : kerr - r - theta-0}\ ] ] it follows that @xmath123 . additionally an orbit lies in the equatorial plane ( @xmath18 ) if @xmath123 ( compare@xcite ) . from we also see the following : * @xmath126 : the geodesics cross @xmath18 but do not cross @xmath37 . * @xmath127 : the geodesics do not cross @xmath18 but it is possible to cross @xmath37 . here the rotating black string shows the same properties as the kerr black hole ( see @xcite ) . as in @xcite we substitute @xmath128 ( with @xmath129 $ ] ) in the function @xmath130 : @xmath131 now we want to determine the number of real zeros of @xmath132 in @xmath133 $ ] . the number of zeros only changes if a zero crosses 0 or 1 , or if a double zero occurs . @xmath134 is a zero of @xmath115 if @xmath135 and therefore @xmath136 since @xmath137 is a pole of @xmath132 for @xmath138 , it is only possible that @xmath137 is a zero of @xmath132 if @xmath49 . @xmath139 yields that @xmath140 . note that due to the new constant of motion @xmath17 ( the momentum in the @xmath33-direction ) the value of @xmath141 is different here in comparison to the kerr spacetime . so @xmath142 and @xmath143 are border cases of the @xmath107-motion . + next we consider the polynomial @xmath144\nu + { \tilde{k}}-({\tilde{a}}e-{\tilde{l}})^2\ ] ] to remove the pole of @xmath132 at @xmath137 . @xmath115 and @xmath145 are related by @xmath146 . double zeros of @xmath132 in @xmath147 and hence of @xmath148 occur if @xmath149 from these conditions we obtain @xmath150 therefore double zeros are only possible if @xmath47 . + with the help of equation and parametric @xmath96-@xmath151-diagrams can be drawn . figure [ pic : kerr - theta - parametric ] shows a typical example of a @xmath96-@xmath151-diagram in the rotating black string spacetime , which resembles the parametric diagrams in the kerr or kerr-(anti- ) de sitter spacetime @xcite . here we recognize four regions . in region ( a ) and ( d ) we have @xmath152 for all @xmath129 $ ] and therefore no geodesic motion is possible . in region ( b ) @xmath132 has one real zero in @xmath133 $ ] , here @xmath126 and the orbit crosses @xmath18 . in region ( d ) @xmath132 has two real zeros in @xmath133 $ ] , here @xmath127 and @xmath18 is not crossed . however , crossing @xmath37 is possible in region ( d ) but not in ( b ) . + in the special case @xmath49 there is a single zero at @xmath153 ( not necessarily in @xmath133 $ ] ) . if additionally @xmath140 then this zero is at @xmath137 and if @xmath154 this zero is at @xmath134 . in the case @xmath155 there is a zero @xmath156 in @xmath133 $ ] if @xmath157 and @xmath158 for all @xmath129 $ ] if @xmath159 . in the case @xmath160 there is a zero @xmath156 in @xmath133 $ ] if @xmath159 but @xmath152 for all @xmath129 $ ] if @xmath157 . , @xmath161 , @xmath162 , @xmath163 : parametric @xmath96-@xmath151-diagram of the @xmath107-motion . in the grey regions ( a ) and ( d ) we have @xmath152 for all @xmath129 $ ] and hence no motion is possible . in region ( b ) @xmath132 has one real zero in @xmath133 $ ] and in region ( d ) @xmath132 has two real zeros in @xmath133$].,width=226 ] the zeros of the polynomial @xmath46 are the turning points of orbits of light and test particles and therefore @xmath46 determines the possible types of orbits . in contrast to the non - rotating black string , here we can consider negatives @xmath32-value too , since there is no singularity for @xmath7 and @xmath102 . orbits that cross @xmath37 are called crossover orbits @xcite . before discussing the @xmath32-motion in detail , we introduce a list of all possible orbits : 1 . _ transit orbit _ ( tro ) with range @xmath164 . this orbit is a crossover orbit . 2 . _ escape orbit _ ( eo ) with range @xmath38 with @xmath165 , or with range @xmath166 $ ] with @xmath167 . two - world escape orbit _ ( teo ) with range @xmath168 where @xmath169 . the teos cross both horizons twice and emerge into another universe . crossover two - world escape orbit _ ( cteo ) with range @xmath168 where @xmath170 . the cteos cross both horizons twice and emerge into another universe . @xmath37 is crossed once . _ bound orbit _ ( bo ) with range @xmath40 $ ] with @xmath171 and 1 . either @xmath172 or 2 . _ many - world bound orbit _ ( mbo ) with range @xmath40 $ ] where @xmath174 and @xmath175 . the mbos cross both horizons several times . each time both horizons are traversed twice the test particles emerge into another universe . 7 . _ terminating orbit _ ( to ) with ranges @xmath34 or @xmath35 $ ] with 1 . either @xmath176 or 2 . + the tos end in the singularity at @xmath37 and @xmath18 . @xmath123 is required . the possible orbit types depend on the number of real zeros of the polynomial @xmath46 . the number of zeros can only change if double zeros occur , i.e. if @xmath178 with the help of equation parametric @xmath96-@xmath151-diagrams can be drawn , where five regions with a different number of real zeros of @xmath46 appear ( see figure [ pic : kerr - r - parametric ] ) . also the @xmath107-equation has to be taken into account . both parametric @xmath96-@xmath151-diagrams for the @xmath107-motion and the @xmath32-motion have to be combined in order to detect all possible behaviour of light and test particles in the rotating black string spacetime ( see figure [ pic : kerr - r - theta - parametric ] ) . , @xmath161 , @xmath179 , @xmath163 : parametric @xmath96-@xmath151-diagram of the @xmath32-motion . in region ( i ) @xmath46 has no real zeros . in the regions ( ii ) and ( v ) @xmath46 has two zeros . four zeros are possible in the regions ( iii ) and ( iv ) . here we see that in the rotating black string spacetime all five regions are present for @xmath54 , whereas in the kerr spacetime region ( iii ) and ( v ) do not exist if @xmath54.,width=226 ] additionally one may define an effective potential consisting of the two parts @xmath180 and @xmath181 by the equation @xmath182 thus @xmath183}}{{\tilde{r}}^2+{\tilde{a}}^2 } \ , .\ ] ] the area between @xmath180 and @xmath181 is a forbidden zone , since there @xmath184 . @xmath180 and @xmath181 meet at the horizons @xmath185 , where @xmath186 . in the limit @xmath59 and @xmath187 the effective potential @xmath188 converges to @xmath189 . if the sign of @xmath96 changes , the effective potential is mirrored at the @xmath32-axis . + taking all informations into acount , we can now determine the possible orbits in the different regions of the parametric diagrams ( below we always assume that @xmath190 ) : 1 . region ( i ) : no real zeros and @xmath191 for all @xmath32 . here only transit orbits are possible which cross @xmath37 . since region ( i ) intersects with region ( d ) , @xmath18 is not crossed . 2 . region ( ii ) : @xmath192 has two real zeros @xmath193 , @xmath194 and @xmath195 for @xmath196 $ ] and @xmath197 . region ( ii ) intersects with region ( b ) and region ( d ) from the @xmath107-motion . in part ( iib ) there is a negative and a positive zero , so that escape orbits are possible for @xmath198 and two - world escape orbits are possible for @xmath199 . here the orbits cross @xmath18 plane . the turning point of the teo can coincide with the inner horizon @xmath200 . in the special case @xmath123 the former positive zero is now at @xmath37 so that the two - world escape orbit turns into a terminating orbit and the motion takes place in the equatorial plane . + in part ( iid ) @xmath46 has two negative zeros , so there is an escape orbit for @xmath198 and a crossover two - world escape orbit . this time @xmath18 is not crossed . 3 . region ( iii ) : @xmath46 has four positive zeros @xmath193 , @xmath194 , @xmath201 , @xmath202 and @xmath195 for @xmath40 $ ] and @xmath203 $ ] . possible orbits are bound orbits with @xmath204 and many - world bound orbits . one or both ( for @xmath49 ) of the turnings points of the many - world orbit can coincide with the horizons . since region ( iii ) intersects with region ( b ) the orbits cross @xmath18 . region ( iv ) : @xmath46 has one negative zero @xmath193 and three positive zero @xmath194 , @xmath201 , @xmath202 . @xmath195 for @xmath166 $ ] , @xmath205 $ ] and @xmath206 . escape orbits with either @xmath198 or @xmath204 and many - world bound orbits are possible . if @xmath141 is small then also bound orbits with @xmath207 are possible instead of the many - world bound orbits . these orbits are hidden behind the inner horizon . in the special case @xmath123 one of the turning points of the ( many - world ) bound orbit is now at @xmath37 so that the orbit turns into a terminating orbit and the motion takes place in the equatorial plane . region ( v ) : @xmath46 has two positive zeros @xmath193 , @xmath194 and @xmath195 for @xmath40 $ ] . only many - world bound orbits are possible . one or both ( for @xmath49 ) of the turnings points can coincide with the horizons . since region ( v ) intersects with region ( b ) the orbit crosses @xmath18 . in the special case @xmath123 one zero is now at @xmath37 so that the orbit turns into a terminating orbit and the motion takes place in the equatorial plane . table [ tab : kerr - type - orbits ] shows all possible orbits types of orbits in the rotating black string spacetime . some examples of energies corressponding to the various orbit types in the effective potential can be seen in figure [ pic : kerr - potential ] . if we consider the original kerr spacetime then the regions ( iii ) and ( v ) are not present for @xmath54 , but in the rotating black string spacetime all five regions are present both for @xmath54 and @xmath208 . that means in the rotating black string spacetime we have bound orbits for light with @xmath204 ( see figure [ pic : kerr - potential - b ] ) which were not possible in the kerr spacetime . orbit + a & 0 & i d & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -4,0)(3.5,0 ) & tro + b & 2 & iib & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -4,0)(-3,0 ) ( -1,0)(3.5,0 ) & eo , teo + b@xmath209 & & & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -4,0)(-3,0 ) ( -0.5,0)(3.5,0 ) & eo , teo + b@xmath64 & & & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -4,0)(-3,0 ) ( -2.5,0)(3.5,0 ) & eo , to + c & 2 & iid & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -4,0)(-3.5,0 ) ( -3,0)(3.5,0 ) & eo , cteo + d & 4 & iiib & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -1,0)(1.5,0 ) ( 2,0)(3,0 ) & mbo , bo + d@xmath210 & & & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -0.5,0)(1,0 ) ( 2,0)(3,0 ) & mbo , bo + d@xmath209 & & & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -0.5,0)(1.5,0 ) ( 2,0)(3,0 ) & mbo , bo + d@xmath211 & & & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -1,0)(1,0 ) ( 2,0)(3,0 ) & mbo , bo + e & 4 & ivb & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -4,0)(-3,0 ) ( -1,0)(1.5,0 ) ( 2,0)(3.5,0 ) & eo , mbo , eo + e@xmath209 & & & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -4,0)(-3,0 ) ( -0.5,0)(1.5,0 ) ( 2,0)(3.5,0 ) & eo , mbo , eo + e@xmath211 & & & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -4,0)(-3,0 ) ( -1,0)(1,0 ) ( 2,0)(3.5,0 ) & eo , mbo , eo + f & 4 & ivb & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -4,0)(-3,0 ) ( -2,0)(-1,0 ) ( 2,0)(3.5,0 ) & eo , bo , eo + f@xmath211 & & & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -4,0)(-3,0 ) ( -2,0)(-1,0 ) ( 1,0)(3.5,0 ) & eo , bo , eo + f@xmath64 & & & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -4,0)(-3,0 ) ( -2.5,0)(-1,0 ) ( 2,0)(3.5,0 ) & eo , to , eo + f@xmath212 & & & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -4,0)(-3,0 ) ( -2.5,0)(-1,0 ) ( 1,0)(3.5,0 ) & eo , to , eo + g & 2 & vb & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -1,0)(1.5,0 ) & mbo + g@xmath210 & & & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -0.5,0)(1,0 ) & mbo + g@xmath209 & & & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -0.5,0)(1.5,0 ) & mbo + g@xmath211 & & & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -1,0)(1,0 ) & mbo + g@xmath64 & & & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -2.5,0)(1.5,0 ) & to + g@xmath212 & & & ( -4,-0.2)(3.5,0.2)(-4,0)(3.5,0 ) ( -2.5,-0.2)(-2.5,0.2 ) ( -0.5,-0.2)(-0.5,0.2 ) ( 1,-0.2)(1,0.2 ) ( -2.5,0)(1,0 ) & to + in this section we present the analytical solution of the equations of motion - . the right hand side of the @xmath32-equation is a polynomial of fourth order with the coefficients : @xmath213\end{aligned}\ ] ] the polynomial @xmath43 can be reduced to cubic order by the substution @xmath214 ( where @xmath215 is a zero of @xmath46 ) : @xmath66 . a further substitution @xmath67 transforms @xmath68 into the weierstra form so that equation turns into @xmath216 where @xmath217 the differential equation is of elliptic type and is solved by the weierstra @xmath0-function @xcite @xmath218 where @xmath219 with @xmath73 . then the solution of acquires the form @xmath220 to solve the @xmath107-equation we substitute @xmath128 ( with @xmath129 $ ] ) in equation : @xmath221 - 4\nu[{\tilde{a}}e/1-\nu)-{\tilde{l}}]^2 \label{eqn : kerr - theta - poly}\ ] ] the right hand side of equation is a polynomial of third order @xmath222 with the coefficients : @xmath223\\ c_2 & = & 4[2{\tilde{a}}e({\tilde{a}}e-{\tilde{l}})-{\tilde{a}}^2(\delta+j^2)-{\tilde{k}}]\\ c_1 & = & 4[{\tilde{k}}-({\tilde{a}}e-{\tilde{k}})^2 ] \ , .\end{aligned}\ ] ] so equation can be transformed into the weierstra form using the substitution @xmath224 : @xmath225 where @xmath226 equation is solved by the weierstra @xmath0-function and the solution @xmath227 of equation can then be obtained by resubstitution : @xmath228 with @xmath229 and @xmath230 . using the @xmath32-equation and the @xmath107-equation , we can write the @xmath75-equation in the following way : @xmath231\frac{{{\rm d}}{\tilde{r}}}{\sqrt{r } } - \frac{1}{\sin^2\vartheta}({\tilde{a}}e \sin^2\vartheta -{\tilde{l}})\frac{{{\rm d}}\vartheta}{\sqrt{\theta } } \ , .\ ] ] so the @xmath75-equation consists of a @xmath32-dependent integral and a @xmath107-dependent integral : @xmath232\frac{{{\rm d}}{\tilde{r}}'}{\sqrt{r } } - \int_{\vartheta_{\rm in}}^{\vartheta } \ ! \frac{1}{\sin^2\vartheta'}({\tilde{a}}e \sin^2\vartheta ' -{\tilde{l}})\frac{{{\rm d}}\vartheta'}{\sqrt{\theta } } = i_{{\tilde{r } } } - i_\vartheta \ , .\ ] ] let us first consider @xmath233 . here we substitute @xmath234 and apply a partial fraction decomposition , so that @xmath233 turns into @xmath235 where @xmath236 and @xmath237 are first order poles of @xmath233 . the sign of @xmath238 depends on the chosen sign in the substitution @xmath234 . @xmath239 are constants that arise from the partial fraction decomposition and depend on the parameters of the metric and the test particle . now we substitute @xmath240 with @xmath241 : @xmath242 the integral @xmath243 can be transformed in the same way by substituting first @xmath128 , then @xmath224 and finally @xmath244 with @xmath245 : @xmath246 where @xmath247 . the integrals @xmath233 and @xmath243 are of elliptic type and can be solved in terms of the elliptic @xmath0- , @xmath1- and @xmath2-function as shown in @xcite . then the final solution of the @xmath75-equation is @xmath248 where @xmath249 , @xmath250 , @xmath241 , @xmath245 and @xmath251 using the @xmath32-equation and the @xmath107-equation , we can write the @xmath33-equation in the following way : @xmath252 like the @xmath75-equation , the @xmath33-equation consist of a @xmath32-dependent part and a @xmath107-dependent part . we integrate and use the same substitutions as in the previous section [ sec : kerr - phisol ] : @xmath253 where @xmath254 . the occurring elliptic intregrals of the third kind can be solved as shown in @xcite . furthermore we use the relation @xmath255 . the final solution of the @xmath33-equation is @xmath256 where @xmath257 . using the @xmath32-equation and the @xmath107-equation , we can write the @xmath86-equation in the following way : @xmath258 \frac{{{\rm d}}{\tilde{r}}}{\sqrt{r } } - { \tilde{a}}({\tilde{a}}e \sin^2\vartheta -{\tilde{l } } ) \frac{{{\rm d}}\vartheta}{\sqrt{\theta } } \ , . \label{eqn : tsplit}\ ] ] like the @xmath75-equation and the @xmath33-equation , the @xmath86-equation consist of a @xmath32-dependent part and a @xmath107-dependent part . we integrate and use the same substitutions as in section [ sec : kerr - phisol ] : @xmath259 where @xmath236 and @xmath237 are first order poles and @xmath260 is a second order pole . @xmath261 are constants arising from a partical fraction decomposition of the @xmath32-dependent part . the occurring elliptic intregrals of the third kind can be solved as shown in @xcite . furthermore we use the relation @xmath255 . the final solution of the @xmath86-equation is @xmath262 where @xmath249 . with these analytical results we have found the complete set of orbits for light and test particles in the rotating black string spacetime . depending on the parameters @xmath12 , @xmath96 , @xmath17 , @xmath263 and @xmath15 various orbits are possible . note that the horizons in the kerr spacetime and in the rotating black string spacetime ( @xmath19-@xmath20-@xmath21-plots ) are not spheres but ellipsoids . a @xmath19-@xmath20-@xmath4-plot of the horizons in the rotating black string spacetime will show cylinders . figure [ pic : kerr - to - particle ] shows a terminating orbit in the rotating black string spacetime , in the @xmath19-@xmath20-@xmath21-plot this orbit lies entirely in the equatorial plane . to hit the ring singularity at @xmath122 the parameters corresponding to the orbit have to fullfill the condition @xmath123 . otherwise it is possible for a geodesic to cross @xmath37 without touching the singularity . like in the kerr spacetime the radial coordinate @xmath32 can take negative values . a transit orbit wich starts at positve @xmath32 , crosses @xmath37 and the continues at negative @xmath32 can be seen in figure [ pic : kerr - tro - particle ] . when @xmath32 changes from positive to negative it looks like the particle is reflected , which can be interpreted as gravity becoming repulsive for negative @xmath32-values ( see @xcite ) . an escape orbit is shown in figure [ pic : kerr - eo - particle ] . if the turning point of an escape orbit lies behind the two horizons the orbit is called two - world escape orbit ( see figure [ pic : kerr - teo - light ] ) . since both horizons are traversed twice , the test particle or light emerges into another universe . it is also possible that the turning point is at negative @xmath32 ( see figure [ pic : kerr - cteo - particle ] ) , then the orbit is called crossover two - world escape orbit since it does not only traverses both horizons , but also crosses @xmath37 . a special feature of the the rotating black string spacetime are stable bound orbits of light ( see figure [ pic : kerr - bo - light ] ) . such orbits do not exist in the ordinary four - dimensional kerr spacetime . like in the kerr spacetime there are also bound orbits hidden behind the inner horizon @xmath200 if the carter constant @xmath141 is rather small ( see figure [ pic : kerr - innerbo - particle ] ) . it is also possible that a particle or light on a bound orbit crosses both horizons several times . each time both horizons are traversed twice , the orbit continues in another universe . such a many - world bound orbit is depicted in figure [ pic : kerr - mbo - particle ] . figure [ pic : zoomwhirl - bo - particle ] shows a so called `` zoom - whirl '' bound orbit ( see also @xcite ) . here the energy of the test particle is very close to a local maximum of the effective potential , so that the particle `` whirls '' around the black string before reaching the periastron and then `` zooms '' out to return to its elliptical orbit where it moves to the apastron . in this article we presented the complete set of analytical solutions of the geodesic equations of test particles and light in the static and the rotating black string spacetime . the static and the rotating black string metric are obtained by adding a compact dimension to the schwarzschild and kerr metric . the geodesic equations can be solved in terms of the elliptic @xmath0- , @xmath1- and @xmath2-functions . using effective potential techniques and parametric diagrams , possible types of orbits were derived . in the static case bos , eos and tos are possible , while in the rotating case bos , mbos , eos , teos , cteos , tros and tos are possible . in contrast to the ordinary four - dimensional schwarzschild and kerr spacetime , bound orbits of light are possible both in the static and the rotating black string spacetime . + hereby the analytic solutions do not only give a proof of the existence of bound orbits of light in the black string spacetime . they also present a usefull tool to calculate the exact orbits and their properties , including observables like the periastron shift of bound orbits , the light deflection of escape orbits , the deflecton angle and the lense - thirring effect . for the calculation of the observables analogous formulas to those given in @xcite can be used . observables can later be compared to observations . + the black strings considered here are so - called uniform black strings , since there is no dependence of the extra dimension . however , uniform black strings are subject to the gregory - laflamme instability @xcite , which is associated with the emergence of a branch of non - uniform black strings , whose horizon size is not constant w.r.t . the compact direction , but depends on the compact coordinate @xcite . at the endpoint of this nonuniform black string branch a horizon topology changing transition should be encountered @xcite . to gain a better understanding of this transition , it is essential to solve the geodesic equations . however , for the construction of nonuniform black strings no analytic techniques are available , and numerical techniques have to be employed . likewise , the geodesic equations must be solved numerically . it is therefore of high relevance , to have analytic solutions of the geodesic equations available as a testbed . moreover , it will be very interesting to see , how the set of analytic solutions changes , as the black strings become more and more nonuniform . also non - uniform rotating and charged black strings have been considered in @xcite and @xcite whose geodesics would be very interesting to study . + it is also interesting to consider higher dimensional ( @xmath264 ) schwarzschild spacetimes , here stable bound orbits are no longer possible @xcite . one may wonder wether bound orbits of light become possible if a compact dimension is added to the metric . but it can be shown that unlike in the ( rotating ) black string spacetime in five dimensions , stable bound orbits of light are not possible if a compact dimension is added to higher dimensional schwarzschild spacetimes . + moreover the bound orbits of light appear neither in the spacetime of an abelian - higgs string @xcite nor in the spacetime of two interacting abelian - higgs strings @xcite . there escape orbits are the only possibility for massless particles . for cosmic superstrings bound orbits of light are possible in melvin spacetimes but not in asymptotically conical spacetimes @xcite . we gratefully acknowledge support by the daad and the dfg , in particular , within the dfg research training group 1620 `` models of gravity '' . also we would like to thank jutta kunz , valeria kagramanova and burkhard kleihaus for helpful discussions . b.k . would like to thank suneeta vardarajan for her guidance . k. schwarzschild , sitzungsber . berlin ( math . phys . ) * 1916 * , 424 ( 1916 ) [ physics/9912033 ] . r. p. kerr , phys . lett . * 11 * , 237 ( 1963 ) . t. kaluza , sitzungsber . berlin ( math . * 1921 * , 966 ( 1921 ) . o. klein , z. phys . * 37 * , 895 ( 1926 ) [ surveys high energ . phys . * 5 * , 241 ( 1986 ) ] . y. hagihara , jpn . j. astron . geophys . * 8 * , 67 ( 1931 ) v. kagramanova , j. kunz , e. hackmann and c. lmmerzahl , phys . rev . d * 81 * , 124044 ( 2010 ) [ arxiv:1002.4342 [ gr - qc ] ] . s. grunau and v. kagramanova , phys . d * 83 * , 044009 ( 2011 ) [ arxiv:1011.5399 [ gr - qc ] ] . v. kagramanova and s. reimers , phys . d * 86 * , 084029 ( 2012 ) [ arxiv:1208.3686 [ gr - qc ] ] . e. hackmann and c. lmmerzahl , phys . lett . * 100 * , 171101 ( 2008 ) . e. hackmann and c. lmmerzahl , phys . d * 78 * , 024035 ( 2008 ) . e. hackmann , v. kagramanova , j. kunz and c. lmmerzahl , phys . d * 78 * , 124018 ( 2008 ) [ erratum - ibid . * 79 * , 029901 ( 2009 ) ] [ arxiv:0812.2428 [ gr - qc ] ] . e. hackmann , c. lmmerzahl , v. kagramanova and j. kunz , phys . d * 81 * , 044020 ( 2010 ) [ arxiv:1009.6117 [ gr - qc ] ] . v. z. enolski , e. hackmann , v. kagramanova , j. kunz and c. lmmerzahl , j. geom . phys . * 61 * , 899 ( 2011 ) [ arxiv:1011.6459 [ gr - qc ] ] . s. grunau , v. kagramanova , j. kunz and c. lmmerzahl , phys . d * 86 * , 104002 ( 2012 ) [ arxiv:1208.2548 [ gr - qc ] ] . s. grunau , v. kagramanova and j. kunz , phys . rev . d * 87 * , 044054 ( 2013 ) arxiv:1212.0416 [ gr - qc ] . a. n. aliev and d. v. galtsov , sov . * 14 * , 48 ( 1988 ) . d. v. galtsov and e. masar , class . quant . grav . * 6 * , 1313 ( 1989 ) . s. chakraborty and l. biswas , class . . grav . * 13 * , 2153 ( 1996 ) . n. ozdemir , class . grav . * 20 * , 4409 ( 2003 ) . f. ozdemir , n. ozdemir and b. t. kaynak , int . j. mod . a * 19 * , 1549 ( 2004 ) . e. hackmann , b. hartmann , c. lmmerzahl and p. sirimachan , phys . d * 81 * , 064016 ( 2010 ) [ arxiv:0912.2327 [ gr - qc ] ] . e. hackmann , b. hartmann , c. lmmerzahl and p. sirimachan , phys . d * 82 * , 044024 ( 2010 ) [ arxiv:1006.1761 [ gr - qc ] ] . b. hartmann and p. sirimachan , jhep * 1008 * , 110 ( 2010 ) [ arxiv:1007.0863 [ gr - qc ] ] . b. hartmann and v. kagramanova , phys . d * 86 * , 045028 ( 2012 ) [ arxiv:1204.0396 [ hep - th ] ] . b. hartmann , c. lmmerzahl and p. sirimachan , phys . d * 83 * , 045027 ( 2011 ) [ arxiv:1012.3285 [ hep - th ] ] . y. mino , phys . d * 67 * , 084027 ( 2003 ) [ gr - qc/0302075 ] . a. i. markushevich , _ theory of functions of a complex variable _ ( prentice - hall , englewood cliffs , nj , 1967 ) , vol . b. carter , phys . rev . * 174 * , 1559 ( 1968 ) . b. carter , phys . rev . * 141 * , 1242 ( 1966 ) . w. schmidt , class . * 19 * , 2743 ( 2002 ) [ gr - qc/0202090 ] . k. glampedakis and d. kennefick , phys . d * 66 * , 044002 ( 2002 ) [ gr - qc/0203086 ] . j. levin and g. perez - giz , phys . d * 79 * , 124013 ( 2009 ) [ arxiv:0811.3814 [ gr - qc ] ] . r. gregory and r. laflamme , phys . lett . * 70 * , 2837 ( 1993 ) [ hep - th/9301052 ] . s. s. gubser , class . grav . * 19 * , 4825 ( 2002 ) [ hep - th/0110193 ] . t. wiseman , class . grav . * 20 * , 1137 ( 2003 ) [ hep - th/0209051 ] . b. kleihaus , j. kunz and e. radu , jhep * 0606 * , 016 ( 2006 ) [ hep - th/0603119 ] . b. kol , jhep * 0510 * , 049 ( 2005 ) [ hep - th/0206220 ] . b. kol , phys . rept . * 422 * , 119 ( 2006 ) [ hep - th/0411240 ] . b. kleihaus , j. kunz and e. radu , jhep * 0705 * , 058 ( 2007 ) [ hep - th/0702053 ]

in the framework of a microscopic string model inclusive charged particle distribution and baryon and antibaryon production are described . the emphasis is put on high energies ( rhic ) where shadowing corrections play a crucial role . some recent developments on suppression at cern - sps are also discussed . possible consequences for the crucial issue of thermal equilibration of the produced system are considered .

nucl-th0303045

this work is a continuation of the one presented in ref @xcite , where i discussed multiparticle production at cern - sps energies . the present article is mainly concerned with higher energies where the efffects of shadowing play a very important role . in the framework of the model presented below , the shadowing corrections can be computed from high mass diffraction practically without any new free parameter . when the effects of shadowing are taken into account , the model describes the inclusive charged particle production at rhic as a function of centrality @xcite . a comparison with the results obtained @xcite in the framework of the saturation model is also presented . another new development concerns net baryon production ( stopping ) . it is shown that both sps and rhic data can be described with the same mechanism ( and the same values of the parameters ) used in @xmath1 interactions . this indicates that there is no evidence for an `` anomalous '' stopping in the heavy ion data @xcite . as already observed at cern - sps , rare processes like strange and multistrange baryon and antibaryon production , can only be described with the introduction of some final state interaction between the produced particles ( comovers interaction ) @xcite . it turns out , however , that the interaction cross - sections required to describe the data are comparatively small ( a few tenths of a mb ) and , in view of the shortness of the interaction time ( @xmath2 fm ) it seem quite improbable that the system can reach thermal equilibrium . in a recent development @xcite reported below , we show that the same formalism of final state interaction used at cern - sps can describe rhic data with the same values of the parameters . predictions for @xmath3 and @xmath4 production have been confirmed by recent star data . predictions for @xmath5 and @xmath6 are also given . finally , we analyze the new na50 data on @xmath0 suppression at cern - sps in the comovers approach @xcite @xcite and discuss expectations at rhic . the dual parton model ( dpm ) @xcite and the quark gluon string model ( qgsm ) @xcite are closely related dynamical models of soft hadronic interactions . they are based on the large-@xmath7 expansion of non - perturbative qcd [ 11 - 13 ] and on gribov s reggeon field theory ( rft ) @xcite . their main aim is to determine the mechanism of multiparticle production in hadronic and nuclear interactions . the basic mechanism is well known in @xmath8 annihilation ( fig . the @xmath8 converts into a virtual photon , which decays into a @xmath9 pair . in the rest system of the virtual photon the quark ( with colour 3 ) and the antiquark ( colour @xmath10 ) separate from each other producing one string ( or chain ) of hadrons , i.e. two back - to - back jets . processes of this type are called one - string processes . in hadron - hadron interactions , a one - string mechanism is also possible but only in some cases , namely when the projectile contains an antiquark ( quark ) of the same type than a quark ( antiquark ) of the target , which can annihilate with each other in their interaction . for instance in @xmath11 , the @xmath12 of @xmath13 can annihilate with the @xmath14 of @xmath15 and a single string is stretched between the @xmath16 of @xmath13 ( colour 3 ) and a diquark @xmath17 of @xmath15 ( colour @xmath10 ) . this mechanism is also possible in @xmath18 interactions ( fig . 2 ) but not in @xmath1 . this already indicates that it can not give the dominant contribution at high energy . indeed , when taking the square of the diagram of fig . 2 ( in the sense of unitarity ) we obtain a planar graph , which is the dominant one according to the large-@xmath7 expansion . however , this only means that this graph has the strongest coupling . since flavour quantum numbers are exchanged between projectile and target , this graph gives a contribution to the total cross - section that decreases as an inverse power of @xmath19 ( @xmath20 . a decrease with @xmath19 is always associated with flavor exchange . for instance , the charge exchange @xmath21 cross - section also decreases as @xmath22 . actually , the diagram in fig . 2 corresponds to the exchange of a secondary reggeon , with intercept close to 1/2 . in order to prevent the exchange of flavour between projectile and target , the @xmath12 and @xmath14 have to stay , respectively , in the projectile and target hemispheres . since they are coloured , they must hadronize stretching a second string of type @xmath12-@xmath14 . we obtain in this way a two - string diagram ( figs . 3 - 5 ) ) . taking the square of this diagram , we obtain a graph with the topology of a cylinder ( fig . it turns out that this is the simplest topology one can construct which does not vanish as @xmath23 due to flavour exchange . therefore , we obtain in this way the dominant graph for hadron - hadron scattering at high energy . the diagram in fig . 6 is called a pomeron ( p ) and the graphs in figs . 3 - 5 a cut pomeron . its order in the large-@xmath7 expansion is @xmath24 [ 12 - 13 ] . note that due to energy conservation the longitudinal momentum fractions taken by the two systems at the string ends have to add up to unity . there are also higher order diagrams ( in the sense of the large-@xmath7 expansion ) with 4 , 6 , 8 strings which give non - vanishing contributions at high energy . an example of the next - to - leading graph for @xmath1 interactions is shown in fig . 7 . it contains four strings the two extra strings are stretched between sea quarks and antiquarks . the square of this graph corresponds to a @xmath25 cut and has the topology of a cylinder with a handle . its order in the large-@xmath7 expansion is @xmath26 . the one with six strings corresponds to a @xmath27 cut and to the topology of a cylinder with two handles ( order @xmath28 ) , etc . the single particle inclusive spectrum is then given by @xcite [ 1e ] & & dn^pp dy(y ) = _ n 1 _ n _ n _ n ( n_n^qq - q_v(y ) + n_n^q_v - qq(y ) + ( 2n-2 ) n_n^q_s-_s(y ) ) + & & n_k^qq - q_v(y ) + n_k^q_v - qq(y ) + ( 2k-2 ) n_k^q_s-_s(y ) where @xmath29 is the average number of inelastic collisions . note that each term consists of @xmath30 strings , i.e. two strings per inelastic collisions . two of these strings , of type @xmath31-@xmath32 , contain the diquarks of the colliding protons . all other strings are of type @xmath32-@xmath33 . the weights @xmath34 of the different graphs , i.e. their contribution to the total cross - section , can not be computed in the large-@xmath7 expansion . however , it has been shown @xcite that there is a one - to - one correspondence between the various graphs in the large-@xmath7 expansion and those in perturbative reggeon field theory @xcite . we use the weights obtained from the latter with the parameters determined from a fit to total and elastic cross - sections @xcite , @xcite . at sps energies we get @xmath35 and at rhic @xmath36 at @xmath37 gev and @xmath38 at @xmath39 gev @xcite . the hadronic spectra of the individual strings @xmath40 are obtained from convolutions of momentum distribution functions , giving the probability to find a given constituent ( valence quark , sea quark of diquark ) in the projectile or in the target , with the corresponding fragmentation functions . the dependence of @xmath40 on the number of collisions appears via the former . it is a result of energy conservation . ( the larger the number of strings , the smaller the average invariant mass of each one ) . momentum distribution and fragmentation functions are largely determined from known regge intercepts @xcite @xcite . the momentum distribution function of a valence quark in a hadron behaves as @xmath41 . as in the parton model , this behaviour results from the intercept 1/2 of a reggeon trajectory . thus , in average , the valence quark in a proton is slow and the diquark is fast due to energy conservation . both momentum distribution and fragmentation functions are assumed to be universal , i.e. the same in all hadronic and nuclear interactions . this property gives to the model a great predictive power . finally , individual strings are assumed to be independent . in this way , the hadronic spectra of a given graph are obtained by adding up the corresponding ones for the individual strings . this leads to a picture , in which , for any individual graph , particles are produced with only short - range ( in rapidity ) correlations . long - range correlations ( and a broadening of the multiplicity distributions ) are due to fluctuations in the number of strings , i.e. to the superposition of different graphs with their corresponding weights . this gives a simple and successful description of the data in hadron - hadron and hadron - nucleus interactions @xcite @xcite @xcite . the generalization of eq . ( [ 1e ] ) to nucleus - nucleus collisions is rather straighforward . for simplicity let us consider the case of @xmath42 collisions and let @xmath43 and @xmath44 be the average number of participants of each nucleus and the average number of binary @xmath45 collisions , respectively . at fixed impact parameter @xmath46 , we have @xcite [ 2e ] & & dn^aa dy(b ) = n_a(b ) + & & + ( n(b ) - n_a(b ) ) 2 k n_(b)^q_s-_s(y ) @xmath47 and @xmath48 are computed from the standard formulae in the glauber model . the physical meaning of eq . ( [ 2e ] ) is quite obvious . the expression in brackets corresponds to a @xmath45 collision . since @xmath43 nucleons of each nucleus participate in the collision , this expression has to be multiplied by @xmath43 . note that in eq . ( [ 1e ] ) the average number of collisions is @xmath49 and the number of strings @xmath50 . in the present case the total average number of collisions is @xmath51 and the number of strings @xmath52 . the second term in eq . ( [ 2e ] ) is precisely needed in order to have the total number of strings required by the model . note that there are @xmath53 strings involving the valence quarks and diquarks of the participating nucleons . the remaining strings are necessarily stretched between sea quarks and antiquarks . the value of @xmath54 is given by @xmath55 with @xmath56 @xmath54 represents the total average number of inelastic collisions suffered by each nucleon ( for more details see section 5 ) . we see from eq . ( [ 2e ] ) that @xmath57 is obtained as a linear combination of the average number of participants and of binary collisions . the coefficients are determined within the model and depend on the impact parameter via @xmath54 . note that the presence of a term proportional to the number of binary collisions is a general feature of rft and is not related to minijet production . as discussed in section 2.1 the average invariant mass of a string containing a diquark at one end is larger than the one of a @xmath32-@xmath33 string since the average momentum fraction taken by a diquark is larger than that of quark . it turns out that the same is true for the central plateau , i.e. : @xmath58 . let us now consider two limiting cases : [ 3e ] if n^q_s-_s(y^ * ~0 ) n^qq - q_v ( y^ * ~0 ) , dn^aa dy(y^ * ~0 ) ~n_a ~a^1 [ 4e ] if n^q_s-_s(y^ * ~0 ) ~n^qq - q_v ( y^ * ~0 ) , dn^aa dy(y^ * ~0 ) ~n ~a^4/3 . in the first case we obtain a proportionality in the number of participants @xmath43 whereas in the second case we obtain a proportionality in the number of binary collisions . since @xmath59 , the latter result implies that @xmath60 , i.e. all unitarity corrections cancel and we obtain the same result as in the impulse approximation ( born term only ) . this result is known as the abramovsky - gribov - kancheli ( agk ) cancellation and is valid for a general class of models which includes the glauber and eikonal ones . it implies that , for the inclusive cross - section , soft and hard processes have the same @xmath61-dependence . however , the agk cancellation is violated by diagrams related to the diffraction production of large - mass states the so - called triple pomeron or enhanced diagrams . these diagrams give rise to shadowing corrections as discussed below . their effect is very important in nuclear collisions since they are enhanced by @xmath62 factors . in appendix a , we discuss the physical content of the agk cutting rules and their practical realization in the probabilistic glauber model . it is shown there that multiple scattering diagrams , resulting from the @xmath19-channel iteration of the born term , give non - vanishing contributions to the total cross - sections ( shadowing ) . however , in the case of the single particle inclusive cross - section , these contributions cancel identically ( agk cancellation ) , provided the measured particle has been produced in an inelastic interaction ( cut pomeron ) . if , on the contrary , the trigger particle is produced in the vertex function ( blob ) of the multiple scattering diagram , one obtains the same shadowing effects than in the total cross - section . this is the physical origin of the agk violations present in dpm ( see section 2.2 ) . it is clear that if the blob has a small extension in rapidity , production from the blob will mainly contribute to the fragmentation region . therefore , at mid - rapidities , and sufficient large energy , the agk cancellation will be valid . let us consider next the contribution to the total cross - section resulting from the diffractive production of large mass states . clearly , this is equivalent to an increase of the rapidity extension of the blob which , in this case , can cover the mid - rapidity region . therefore , shadowing corrections to the single particle cross - section will be present in this case , provided the measured particle is part of the diffractively produced system . as shown in appendix a , the shadowing correction is just given by the diffractive cross - section with negative sign . ( this is exactly true only for purely imaginary amplitudes ) . the theoretical expression of the diffractive cross - section is well - known . an important part is given by the triple - pomeron term . it has also been measured experimentally and , thus , the shadowing corrections can be computed with no free parameters . considering for simplicity only the contribution of the triple pomeron term , the effect of the shadowing corrections is obtained @xcite by multiplying eq . ( [ 2e ] ) by [ 5e ] r_ab(b ) = d^2s f_a(s ) f_b(b - s ) t_ab(b ) where [ 6e ] f_a(b ) = t_a(b ) 1 + a f(s ) t_a(b ) . here the function @xmath63 is given by the integral of the ratio of the triple pomeron cross - section @xmath64 at @xmath65 to the single pomeron exchange cross - section @xmath66 : [ 7e ] . f(s ) = 4 _ y_min^y_max dy 1 _ p(s ) d^2 ^ppp dy dt |_t=0 = c with @xmath67 , where @xmath68 is the squared mass of the diffractive system . for a particle produced at @xmath69 , @xmath70 and @xmath71 . @xmath72 and @xmath73 is a constant proportional to the triple pomeron coupling . @xmath74 is the nuclear radius , @xmath75 the nuclear profile function and @xmath76 . ( [ 5e ] ) to ( [ 7e ] ) can be derived only when the triple pomeron coupling is small and , thus , the second term in the denominator of ( [ 5e ] ) is small compared to the first one . in this case , we have @xmath77^{-1 } \sim 1 - af(s ) t_a(b)$ ] , and only the contribution of the triple pomeron graph is involved in the shadowing . in general higher order rescatterings are also present . they are model dependent . the denominator in eq . ( [ 5e ] ) correspond to the sum of all `` fan '' diagrams with pomeron branchings ( generalized schwimmer model @xcite ) . it is interesting to study the @xmath61-dependence of the shadowing corrections in the limit of large triple pomeron coupling ( when the first term in the denominator of ( [ 6e ] ) can be neglected ) . in this case we find @xmath78 , i.e. the @xmath79 behaviour resulting from the agk cancellation is reduced to @xmath80 . this limit was considered by kancheli many years ago @xcite . note that shadowing corrections to inclusive spectra are not especific to soft processes . the triple pomeron terms described above are also responsible for shadowing in hard processes . at sps energies the limit given by eq . ( [ 4e ] ) is not reached , and eq . ( [ 2e ] ) leads to an @xmath61 dependence of @xmath57 at @xmath82 in @xmath83 with @xmath84 only slightly above unity . ( @xmath85 between 2 and 370 participants ) . on the other hand , shadowing corrections are small due to phase space limitations ( @xmath86 in eq . ( [ 7e ] ) ) . the results @xcite for @xmath87 @xmath87 collisions at @xmath88 gev are shown in fig . we see that both the absolute values and the centrality dependence are well reproduced . when the energy increases , eq . ( [ 4e ] ) shows that the value of @xmath84 should increase towards @xmath89 , in the absence of shadowing corrections . however , the effect of the latter is increasingly important and , as a result , the value of @xmath84 varies little with @xmath19 . at @xmath37 gev , without shadowing corrections the @xmath61-dependence is @xmath83 , with @xmath90 in the same range of @xmath91 a value which is not far from the maximal one , @xmath92 from eq . ( [ 4e ] ) . with the shadowing corrections the @xmath61-dependence is much weaker ( lower line of the shaded area in fig . 9 ) @xcite . the @xmath61-dependence is now given by @xmath83 with @xmath93 always in the range of @xmath91 from 2 to 370 . as we see , the increase of @xmath84 from sps to rhic energies is rather small . this value of @xmath84 is predicted to change very little between rhic and lhc , where @xmath94 . for , the increase from @xmath90 to @xmath95 obtained in the absence of shadowing is compensated by an increase in the strength of the shadowing corrections , leaving the effective value of @xmath84 practically unchanged . let us define the ratio [ 8e ] r_aa(b ) = d^2n^aa / dy dp_t n(b ) d^2n^pp / dy dp_t . for central @xmath97 @xmath97 collisions at @xmath98 , we find @xmath99 when numerator and denominator are integrated over @xmath96 . clearly this value corresponds to small values of @xmath96 which give the dominant contribution to @xmath100 . this result is in agreement with the measured values of @xmath101 at @xmath102 @xcite . this was to be expected from the results presented in fig . it is interesting that these data , as well as phenix ones @xcite show approximately the same value of @xmath101 at large @xmath96 . more precisely , the data show a small increase of @xmath101 at moderate @xmath96 and , at large @xmath96 , they show a scaling in the number of participants rather than in the number of binary collisions . such a result is expected in the present approach . indeed , at large @xmath96 the shadowing corrections strongly decrease due to the increase of @xmath103 in @xmath104 ( eq . ( [ 7e ] ) ) . however , the larger threshold at large @xmath96 affects mostly the @xmath32-@xmath33 strings which have a smaller invariant mass than the @xmath31-@xmath32 ones . thus , with increasing @xmath96 we approach the limit in eq . ( [ 3e ] ) leading to a scaling in the number of participants . the small increase of @xmath101 at moderate @xmath96 is probably due to the cronin effect or to a combination of this effect and jet quenching . recently , it has been shown @xcite that the cronin effect is considerably smaller at rhic and lhc energies than at sps ones due to the change with energy of the coherence length . in the saturation model , the @xmath61-dependence of charged particle inclusive spectrum in the saturation regime ( i.e. @xmath105 , where @xmath106 is the saturation scale ) is given by @xcite [ 9e ] dn dy d^2p_t ~a^2/3 _ s ( q_s ) . comparison with previous section results indicates that , apart from logarithmic factors due to @xmath107 , we obtain the same result as in rft with maximal shadowing . as discussed in the previous section , this result is in violent disagreement with rhic data . the question is then how a reasonable description of the data has been obtained in @xcite . the answer is the following . the authors have considered @xmath100 rather than @xmath108 . by integrating over @xmath109 up to @xmath106 and assuming a @xmath96-broadening corresponding to @xmath110 they have gained one power of @xmath62 . furthermore , the factor @xmath111 increases the effective power slightly above @xmath112 reaching agreement with experiment . the problem with this explanation is that a @xmath96-broadening in @xmath62 is much larger than the one seen in the data ( which is of the order of 30 % between peripheral and central collisions ) . in order to describe it a parametrization of the saturation scale has been introduced in @xcite of the type : [ 10e ] p^2_so . with this definition , @xmath113 is the value of the saturation scale for the most central collisions ( corresponding to @xmath114 participants ) . with such an expression of the saturation scale , the @xmath61-dependence of @xmath115 is the same as that of @xmath108 within 30 % and , with the mild @xmath61-dependence of @xmath116 used before , it is not possible to describe the data . in view of that the authors use instead the following expression [ 11e ] . _ s^-1 ~ where @xmath117 is chosen to be 0.6 . note the fine tuning between the 0.61 in the numerator and 0.6 in the denominator of ( [ 11e ] ) . as a consequence , the value of @xmath116 for peripheral collisions is extremely small and @xmath116 increases by a large factor between peripheral and central collisions . with this fine tuning agreement with experiment is recovered . note that for peripheral collisions ( @xmath118 ) one is practically sitting on the landau pole ( i.e. the argument of the @xmath119 is very close to unity ) . note also that with @xmath120 mev , the value of the saturation scale for the most central collisions is very small ( 260 mev ) . in the saturation model the @xmath61-dependence of @xmath108 grows larger with increasing @xmath96 and it turns out that the model can reproduce @xcite the measured values of the ratio @xmath101 , eq . ( [ 8e ] ) , in the range 1.5 gev @xmath121 gev . however , as discussed above the model has an @xmath61-dependence at lower @xmath96 which is too weak . thus , it is perhaps not surprising that it has the right @xmath61-dependence for larger values of @xmath96 in some range . the above considerations indicate that saturation is not reached at rhic energies . the considerations in the previous section based on rft suggest that it will not be reached at lhc either . in @xmath1 collisions the net proton ( @xmath15-@xmath122 ) distribution is large in the fragmentation regions and has a deep minimum at mid - rapidities . in contrast to this situation a much flatter distribution has been observed @xcite in central @xmath87 @xmath87 collisions at cern - sps . in view of that , several authors have claimed that the stopping in heavy ion collisions is anomalous , in the sense that it can not be reproduced with the same mechanism ( and the same values of the parameters ) used to describe the @xmath1 data . in a recent paper @xcite it has been shown that this claim is not correct . in the model described in previous sections , the net baryon can be produced directly from the fragmentation of the diquark . another possibility is that the diquark splits producing a leading meson in the first string break - up and the net baryon is produced in a further break - up . clearly , in the first case , the net baryon distribution will be more concentrated in the fragmentation region than in the second case . the corresponding rapidity distributions are related to the intercepts of the relevant regge trajectories , @xmath123 and @xmath124 , respectively , i.e. they are given by @xmath125 . here @xmath126 is the difference between the rapidity of the produced net baryon and the maximal one . in the case of the first component , in order to slow down the net baryon it is necessary to slow down a diquark . the corresponding regge trajectory is called baryonium and its intercept is known experimentally to be @xmath127 . for the second component , where a valence quark is slowed down , we take @xmath128 . @xcite or @xmath129 @xcite . we find no evidence for such a component from the existing @xmath1 and @xmath42 data . its smallness could be related to the fact that it produces an extra string of hadrons and , thus , does not correspond to the dominant topology . ] in this way we arrive to the following two component model for net baryon production out of a single nucleon [ 12e ] & & dn_(b ) dy(y ) = a c _ z_+^1-_q(0)(1 - z_+)^(b ) - 3/2 + n_sq(_(0)-_(0 ) ) + & & + ( 1 - a ) c _ z_+^1- _ qq ( 0 ) ( 1 - z_+)^(b ) - 3/2 + c + n_sq(_(0 ) - _ ( 0 ) where @xmath130 is the number of strange quarks in the hyperon @xmath131 @xmath132 , @xmath133 , @xmath104 is the maximal value of the baryon rapidity and @xmath54 is the average number of inelastic collisions suffered by the nucleon at fixed impact parameter @xmath46 ( see section 2.2 ) . the constants @xmath134 and @xmath135 are obtained from the normalization to unity of each term . the small @xmath136 behaviour is contro lled by the corresponding intercept . the factor @xmath137 is obtained by requiring that the @xmath136-fractions of all quarks at the ends of the strings , other than the one in which the baryon is produced , go to zero [ 9 - 10 ] . following conventional regge rules @xcite an extra @xmath138 is added to the power of @xmath139 for each strange quark in the hyperon . the fraction , @xmath140 , of the db breaking component is treated as a free parameter . the same for the parameter @xmath141 in the dp component which has to be determined from the shape of the ( non - diffractive ) proton inclusive cross - section in the baryon fragmentation region . it can be seen from eq . ( [ 12e ] ) that stopping increases with @xmath54 , i.e. with the total number of inelastic collisions suffered by each nucleon . this effect is present in the two terms of ( [ 12e ] ) and is a consequence of energy conservation . the question is whether or not this `` normal '' stopping is sufficient to reproduce the data . in other words whether or not the data can be described with a universal value of @xmath140 , i.e. independent of @xmath142 and the same for all reactions . ( [ 12e ] ) gives the total net baryon density , but it does not allow to determine the relative densities of different baryon species . in order to do so we use the simple quark counting rules described in appendix b. a good description of the data on the rapidity distribution of @xmath143 both at @xmath144 gev and @xmath145 gev is obtained from eq . ( [ 12e ] ) with @xmath146 , @xmath147 , @xmath148 and @xmath149 . the results are shown in table 1 at three different energies , and compared with the data . as we see the agreement is reasonable . for comparison with the nucleus - nucleus results , all values in table 1 have been scaled by the number of participants pairs in central @xmath87 @xmath87 and @xmath97 @xmath97 collisions ( @xmath150 ) . as it is well known , a pronounced minimum is present at @xmath151 . there is also a substantial decrease of the mid - rapidity yields with increasing energy . also , the mid - rapidity distributions get flatter with increasing energy since the net proton peaks are shifted towards the fragmentation regions . it is now possible to compute the corresponding net baryon production in heavy ion collisions and to check whether or not the data can be described with eq . ( [ 12e ] ) using the same set of parameters as in @xmath1 . the results @xcite for net proton @xmath152 and net baryon @xmath153 production in central @xmath87 @xmath87 collisions at @xmath144 gev and central @xmath97 @xmath97 collisions at @xmath37 gev are given in table 2 . the centrality is defined by the average number of participants @xmath154 in both cases . experimental results are given in brackets . the comparison of column 2 with the @xmath1 results in table 1 at the same energy , shows the well known change in the shape of the rapidity distribution between @xmath1 and central @xmath87 @xmath87 collisions at sps . the minimum at @xmath151 is much less pronounced in @xmath87 @xmath87 and the net proton peaks in the @xmath1 fragmentation regions are shifted to @xmath155 . more interesting are the results in columns 4 and 5 which contain the predictions for @xmath97 @xmath97 at rhic . we see that the shape of the rapidity distribution is very different from the one at sps . in conclusion , we have found that `` anomalous '' stopping is not needed in order to describe the present data . related models have been proposed in [ 29 - 31 ] . the results for heavy ion collisions are rather similar to the ones obtained from eq . ( [ 12e ] ) . however , in these models there is some increase in the size of the second component with the number of inelastic collisions , i.e. some anomalous stopping is present . strange particle production and , in particular , of multistrange hyperons has been proposed as a signal of quark gluon plasma formation . flavor equilibration is very efficient in a plasma due to large gluon densities and low thresholds . an analysis of the results at sps in the framework of the present model has been presented in @xcite . in the following we concentrate on rhic results . a general result in dpm is that the ratios @xmath156 and @xmath157 of baryon and antibaryon yields over negatives decrease with increasing centralities . this is easy to see from eq . ( [ 2e ] ) . the production from @xmath158 strings scales with the number of binary collisions . these strings have a smaller ( average ) invariant mass than the @xmath31-@xmath32 strings and , thus , are more affected by the thresholds needed for @xmath159 pair production . as a consequence , the centrality dependence of @xmath160 and @xmath161 production will be smaller than the one of negatives . ( the same effect was discussed in section 3.2 in connection with large @xmath96 production ) . the effect is rather small at rhic energies . however , it is sizable and increases with the mass of the produced baryon . in contrast with this situation , the data for @xmath162 s show no such decrease and an increase is present for @xmath3 production . data on @xmath5 production are not yet available . however , sps data clearly show a hierachy in the sense that the enhancement of baryon production increase with the mass ( or strange quark content ) of the produced baryon . the only way out we have found is to give up the assumption of string independence . until now we have assumed that particles produced in different strings are independent from each other . in the following we allow for some final state interactions between comoving hadrons or partons ( see section 8) . we proceed as follows . the hadronic densities obtained in section 2 are used as initial conditions in the gain and loss differential equations which govern final state interactions . in the conventional derivation @xcite of these equations , one uses cylindrical space - time variables and assumes boost invariance . furthermore , one assumes that the dilution in time of the densities is only due to longitudinal motion , which leads to a @xmath163 dependence on the longitudinal proper time @xmath164 . these equations can be written @xcite @xcite [ 13e ] d_i d = _ k _ k _ k _ - _ k _ ik _ i _ k . the first term in the r.h.s . of ( [ 13e ] ) describes the production ( gain ) of particles of type @xmath165 resulting from the interaction of particles @xmath49 and @xmath166 . the second term describes the loss of particles of type @xmath165 due to its interactions with particles of type @xmath49 . in eq . ( [ 13e ] ) @xmath167 are the particles yields per unit rapidity and per unit of transverse area , at fixed impact parameter . they can be obtained from the rapidity densities ( [ 2e ] ) using the geometry , i.e. the @xmath19-dependence of @xmath43 and @xmath44 . the procedure is explained in detail in @xcite where the pion fragmentation functions are also given . those of kaons and baryons can be found in @xcite . @xmath168 are the corresponding cross - sections averaged over the momentum distribution of the colliding particles . equations ( [ 13e ] ) have to be integrated from initial time @xmath169 to freeze - out time @xmath170 . they are invariant under the change @xmath171 and , thus , the result depends only on the ratio @xmath172 . we use the inverse proportionality between proper time and densities and put @xmath173 . here the numerator is given by the dpm particles densities . we take @xmath174(dn^-/dy)_{y^ * \sim 0 } = 2 $ ] @xmath175 , which corresponds to the density of charged and neutrals per unit rapidity in a @xmath1 collisions at @xmath37 gev . this density is about 70 % larger than at sps energies . since the corresponding increase in the @xmath42 density is comparable , the average duration time of the interaction will be approximately the same at cern sps and rhic about 5 to 7 fm . next , we specify the channels that have been taken into account in our calculations . they are [ 14e ] n k ( ) , ( ) k , k we have also taken into account the strangeness exchange reactions [ 15e ] ( ) k n , k ( ) , k as well as the channels corresponding to ( [ 14e ] ) and ( [ 15e ] ) for antiparticles ) , we have only kept those involving the annihilation of a light @xmath32-@xmath33 pair and production of an @xmath19-@xmath176 in the @xmath19-channel . the other reactions , involving three quarks in the @xmath177-channel intermediate state , have substantially smaller cross - sections and have been neglected . all channels involving @xmath178 have been taken with cross - section @xmath179 since only one of the @xmath180 and @xmath181 components of @xmath178 can participate to a given charge combination . for details see the first paper of @xcite . ] . we have taken @xmath182 mb , i.e. a single value for all reactions in ( [ 14e ] ) and ( [ 15e ] ) the same value used in ref . @xcite to describe the cern sps data . before discussing the numerical results and the comparison with experiment let us examine the qualitative effects of comovers interaction . as explained in the beginning of this section , without final state interactions all ratios @xmath183 , @xmath156 and @xmath157 decrease with increasing centrality . the final state interactions ( [ 14e ] ) , ( [ 15e ] ) lead to a gain of strange particle yields . the reason for this is the following . in the first direct reaction ( [ 14e ] ) we have @xmath184 , @xmath185 , @xmath186 . the same is true for all direct reaction ( [ 14e ] ) . in view of that , the effect of the inverse reactions ( [ 14e ] ) is small . on the contrary , in all reactions ( [ 15e ] ) , the product of densities in the initial and final state are comparable and the direct and inverse reactions tend to compensate with each other . baryons with the largest strange quark content , which find themselves at the end of the chain of direct reactions ( [ 14e ] ) and have the smallest yield before final state interaction , have the largest enhancement . moreover , the gain in the yield of strange baryons is larger than the one of antibaryons since @xmath187 . furthermore , the enhancement of all baryon species increases with centrality , since the gain , resulting from the first term in eq . ( [ 13e ] ) , contains a product of densities and thus , increases quadratically with increasing centrality . all our results refer to mid - rapidities . the calculations have been performed in the interval @xmath188 . in fig . 10a-10d we show the rapidity densities of @xmath160 , @xmath161 and @xmath189 versus @xmath190 and compare them with available data [ 33 - 35 ] . we would like to stress that the results for @xmath3 and @xmath4 were given @xcite before the data @xcite . this is an important success of our approach . in first approximation , the yields of @xmath15 , @xmath122 , @xmath162 and @xmath191 yields over @xmath192 are independent of centrality . quantitatively , there is a slight decrease with centrality of @xmath193 and @xmath194 ratios , a slight increase of @xmath195 and @xmath196 and a much larger increase for @xmath3 ( @xmath197 and @xmath5 ( @xmath198 . this is better seen in fig . 11a and 11b where we plot the yields of @xmath160 and @xmath161 per participant normalized to the same ratio for peripheral collisions versus @xmath91 . the enhancement of @xmath160 and @xmath161 increases with the number of strange quarks in the baryon . this increase is comparable to the one found at sps between @xmath199 and central @xmath87 @xmath87 collisions , especially for antibaryons . the ratio @xmath200 increases by 30 % in the same centrality range , between 0.11 and 0.14 in agreement with present data . the ratios @xmath201 have a mild decrease with centrality of about 15 % for all baryon species which is also seen in the data . our values for @xmath202 are : @xmath203 , @xmath204 , @xmath205 , @xmath206 , to be compared with the measured values @xcite : @xmath207 the ratio @xmath208 and has a mild increase with centrality , a feature also seen in the data . note that a single parameter has been adjusted in order to determine the absolute yields of @xmath159 pair production , namely the @xmath122 one which has been adjusted to the experimental @xmath122 value for peripheral collisions . the yields of all other @xmath159 pairs has been determined using the quark counting rules given in appendix b. the experimental data in fig . 10 are not corrected for feed - down from weak decays . if these corrections were the same ( in percentage ) for all baryon species , our results should be compared with uncorrected yields . this seems to be the case for @xmath15 , @xmath122 , @xmath162 and @xmath191 where the feed - down corrections are of the order of 20 % . as a consequence , our predictions for @xmath3 , @xmath4 , @xmath5 and @xmath6 have a 20 % uncertainty . although the inverse slopes ( `` temperature '' ) have not been discussed here , let us note that in dpm they are approximately the same for all baryons and antibaryons both before and after final state interaction the effect of final state interaction on these slopes being rather small @xcite . the na38-na50 collaboration have observed a decrease of the ratio of @xmath0 to dimuon ( dy ) cross - sections with increasing centrality in @xmath209 and @xmath87 @xmath87 collisions . the same phenomenon has been observed in @xmath199 collisions with increasing values of @xmath61 . in this case , it is interpreted as due to the interaction of the pre - resonant @xmath210 pair with the nucleons of the nucleus it meets in its path ( nuclear absorption ) . as a result of this interaction , the @xmath210 pair is modified in such a way that , after interaction , it has no projection into @xmath0 ( a @xmath211 pair is produced instead ) . the @xmath0 survival probability @xmath212 is well known ( see for instance eq . ( 7 ) of @xcite ) and depends on a single free parameter @xmath213 , i.e. the absorptive @xmath214 cross section . the na50 collaboration has shown that the @xmath0 suppression in @xmath87 @xmath87 collisions has an anomalous component , i.e. it can not be reproduced using nuclear absorption alone . two main interpretations have been proposed : deconfinement and comovers interaction . the latter mechanism has been described in section 6 for strange particle production . in the case of @xmath0 suppression , a single channel is important namely @xmath210 ( or @xmath0 ) interacting with comoving hadrons and producing a @xmath211 pair . in this case , ( [ 13e ] ) can be solved analytically . the expression of the survival probability @xmath215 can be found in @xcite ( see eq . it depends on a free parameter @xmath216 , i.e. the effective cross - section for the comovers interaction . two important sets of new data have been presented recently by the na50 collaboration on @xmath199 @xcite and @xmath87 @xmath87 collisions @xcite . before these data were available , the na50 interpretation of the data was as follows . the @xmath199 , @xmath209 and peripheral @xmath87 @xmath87 data can be described with nuclear absorption alone , with @xmath217 mb . at @xmath218 gev there is a sudden onset of anomalous suppression with a steady fall off at large @xmath219 . however , at variance with this view , the most peripheral @xmath87 @xmath87 points lied above the nuclear absorption curve which extrapolates @xmath199 and @xmath209 data . the new @xmath199 data indicate a substantially smaller value of the absorptive cross - section . however , within errors , @xmath199 and @xmath209 data can still be described with @xmath220 mb @xcite . the new @xmath87 @xmath87 preliminary data , taken in 2000 with a target under vacuum , are consistent with previous ones except for the most peripheral ones which are now lower and consistent with the nuclear absorption curve @xcite . in this way , the na50 interpretation remains valid . however , the new data lend support to the interpretation based on comovers interaction . indeed , due to the smaller value of @xmath213 there is more room for comovers interaction ( i.e. for anomalous suppression ) in @xmath209 . actually , before the new data were available , it has been argued @xcite that a value of @xmath221 mb is also consistent within errors with the old @xmath199 data . using this value and @xmath222 mb it has been possible to describe all available data within the comovers scenario @xcite there was , however , a caveat , as pointed out in @xcite . indeed , there was a mismatch of about 30 % between the absolute normalizations in @xmath209 and @xmath87 @xmath87 . actually , the ratio of the first normalization to the second one is only @xmath223 @xcite . ( this factor takes into account both the isospin correction in @xmath209 and the rescaling in energy ) . this mismatch was induced by the high values of the most peripheral @xmath87 @xmath87 data in the old na50 data . indeed , since the relative contribution of the comovers to @xmath0 suppression is larger for central collisions , the centrality dependence of the @xmath0 suppression gets flatter ( steeper ) with decreasing ( increasing ) values of @xmath216 ( at fixed @xmath213 ) . in order to reproduce the shape of the @xmath0 over @xmath224 ratio , from very peripheral to central collisions , in the old na50 analysis , a value of @xmath225 mb was required . on the other hand the new data are described with a smaller value @xmath226 mb . this decrease of @xmath216 leads to a decrease of the absolute normalization , which is now consistent with the @xmath209 one ( @xmath227 mb ) was obtained in @xcite from an analysis of @xmath209 data and old @xmath87 @xmath87 data ( which covered a much smaller centrality range ) . in @xcite the absolute normalization in @xmath209 and @xmath87 @xmath87 were in good agreement with each other . ] . the results @xcite of the comover interaction model with @xmath221 mb and @xmath226 mb are presented in fig . 12 . as in ref . @xcite the steady fall - off of the @xmath0 over @xmath224 ratio at large @xmath219 is obtained introducing the @xmath219 fluctuations . the agreement with the new na50 data @xcite is quite satisfactory . the absolute normalization is 47 . the corresponding one in @xmath209 is 45 in perfect agreement with the expectations discussed above . it is interesting to note that the data obtained using the @xmath219 calorimeter and the zero - degree calorimeter ( zdc ) analysis are consistent with each other when using the measured @xmath228 correlation . this result was predicted in ref . @xcite . next , i would like to discuss briefly the expectations for @xmath0 supression at rhic in the comovers interaction model . the calculation of the survival probability @xmath215 is quite safe . indeed , since @xmath216 is a cross - section near threshold , the same value obtained at sps should be used at rhic . the situation is quite different for @xmath212 . many authors assume that @xmath213 is the same at rhic and at sps . it has also been suggested that it can be significantly larger at rhic . however , it seems plausible that at mid - rapidities , nuclear absorption at rhic is small due to the fact that , contrary to sps , the @xmath210 pair is produced outside the colliding nuclei . it is therefore crucial to have data on @xmath0 production in @xmath199 interactions at rhic . if @xmath229 the @xmath0 suppression at rhic and sps will be comparable since the smallness of the nuclear absorption will be approximately compensated by the increase of the comovers suppression due to a larger comovers density at rhic . very preliminary data tend to indicate that this is indeed the case . a quantitative analysis of the new na50 data in the deconfining scenario is still missing . on the other hand , the centrality dependence of the average @xmath96 of @xmath0 is better described in the comovers approach than in a deconfining scenario @xcite . at rhic energies , a small nuclear absorption in @xmath199 collisions ( i.e. @xmath229 ) , would be a very interesting situation in order to discriminate the comovers interaction model from a deconfining scenario . indeed , in the latter , the shape of the centrality dependence would be almost flat for peripheral collisions ( below the deconfining threshold ) and would decrease above the threshold . such a behaviour would be a clear signal of deconfinement . on the contrary , in the comovers scenario , the fall - off would be continuous , from peripheral to central collisions , and determined by the same value of @xmath216 obtained from cern sps data . quark gluon plasma ( qgp ) formation is obtained in statistical qcd , i.e. qcd applied to a system in thermal equilibrium . therefore , one of the main issues in heavy ions physics is to determine whether or not the produced final state reaches thermal equilibrium . an argument in favor of equilibrium is the fact that particle abundances are well described in terms of statistical models . however , one should take into account that statistical models are also very successful in @xmath1 and even @xmath8 interactions . therefore , it is important to study whether or not particle abundances can be obtained in a microscopic model such as dpm . as a starting point we have assumed that particles produced in different strings are independent ( see section 2 ) . in this case thermal equilibrium can not be reached no matter how large the energy density is . indeed , in this case a large energy - density is the result of piling up a large number of independent strings . the assumption of independence of strings works remarkably well in @xmath230 and @xmath231 interactions @xcite @xcite even in the case of event samples with 5 or 6 times the average multiplicity indicating that no sizable final state interaction is present in these reactions . in nucleus - nucleus collisions , we have described charged particle inclusive production and its centrality dependence . the model exhibits a term proportional to the number of binary collisions which has been seen in the data both at sps and rhic . the presence of such a term is required by unitarity and is not due to minijets . however , it is clear that in heavy ion collisions , where several strings occupy a transverse area of 1 @xmath232 , the assumption of string independence has to break down . this is indeed the case . as we have seen , some data can not be described without final state interaction . it could have happened that this final state interaction is so strong that the string picture breaks down and becomes totally useless . this does not seem to be the case . on the contrary , present data can be described using the particle densities computed in the model as initial conditions in the gain and loss ( transport ) equations governing the final state interaction . the interaction cross - section turns out to be small ( of the order of a few tenths of a mb ) . due to this smallness and to the limited interaction time available , final state interaction has an important effect only on rare processes , in particular @xmath3 , @xmath5 and @xmath0 production . the bulk of the final state is not affected . of course it is not possible to conclude that thermal equilibrium has not reached . however , particle abundances not only do not allow to conclude that it has been reached , but , on the contrary , their centrality dependence tends to indicate that this is not the case . let us consider for instance @xmath15 and @xmath122 production . in our model , their yields are practically not affected by final state interaction , i.e. they are practically the same assuming string independence . yet , the model reproduces the data , from very peripheral to very central interaction . this success would be difficult to understand in a qgp scenario in which for peripheral collisions ( below the critical density ) there is strong , non - equilibrated , @xmath233 annihilation , which becomes equilibrated for central ones , above the critical density . more generally , the qgp scenario would be strongly supported if some kind of threshold would be found in the strange baryon yields around the critical density value . at sps energies , evidence for such a threshold in the @xmath4 yield has been claimed by the na57 collaboration @xcite . unfortunately , these data only cover a limited range of centrality . in contrast to this situation the rhic data explore the whole centrality range from very peripheral to very central collisions and the centrality dependence of the yields of @xmath15 , @xmath162 , @xmath3 and their antiparticles shows no structure whatsoever . if the same happens for @xmath5 and @xmath6 production ( as predicted in our approach ) the case for qgp formation from strange baryon enhancement will be rather weak . finally , it should be stressed that the final state interaction of comovers in our approach is by no means a trivial hadronic effect . indeed , the interaction of comovers starts at the early times where densities , as computed in dpm , are very large . in this situation the comovers are not hadrons ( there are several of them in the volume occupied by one hadron , and , moreover , at these early times hadrons are not yet formed ) . this is probably the reason why in our approach the comover interaction cross - sections required to describe the data are smaller than in a hadron gas model . before concluding , i would like to say that it is an honor and a pleasure to contribute to this special issue of acta physica polonica in hommage to my friend jan kwiecinski i would also like to acknowledge his important contributions to the model presented here , realized during his ( too rare ) visits to orsay . in particular he played an important role in the generalization of the dual parton model to heavy ions collisions @xcite and also in introducing @xcite a semi - hard component in the model . the reggeon calculus or reggeon field theory ( rft ) @xcite provides a field theoretical formulation of the eikonal ( for @xmath230 collisions ) or the glauber ( for @xmath231 and @xmath234 ) models , valid at high energies . the main difference between the rft and the glauber model is that , at high energies , the coherence length is large and the whole nucleus is involved in the interaction . moreover , due to the space - time development of the interaction , when , at high energy a projectile interacts inelastically with a nucleon of the nucleus , the formation time of ( most of ) the produced particles is larger than the nuclear size and , thus , particles are produced outside the nucleus . therefore , planar diagrams give a vanishing contribution at high energy . the relevant diagrams are non - planar , describing the `` parallel '' interactions of constituents of the projectile with the target nucleons ( in the case of an @xmath231 collision ) . this picture is in clear contrast with the glauber model , in which the projectile undergoes successive ( billiard ball type of ) collisions with the nucleons of the target . in spite of these differences , one recovers the glauber formula in first approximation . this formula corresponds to the contribution of the initial state ( on - shell projectile pole ) to the various rescattering terms . in rft one has , besides these contributions , also the contributions due to low mass and high mass diffractive excitations of the projectile . the latter are very important since , as we have seen in section 2.3 , they give rise to shadowing corrections . an important feature of rft is that it obeys to the so - called agk cutting rules @xcite . these rules allow to relate to each other the different @xmath19-channel discontinuities of a given graph , and also to relate them to the contribution of this graph to the total cross - section . in this way , they provide a powerful link between total cross - section and multiparticle production . in order to illustrate these rules , let us consider the case of an interaction of a hadron @xmath235 with two different nucleons of the target nucleus @xmath61 ( with @xmath236 spectators ) , and let us assume that the object exchanged in the @xmath177-channel of each collision is purely imaginary ( pomeron ) . let us consider the cutting by a plane in between the two interactions ( i.e. in between the two pomerons ) . we obtain in this way a diffractive intermediate state containing a large rapidity gap . let us call @xmath237 its contribution to @xmath238 . from the cutting rules we find that the inelastic contribution obtained cutting through one of the interactions ( an interference term ) has a weight @xmath239 relative to the previous one . since there are two interactions one can cut through , one obtains @xmath240 . finally , cutting by a plane through the two interactions ( which is possible since the graph is non - planar ) has a relative weight @xmath241 . this last contribution is also inelastic and has an average multiplicity which is twice that of the previous one . the total contribution of this double scattering , to @xmath238 is thus equal to @xmath242 , a negative contribution . the total contribution to the non - diffractive inelastic cross - section is @xmath243 . we see in this way that the ( negative ) contribution of a double interaction to @xmath244 is two times larger , in absolute value that its contribution to @xmath238 . in the case of @xmath44 collisions the corresponding factor is @xmath245 . let us now consider the contribution of a double interaction to the non - diffractive single particle inclusive cross - section @xmath246 . this contribution is @xmath247 . indeed , in the case of the cut through the two interactions the contribution to @xmath246 has an extra factor 2 since the triggered particle can be produced in either of the two interactions . it turns out that such a cancellation is true to all orders in the number of interactions . we obtain in this way the so - called agk cancellation . all rescattering corrections of the glauber type cancel identically in @xmath246 . only the term with a single interaction is left which is proportional to @xmath112 in @xmath199 interactions . note that the crucial ingredient in obtaining the agk cancellation is the fact that the triggered particle has been produced in a cut interaction which gives the extra factor 2 for two cut interactions . the other possibility is that , the trigger particle is emitted from the ( cut ) vertex function ( blob ) . clearly , in this case the extra factor 2 is absent and the agk cancellation is not valid . in this case the shadowing corrections are the same as in the total cross - section . the agk cutting rule described above are quite general . they are valid in any field theory in which the vertex functions obey the general properties of unitarity , crossing and large @xmath96 damping . the glauber model is a particular example in which the agk rules are valid . their derivation in this case is straightforward , as discussed below . let us consider for simplicity @xmath199 scattering . the main formula of the probabilistic glauber model is the one that gives the cross - section @xmath34 for @xmath44 inelastic collisions of the projectile with @xmath44 nucleons of the target nucleus , at fixed impact parameter @xmath46 : where @xmath249 is the proton - nucleon inelastic cross - section and @xmath75 is the nuclear profile function . this equation is just the bernoulli s formula for composite probabilities . the first factor is a trivial combinatorial factor corresponding to the different ways of choosing @xmath44 nucleons out of @xmath61 . the second one gives the probability of having @xmath44 inelastic @xmath250 collisions at given @xmath46 . the third one is the probability that the remaining @xmath251 nucleons do not interact inelastically . let us consider first a term with two collisions both of which are inelastic . the corresponding cross - section is @xmath252 i.e. a positive term . let us now consider the case of two collisions only one of which is inelastic . the corresponding ( interference ) term is @xmath253 obtained from eq . ( a.1 ) by putting @xmath254 and taking the second term in the expansion of the last factor . we get @xmath255 . we see that @xmath256 . thus , a rescattering term containing two collisions gives a negative contribution to @xmath238 . let us now consider the contribution to @xmath257 . it is given by @xmath258 . indeed , in the case of a double inelastic collision , the triggered particle can be emitted in either of them hence an extra factor 2 . this is just the agk cancellation . it is easy to see that it is valid order by order in the total number of collisions . this can also be seen as follows . the total inelastic cross - section for @xmath199 collision in the glauber model is given by the well known expression this expression contains a term in @xmath112 ( born term or impulse approximation ) . it also contains contribution from multiple scattering with alternate signs . numerically , it behaves as @xmath83 with @xmath260 . the single particle inclusive cross - section is given by we see that here multiple - scattering contributions cancel identically and only the born term is left . as a consequence of this agk cancellation the @xmath61-dependence of @xmath246 in @xmath199 interactions behaves as @xmath112 . in the case of @xmath234 collisions it behaves as @xmath234 and @xmath262 is proportional to the number of binary collisions rather than to the number of participants . we see in this way that the agk rules are trivially satisfied in the glauber model . as mentioned in section a.1 in the glauber model only the initial state is present in the vertex function ( blob ) . thus a secondary can only be produced in an interaction and the agk cancellation is exact . in a general theory with a more complicated vertex function , the triggered particle may be produced in the blob . as discussed in section a.2 this gives rise to a violation of the agk cancellation which is responsible for the shadowing corrections to the inclusive spectra . in order to get the relative densities of each baryon and antibaryon species we use simple quark counting rules @xcite @xcite . denoting the strangeness suppression factor by @xmath263 ( with @xmath264 ) , baryons produced out of three sea quarks ( which is the case for pair production ) are given the relative weights for net baryon production , we have seen in section 5 that the baryon can contain either one or two sea quarks . the first case corresponds to direct diquark fragmentation described by the second term of eq . ( [ 12e ] ) . the second case corresponds to diquark splitting , described by the first term of ( [ 12e ] ) . in these two cases , the relative densities of each baryon species are respectively given by in order to take into account the decay of @xmath275 into @xmath276 , we redefine the relative rate of @xmath162 s and @xmath277 s using the empirical rule @xmath278 ) keeping , of course , the total yield of @xmath162 s plus @xmath277 s unchanged . in this way the normalization constants of all baryon species in pair production are determined from one of them . this constant , together with the relative normalization of @xmath279 and @xmath280 , are determined from the data for very peripheral collisions . in the calculations we use @xmath281 @xmath282 . calculated values @xcite of the rapidity distribution of @xmath283 at @xmath144 gev and 27.4 gev ( @xmath35 ) and @xmath284 gev ( @xmath36 ) . ( in order to convert @xmath246 into @xmath100 a value of @xmath285 mb has been used ) . for comparison with the nucleus - nucleus results , all values in this table have been scaled by @xmath150 the number of participant pairs in central @xmath87 @xmath87 and @xmath97 @xmath97 collisions . data are in brackets . : the values of @xmath287 for @xmath97 @xmath97 collisions at @xmath37 gev computed @xcite from eq . ( 2 ) including shadowing corrections are given by the dark band in between solid lines . the phenix data are also shown ( black circles and shaded area ) . : ( a ) calculated values @xcite of @xmath100 of @xmath15 ( solid line ) @xmath122 ( dashed line ) , and @xmath288 ( dotted line ) at mid rapidities , @xmath289 , are plotted as a function of @xmath290 , and compared with phenix data [ 33 ] ; ( b ) same for @xmath162 and @xmath191 compared with preliminary star data [ 34 ] ; ( c ) same for @xmath268 and @xmath291 compared to preliminary star data [ 35 ] ; ( d ) same for @xmath5 and @xmath6 . : the ratio of @xmath0 over @xmath224 cross - sections in @xmath87 @xmath87 collisions a 158 gev / c versus @xmath219 obtained @xcite in the comovers interaction model with @xmath221 mb and @xmath295 mb . the absolute normalization is 47 . the preliminary data are from [ 39 ] .

we present a study of the x - ray properties of a volume - limited sample of optically selected emission - line galaxies . the sample is derived from a correlation between the kpno international spectroscopic survey ( kiss ) , an-selected objective - prism survey of agns and starbursting galaxies , and the rosat all - sky survey ( rass ) . after elimination of all spurious matches , we identify 18 rosat - detected x - ray sources within the kiss sample in the kev band . due to soft x - ray selection effects , the majority of the rosat sources are seyfert 1 galaxies . the majority ( 54% ) of the rosat - kiss seyferts are classified as narrow - line seyfert 1 objects , a relatively high percentage compared to previous objective - prism - selected seyfert galaxy samples . we estimate the x - ray luminosities of the rosat - detected kiss objects and derive volume emissivities of and for the and strips , respectively . for those kiss agns not detected by rass , we use the median ratio derived from a previous study to estimate . the total kev volume emissivity we predict for the overall kiss agn sample is sufficient to account for of the soft x - ray background ( xrb ) , averaged over both survey strips . the kiss agn sample is made up predominantly of intermediate - luminosity seyfert 2 s and liners , which tend to be weak soft x - ray sources . they may , however , represent a much more significant contribution to the hard xrb .

astro-ph0210101

active galactic nuclei ( agns ) are among the most energetic objects in the universe , with energy output spanning the entire electromagnetic spectrum . it is currently believed that the majority of soft x - ray emission by agns arises from superheating of the accretion disk surrounding the central supermassive black hole . however , the x - ray emission actually observed from agns depends strongly on the degree to which the `` central engine '' is obscured by surrounding material ( e.g. , the dusty torus model ) . the observed spectrum ( i.e. , hardness ) of the source also depends on the amount of obscuration present . this paper seeks to study the soft x - ray properties of cosmologically local ( @xmath11 ) agns . towards this end , we have constructed an optically selected sample of such agns detected as part of the rosat all - sky survey ( rass ) . in order to better understand the x - ray emission properties of local agns , we have compiled a relatively unbiased sample of these objects for study . the kpno international spectroscopic survey ( kiss ; salzer et al . 2000 , 2001 ) provides a unique sample of galaxies ideally suited for statistical analyses . the kiss sample includes all emission - line galaxies ( elgs ) with detectable h@xmath12 emission ( ew @xmath13 ) . since it is a line - selected sample with an imposed maximum wavelength , there is an implicit redshift limit associated with the survey . this limit , along with the great sensitivity of the survey , means that kiss is effectively volume - limited for the more luminous portion of the elg sample , which includes the majority of the agns . because of the unique way in which kiss galaxies are selected , our rosat - kiss comparison differs in a fundamental way from previous x - ray / optical correlation studies . most of these are x - ray selected ( e.g. , ueda et al . 1999 , griffiths et al . 1996 ) and focus on the optical properties of objects detected in the x - ray . here , we are studying the x - ray properties of an optically selected sample . this approach yields a sample of galaxies significantly different from those derived by previous authors ( see section [ properties ] ) . studies by @xcite and @xcite also use optical catalogs as the primary basis for selection . their galaxy samples are drawn from the palomar survey of nearby galaxies , a magnitude - limited survey which has full coverage of the northern sky . the sample constructed from this survey is volume - limited and complete to a comoving distance of @xmath14 18 mpc . although the sky coverage of kiss is only 128 @xmath15 in total , the sample volume is quite deep : the comoving distance limit of the survey is @xmath16 mpc . the current study complements those of halderson et al . and ho et al . by exploring the x - ray properties of a much more distant ( but still cosmologically `` local '' ) sample . on average , the rosat - kiss sample is @xmath14 100 times more luminous than the halderson et al . sample in the soft x - ray band . emission - line galaxies similar to those discovered by kiss are important components of the cosmic x - ray background ( xrb ) . the xrb was discovered 40 years ago @xcite in the 26 kev range , and has since been demonstrated to extend to both higher and lower energies ( see the review by fabian & barcons 1992 ) . in deep images obtained recently with the _ chandra x - ray observatory _ , @xmath17 8090% of the background in the soft ( 0.52 kev ) and hard ( 210 kev ) x - ray bands has been resolved into discrete sources ( e.g. , rosati et al . models ( comastri et al . 1995 ; gilli et al.2001 ) and observations of both nearby ( moran et al . 2001 ) and distant ( tozzi et al . 2001 ) x - ray sources have indicated that the majority of the hard xrb is produced by type 2 agns whose soft x - ray fluxes are absorbed by dense circumnuclear material . in contrast , at least 50% of the xrb at lower energies is produced by high - luminosity , unobscured ( type 1 ) agns and quasars ( hasinger et al . 1998 ; schmidt et al . 1998 ) . a significant fraction of the remaining soft xrb appears to be associated with optically bright galaxies of relatively low x - ray luminosity ( hornschemeier et al . 2001 ; tozzi et al . it is unclear whether these objects are low - luminosity agns , non - active galaxies , or a combination of the two . a major aim of this study , therefore , is to constrain the soft xrb contribution of moderately luminous agns . we describe the kiss and rosat catalogs and the results of the x - ray - optical matching in section 2 . the properties of the galaxies in the correlation sample are discussed in detail in section 3 . we present xrb calculations in section 4 ; we take into account contributions from both rosat - kiss objects and those whose 0.1 - 2.4 kev emission lies below the rosat detection threshold . we close with a discussion of the implications of our analysis in the context of current xrb research . kiss is a line - selected , objective - prism survey for extragalactic emission - line objects @xcite . kiss is the first purely digital spectroscopic survey for emission - line galaxies ( elgs ) . it therefore represents a significant advance beyond slitless spectroscopic surveys carried out using photographic plates . examples of such photographic surveys include the markarian @xcite , michigan ( um ; @xcite ) , case @xcite and ucm @xcite surveys . its digital nature allows kiss to reach much fainter flux limits than previous studies , and therefore to reach a substantially greater galaxian surface density ( 18.1 @xmath18 over the @xmath5 strip ; see below ) . kiss is being carried out over several `` strips '' , each @xmath19 wide in declination @xcite . the first survey area ( kr1 ; hereafter the @xmath5 strip ) covers the region ra = 12h 15 m - 17h 0 m , @xmath20 = @xmath21 - @xmath5 , coinciding with part of the cfa century redshift survey @xcite . the second ( kr2 ; hereafter the @xmath6 strip ) runs through the bootes void : ra = 11h 55 m - 16h 15 m , @xmath20 = @xmath22 - @xmath23 . the total sky coverage of these two survey regions is 128.1 @xmath15 . in the @xmath5 strip , 1128 elgs have been cataloged @xcite , while a total of 1029 have been identified in the @xmath6 strip @xcite . to date , high - resolution follow - up spectra have been obtained for 64% of the objects in the @xmath5 strip and 14% of those in the @xmath6 strip , allowing precise determination of redshift and elg activity type ( i.e. , starburst vs. agn ) . survey data were obtained in two distinct spectral regions . the blue survey ( salzer et al . 2002 ) selects elgs by the presence of the [ o iii ] @xmath245007 emission line ; the red survey ( see salzer et al . 2001 ) selects _ via _ the h@xmath12 line . this paper focuses on the red spectral data : only the h@xmath12-selected portion of the kiss survey data is considered . the spectral range of the red survey data extends from 6400 to 7200 , beginning slightly below the rest wavelength of the h@xmath12 line and cutting off before the oh molecular band at about 7240 . this implies that the h@xmath12 line is detectable to a redshift of @xmath25 . in a minority of cases a high - redshift emission - line object will be detected by kiss when another strong spectral line ( e.g. , [ o iii ] , h@xmath26 ) redshifts into the kiss wavelength range . several qsos and seyfert galaxies have been detected by kiss in this manner . because of its well - defined redshift cutoff , kiss is essentially volume - limited for the more luminous portion of the sample . some incompleteness arises for galaxies with faint h@xmath12 line strengths . however , the majority of the kiss agns are included in the volume - limited portion of the sample . for a detailed discussion of kiss completeness limits , see @xcite . due to its depth and emission line selection method , kiss is ideal for performing relatively unbiased statistical studies of the local population of star - forming galaxies and agn . the rass resulted in a sample of soft ( 0.1 - 2.4 kev ) x - ray sources which is a factor of 20 deeper than previous x - ray surveys such as the einstein medium sensitivity survey @xcite , and which has nearly full sky coverage . at a brightness limit of 0.1 cts / s , the survey covers 92% of the sky @xcite . the large effective area and short focal length of rosat , combined with the low background noise of the pspc , resulted in an instrument that was designed to study x - ray point sources , but which was nearly ideal for studying diffuse x - ray emission as well . as a result , the survey yielded the most accurate maps to date of the diffuse xrb in the 0.1 - 2.4 kev energy band @xcite . these maps have an angular resolution of approximately @xmath27 . the resolution of these maps was later improved to 12 @xcite . deep surveys later conducted by rosat @xcite revealed that 70 to 80% of this diffuse background could in fact be resolved into discrete sources to flux levels on the order of @xmath28 @xmath29 @xmath30 @xmath31 . the rass was performed using the position sensitive proportional counter ( pspc ) , an instrument with a positional accuracy on the order of 10 arcsec ( correlation with the tycho catalog of bright stars shows that 68% of bright source catalog objects are found within 13 arcsec of the optical position ; voges et al . 1999 ) . catalogs derived from rass provide detailed information about each source , including position , positional error , source count rate and error , and exposure time . in constructing the rosat - kiss database , we have combined all relevant x - ray source information with the optical data contained in the kiss tables @xcite . the x - ray data used in this paper is drawn from the bright and faint source catalogs , which are derived directly from processing of the all - sky survey data @xcite . the bright source catalog ( bsc ) consists of the brightest objects in the rass - ii database ( resulting from the second rass processing ; see voges et al . the limiting pspc count rate is @xmath32 cts / s with at least 15 source counts required for inclusion , or a detection likelihood of at least 15 . detection likelihood is defined as @xmath33 , where p is the probability that the observed photon distribution is due to random background fluctuations . these two criteria result in a source catalog of 18,811 objects . the second and larger catalog constructed from all - sky survey data is the faint source catalog ( fsc ) . the fsc is an extension of the bsc to fainter limiting count rates ; it represents the full measure of the non - spurious sources detected in the rass - ii processing , containing 145,060 objects . the criterion for inclusion in the fsc is at least 7 source counts , with a detection likelihood of 7 or greater . the median count rate for fsc sources in the two kiss areas is approximately @xmath34 cts / s . ( see section [ predictions ] for discussion of the fsc flux threshold . ) previous rass correlations have been performed with catalogs of sources at various wavelengths . these include correlations with the rosat catalog of pointed observations ( wgacat ; white et al . 1994 ) , the simbad and ned public databases , the first radio survey @xcite , and the iras catalogs @xcite . a positional - coincidence test was used to conduct the correlation between the kiss database and the bsc / fsc , based on correspondences between the coordinates of objects in x - ray and optical wavelengths . a uniform search radius of 60 arcsec was adopted : all rosat sources within 60 arcsec of a kiss elg were considered to be possible matches . the median formal uncertainty in rosat source position recorded in the bsc / fsc for rosat - detected kiss objects is @xmath14 10 arcsec , with errors as large as 25 arcsec present in the final dataset . there were a total of 22 matches in the first - pass correlation table , which includes all matches , legitimate and spurious , for rosat bsc and fsc sources appearing within 60 arcsec of a kiss object . the median difference between x - ray and optical source positions was 10.9 arcseconds , with a standard deviation of 12.0 ; any source with a positional offset greater than 30 arcseconds was therefore flagged as a possible spurious match . each source pair was verified visually using the skyview database , to rule out chance correspondences . x - ray contour images drawn from the archive and overlaid on the online digitized sky survey ( dss ) images were used to confirm positional matchesr extraterrestrische physik ( mpe ) at garching , germany . ] . for a number of sources the kiss object is too faint to be visible on the dss images . in these cases the location of the x - ray contours relative to other objects in the dss field was compared with the finder charts available for all kiss elgs ( which are substantially deeper than the dss images ) in order to verify the match . three kiss emission - line galaxies with positional offsets of greater than 30 arcsec were discarded after visual examination ( kissr 624 , 662 , and 2128 , with 57.2 , 33.7 and 48.8 arcsec positional offsets , respectively ) . another matched galaxy , kissr 229 , was found not to be an elg after acquisition of a follow - up spectrum . furthermore , the positional offset between the rosat source and kissr 229 is 24.7 arcsec . thus , kissr 229 is not likely to be the true optical counterpart for this x - ray source , since the high density of non - elgs in these fields greatly increases the odds of chance correspondence with an x - ray source position . the remaining 18 objects were judged to be likely matches after visual examination . the median x - ray / optical positional offset for these galaxies is 9.6 arcsec . this is roughly consistent with the results of the correlation by brinkman et al . ( 2000 ) of the rosat and first ( 20 cm ) catalogs . this study finds that only 2% of all rosat - first matches are chance coincidences , to an angular separation of 30 arcsec . the 18 objects described above comprise the rosat - kiss sample . their distribution between the two kiss areas and the bsc / fsc are summarized in table [ table1 ] , while their optical and x - ray characteristics can be found in table [ table2 ] . comparisons were also made between the kiss catalogs and the rosat public archive of pointed pspc observations ( wgacat ) . the correlation was conducted remotely via the heasarc server , again using a search radius of 60 arcsec . only two rass - detected sources were found to have corresponding pointed observations : kissr 281 , a seyfert 1 galaxy , and the nearby liner kissr 29 ( ngc 4278 ; for details , see section [ properties ] ) . observations of the latter object can also be found in the public archive of rosat high resolution imager ( hri ) images ; further search of this latter catalog revealed that no other kiss galaxies had been detected . for the sake of uniformity , we use only the rass pspc fluxes for these two sources in our analysis . -0.5 in = 7.0 in -0.3 in -0.5 in = 7.0 in -0.3 in -0.5 in = 7.0 in -0.3 in 0.2 in = 4.4 in 1.3 in the rosat - kiss sample consists of 18 x - ray detected sources . the basic properties of these objects are listed in table [ table2 ] . column 1 gives the kissr catalog number of each source , while column 2 lists the corresponding designation in the appropriate rass catalog . columns 3 and 4 list the ra and dec coordinates of each source ( equinox j2000 ) , taken from the kiss catalogs @xcite . the symbol @xmath35 in column 5 of the table represents the offset in arcseconds between the optical and x - ray positions of the object . in column 6 , the elg type is given for each object . the absolute and apparent b magnitudes of all rosat - kiss sources are listed in columns 7 and 8 . the b - v color of all objects is recorded in column 9 , while their redshifts can be found in column 10 . all photometric and spectroscopic data come from the kiss database . the pspc count rate for each source is given in column 11 , while the 0.1 - 2.4 kev luminosities derived from this count rate ( using appropriate spectral parameters ; see section 4.1 ) make up column 12 . follow - up optical spectra have been obtained for approximately 900 galaxies in the overall kiss sample , during the period from 1998 - 2001 . results of spectral observations are given in @xcite , @xcite , and @xcite . the spectroscopic properties of the kiss elg sample as a whole can be found in @xcite . follow - up optical spectra are available for all 18 rosat - kiss objects , making activity type classifications ( agn / starburst ) and accurate redshift determination possible in all cases . the spectra for all objects in the rosat - kiss sample are shown in figure [ figure1 ] . knowledge of the spectroscopic properties of the sample is extremely useful for establishing the nature of each source . accurate redshift data is also essential in xrb calculations requiring knowledge of the emissivity within a certain volume ( section [ xrb ] ) , where it becomes necessary to exclude luminosity contributions from sources outside the limiting h@xmath12 redshift of kiss . the spectral characteristics of the x - ray selected sample vary widely , from stellar absorption lines to broad - line qso emission . thirteen of the 18 objects in the rosat - kiss sample have been spectroscopically identified as seyfert 1 galaxies . this high percentage of seyfert 1s is most likely due to the fact that the rosat pspc detection threshold corresponds to a relatively high flux level , meaning that only the brightest soft x - ray sources will be reliably detected . the low degree of obscuration associated with broad - line objects implies that they will be more readily detected in the soft x - ray band than will narrow - line agns . four of the remaining five objects are other types of agn , including one seyfert 2 galaxy , one qso , and two liners . the remaining source , kissr 817 , is identified as a stellar x - ray source ( see section [ indiv ] ) . we note in passing that although kiss contains a sample of @xmath14 2000 starbursting galaxies , none of these objects were detected by rosat . figures [ figure2 ] and [ figure3 ] show the distribution of apparent and absolute magnitudes for both the general kiss sample and the rosat - kiss sample . the median in each case is brighter for the rosat - detected objects than for the general kiss objects . for the kiss sample as a whole , the median absolute magnitude @xmath36 . this is more than a full magnitude less luminous than the median for the rosat - kiss sample ( @xmath37 ) . the same is true for apparent magnitude , where the median @xmath38 for the general kiss population , but is @xmath39 for the rosat - kiss objects . the general kiss elg sample is heavily dominated by starbursting galaxies . for example , @xmath40 of the 663 kiss elgs with follow - up spectra from the @xmath5 strip are starbursting galaxies . only 5.3% of these galaxies are seyferts , and 5.7% are liners . of the seyfert galaxies , 34.3% exhibit broad lines ( seyfert 1s ) . there are only 19 seyfert 1 galaxies identified to date in the whole of the sample . a large fraction ( 68% ) of the seyfert 1s detected by kiss belong to the rosat - detected sample . these galaxies are among the most optically luminous objects in the kiss database ( median @xmath41 ) . an interesting aspect of the rosat - kiss seyfert 1 subsample is the high proportion of `` narrow - line '' seyfert 1 ( nls1 ) galaxies present . inspection of figure 1 reveals that many of the seyfert 1 s exhibit small fwhm values for both h@xmath26 and h@xmath12 , as well as substantial fe ii emission on either side of the h@xmath26 and [ o iii ] lines . examples include kissr 281 , 350 , 489 , 531 , 1511 , 1755 , and 1903 . these galaxies all have h@xmath26 fwhms of less than 1500 km / s and [ o iii]/h@xmath26 flux ratios of less than @xmath42 , consistent with the emission - line criteria for nls1 s described by osterbrock & pogge ( 1985 ) and goodrich ( 1989 ) . nls1 s have generally steeper soft x - ray spectra than classical seyfert 1 s ( boller , brandt , & fink 1996 ) ; as a result , previous soft x - ray selected samples of agns have included high fractions of these objects ( e.g. , stephens 1989 ; bade et al . 1995 ; grupe et al . 1999 ) . 0.2 in = 4.4 in 1.3 in remarkably , not only does a large fraction of the rosat - detected seyfert 1 sample of kiss galaxies exhibit narrow lines ( 7 out of 13 , or 54% ) , but _ a large fraction of all kiss seyfert 1 galaxies are nls1s_. in total there are 16 seyfert 1 galaxies in the kiss catalog within the h@xmath12-selected redshift region ( @xmath43 ) . of these , at least 7 ( 44% ) would qualify as nls1 s . while the fraction of sources among the general population of seyfert 1 galaxies possessing narrow lines is not well known , the numbers for the kiss sample ( which is selected in a relatively unbiased way , at least with regard to line widths ) suggest that it may be quite high . further analysis of this issue will await more complete spectroscopic follow - up of the kiss elgs . in order to find the x - ray luminosities of the objects in the rosat - kiss sample , we first must convert rass count rates to fluxes in the 0.1 - 2.4 kev band . to do this , we apply an absorbed power - law model with @xmath44 galactic + @xmath45 @xmath31 to convert pspc source counts to x - ray flux . these spectral parameters are consistent with the methods of @xcite , who determined @xmath44 galactic + @xmath45 @xmath31 through modeling of rosat pspc spectra for a low - luminosity , distance - limited agn sample . @xcite use @xmath46 , with @xmath47 equal to the galactic value , for flux calculations ( found in the electronic form of the rass correlation catalogs ) . we make use of pimms in performing the conversion between rosat pspc source counts and the @xmath1 kev flux . combining the calculated x - ray fluxes from the above spectral model with the distances calculated from measured redshifts allows us to calculate the absorption - corrected @xmath1 kev luminosity of each source . this information is given in table [ table2 ] . we use the _ _ colden _ _ software package , in which column densities are drawn from the nrao dataset @xcite , to calculate the galactic h i column density along the line of sight to each galaxy . for the rosat - kiss objects in the @xmath5 strip , the galactic component of @xmath47 has a median value of @xmath48 @xmath31 ; @xmath6 strip objects have a median galactic component of @xmath49 @xmath31 . 0.2 in = 4.4 in 1.3 in knowledge of the total x - ray emissivity over the kiss volume is necessary for an understanding of the xrb contribution made by the low - luminosity agns in the kiss sample ( section 4 ) . however , the majority of the kiss sample is not detected by rosat . since the optical luminosities of the rosat - kiss sample are not drastically different from those of the general sample ( section [ properties ] ) , we expect that the contribution to the xrb made by kiss objects below the rosat threshold will be non - negligible . this implies that an estimate of the x - ray luminosity of the non - detected kiss galaxies is necessary for an accurate estimate of the xrb contribution from the full kiss agn sample . in an attempt to constrain the x - ray luminosities of the kiss seyfert galaxies not included in the rosat - kiss sample , we derive an @xmath50 relation using the 15 lower redshift extragalactic objects for which x - ray data are available . the h@xmath12 luminosities of these galaxies are calculated using the objective - prism spectrum of each galaxy . the objective - prism fluxes are used , rather than the values from the follow - up slit spectra , because the former are calibrated in a uniform , homogeneous way ( see kr1 and gronwall et al . since the @xmath0 and [ n ii ] ( @xmath516548 ) lines are blended in the objective - prism spectra , the observed [ n ii]/@xmath0 ratio from the follow - up spectra are used to correct the objective - prism fluxes for the contribution due to [ n ii ] . in addition , the fluxes of the seyfert 2 and liner galaxies are corrected for intrinsic absorption using reddening coefficients calculated from the balmer decrement . this latter correction can be especially significant for the liners in our sample . the resulting @xmath50 relation is shown in figure 4 . in the figure seyfert 1 galaxies are plotted as filled squares , the lone seyfert 2 as a filled triangle , and the two liners as open triangles . there appears to be a fairly well - defined linear relation between @xmath8 and @xmath52 for these galaxies , at least for the seyfert 1s . the median @xmath7 ratio for the volume - limited sample is 20.7 , which is illustrated by the solid line in figure 4 . although a significant portion of the kiss agns have a soft @xmath8 too faint to allow detection by rosat , this does not eliminate the possibility that these objects may be responsible for a non - trivial xrb contribution . 0.2 in = 4.4 in 1.3 in as mentioned above , many kiss agns have not been detected by rass . in order to derive an accurate assessment of the contribution of kiss - type agns to the soft xrb , the x - ray luminosities of the undetected objects must be estimated . to do this , we employ the l@xmath53/l@xmath54 ratio and the measured h@xmath12 luminosities of the galaxies . our initial plan had been to utilize the value derived in the previous subsection for the rosat - kiss sample . however , examination of figure 4 shows that our median value for the @xmath7 ratio is dominated by the seyfert 1 galaxies in the sample , while the seyfert 2s and liners are systematically lower . since the vast majority of non - rosat - detected objects are seyfert 2s and liners , we wanted to use an @xmath7 ratio that was representative of these classes of agn . however , we did not feel comfortable using the small sample of narrow - lined rosat - kiss galaxies ( three ) to determine a median @xmath7 ratio to be used in our computations . therefore , we decided to adopt the @xmath7 ratio determined by @xcite for their local sample of mainly narrow - lined agns . these authors found a median @xmath7 of 7 . it is no great surprise that the median @xmath7 we calculate for the rosat - kiss sample is larger than that found by @xcite , since our sample is dominated by seyfert 1 galaxies . because of the greater obscuration expected for type 2 seyferts in the dusty torus model , unabsorbed emitters such as seyfert 1s are expected to have a significantly higher @xmath7 in the soft x - ray band . this is precisely what we observe in our sample . one can not simply use the value of the @xmath7 ratio to estimate @xmath8 . examination of the distribution of fsc sources with flux reveals a loss of completeness beginning at approximately @xmath55 cts / s . we therefore adopt this value as an upper limit on the pspc count rate for our non - detected objects . computing @xmath8 using the average @xmath7 ratio described above will , in some cases , lead to an estimate of the x - ray flux for an object which lies above the fsc flux limit . we therefore adopt the upper limit @xmath1 kev count rate of 0.015 cts / s for all such objects , rather than the value inferred from the @xmath7 ratio . this method yields , in turn , an upper limit for the soft xrb contribution of these non - rosat - detected kiss galaxies . figure [ figure5 ] shows the @xmath56 relation with the non - rosat kiss galaxies included . the non - detected seyfert galaxies are plotted as filled circles , while non - detected liners are open circles . objects whose @xmath8 estimate is based on the fsc flux limit , as descibed above , are indicated by an upper - limit arrow . many of the non - detected seyfert galaxies fall into this group . all other non - detections have had their x - ray luminosities predicted using the relation @xmath57 . these objects , which include many of the liners , fall along a straight line in the plot . there are a total of 65 seyfert galaxies in the kiss database . fourteen of these galaxies are detected by rosat , leaving 51 seyferts with no x - ray flux data . after elimination of those galaxies outside the kiss completeness limit ( @xmath58 ; see section 4.1 ) , 36 non - rosat - detected seyfert galaxies remain . kiss also detects a large population of liners - there are 55 liners in the combined @xmath5 and @xmath6 kiss database . most of these objects have not been detected by rosat , and nearly all of them lie within the @xmath58 completeness limit ; we predict @xmath8 using the above method for 52 liners in total . since the rosat - kiss sample consists of just 18 objects , it is possible for us to examine each one in some detail , while still making generalizations about the sample as a whole . this section details the properties of selected rosat - kiss objects . _ kissr 29 : _ there are only two liners in the rosat - kiss sample ( the other example being kissr 2059 ) . kissr 29 ( ngc 4278 ) is remarkable in that it is one of the optically brightest sources in the kiss survey ( @xmath59 ) ; this galaxy is extremely nearby ( only 8 mpc distant ) . kissr 29 is the only rosat - kiss object for which archival hri images are available . however , in the interests of uniformity we have not made use of the pointed data in our calculations . @xcite use pointed pspc observations of kissr 29 to derive an estimate of the total soft x - ray flux ( 0.1 - 2.4 kev ) of @xmath60 ergs @xmath30 @xmath31 , assuming a photon index of @xmath61 and column density @xmath62 galactic + @xmath63 @xmath31 . using the same spectral parameters , we find that the 0.1 - 2.4 kev flux for this object is @xmath64 ergs @xmath30 @xmath31 using the rass detection . _ kissr 817 : _ this is the only stellar x - ray source in our sample . kissr 817 is believed to be a cataclysmic variable star serendipitously detected by the all - sky survey ; it was detected as part of our sample due to its relatively high soft x - ray flux . the follow - up optical spectrum detected no @xmath65 emission . its spectrum resembles that of a hot white dwarf , and it is quite blue in color ( @xmath66 ) . this is consistent with the behavior of a cataclysmic variable ; the star may have been in a high state when the objective - prism survey data were gathered . this would explain the presence of the emission line in the survey data , of which no evidence was seen in follow - up observations . _ kissr 864 : _ this object is one of the five qsos identified to date in the kiss catalog . as with all high - redshift kiss objects , it was detected via the presence of a strong emission line other than h@xmath12 which had redshifted into the relevant wavelength range . in this case , the emission line was @xmath67 . kissr 864 is the most luminous object in the rosat - kiss sample , both in the optical ( @xmath68 ) and in the x - ray ( @xmath69 ) . _ kissr 1494 : _ one kiss seyfert 2 galaxy , kissr 1494 , was detected by the all - sky survey . this galaxy is one of the more luminous sources in our rosat sample , with an absolute magnitude of -20.89 . the median absolute magnitude for the overall sample of 46 kiss seyfert 2 galaxies is @xmath70 . _ kissr 1802 _ : this object is the only high - redshift seyfert 1 galaxy detected by rosat - kiss . the h@xmath26 and [ o iii ] emission lines are redshifted into the kiss bandpass , revealing a redshift of @xmath71 . as one might expect , this object is a very powerful x - ray source , with @xmath72 , making it the second most luminous source in the rosat - kiss sample . _ seyfert 1 objects : _ for our purposes , all galaxies exhibiting broad balmer emission lines and with @xmath73 are termed seyfert 1 galaxies . this includes a few `` intermediate '' seyferts , such as the seyfert 1.5 galaxies kissr 1503 and kissr 1802 . the thirteen seyfert 1 galaxies in the rosat - kiss sample have a very high median luminosity : @xmath74 as compared to the general median @xmath75 for all kiss objects . the median @xmath76 for the rosat - kiss sy 1 s is statistically indistinguishable from the value for the general kiss sy 1 sample . as discussed above , the proportion of narrow - line seyfert galaxies ( h@xmath26 fwhm @xmath77 km / s , [ o iii]/h@xmath78 3 ) in the rosat - kiss sample seems to be significantly higher than that found in previous optically - selected agn samples . determination of the relative contributions of various source populations to the x - ray background is of major importance . as summarized in section 1 , a modest fraction of the soft xrb may well be due to a population of moderate - luminosity agns . in this section , we perform comoving volume and emissivity calculations in order to estimate the xrb contribution from the types of agns detected locally by kiss . we assume a cosmology of @xmath79 @xmath80 , @xmath81 , @xmath82 for all computations . computation of the total volume covered by kiss ( necessary for volume emissivity determination ; see section [ xrb ] ) requires us to set a maximum redshift for the survey . the redshift cutoff for kiss objects does not occur instantaneously : we experience significant `` edge effects '' for objects near the cutoff redshift , due to loss of flux from the relevant emission line ( h@xmath12 ) . studies @xcite indicate that this effect begins to become significant at approximately @xmath83 . we therefore do not consider any kiss objects with spectroscopic redshifts larger than @xmath58 . this results in elimination of one source ( kissr 1903 ; z=0.094 ) . in addition , we do not consider the two high - redshift objects noted in section [ indiv ] ( kissr 864 and 1802 ) during computation of volume emissivities . edge effects are present not only at the outskirts of a survey volume , but at the inner limit as well ; the distribution of extremely nearby sources will be heavily influenced by the local structure of the universe . we have therefore excluded from our calculations the only rosat - kiss source close enough to experience these effects - the liner , kissr 29 ( ngc 4278 ) . after the elimination of all sources outside the relevant volume , we are left with a sample of 13 galaxies . this sample , selected from a nearly volume - limited survey , should be representative of the local x - ray - emitting galaxy population as a whole . accurate estimates of the emissivity produced by the objects in the kiss volumes are necessary to obtain a knowledge of the xrb contribution from these objects . once the emissivity over the actual volume is known , it can be generalized to all redshifts using the appropriate evolution model ( see below ) in order to estimate the x - ray intensity over the sky area in question , which can then be compared with the known intensity of the xrb . the areal coverage of the kiss strips is , in fact , fairly small : the @xmath5 strip covers an area of 62.16 @xmath84 , while the @xmath6 strip covers 65.86 @xmath84 . the volume encompassed by the survey is nevertheless quite large relative to magnitude - limited surveys , because of the comparatively large redshift limit . the comoving distance limit corresponding to our adopted @xmath85 is @xmath14 343 mpc . this leads to volume emissivities in the @xmath1 kev band of @xmath86 ergs @xmath30 @xmath87 for the @xmath5 strip and @xmath88 ergs @xmath30 @xmath87 for the @xmath6 strip . the quoted errors are simply poissonian , based on the numbers of objects detected in the two survey strips . we have opted to employ poisson statistics in computing our uncertainties since small number statistics will dominate over the uncertainties in deriving the x - ray luminosities for the individual objects . our approach leads to more conservative estimates of the uncertainties . it is worthwhile to note that the volume emissivities from the two kiss strips differ by a factor of over four . this is largely a result of the small number of objects in each sample . in fact , much of the difference between the xrb contributions from the rosat - kiss objects in the two strips is due to a single , high - luminosity source in the @xmath5 strip ( kissr 924 ) . the above numbers represent the luminosity contributions from the rosat - detected subsample of the kiss objects . however , as discussed in section [ predictions ] , it is possible that those objects which are too faint to have been detected by rosat may nonetheless be significant contributors to the total volume emissivity . using x - ray luminosities estimated from the observed h@xmath12 luminosities and the adopted @xmath7 ratio , we find that inclusion of the non - detected kiss galaxies raises the volume emissivity to @xmath89 ergs @xmath30 @xmath87 in the @xmath5 strip , and to @xmath90 ergs @xmath30 @xmath87 in the @xmath6 strip . again , the errors are poissonian in nature , based on n = 66 agns in the @xmath5 strip and n = 36 in the @xmath6 strip . the inclusion of objects below the rosat detection threshold increases the volume emissivity over the kiss volumes by nearly a factor of 2 , indicating that these objects do indeed have a significant effect on the global soft x - ray emission from kiss - type objects . of course , because of the prediction method employed , the above values represent upper limits on the volume emissivity from the kiss galaxies spectroscopically identified as agns . one additional concern remains to be considered . since we do not have follow - up optical spectra for all galaxies in the kiss database , selection effects are introduced into our xrb calculations . currently , 64% and 14% of the elg sample possess follow - up spectra in the @xmath5 and @xmath6 strips , respectively . this incompleteness results in an underestimate of the xrb contribution from the kiss elgs . we use the properties of the @xmath5 and @xmath6 samples with follow - up spectra to correct for incompleteness effects . assuming the proportion of agns in the total galaxian population is constant , we use the observed number of galaxies with follow - up spectra to correct for the number of additional agns expected after completion of spectral identification . after performing this correction , we find that the total volume emissivity contributions predicted from the inclusion of non - rass detected seyfert galaxies and liners in the @xmath5 and @xmath6 strips are @xmath91 ergs @xmath30 @xmath87 and @xmath92 ergs @xmath30 @xmath87 , respectively . the errors quoted are based on the poissonian uncertainties for the known agns in the two strips , as indicated in the preceding paragraph . given these results for the emissivity within the survey volumes , we then calculate the x - ray intensities using the results of @xcite : @xmath93 we have ignored the possibility of redshift evolution of the x - ray luminosity function ( xlf ) in our calculations , which introduces significant uncertainties into our result . @xcite and @xcite use population synthesis models to calculate the evolution of the soft x - ray agn luminosity function from rosat data . the inclusion of evolutionary effects will increase the integrated x - ray intensity ( see below ) . the fact that upper limits have been used to predict the total x - ray intensity over the kiss areas due to objects not detected by rosat will tend to compensate for the lack of consideration of evolution in our calculations , but it is not possible to determine precisely to what degree this is true . using the power - law model with spectral parameters discussed in section [ refine ] , we then convert the 0.1 - 2.4 kev intensities from each survey strip to the 0.5 - 2 kev band . we can then apply the result of @xcite , namely , that the total 0.5 - 2 kev xrb intensity is @xmath94 ergs @xmath30 @xmath31 @xmath95 . using this value for the xrb intensity along with our corrected volume emissivities , we find that the total xrb contribution in the @xmath5 strip is @xmath96% , and is @xmath97% in the @xmath6 strip . the average of these two values is @xmath98 . the large uncertainty we quote in this final estimate reflects the scatter between the measurements from the two survey areas , rather than the poissonian uncertainty used in the errors for the xrb contributions from the individual survey strips . we feel that this larger value is likely to be more indicative of the actual uncertainty in our result . we stress that the predicted xrb contribution for the types of agns detected in local samples like kiss does not take into account evolution of the agn population with lookback time . however , @xcite found that application of the pure density evolution model derived for luminous seyfert 1s by @xcite leads to an _ overproduction _ of the 0.5 - 2 kev xrb by low - luminosity agns , given the non - evolving contribution of 9 - 11% they measured for such objects . since the xrb contribution found for the rosat - kiss sample is somewhat higher than the halderson et al . value , application of the miyaji evolution model will also overproduce the xrb for our sample in this energy range . while the result from this particular evolutionary model is not useful for tightening estimates of the xrb produced by low - luminosity agns , it should serve as a mild constraint on models of the evolution of x - ray luminosity from kiss - type objects . the miyaji et al . models are based primarily on high luminosity sources , for which the evolution with redshift may well differ from that of galaxies with lower @xmath8 . the depth and completeness of kiss allows the survey to make unique contributions to the study of emission - line galaxies . due to its nearly volume - limited nature , this sample will allow for an accurate determination of the statistical properties of agns . the global properties of the kiss sample should generalize with a reasonable degree of accuracy to the low - redshift agn population as a whole . thus we can use the ( statistically small ) sample of rosat - detected kiss agns to study the general properties of local active galaxies . the use of the rosat all - sky survey in correlation with kiss gives rise to certain selection effects . rosat focuses on the soft ( @xmath1 kev ) x - ray band . thus , it is to be expected that such a survey will preferentially detect larger numbers of unabsorbed seyfert 1 galaxies . this expectation is reflected in the rosat - kiss sample . nearly all of the 19 seyfert 1 galaxies present in the kiss sample have been detected by the all - sky survey ; only 6 kiss sy 1 s are not . only a very small fraction of the known kiss narrow - line agns ( seyfert 2s and liners ) are rosat - detected . in addition , none of the kiss starburst galaxies are detected . our estimate for the soft xrb contribution by rosat - kiss objects is roughly a factor of two higher than that found by @xcite for lower - luminosity agns with @xmath99 ergs @xmath30 . the different values for the soft - xrb contributions found for the two studies are due , in part , to the fact that the kiss agns are on average more luminous ( both optically and in the x - rays ) than the galaxies in the halderson et al . sample . the kiss agns are detected over a much larger volume than the very local halderson et al . sample , meaning that the intrinsically rare luminous agns have a much better chance of showing up in kiss . another factor is that , unlike halderson et al . who worked with a fairly complete sample of x - ray - detected agns , our calculations of the total volume emissivity and soft xrb contribution from kiss galaxies include predictions ( often upper limits ) of the x - ray luminosities of non - detected kiss objects . the inclusion of non - detected seyfert and liner - type objects in calculations of the xrb contribution from rosat - kiss leads to a significant increase in our derived value . the total contribution to the 0.5 - 2.0 kev xrb is approximately @xmath10 after corrections have been performed , representing the average of the values from the @xmath5 and @xmath6 strips ( @xmath100 and @xmath101 , respectively ) . the difference between the xrb contributions from the @xmath5 and @xmath6 strips remains large ( more than a factor of 2 ) even after correcting for the incomplete nature of kiss follow - up spectral data . this variation may be due in part to differences in large - scale structure between these two regions of the sky , but is primarily due to the presence of one or two very luminous sources in the @xmath5 strip , with no corresponding luminous objects detected in the @xmath6 strip . our average xrb contribution represents an upper limit to the 0.5 - 2 kev x - ray xrb contribution since a significant number of the kiss objects were not detected by rosat and were therefore assigned upper - limit values for @xmath8 . our results are consistent with recent research showing that high - redshift , more luminous agns are the dominant source of xrb emission in both the hard and soft energy bands . although objects like those in the kiss catalog can not be the dominant source of the soft xrb , it is possible that their contribution may become much more important when viewed in hard x - rays , where the effects of absorption are far less significant . because kiss is a local sample of galaxies , it is expected that the agns in kiss will be less luminous than a similar sample at higher redshift , due to a combination of sensitivity and evolution effects . we have not included such evolution in our calculations , a fact which may have significant impact on our results . several authors @xcite show that the luminosity function of agns does in fact evolve significantly with redshift , and @xcite succeed in correcting for evolutionary effects when generalizing local volume emissivity to xrb contribution . however , there is some evidence that the evolution with redshift is strongest for the most luminous sources , and hence may not be directly applicable to our sample . in the absence of an evolutionary model which does not overproduce the xrb when applied to the rosat - kiss sample , our estimate for the soft xrb contribution is a reasonable first - order result . we have conducted a positional correlation between the kpno international spectroscopic survey and the rosat all - sky survey databases . there are 18 non - spurious matches in total , 72% ( 13 ) of which are spectroscopically identified as seyfert 1 objects using follow - up spectral observations . accurate , high - resolution optical spectra are available for all rosat - kiss objects . the fact that most soft x - ray detected kiss objects are identified as seyfert 1 galaxies is explained _ via _ selection effects . type 1 seyferts , being less highly absorbed , are expected to be far more luminous in the soft x - ray , and their preferential detection by rosat - kiss comes as a natural consequence of the sample selection . a large proportion ( 54% ) of the seyfert 1s in the rosat - kiss sample are in fact narrow - line seyfert 1 galaxies ; this is a reflection of the high fraction of nls1s in the general kiss sample ( 44% of seyfert 1s with follow - up spectra available to date ) . notable non - seyfert 1 objects within the rosat - kiss sample include one seyfert 2 galaxy and two liners ; the liners are intrinsically faint objects , which may have been detected due to their proximity . several rosat - kiss sources lie outside our adopted volume limits ; these objects have been discarded from subsequent calculations . rosat - kiss includes one of the five qsos in the kiss database , kissr 864 . we also note the presence of a stellar object , kissr 817 , which is bright in the soft x - ray . kiss provides a nearly complete sample of elgs over a well - defined volume in space . we therefore expect that the xrb contributions estimated from the rosat - kiss correlation will yield an accurate representation of the local contribution of agns to the soft xrb . the locality of the kiss volume may imply inaccuracies in our result due to the fact that evolution has not been taken into account , but application of the density evolution model of @xcite result in overproduction of the soft xrb . we estimate upper limits for the 0.1 - 2.4 kev @xmath8 for kiss objects not detected as x - ray sources by rosat , using the median @xmath7 from a previous study . we also predict the number of additional agns expected to be identified after follow - up spectra have been obtained for all kiss objects , and adjust the emissivity accordingly . we find that the upper limit for the xrb contribution from kiss agns is @xmath10 , the average over the two kiss areas . this result is a factor of two higher than the 9 - 11% contribution estimated by @xcite for low - luminosity agns . the differences between the two studies is primarily the depth , and hence luminosity ranges , sampled by the two surveys . it should be remembered that the kiss - dervied xrb contribution is an upper limit , due to the manner in which the x - ray luminosities are estimated for our non - rosat - detected sources . this project has taken a different approach to the study of the xrb . instead of studying the optical properties of an x - ray selected sample , we examine the x - ray properties of a sample of optically luminous galaxies . future studies of the global x - ray emission properties of volume - limited galaxy samples may prove highly significant in constraining the xrb contribution from individual galaxian populations , as well as the redshift evolution of the x - ray luminosity function . the kiss agn sample would , in fact , be an excellent one to use for establishing the x - ray characteristics for the low - redshift , low- and intermediate - luminosity agn population . our sample size is large enough to construct an x - ray luminosity function to be used as a template for future studies of the evolution of x - ray sources . currently , the biggest limitation is the lack of quality x - ray observations for the bulk of the kiss agn sample . we gratefully acknowledge financial support for the kiss project from an nsf presidential faculty award to jjs ( nsf - ast-9553020 ) , as well as continued support for our ongoing follow - up spectroscopy campaign ( nsf - ast-0071114 ) . we also thank wesleyan university for providing additional funding for several of the observing runs during which the important follow - up spectral data were obtained . we thank the numerous kiss team members who have participated in these spectroscopic follow - up observations during the past few years , particularly caryl gronwall , drew phillips , gary wegner , laura chomiuk , jason melbourne , jeff van duyne , kerrie mckinstry , and robin ciardullo . we appreciate help from eli beckerman and paul green regarding accessing the rosat data as well as for discussions concerning the x - ray properties of agn . several useful suggestions by the anonymous referee helped to improve the presentation of this paper . finally , we wish to thank the support staffs of kitt peak national observatory , lick observatory , the hobbey - eberly telescope , mdm observatory , and apache point observatory for their excellent assistance in obtaining both the survey data as well as the spectroscopic observations . without their assistance this project would have been impossible . bade , n. , et al . 1995 , , 110 , 469 boller , t. , bertoldi , f. , dennefeld , m. , & voges , w. 1998 , a&as , 129 , 87 boller , t. , brandt , w. n. , & fink , h. 1996 , a&a , 305 , 53 brinkmann , w. , et al . 2000 , , 356 , 445 comastri , a. , setti , g. , zamorani , g. , & hasinger , g. 1995 , , 296 , 1 dickey , j. m. & lockman , f. j. 1990 , , 28 , 215d fabian , a. c. , & barcons , x. 1992 , , 30 , 429 geller , m. j. et al . 1997 , , 114 , 2205 giacconi , r. , gursky , h. , paolini , f. r. , & rossi , b. b. 1962 , , 9 , 439 gilli , r. , salvati , m. , & hasinger , g. 2001 , , 366 , 407 gioia , i.m . , et al . 1990 , , 72 , 567 goodrich , r. w. 1989 , , 342 , 224 griffiths , r. e. , della ceca , r. , georgantopoulos , i. , boyle , b. j. , stewart , g. c. , shanks . t. , fruscione , a. 1996 , , 281 , 71 gronwall , c. , salzer , j. j. , sarajedini , v. l. , jangren , a. , chomiuk , l. b. , moody , j. w. , frattare , l. m. , & boroson , t. a. 2002a , in preparation ( kr2 ) gronwall , c. , salzer , j. j. , jangren , a. , wegner , g. a. , phillips , a. c. , melbourne , j. , ciardullo , r. , & mckinstry , k. 2002b , in preparation gronwall , c. et al . 2002c , in preparation grupe , d , beuermann , k. , mannheim , k. , & thomas , h .- c . 1999 , , 350 , 805 halderson , e. l. , moran , e. c. , filippenko , a. v. , & ho , l. c. 2001 , , 122 , 637 hasinger , g. , et al . 1998 , , 329 , 482 ho , l. c. , et al . 2001 , , 549 , 51 hornschemeier et al . 2001 , , 554 , 742 macalpine , g. m. , smith , s. b. , & lewis , d. w. 1977 , , 34 , 95 macalpine , g. m. , & williams , g. a. 1981 , , 45 , 113 markarian , b. e. 1967 , astrofizika , 3 , 55 markarian , b. e. , lipovetskii , v. a. , & stepanian , d. a. 1981 , astrofizika , 17 , 619 ( english transl . astrophysics , 19 , 14 ) melbourne , j. , phillips , a. c. , salzer , j. j , gronwall , c. , & sarajedini , v. l. 2002 , in preparation miyaji , t. , hasinger , g. , & schmidt , m. 2000 , , 353 , 25 miyaji , t. , hasinger , g. , & schmidt , m. 2001 , , 369 , 49 moran , e. c. , kay , l. e. , davis , m. , filippenko , a. v. , barth , a. j. 2001 , , 556 , l75 osterbrock , d. e. , & pogge , r. w. 1985 , , 297 , 166 pesch , p. , & sanduleak , n. 1983 , , 51 , 171 rosati , p. , et al . 2002 , , 566 , 667 salzer , j. j. , et al . 2000 , , 120 , 80 salzer , j. j. , et al . 2001 , , 121 , 66 ( kr1 ) salzer , j. j. , et al . 2002 , , 123 , 1292 salzer , j. j. , gronwall , c. , jangren , a. , & mckinstry , k. 2003 , in preparation schmidt , m. , et al . 1998 , , 329 , 495 snowden , s. l. , et al . 1995 , , 454 , 643 snowden , s. l. , et al . 1997 , 485 , 125 soltan , a. m. , hasinger , g. , egger , r. , snowden , s. , truemper , j. 1996 , , 305 , 17 stephens , s. a. 1989 , , 97 , 10 stephenson , c. b. , pesch , p. , & macconnell , d. j. 1992 , , 82 , 471 tozzi , p. , et al . 2001 , , 562 , 42 ueda , y. , et al . 1999 , , 518 , 656 voges , w. , et al . 1999 , , 349 , 389 wegner , g. a. , et al . 2001 , , 122 , 2893 wegner , g. a. , salzer , j. j. , jangren , a. , gronwall , c. , & melbourne , j. 2002 , in preparation white , n. e. , giommi , p. , angelini , l. 1994 , , 6100 , 1 zamorano , j. , gallego , j. , rego , m. , vitores , a. g. , & alonso , o. 1996 , , 105 , 343 rrllrcrrrrrr 29&j122006.2 + 291643&12:20:06.8&29:16:50.3&12.1&liner&-18.06&11.53&0.99&0.00203&0.036&39.48 + 281&j132438.5 + 291020&13:24:38.7&29:10:11.9&8.6&sy 1&-19.43&17.95&0.72&0.07257&0.050&42.72 + 350&j134607.5 + 293814&13:46:08.1&29:38:10.7&9.5&sy 1&-20.97&16.51&0.47&0.07663&0.261&43.49 + 360&j135020.1 + 291005&13:50:19.7&29:10:03.2&6.3&sy 1&-18.05&19.13&1.20&0.06680&0.021&42.27 + 489&j141434.6 + 293439&14:14:34.5&29:34:28.7&10.4&sy 1&-19.61&17.85&0.50&0.07590&0.041&42.68 + 531&j142219.8 + 294257&14:22:20.2&29:42:55.5&6.1&sy 1&-20.67&15.99&0.60&0.05289&0.105&42.78 + 583&j145109.1 + 292633&14:51:09.4&29:26:26.0&8.0&sy 1&-20.19&17.30&1.22&0.07563&0.137&43.22 + 632&j150620.7 + 294005&15:06:20.8&29:39:57.3&8.4&sy 1&-19.57&17.36&0.93&0.05895&0.028&42.33 + 817&j153619.3 + 292146&15:36:18.6&29:21:50.9&11.4&cv ( ? ) & ....... & 19.34&-0.09&0.00000&0.058 & ....... + 864&j154842.8 + 290901&15:48:43.7&29:08:59.8&13.3&qso&-24.33&17.70&0.06&0.55010&0.049&44.61 + 924&j155643.0 + 294838&15:56:42.9&29:48:48.0&9.7&sy 1&-21.76&16.08&0.68&0.08554&0.457&43.91 + 1494&j131325.3 + 433219&13:13:25.8&43:32:14.2&9.6&sy 2&-20.89&15.95&0.82&0.05732&0.015&42.02 + 1503&j131406.3 + 434649&13:14:07.1&43:46:34.1&19.7&sy 1&-19.40&18.36&1.14&0.08723&0.026&42.62 + 1511&j131511.5 + 432544&13:15:10.1&43:25:46.8&21.8&sy 1&-20.16&17.58&0.62&0.08614&0.024&42.58 + 1755&j141755.1 + 431152&14:17:55.5&43:11:55.2&6.8&sy 1&-19.90&17.66&0.57&0.07902&0.201&43.41 + 1802&j143239.7 + 430021&14:32:41.1&43:00:39.8&28.5&sy 1&-21.78&19.57&0.11&0.42133&0.038&44.20 + 1903&j150506.8 + 435002&15:05:07.3&43:50:05.2&8.3&sy 1&-20.21&17.74&1.00&0.09359&0.069&43.12 + 2059&j155339.7 + 434410&15:53:40.4&43:44:04.3&11.9&liner&-19.67&16.48&0.90&0.04048&0.017&41.77 +

an overview of the recent progress in the studies of nuclear multifragmentation is presented . special emphasis is put on the exploration of isotopic trends in nuclear multifragmentation and the possibilities to extract physical information related to the nuclear equation of state . relevant experimental methods of isotope identification are described . the isotopic composition of fragments is used to extract the values of thermodynamical observables of the system undergoing multifragmentation such as temperature and chemical potentials . various methods for extraction of thermodynamical variables are analyzed . an overview of methods of isotope thermometry , exploring the sensitivity of various yield ratios to temperature , is presented . an exponential scaling of relative isotopic yields from reactions with different neutron content , called isoscaling , is used to explore the evolution of the isospin degrees of freedom of the system . finally , the nuclear equation of state and the isospin - asymmetric liquid - gas phase transition in the nuclear matter are discussed .

nucl-ex0407002

experimental studies , carried out during last two decades ( for review see @xcite and for update of the recent progress see e.g. @xcite ) , demonstrate that , in nuclear reactions such as the nucleus - nucleus collisions in the fermi energy domain or the collisions of a high - energy light particle with a heavy target , the amount of kinetic energy dissipated into the internal excitation energy is large enough to produce highly excited ( hot ) nuclei which may further de - excite via multifragmentation , a decay mode where the hot nucleus disintegrates into many fragments . in the nature , nuclear processes of such type can be observed in the interaction of cosmic rays with terrestrial and interplanetary matter . such spallation reactions contribute to the observed isotopic composition of the interstellar matter which can be used to study the properties of the cosmic rays and of their astrophysical sources ( for review see e.g. @xcite ) . with the advent of powerful accelerators of high - energy particles and , more recently , of heavy ions of intermediate and high energy , it became possible to study the process of multifragmentation in the laboratory . extensive experimental studies of multifragmentation are conducted since eighties of the last century . as can be deduced from observed experimental data @xcite , during the process of multifragmentation the nuclear medium becomes hot , expands and is supposed to undergo a phase transition where two phases are created , a dense isospin - symmetric phase ( liquid ) and a dilute isospin - asymmetric phase ( vapor or gas ) . such a phase transition was predicted by theory ( for review see @xcite ) . the most common theoretical description of multifragmentation is based on the statistical model of multifragmentation @xcite , typically assuming a thermally equilibrated freeze - out configuration with hot ( excited ) or cold fragments . the probabilities of various fragment partitions can be evaluated analytically using the grand - canonical approach . while the statistical model of multifragmentation is quite successful in describing a wide range of fragment observables , it does not provide a specific information on the mechanism of the underlying process of fragment formation , apart from the assumption that at the freeze - out configuration a thermal equilibrium among the fragments is reached . the underlying process of cluster formation can be related to the equation of state of asymmetric nuclear matter at sub - saturation densities , where the phase transition is supposed to occur . the coexistence line , typical for the one - component system such as symmetric nuclear matter , develops with introduction of isospin asymmetry into a complex coexistence region @xcite , typically suggesting a coexistence of the dilute isospin - asymmetric phase and isospin - symmetric dense phase . further theoretical investigations @xcite suggest that instability modes responsible for the phase transition are typically a combinations of mechanical and chemical instability modes . in this article , we present an overview of the recent progress in the studies of nuclear multifragmentation . we place a special emphasis on the exploration of isotopic trends in nuclear multifragmentation and the possibilities to extract physical information related to the nuclear equation of state . experimental studies of isotopic trends in nuclear multifragmentation have been made possible by use of the state - of - the - art detection techniques , which allow to identify individual isotopes of light charged particles and intermediate mass fragments in wide angular ranges . the isotopic composition of fragments can be used to extract the values of thermodynamical observables of the system undergoing multifragmentation such as temperature and chemical potentials ( free nucleon densities ) . various methods for extraction of temperature from isotopic composition are analyzed . an exponential scaling of the yields of specific isotopes from different reactions with the isotope neutron and proton numbers ( isoscaling ) is used to study the evolution of the isospin degrees of freedom of the system . finally , the nuclear equation of state and the isospin - asymmetric liquid - gas phase transition in the nuclear matter are discussed . multifragmentation is a process where the final state consists of multiple fragments which are the remnants of a hot nucleus . the aim of the multifragmentation theory is to relate the properties of final fragments to properties of the original hot nucleus . statistical phase - space models of multifragmentation typically assume that the properties of final fragments are determined in the so - called `` freeze - out '' configuration , where the typical distance of pre - fragments becomes , due to thermal expansion , larger than radius of the short - range nuclear interaction . it is , however , difficult to study the properties of the freeze - out configuration using the experimentally observed final fragments without additional model assumptions reflecting the reaction mechanism by which the hot nucleus was created . a detailed knowledge of the reaction mechanisms contributing to production of hot nuclei is necessary to properly characterize initial conditions of multifragmentation . the production of hot nuclei can be treated separately as a stage prior to multifragmentation , assuming that statistical equilibrium will be reached at later stage ( typically at freeze - out ) . in a microscopic description of nuclear medium , multifragmentation can be related to a phase transition into dense and dilute phase . isospin asymmetry of two phases may be different . furthermore , multifragmentation can be treated dynamically using the transport theory . such treatment offers the possibility to study the evolution of the nuclear medium during the whole process , including the conditions for the isospin - asymmetric phase transition . in this section we briefly review the most relevant aspects of the multifragmentation theory and related reaction theory . a main assumption of the statistical phase - space models describing various processes is that a certain stage of the process can be identified where the available phase - space determines the final probability distribution of exit channels . for instance , in statistical description of the fully equilibrated compound nucleus , the probabilities of various evaporation channels are determined by available phase space of the corresponding evaporation residue and the fission probability is determined by available phase space at the saddle - point on top of the fission barrier . in the case of simultaneous multifragmentation , the phase - space is explored for the possible multi - fragment configurations ( partitions ) , typically at the stage where short - range nuclear interaction freezes out and the identities of final fragments are determined . this is the so - called `` freeze - out '' configuration . the probabilities of various partitions can be determined statistically , exactly using the micro - canonical ensemble @xcite , or approximately using the canonical and grand - canonical ensembles @xcite . canonical and grand - canonical approximations typically reduce the mathematical complexity while raising questions on the applicability of such approximations since the nuclei are rather small systems . detailed discussion of the application of statistical mechanics to small systems can be found in @xcite . here we briefly review the most commonly used phase - space model of multifragmentation , the statistical model of multifragmentation ( smm ) @xcite . the model uses the grand - canonical approximation , the freeze - out volume depends on fragment multiplicity and internal excitation of fragments in the freeze - out configuration is considered , thus producing `` hot '' fragments . analogous models with `` cold '' fragments also exist @xcite , typically the freeze - out volume is chosen approximately twice larger than in the case of hot fragments . yield of fragments with neutron and proton numbers n and z can be , within grand - canonical approach , characterized by formula @xcite : @xmath0 where @xmath1 is the ground state spin degeneracy , @xmath2 is the free volume ( @xmath3 where @xmath4 and @xmath5 are the volumes at freeze - out and ground state ) , @xmath6 is the thermal wavelength , @xmath7 is the intrinsic free energy of the fragment , @xmath8 and @xmath9 are the free neutron and proton chemical potentials , and @xmath10 is the temperature . the free volume at freeze - out is determined as @xmath11 . the parameter @xmath12 is determined as @xmath13 where @xmath14 is the fragment multiplicity , @xmath15 is nuclear diffuseness , @xmath16 is the mass of the system and @xmath17 is the nuclear radius parameter . it is essentially a product of the additional fragment surface area and nuclear diffuseness length . the values of @xmath12 range from 0.2 to 2 . the fragment free energy is at most part determined using the liquid drop model . the component of free energy corresponding to coulomb interaction is determined using the wigner - seitz approximation . the wigner - seitz clusterization energy is calculated at volume @xmath18 . the value of @xmath19 can be related to average coulomb barrier and thus is generally different from @xmath12 , which controls the freeze - out of strong interaction . typical value is @xmath20 which is between the values for closely packed fragments ( @xmath21 ) and for fission ( @xmath22 ) . the statistical model of multifragmentation was used for extensive comparisons with experimental multifragmentation data , typically obtaining a consistent description of a wide range of experimental observables @xcite . as an alternative to models of simultaneous multifragmentation , the traditional model of compound nucleus decay based on hauser - feshbach approximation @xcite was extended to describe the emission of intermediate mass fragments ( imf ) @xcite . the emission of such complex fragments is described as a binary split , essentially an asymmetric fission where the imf is accompanied by a heavy residue . the emission probability is determined by a value of the mass - asymmetric fission barrier height . the model of fission - like binary split describes reasonably well the onset of fragment emission , thus indicating that at the onset the complex fragment emission is a fission - like process @xcite . success of such approach prompted the attempts to describe multifragmentation using the model of sequential binary decay ( sbd ) . several implementations ( e.g. the gemini code @xcite ) have been used extensively for comparisons with experimental multifragmentation data , with reasonable success in description of e.g. inclusive mass and charge distributions . however , the model of sequential binary decay fails to describe the results of the exclusive measurements at the zero angle @xcite allowing to select a narrow region around the multifragmentation threshold while smm performs rather well . an explanation is suggested by comparison of the model of sequential binary decay to transitional - state model of multifragmentation @xcite where even for binary exit channel the number of available degrees of freedom is different . furthermore , the model of sequential binary decay also fails to reproduce the observed fragment correlations @xcite . phase - space models of multifragmentation typically explore the properties of final fragments in the freeze - out configuration where nuclear interaction between the fragments does not play a role . the underlying mechanism leading to multifragmentation is nevertheless closely related to properties of the nuclear medium , which can be studied microscopically . over the recent decades , thermal properties of the nuclear medium were studied theoretically using microscopic approaches such as temperature - dependent hartree - fock @xcite or relativistic mean - field @xcite approximations . the microscopic theory of the nuclear matter predicts a phase transition of the liquid - gas type . below a critical temperature , the two phases can coexist . the spatial distribution of phases can be expected to develop from vapor bubbles in liquid at high density toward liquid droplets in vapor at low densities @xcite . a hot nucleus undergoing multifragmentation can be envisioned as rapidly expanding until it enters the spinodal region . at that stage multifragmentation occurs due to rapidly developing dynamical instabilities . when considering the isospin - asymmetric nuclear matter @xcite , the phase transition is supposed to lead to separation into a symmetric dense phase and asymmetric dilute phase . it has been discussed in literature @xcite to what extent such a phase transition is generated by fluctuations of density or concentration , typically suggesting a coupling of different instability modes . the properties of clusters formed in the phase transition can be described by a variety of random cluster models . simple model of clusterization was proposed by fisher @xcite , where the liquid drop expansion of the gibbs free energy leads to the mass distribution of clusters @xmath23 where the surface term @xmath24 decreases monotonically toward zero as temperature approaches critical temperature @xmath25 . at critical temperature the mass distribution becomes a power law with critical exponent @xmath26 . similar properties are obtained using percolation models , lattice models and cellular model of fragmentation ( an extensive overview of these models in the context of multifragmentation can be found in @xcite , sections 4.,5 . ) . calculations based on the microscopic nuclear theory @xcite suggest that nuclear matter undergoes liquid - gas phase transition which is of the first order , the density difference between phases being the order parameter . similar conclusions can be obtained also in the context of smm @xcite . however , the phase transition in the percolation models is of the second order and lattice models such as ising model @xcite exhibit a complex phase diagram with both first and second order phase transitions . furthermore , since theory of nuclear phase transitions is formulated for infinite systems while the experiments investigate multifragmentation of finite nuclei , it is necessary to assess the effect of finite size . as discussed in @xcite , the critical behavior observed in random - cluster models can be related to scaling properties and self - similarity of the system at different scales , what suggests that observed properties of finite system represent the properties of infinite media . on the other hand , finiteness constraints of small systems can lead to a behavior analogous to criticality @xcite . obviously , for any experimental observation the effect of finite size should be assessed . the statistical multifragmentation theory typically does not explicitly consider the evolution of the system prior to the freeze - out ( or underlying phase transition ) . usually it is difficult to determine the properties of the freeze - out configuration using the experimentally observed final fragments without additional model assumptions reflecting the reaction mechanism by which the hot nucleus was created . a detailed knowledge of the reaction mechanisms contributing to production of hot nuclei is necessary to properly characterize initial conditions of multifragmentation and to verify the model assumptions used by multifragmentation theory . in order to undergo multifragmentation , intrinsic excitation energy of the hot massive nuclei should exceed 2 - 3 amev . such excitation energies can be reached in nucleus - nucleus collisions with the nuclear beams of intermediate energies ( 20 - 100 amev ) . in collisions of a high - energy light particle with a heavy target the necessary energies range from hundreds of amev to agev . these are the reaction domains which have been used extensively for multifragmentation studies . a large variety of reaction mechanisms has been observed in nucleus - nucleus collisions in the fermi energy domain ( 20 - 50 amev ) depending on the impact parameter , projectile - target asymmetry and the projectile energy . the reaction mechanisms typically observed are @xcite : + - peripheral elastic and quasi - elastic ( qe ) scattering / transfer reactions around the grazing impact parameter . - deep inelastic transfer ( dit ) reactions at semi - peripheral impact parameters with partial overlap of the projectile and target and a significant part of the relative kinetic energy transferred into internal excitation energies of the projectile and target . - incomplete fusion ( icf ) reactions at central impact parameters with a typical participant - spectator scenario . at energies around the fermi energy the participant zone typically fuses with one of the spectator zones ( typically heavier ) thus creating a highly excited composite nucleus . at most central impact parameters the complete fusion ( cf ) can occur . - pre - equilibrium emission of direct particles , caused by onset of two - body nucleon - nucleon collisions at central and mid - peripheral impact parameters . pre - equilibrium emission typically precedes the icf / cf and dit reactions . + while such a scheme is rather well justified by experimental data @xcite , there are still many open questions related to the reaction mechanism , specifically concerning the development of isospin - asymmetry in nucleus - nucleus collisions in the fermi energy domain . studies of isospin degrees of freedom in the nucleus - nucleus collisions in the fermi energy domain reveal many interesting details of the reaction scenario . recently , an enhancement in the production of neutron - rich nuclei was observed in peripheral nucleus - nucleus collisions @xcite , which can be related to the effect of neutron - rich surface of the target . at projectile energies of hundreds of amev and above the impinging light particle initiates an intra - nuclear cascade @xcite of nucleon - nucleon collisions which is the main mechanism leading to production of hot nuclei . the heavy remnant after the intra - nuclear cascade is highly excited and typically undergoes multifragmentation at wide range of impact parameters @xcite . in the nucleus - nucleus collisions at similar energies the participant zone develops independently of two spectator zones and typically disintegrates due to intense cascade of nucleon - nucleon collisions , with possibility to produce sub - nuclear particles via excitation of hadronic degrees of freedom . the spectator zones are relatively colder , nevertheless typically hot enough to undergo multifragmentation @xcite . as an alternative to two - step reaction scenario , where reaction dynamics is attributed to the early stage of nucleus - nucleus collision while multifragmentation of the hot nucleus is described statistically , the whole process can be described in unified fashion by dynamical theory with explicit time - dependence . in the microscopic time - dependent hartree - fock theory @xcite , an extension of the hartree - fock theory of the ground state , the nucleus - nucleus collision is represented by time evolution of the one - body mean field , which defines the collective degrees of freedom of the system . the time - dependent hartree - fock theory does exhibit collective behavior which reminds many experimentally observed phenomena @xcite , it nevertheless does not include two - body dissipation via collisions of nucleons which is supposed to play significant role at energy regimes where the hot nuclei are produced . two - body interaction can be included in the framework of quantum kinetic theory as represented by the boltzmann transport equation , which describes the reaction dynamics in terms of both mean - field and two - body dissipation . various semi - classical methods featuring the concept of pseudo - particle are used to simulate the solution of the boltzmann transport equation . the most common approximations are vlasov - uhling - uhlenbeck ( vuu ) @xcite , boltzmann - uhling - uhlenbeck ( buu ) @xcite , landau - vlasov ( lv ) @xcite , boltzmann - nordheim - vlasov ( bnv ) @xcite and quantum molecular dynamics ( qmd ) @xcite methods . results of such simulations typically offer a qualitative description of the main features of reaction dynamics including multifragmentation . using the transport theory , the properties of the nuclear equation of state such as nuclear incompressibility and symmetry energy can be investigated phenomenologically . array nimrod @xcite . , width=359,height=207 ] experimental studies of isotopic trends in nuclear multifragmentation were enabled by the recent progress in acceleration and detection techniques . a crucial condition is unambiguous isotopic identification of detected fragments , which can be achieved using high - resolution particle telescopes . in order to reconstruct the multifragmenting system and to understand the production mechanism of the hot nuclei it is desirable to detect the charged particles in a wide angular range , ideally in the 4@xmath27 geometry . in this section we briefly review the experimental devices ( using the facilities installed at texas a&m university as examples ) and essential methods relevant to the studies of isospin degrees of freedom in nuclear multifragmentation . he , @xmath28be and @xmath29n lines used for calibration are shown as red lines . the data from reaction @xmath30sn+@xmath30sn at 28 amev were collected in ring 8 of nimrod . , width=340,height=264 ] during the process of multifragmentation , multiple charged particles are emitted in a wide angular range . angular distribution of fragments in the laboratory frame is determined by the production mechanism of the hot nucleus and by the choice of the projectile and target nuclei . the design of arrays of charged particle detectors must be optimized in order to detect and resolve emitted charged particles from the expected multifragmentation source . this can be achieved by choice of angular coverage , granularity and detection thresholds . a state - of - the - art forward array , the multidetector array faust @xcite , installed at the cyclotron institute of texas a&m university , covers forward angles between 2 - 35 @xmath31 where the fragments originating from the projectile - like source can be expected . it consists of 68 charged particle telescopes arranged into 5 rings , as shown in fig . [ ffaust ] . each particle telescope consists of 300 @xmath32 m thick silicon detector followed by 3 cm thick csi(tl ) crystal . angular ranges covered by the rings are chosen in order to distribute the multiplicity of detected particles evenly , thus avoiding the cases when one telescope is hit by multiple particles . the mass and atomic number of the detected charged particles can be identified up to z=6 . forward arrays such as faust are optimized for the study of projectile multifragmentation , specifically for asymmetric reactions where target nucleus is much heavier than the projectile . in order to study symmetric reactions where a hot composite source is created it is necessary to cover largest possible angular range , ideally 4@xmath27 . for instance , the multidetector nimrod @xcite at the cyclotron institute of texas a&m university provides such a coverage using 96 detector modules arranged in 12 rings ( see fig . [ fnimrod ] ) . light charged particles ( z@xmath332 ) are detected and identified in the whole range while a subset of modules ( 2 per ring ) offer isotopic identification up to z=8 . furthermore , atomic number of heavy residues can be resolved using ionization chambers and neutron multiplicity can also be determined since the array is surrounded by a 4@xmath27 neutron multiplicity counter ( neutron ball . ) . . the isotopic lines used for identification are shown . , width=340,height=264 ] an important step in the off - line analysis of data from the multidetector arrays with large angular coverage such as faust and nimrod is to identify as many fragment species as possible . the method of isotope identification is based on well known particle telescope technique in which the isotopes are resolved in two - dimensional @xmath34e - e spectra of energy losses in two detectors ( typically a thin one followed by a thick one ) . we developed and successfully used @xcite a method which enables to carry out identification and energy calibrations simultaneously using a minimization procedure applied to two - dimensional spectra . in the experimental spectra , the lines for three known isotopes ( typically the most characteristic isotopes such as @xmath35h , @xmath36he , @xmath28be ) are assigned and calibration is carried out by a minimization procedure where these lines are fitted to corresponding calculated energy losses for a given @xmath34e - e telescope . the calibrations coefficients are thus obtained as optimum values of minimization parameters . he , @xmath37li and @xmath28be lines were used in calibration . the data from reaction @xmath38ca+@xmath39al at 45 amev were collected in ring e of faust . , width=340 ] sn+@xmath30sn at 28 amev were collected in the ring 8 of nimrod @xcite . , width=396,height=264 ] sn+@xmath30sn at 28 amev were collected in the ring 8 of nimrod @xcite . , width=396,height=264 ] for si - si calibration , the formula @xmath40 was used for both si detectors ( @xmath41 being the amplitude of electronic signal ) . additional constraints were applied in the minimization functional in order to position correctly the punch - through points . figure [ fsisiraw ] shows an example of the experimental spectra with selected @xmath36he , @xmath28be and @xmath29n lines and the calibrated spectrum with calculated @xmath34e - e lines for a wide range of isotopes is shown in fig . [ fsisi ] . the data used were collected in ring 8 of the nimrod array in the reaction @xmath30sn+@xmath30sn at 28 amev . one can see that the overall agreement in the shape of calculated and calibrated energy losses is very good , both interpolating within and extrapolating beyond the range of selected isotopes ( at least to the neighboring element ) . the way to identification of different isotopes by linearization of experimental spectra and comparison to calculated lines is straightforward . for si - csi(tl ) calibration a similar procedure can be carried out . the difficulty when compared to si - si telescope is caused by the fact that the measured signal in csi(tl ) crystal is the light output which depends on particle energy non - linearly ( including explicit dependence on particle mass and charge ) . an approximate formula of tassan - got @xcite for the energy calibration of csi(tl ) detector was used @xmath42 where l is the light output and @xmath43 where @xmath44 is a calibration parameter . an example of obtained agreement of experimental and calculated lines is given in figure [ fsicsi ] using the data detected in the faust array in the reaction @xmath38ca+@xmath39al at 45 amev . the calibration of si detector was obtained using radioactive source of @xmath45-particles and the calibration coefficients for csi(tl ) detector were obtained using minimization procedure . due to explicit mass and charge dependence the calibration must be preceded by identification , so in this case the calculated @xmath34e - e lines are de - calibrated using calibration coefficients and projected onto experimental spectrum . again the overall agreement is quite good for both interpolation and extrapolation . the examples of obtained isotope resolution ( after linearization ) are given in figs . [ fgident],[fgident2 ] . data from the reaction @xmath30sn+@xmath30sn at 28 amev were collected in the ring 8 of nimrod . one isotope line for each z was used in linearization . the examples of particle spectra calibrated using the above methods are given in fig . [ fgspec ] . data used in fig . [ fgspec ] were collected in the reaction @xmath30sn+@xmath46si at 28 amev using nimrod @xcite . the experimental spectra are compared to simulations @xcite described in previous section . modern multidetector arrays with large angular coverage offer a possibility to study not only the inclusive observables such as isotope yields and particle spectra but allow to study development of the fragment properties with evolving reaction dynamics . typical goal of the exclusive study is to identify dynamical properties of the hot multifragmentation source being created in the early stage of the nucleus - nucleus collision . here we provide some examples for both peripheral and violent collisions . using the forward arrays like faust one typically observes the multifragmentation of the hot quasiprojectile which originates from peripheral and mid - peripheral nucleus - nucleus collision . we present here an example of the projectile multifragmentation in the reaction of @xmath46si beam with @xmath47sn and @xmath30sn targets @xcite . a hot multifragmenting source was identified using the calorimetry technique for the events where all emitted fragments were isotopically identified . the velocity distributions of the reconstructed quasiprojectiles with total charge @xmath48 were determined . resulting velocity distributions for the projectile energies 30 and 50 amev are given in fig . [ vqpex]a , b . solid squares represent the reactions with @xmath47sn target and open squares reactions with @xmath30sn . for a given projectile energy , the mean velocities and widths of distributions are practically identical for both targets . the shapes of velocity distributions are close to gaussian . velocity distributions are symmetric and no significant low or high energy tails are observed . thus , the reconstructed quasiprojectiles appear to originate from a single source which can be identified with the quasiprojectile . @xcite ( solid and open squares mean @xmath47sn and @xmath30sn target , ( a ) and ( b ) the projectile energy 30 and 50 amev , respectively ) . solid lines mean gaussian fits . , width=321 ] ( solid circles ) and @xmath48 ( solid squares ) . solid histograms mean simulated distributions for corresponding data sets , ( a ) - @xmath46si(30 amev)+@xmath47sn , ( b ) - @xmath46si(30 amev)+@xmath30sn , ( c ) - @xmath46si(50 amev)+@xmath47sn , ( d ) - @xmath46si(50 amev)+@xmath30sn.,width=283 ] -particles ( top ) from reaction @xmath30xe+@xmath30sn and spectra for 6 centrality bins ordered from top to bottom by increasing centrality @xcite . ( b ) - centrality cuts in the m@xmath49 vs. m@xmath50 histogram . skewed lines indicate centrality cuts used in ( a ) , horizontal and vertical lines indicate one - dimensional cuts in m@xmath49 and m@xmath50 histograms.,width=359 ] an apparent charged particle excitation energy of the quasiprojectile can be reconstructed for each projectile fragmentation event from the energy balance in the center of mass frame of the quasiprojectile @xmath51 where @xmath52 is the kinetic energy of the fragment in the quasiprojectile frame and @xmath53 and @xmath54 are mass excesses of the fragment and quasiprojectile , respectively . even if emitted neutrons are not included in this observable , it is useful for the light fragmenting systems , where the neutron emission is not expected to dominate . the distributions of the apparent quasiprojectile excitation energies reconstructed from fully isotopically resolved events are shown in fig . [ exsim ] for both @xmath55 and @xmath48 ( solid circles and squares respectively ) along with the distributions obtained using dit / smm simulation @xcite ( solid histograms ) . the agreement of the simulated and experimental apparent quasiprojectile excitation energy distributions is quite good . the onset of multifragmentation into channels with @xmath56 in the low energy part is described with good precision for all sets of data . while it is possible to fully reconstruct the projectile - like sources in the mass range a=20 - 30 in peripheral collisions , with increasing violence of the nucleus - nucleus collision a hot nucleus is formed from the parts of both the projectile and target . the excitation energy of such a source is high enough for emission of several intermediate mass fragments ( imf ) . information on both mass and charge of emitted imfs can be crucial to determine the emission mechanism and thus properties of the emitting source . various techniques have been developed in order to characterize the impact parameter of the collision ( for a detailed study see @xcite ) . a simple method which is nevertheless widely used to determine the centrality of collision is based on the assumption that the particle multiplicity increases monotonously with impact parameter @xcite . however , such assumption may be too simplistic and more sophisticated methods using the combinations of several observables are developed @xcite . as an example , we present here a method @xcite where the centrality bins are selected in the two - dimensional histogram of imf multiplicities ( m@xmath49 ) vs multiplicities of charged particles ( m@xmath50 ) by parallel lines chosen so that the most central bin integrates the events with highest multiplicities of charged particles and highest imf multiplicities ( see fig . [ nimspec]b ) with the intermediate event classes , thus reflecting the effect of finite size . [ nimspec]a shows that the shapes of spectra develop with increasing centrality . event shape analysis was carried out for two - dimensional m@xmath49 vs. m@xmath50 cuts as well as for the one - dimensional m@xmath49 and m@xmath50 cuts . all types of centrality cuts exhibit similar trends in terms of sphericity and coplanarity ( see fig . [ fgcnt ] ) , nevertheless some differences can be observed in the most central bins . when considering the total number of events selected by various criteria , the statistics in the most central bin of the two - dimensional selection increased by two orders of magnitude compared to the traditional m@xmath50 selection ( see fig . [ fgcnt ] ) . such a gain in statistics is caused by better separation of central and peripheral events . the one - dimensional cuts typically admix high - multiplicity peripheral events to a smaller number of genuine central events , thus decreasing the overall centrality . the use of two - dimensional criteria allows more detailed studies of central events . vs. m@xmath50 cuts ( triangles ) as well as for the one - dimensional m@xmath49 ( squares ) and m@xmath50 ( circles ) cuts @xcite . , width=283 ] the isotopic composition of the multifragmentation products such as light charged particles ( lcp ) , intermediate mass fragments ( imf ) and heavy residues is an experimental information which can be used to trace the dynamics of isospin degrees of freedom during the multifragmentation and , using grand - canonical picture , to extract the values of thermodynamical observables of the system undergoing multifragmentation such as temperature and free nucleon densities . the isotopic composition can be studied both as a bulk or in terms of yield ratios of specific isotopes . a variation of the neutron content of reaction products can be used to characterize the response of the isotopic composition to variation of initial conditions . a wide variety of fragment species is typically produced in multifragmentation of heavy nuclei . in order to obtain information on isospin dynamics , an obvious choice is to study the behavior of isotopic chains . as an example we provide in fig . [ fgyazn ] isotopic distributions of fragments with z=2 - 8 from four reactions of @xmath30sn,@xmath30xe beams with @xmath57sn targets at 28 @xcite studied at the cyclotron institute of texas a&m university using the 4@xmath27 multi - detector array nimrod @xcite . the isotopic distributions from different reactions are distinguished by different symbols . each panel shows the isotopic distributions measured at three angular ranges corresponding to rings 2 , 5 and 8 ( from top to bottom ) . sn,@xmath30xe beams with @xmath57sn targets at 28 @xcite , measured using the 4@xmath27 multi - detector array nimrod @xcite at three angular ranges . ] as one can see , the overall shapes remain quite similar at different angular ranges . it is interesting to note that the isotopes with maximum yield for z@xmath583 are typically those with a=2z+1 for both even and odd z. the effect of different reactions is most pronounced at the proton and neutron - rich tails of distributions , neutron - rich isotopes being more abundant in neutron - rich systems and vice versa , while the central part corresponding to @xmath59-stable nuclei is relatively insensitive . in any case , such comparisons of the shapes of isotopic distributions are rather qualitative . a quantitative analysis can be carried out in terms of integral observables such as momenta of distributions . to express the effect of experimental conditions such as projectile - target combination or detection angle quantitatively one can evaluate the mean n / z- ( neutron to proton ) ratio for each isotopic chain . in fig . [ fgyz ] are given the mean n / z - ratios of emitted lcps and imfs with z=2 - 8 from four reactions obtained at different angles @xcite . the overall dependence is similar at all angular ranges . he - isotopes are dominated by @xmath45-particles . mean n / z ratios are highest for li- and be - isotopes and decrease gradually with increasing atomic number of imfs . when comparing mean n / z ratios from four reactions two characteristic patterns can be distinguished . at most forward angles the mean n / z ratios of li- and be - isotopes appear to track with the isospin asymmetry of the composite projectile - target system what leads to grouping of experimental point into a typical 1 - 2 - 1 pattern . on the other hand the isotopes with z=5 - 8 appear to track with the isospin asymmetry of the target nucleus what creates a 2 - 2 pattern . at angles 11.0 - 14.9@xmath31 the 1 - 2 - 1 pattern can be identified for li- and be - isotopes while for heavier fragments the patterns vary . at angles 28.6 - 35.5@xmath31 one can recognize the 2 - 2 pattern for li - isotopes while at heavier fragments the 1 - 2 - 1 pattern dominates . in principle , one can identify the 1 - 2 - 1 pattern with the fragments emitted from a hot composite source and 2 - 2 pattern to the emission from a quasiprojectile . the fragments with z=5 - 8 at forward angles possibly originate from the binary de - excitation channels of the quasiprojectile while the 2 - 2 pattern of li - isotopes at the most central angles suggests either backward emission from the quasiprojectile or emission from quasitarget . similar isospin asymmetry patterns can be identified also in the isotopic yield distributions in fig . [ fgyazn ] . thus , the mean values of n / z ratio appear to exhibit an interesting information on isospin dynamics . the relative variation of the mean n / z values is nevertheless rather small when compared to the variation of isospin - asymmetry of the projectile - target combination and more sensitive isospin tracers should be identified . as demonstrated in the previous subsection , variation of the projectile and target nuclei influences the integral characteristics of isotopic distributions such as mean values of n / z ratio rather weakly . nevertheless , when observing the evolution of isotopic distributions in different reactions and angular ranges , one can notice that the relative yields of neighboring isotopes are rather sensitive ( especially at the tails of distributions ) . such yield ratios have been widely used in experimental studies of multifragmentation data during last 15 years @xcite . sn,@xmath30xe beams with @xmath57sn targets at 28 @xcite at three different angular ranges . the data from different reactions are characterized by a n / z - ratio of the composite system ] as an example , we present in fig . [ fgisotop ] the values of three isotopic yield ratios , again for the multifragmentation data @xcite from four reactions of @xmath30sn,@xmath30xe beams with @xmath57sn at 28 . in most cases one can observe an approximately exponential dependence of the isotopic yield ratio on the n / z ratio of the composite system ( n / z@xmath60 ) , which can be expected according to formula ( [ eqnynz ] ) when assuming linear dependence of neutron chemical potential on n / z@xmath60 . the yield ratios depend rather strongly on the angular range . such a dependence can be caused by different isospin dynamics at different angles . sn,@xmath30xe beams with @xmath57sn targets at 28 @xcite at three different angular ranges . the data from different reactions are characterized by a n / z - ratio of the composite system ( n / z@xmath60 ) . ] furthermore , we present in fig . [ fgisobar ] the values of three isobaric yield ratios for the same multifragmentation data . specifically , the isobaric ratios are evaluated for pairs of mirror nuclei . the sensitivity of isobaric ratios to n / z@xmath60 is significantly stronger than in the previous case , especially for the ratios @xmath61 and @xmath62 where the fragments @xmath63 and @xmath64 are populated rather weakly . the overall behavior is nevertheless rather irregular . especially the observed pattern at most forward angles suggests a contribution from the cold process such as projectile break - up , which appears to contribute considerably to production of isospin - asymmetric species at forward angles . such observation is in agreement with available experimental data @xcite . h)/y(@xmath65he ) on n / z ratio of the isotopically resolved quasiprojectiles with @xmath66 for nine bins of @xmath67 @xcite . ] a behavior of the yield ratio of mirror nuclei @xmath68 was investigated in the quasiprojectile multifragmentation of a @xmath46si beam with @xmath47sn and @xmath30sn targets at 30 and 50 @xcite . as described in experimental section , the properties of the system which actually undergoes multifragmentation , including n / z ratio ( @xmath69 ) , are known with good precision . [ fgy3rat ] shows dependence of the isobaric yield ratio y(@xmath65h)/y(@xmath65he ) on @xmath69 for nine bins of the apparent excitation energy per mass unit of the quasiprojectile ( @xmath67 ) . the experimental data are represented as squares and the lines represent the exponential fits . the observed exponential dependence in all excitation energy bins is in agreement with formula ( [ eqnynz ] ) . the slopes are steepest at low excitation energies and become flatter with increasing excitation energy . the apparent compliance of the yield ratio behavior to the grand - canonical picture as represented by the formula ( [ eqnynz ] ) suggests the possibility to extract the values of thermodynamic observables , in particular temperature . the procedure for extraction of the temperature was suggested by albergo @xcite . when expressing the isotopic yield within the equilibrium limit of the grand - canonical ensemble , a simplified analogue of the formula ( [ eqnynz ] ) is obtained @xcite @xmath70 the isotopic yield is thus essentially governed by free neutron and proton chemical potentials @xmath71 , temperature @xmath10 and fragment binding energy @xmath72 . the factor @xmath73 encompasses information on the global properties of the fragment partition . the ground state spin degeneracy factor @xmath74 represents the internal state sum of the fragment , which is restricted to ground state only . in oder to extract temperature , a double isotope ratio is constructed @xcite @xmath75 and the temperature @xmath10 can be obtained as @xmath76 where @xmath77 and @xmath78 . thus an estimate of the grand - canonical temperature can be determined using the isotope yields and some of their ground state characteristics . from four reactions plotted against the difference of binding energies @xcite . ] the formula ( [ tempalb ] ) was used extensively in experimental studies during the last decade @xcite . typically , the temperature determined using various thermometers ( sets of isotope yields ) exhibited large spread , possibly due to side - feeding from secondary decay . a method of correction for secondary decay was proposed by tsang et al . @xcite . in @xcite we employed a graphical method to extract the average temperature . the isotopic yield ratios , corrected for ground state spin degeneracy , are plotted as a function of binding energy difference ( see fig . [ nimtemp ] ) . the data from four reactions of @xmath30sn,@xmath30xe beams with @xmath57sn targets at 28 , used already in figs . [ fgisotop],[fgisobar ] , are investigated . individual double isotope ratio thermometers can be represented ( in the logarithmic scale ) when connecting the two points corresponding to used isotope ratios by a straight line . the inverse of the slope of such line is then the temperature . when assuming that the effect of side - feeding in the secondary decay is distributed randomly , an average temperature can be obtained by a linear fit of the systematics of all isotopic yield ratios . an information on the temperature can be obtained also from the slope of dependence of the isobaric yield ratio y(@xmath65h)/y(@xmath65he ) on neutron to proton ratio of the quasiprojectile @xmath69 . the data from reactions of a @xmath46si beam with @xmath47sn and @xmath30sn targets at 30 and 50 ( see fig . [ fgy3rat ] ) were used @xcite and the extracted temperatures compared well with the temperatures obtained using the double isotope ratio thermometer ( @xmath79h/@xmath65h)/(@xmath65he/@xmath36he ) . an assumption was employed that the values of chemical potentials can be approximated by the values of separation energies . the reason why such approximation is applicable for hot expanded nucleus can be in the possible interplay of the effect of expansion ( reducing absolute values of chemical potentials ) and separation of the isospin - asymmetric free nucleon gas from the isospin - symmetric heavier liquid phase @xcite , as indicated by observed inhomogeneous isospin distribution among light charged particles and heavier fragments @xcite . the isotopic yield can be , within the equilibrium limit of the grand - canonical ensemble , expressed in terms of the density of the free nucleon gas @xcite @xmath80 where @xmath81 . the values of yield ratios of mirror nuclei with @xmath82 then can be expressed as @xmath83 thus relating the yield ratio to the density ratio of the free neutron and proton gas . the yield ratios of the mirror nuclei @xmath68 , @xmath84 and @xmath62 were used in work @xcite to determine the ratio @xmath85 by an exponential extrapolation toward @xmath86 , yielding the values approximately 2 - 5 . while such approximation is valid for grand - canonical ensemble of infinite systems , when dealing with the finite systems such as nuclei the values of free nucleon chemical potentials depend on the fragment partition in a way guaranteeing conservation of the mass and charge of the system . thus the value of the ratio @xmath85 may differ strongly for different event classes and no globally applicable value may exist in the finite system . furthermore , the effect of secondary emission may also contribute to uncertainty in determination of @xmath85 . as shown in the previous subsection , the fragment yield ratios can be used to extract thermodynamical observables of the fragmenting system such as temperature and chemical potential . in the context of isotopic distributions , the fragment yield ratios represent the details of the distribution sensitive to isospin degrees of freedom . similar sensitivity can be explored globally by investigating the ratio of isotopic yields from two processes with different isospin asymmetry , essentially dividing the two isotopic distributions in point - by - point fashion . when employing the formula ( [ ynzalb ] ) , such a ratio will depend on n and z as follows @xcite @xmath87 where @xmath45 = @xmath88/t and @xmath59 = @xmath89/t , @xmath88 and @xmath89 are the differences in the free neutron and proton chemical potentials of the fragmenting systems . c is an overall normalization constant . alternatively @xcite the dependence can be expressed as @xmath90 thus introducing the parameters which can be related to the isoscalar and isovector components of free nucleon chemical potential since @xmath91 = @xmath92/2 t and @xmath93 = @xmath94/2 t . sn+@xmath30sn and @xmath30xe+@xmath47sn at 28 @xcite at three different angular ranges . ] an exponential scaling of @xmath95 with the isotope neutron and proton numbers was observed experimentally in the multifragmentation data from the reactions of high - energy light particle with massive target nucleus @xcite or from the reactions of mass symmetric projectile and target at intermediate energies @xcite and such behavior is called isoscaling @xcite ( the parameters @xmath96 being called isoscaling parameters ) . isoscaling behavior was further reported in the heavy residue data @xcite and also in fission data @xcite . it was shown that the values of isoscaling parameters can be related to symmetry energy @xcite , to the level of isospin equilibration @xcite and to the values of transport coefficients @xcite . an example of isoscaling behavior of the fragments with z=3 - 8 detected in the reactions of @xmath30sn+@xmath30sn and @xmath30xe+@xmath47sn at 28 @xcite is given in fig . [ fgnimiso ] . the isoscaling behavior is rather regular at all three angular ranges , the isoscaling parameter @xmath45 increases with the detection angle . such a trend of the isoscaling parameter @xmath45 can be possibly related to the fact that with increasing angle the fragments originate from still more damped collisions , with increasing level of isospin equilibration between projectile- and target - like fragment @xcite . si+@xmath97sn at 50 amev for the full set of isotopically resolved quasi - projectiles and for five bins of quasi - projectile excitation energy per nucleon @xcite . , width=302,height=302 ] in fig . [ iso50 ] are presented the isoscaling data from statistical decay of hot quasiprojectiles from the reactions @xmath46si+@xmath97sn at projectile energy 50 amev @xcite . observed charged particles with z@xmath335 were isotopically resolved and total observed charge was close to the charge of the projectile ( z=12 - 15 ) @xcite . due to the restricted set of identified fragment species , the isoscaling plots are presented , according to formula ( [ r21isobt ] ) , as a function of isospin asymmetry . the isoscaling plots are presented not only for the full data but also for five bins of excitation energy . the slopes depend on the excitation energy in a similar way as the slope of the dependence of isobaric ratio y(@xmath65h)/y(@xmath65he ) on the quasi - projectile n / z observed in @xcite ( see fig . [ fgy3rat ] ) . when compared to similar results at 30 amev , the isoscaling parameters @xmath93 for individual excitation energy bins do not depend on the projectile energy and the difference of the slopes of the inclusive data is caused by different excitation energy distributions of the hot quasi - projectiles at two projectile energies ( see fig . [ exsim ] ) . as demonstrated , isoscaling appears to be a global feature of the multifragmentation data and the values of isoscaling parameters show sensitivity to properties of the hot source . it is of interest to explore such a sensitivity in the studies of thermodynamical properties of the multifragmenting source . in the previous section we presented an overview of the methods by which the isotopic trends of the fragment data can be used to extract thermodynamical observables of the hot multifragmenting source . here we present a systematic investigation of thermodynamical properties of the well characterized hot quasi - projectiles with mass a=20 - 30 , originating from the reactions @xmath46si+@xmath97sn at projectile energy at 30 and 50 amev @xcite . signatures and argumentation supporting the conclusion that multifragmentation of such quasiprojectiles is a statistical process will be presented . furthermore , the signals for the isospin - asymmetric liquid - gas phase transition in the nuclear matter as a process underlying to multifragmentation will be investigated and implications for the nuclear equation of state at sub - saturation densities will be discussed . in order to establish whether the fragment data from reactions @xmath46si+@xmath97sn at projectile energy 50 and 30 amev @xcite originate from statistical decay of hot quasiprojectiles , it is necessary to investigate the dynamical properties of the projectile - like source and to determine the production mechanism . the observed fragment data @xcite provide full information ( with the exception of emitted neutrons ) on the decay of thermally equilibrated hot quasi - projectiles with known mass ( a=20 - 30 ) , charge , velocity and excitation energy . detailed investigation of the reaction mechanism @xcite allowed to establish a dominant reaction scenario . excellent description of the fragment observables was obtained using the model of deep - inelastic transfer ( dit ) @xcite for the early stage of collisions and the statistical multifragmentation model ( smm ) @xcite for de - excitation . the model describes well the dynamical properties of the reconstructed quasi - projectile such as center of mass velocity , excitation energy ( see fig . [ exsim ] ) and isospin - asymmetry ( see fig . [ fgnzqp ] ) . the fragment observables such as multiplicity , charge distributions and mean values of n / z for a given charge were also reproduced reasonably well @xcite . thus the data can be considered as well understood in terms of reaction mechanism . @xcite , ( a ) - @xmath46si(30amev ) + @xmath47sn , ( b ) - @xmath46si(30amev ) + @xmath30sn , ( c ) - @xmath46si(50amev ) + @xmath47sn , ( d ) - @xmath46si(50amev ) + @xmath30sn . , width=302,height=302 ] the observation that a wide range of observables is in agreement with the dominant reaction scenario of the deep inelastic transfer ( binary collision with intense nucleon exchange between the projectile and target nuclei , each of them represented by a statistically equilibrated fermi gas ) followed by statistical multifragmentation ( assuming statistically equilibrated freeze - out configuration ) allows to conclude that the assumption of statistical equilibrium is applicable and that thermodynamical observables can be defined for the hot quasi - projectile . the experimentally observed exponential scaling behavior shown in figs . [ fgy3rat ] and [ iso50 ] further demonstrates compliance with grand - canonical picture . the contribution from non - equilibrium processes such as pre - equilibrium emission was shown to be weak @xcite . the number of emitted neutrons , which are not detected , is , according to simulation , between one and two per event and thus no significant distortion of the picture can be expected . the observed quasi - projectile multifragmentation data can thus be used for investigation of thermodynamical properties of the multifragmentation source and of the possible phase transition . in order to investigate thermodynamical properties of the hot multifragmentation source , one needs to characterize a set of observables , including extensive ones such as thermal excitation energy ( heat ) and intensive ones such as temperature and chemical potentials . on excitation energy @xmath67 . solid squares - temperature determined using the double isotope ratio thermometer . open squares - temperature determined from the dependence of the yield ratio y(@xmath65h)/y(@xmath65he ) on the n / z ratio of the quasiprojectile @xcite . , width=302,height=264 ] for the quasi - projectile data from the reactions @xmath46si+@xmath97sn at projectile energy 30 and 50 amev , temperature was determined both using double isotope ratio method and the n / z - dependence of the isobaric yield ratio y(@xmath65h)/y(@xmath65he ) @xcite . the results are shown in fig . [ sisntemp ] . both methods seem to provide consistent results . double isotope ratio temperature exhibits a short plateau starting around @xmath67=4.5 amev , while the y(@xmath65h)/y(@xmath65he ) temperature seems to exhibit similar feature at somewhat higher excitation energy . the presence of a plateau in the caloric curve was interpreted in the literature @xcite as a signature of the first - order phase transition , nevertheless the influence of additional effects such as non - thermal flow or variation of the source size may affect the shape of the observed caloric curve @xcite and such a signal can not be considered as an unambiguous signature without providing further evidence . for the data in fig . [ sisntemp ] the influence of non - thermal flow and variation of the source size can be excluded due to well understood reaction mechanism . ( @xmath98 ) on relative excitation energy of the quasiprojectile @xcite ( full squares ) . horizontal line shows estimate of @xmath99 obtained when assuming @xmath100 . dashed curve represents the expected trend for homogeneous system . ( b ) corresponding caloric curve @xcite . vertical line indicates the position of turning point . , width=302,height=302 ] of the two methods for temperature determination , the y(@xmath65h)/y(@xmath65he ) temperature relies on additional assumption concerning chemical potentials . the insight into behavior of the chemical potentials can be obtained using the isoscaling analysis . the values of the isoscaling parameter @xmath93 for the data in fig . [ iso50 ] can be related , according to formula ( [ r21isobt ] ) , to the isovector component of the chemical potential . in fig . [ isotmp]a we present an estimate of the difference of isovector chemical potential using the observable @xmath99 , canceling out a trivial 1/t - dependence of the isoscaling parameter @xcite . the double isotope ratio temperature @xcite from the isotopic ratios y(@xmath79h)/y(@xmath65h ) and y(@xmath65he)/y(@xmath36he ) , which is independent of assumptions concerning chemical potentials , was used . the horizontal line represents an estimate of the zero temperature value of @xmath101 by the proton and neutron separation energies ( @xmath100 ) of reconstructed quasi - projectiles , using the known n / z equilibration and the correction for neutron emission from back - tracing of dit / smm simulations @xcite . the zero temperature estimates of @xmath101 for subsets of data in all excitation energy bins are consistent with the horizontal line . any experimental deviations from the horizontal line can be understood as a non - trivial dependence representing the details of de - excitation . the experimental dependence of @xmath99 indicates initial decrease at low excitation energies ( consistent with expansion of the homogeneous hot source ) , further it exhibits a turning - point at 4 amev followed by increase toward the zero temperature value at 6 - 7 amev . the estimate of @xmath99 for homogeneous system ( dashed curve ) was obtained assuming the fermi - gas @xmath102-dependence and using the estimate of free volume by formula ( [ eqnchi ] ) for each excitation energy bin . the trend explains a reasonable success of the @xmath100 approximation used to extract the mirror nucleus temperature in fig . [ sisntemp ] , where the viability of the approximation at high temperatures was supported by a possible counterbalance of two effects , namely expansion of the hot nucleus ( leading to a decrease of the absolute values of chemical potentials ) and chemical separation into an isospin symmetric heavy fraction and an isospin asymmetric nucleon gas , leading to increase of @xmath103 of the dilute phase ( nucleon gas ) . since the isoscaling parameter depends directly on the free nucleon chemical potentials , the turning - point at 4 amev can be understood as a signal of the onset of chemical separation which reverts the decrease of the free nucleon chemical potential consistent with expansion of the homogeneous system . as demonstrated by comparison with the caloric curve in fig . [ isotmp]b ( obtained using double isotope ratio thermometer without assumption concerning chemical potentials ) the onset of chemical separation is correlated to the onset of the plateau in the caloric curve , thus signaling that chemical separation is accompanied by latent heat . such a behavior can be expected in the first - order phase transition . the dotted lines in fig . [ isotmp]b indicate the discontinuity in the heat capacity of the system . . corresponding lines represent the dit / smm calculation . ( b ) - experimental ratio of the mean values of n / z of lcps and imfs versus the n / z@xmath104 . data are given for the reaction @xmath46si(50amev)+@xmath47sn @xcite . , width=340,height=226 ] the behavior , observed in fig . [ isotmp ] , is consistent with the first - order phase transition . several signals of the first - order phase transition have been reported in the recent experimental studies , such as negative micro - canonical heat capacity @xcite , characterized by abnormal kinetic energy fluctuation , and the `` fossil '' signal of spinodal decomposition @xcite , characterized by abnormal production of equally sized fragments . these indications of the first - order phase transition are based on careful event - by - event analysis of the multifragmentation data combined with model assumptions and results of theoretical simulations . the isospin - asymmetry of the fragments is not explored in such analyzes . the behavior in fig . [ isotmp ] relates the signal of chemical separation to the plateau of the caloric curve , thus providing a signal of isospin - asymmetric first - order phase transition . further support for such conclusion can be deduced from fig . [ mcpnz]b , demonstrating the inhomogeneous distribution of isospin - asymmetry @xcite among dense and dilute phases represented by light charged particles and intermediate mass fragments , respectively . the effect is strongly dependent on the n / z ratio of the quasiprojectile . so are multiplicities of lcps and imfs in fig . [ mcpnz]a which further indicate that with increasing isospin - asymmetry of the quasiprojectile the dilute phase becomes more isospin - asymmetric and abundant . the observations related to figs . [ isotmp ] , [ mcpnz ] are consistent with theoretical predictions of the isospin - asymmetric phase transition in nuclear media . the estimated density of the homogeneous system at the turning point in fig . [ isotmp]a is consistent with theoretical estimates of the position and shape of the spinodal contour , typically at almost constant total density 0.6@xmath105 in a wide range of asymmetries @xcite . thus , inside the spinodal region the homogeneous nuclear medium is quickly replaced by both isospin and spatially inhomogeneous system . such a conclusion is further supported by a comparison of the temperature obtained using the double isotope ratio thermometer @xcite with kinematic temperatures of p , d , t,@xmath65he,@xmath36he obtained from maxwellian fits of the particle kinetic energy spectra in the quasiprojectile frame ( see fig . [ fgslptmp ] ) . while at low excitation energies the double isotope ratio thermometer represents essentially an average of the kinematic temperatures of individual species , in the plateau region occurs a transition to the regime where the double isotope ratio temperature represents the kinematic temperature of protons ( and possibly of the nucleon gas ) . such a transition , just above the turning - point where the isovector part of chemical potential deviates from the trend of homogeneous system ( as shown in fig . [ isotmp ] ) , suggests that the isotopic composition of fragments is in equilibrium with the free nucleon gas what is a basic assumption used in grand - canonical models such as smm @xcite . the higher values of kinematic temperatures of d , t,@xmath65he,@xmath36he suggest that the clusters are created prior to such equilibration ( possibly by spinodal decomposition ) . on the other hand , the low energy behavior can be explained by an assumption that the isotopic composition is determined at the instant when clusters are emitted what is in good agreement with our understanding of de - excitation of the compound nucleus @xcite . the two different values of heat capacities in low- and high - energy parts of the caloric curve , suggested in fig . [ isotmp]b , can thus be understood as representing a homogeneous nuclear liquid and an isospin - asymmetric free nucleon gas , respectively . he,@xmath36he in the c.m . frame of the quasi - projectiles . , width=302,height=226 ] apart from signatures of the first - order phase transition , several experimental works @xcite , exploring observed mass ( charge ) distributions , report a critical behavior in the multifragmentation data , which can be understood as a signal of the second - order phase transition at the critical temperatures between 5 - 10 mev @xcite . however , multifragmentation studies employing the statistical model analysis of charge distribution @xcite suggest the value of critical temperature above 15 mev . besides of raising obvious questions concerning the applicability of used methods of analysis , such contradictory observations may in principle suggest a complex phase diagram , possibly analogous to the ising model , where both first - order and second - order can be observed . it is also possible that the observed critical behavior is a specific feature of the finite system , as suggested in theory @xcite . the recent experimental work where the critical behavior is related to the size of largest fragment as an order parameter @xcite supports such assumptions . the finite - size effect can play role also in the multifragmentation of quasiprojectiles with a=20 - 30 and thus similar analysis for heavier sources is necessary to evaluate the role of such effect and establish the link to the phase transition in the isospin - asymmetric nuclear matter at sub - saturation densities . it may be however noted that theoretical studies of small systems @xcite suggest that van der waals - like loops observed in infinite systems are affected in transition to finite system only by small terms and thus the conclusions for finite systems can be in principle applicable also for infinite case . in this article we presented an overview of the recent progress in the studies of nuclear multifragmentation . theoretical and experimental concepts relevant to investigation of isotopic trends in nuclear multifragmentation were described and the possibilities to extract the physical information on the nuclear equation of state were outlined . a special emphasis was put on the methods how the isotopic composition of fragments can be used to extract the values of thermodynamical observables of the system undergoing multifragmentation such as temperature and chemical potentials ( free nucleon densities ) . various methods for extraction of thermodynamical variables were presented and possibilities of their use were documented on examples of multifragmentation data . it was shown that the yield ratios of specific isotopes from statistical multifragmentation typically exhibit an exponential scaling with various observables . the logarithmic slopes of such dependences can be related to thermodynamical observables such as grand - canonical temperatures ( isotopic thermometers ) or chemical potentials ( isoscaling ) . finally , an extensive overview of investigations of multifragmentation of the fully reconstructed quasi - projectiles is presented . the dominant reaction mechanism is determined as binary nucleus - nucleus collisions with intense dissipation and heating via nucleon exchange . the implications of the fragment data concerning the isospin - asymmetric liquid - gas phase transition in the nuclear matter are discussed . multiple correlated signals point toward observation of a first - order phase - transition in the quasi - projectile multifragmentation data . the author acknowledges fruitful collaboration with research groups of professors s.j . yennello and j.b . natowitz during his stay at cyclotron institute of texas a&m university . examples from the analysis of experimental data acquired using faust and nimrod arrays are used throughout this work to illustrate the use of reviewed methods . insightful comments from a.i . sanzhur and p.i . zarubin are gratefully acknowledged . this work was supported by the slovak scientific grant agency through grant vega-2/1132/21 . li , c. ko // p.498 ; v. baran et al . p.4492 ; m. colonna et al . // p.122701 ; j. marqueron , p. chomaz // phys . rev . c. 2003 . p.41602(r ) . m. p. guthrie , r. g. alsmiller , h. w. bertini // instrum . meth . p.29 ; v.s . barashenkov , v.d . toneev // high energy interactions of particles and nuclei with nuclei ( in russian ) , 1972 . h. kruse , b.v . jacak , h. stoecker // p.289 ; j. molitoris , h. stoecker // phys . rev . p.346 ; j. molitoris , h. stoecker , b. winer // phys . rev . p.220 . n. marie et al . // progress in research , 19971998 , cyclotron institute , texas a&m university , college station , 1998 . p.v-19 ; r. wada et al . // progress in research , 19981999 , cyclotron institute , texas a&m university , college station , 1999 . p.v-15 . r. wada et al . // p.1829 ; h.w . barz et al . // p.10 ; s.j . yennello et al . // p.14 ; h. johnston et al . // p.186 ; e. ramakrishnan et al . // phys . rev . v.57 . p.1803 .

the hydrogen plasma is studied in the very high density ( atomic and metallic ) regime by extensive ab initio molecular dynamics simulations . protons are treated classically , and electrons in the born- oppenheimer framework , within the local density approximation ( lda ) to density functional theory . densities and temperatures studied fall within the strong coupling regime of the protons . we address the question of the validity of linear screening , and we find it to yield a reasonably good description up to , but already too crude for ( with the ion sphere radius ) . these values are typical of jovian planets interiors . finite - size and brillouin zone sampling effects in metallic systems are studied and shown to be very delicate also in the fluid ( liquid metal ) phase . we analyse the low - temperature phase diagram and the melting transition . a remarkably fast decrease of the melting temperature with decreasing density is found , up to a point when it becomes comparable to the fermi temperature of the protons . the possible vicinity of a triple point bcc - hcp(fcc)-liquid is discussed in the region of and . the fluid phase is studied in detail for several temperatures . the structure of the fluid is found to be reminiscent of the underlying bcc ( solid ) phase . proton - electron correlations show a weak temperature dependence , and proton - proton correlations exhibit a well - defined first coordination shell , thus characterizing fluid h in this regime as an atomic liquid . diffusion coefficients are computed and compared to the values for the one - component plasma ( ocp ) . vibrational densities of states ( vdos ) show a plasmon renormalization due to electron screening , and the presence of a plasmon - coupled single - particle mode up to very high temperatures . collective modes are studied through dynamical structure factors . in close relationship with the vdos , the simulations reveal the remarkable persistence of a weakly damped high - frequency ion acoustic mode , even under conditions of strong electron screening . the possibility of using this observation as a diagnostic for the plasma phase transition to the fluid molecular phase at lower densities is discussed .

mtrl-th9602004

the possibility of hydrogen metallization under high pressure was first discussed by wigner and huntington in the thirties @xcite . this particular subject was included in the more general conjecture that any system becomes metallic if sufficient pressure is applied . electrons , which are bound in the isolated atomic or molecular species , hybridize in a condensed phase to form extended states which gather in energy bands . when pressure is applied , the bands widen due to the enhancement of the overlap between neighboring atoms orbitals , up to the point at which the energy gap between the valence and the conduction band vanishes , thus giving rise to metallic behavior . this enhancement of the overlap can also be viewed as the growing importance of the electronic kinetic energy relative to the ion - electron potential ; the latter tends to bind electrons to the ions ( also in the form of interatomic , intramolecular or intermolecular bonds ) . in the case of pure hydrogen , metallization was estimated to occur around 2.5 mbars ( 1 mbar = 100 gpa ) , i.e. at pressures that have become accessible to experiment in diamond anvil cells only very recently @xcite . the absence of clear signs of metallic behaviour up to 2.9 mbar @xcite certainly adds to the challenge , and stimulates the continuous improvement of experimental setups in the search for the _ reluctant _ metallization . an additional difficulty in the description of this particular metal - insulator phase transition arises from the fact that at low pressures the low - temperature phase is not a monoatomic but a molecular solid , i.e. an hexagonal close packed ( hcp ) arrangement of h@xmath5 molecules @xcite . in this sense , h behaves more like a halogen than like an alkali metal @xcite . up to pressures of the order of 1 mbar , experiment indicates that the hcp arrangement is preserved . at pressures of the order of 110 gpa , the low - pressure free - rotator phase of para - h@xmath5 freezes into an orientationally ordered phase , whose nature is not yet fully understood . the observation of more than one vibron mode @xcite precludes the hcp structure with all the molecules pointing along the c - axis . however , other more complex hcp structures with tilted molecules ( herringbone - like ) are compatible with experiment @xcite . it is also possible that orientational order is not complete , thus giving rise to a sequence of weak phase transitions until the perfectly ordered phase is reached . an intriguing phase transition at 150 gpa , signalled by a discontinuity in the molecular vibron @xcite , has been ascribed either to a relative reorientation of the h@xmath5 groups @xcite , to metallization arising from an electronic band - overlap mechanism without molecular reorientation @xcite , and to both occuring at the same time @xcite . density functional ( dft ) calculations indicate that arrangements with complex molecular orientations ( non parallel ) are energetically preferred in the high pressure phase @xcite . another study @xcite concentrated on other classes of structures including cubic @xmath6 and rutile , and also two hcp structures , namely @xmath7 and @xmath8 . the hexagonal @xmath7 was found to be the favorable one above the 150 gpa transition . very recently , tse and klug performed ab initio simulated annealing calculations with 96 h atoms in the supercell , and found as a ground state an orthorombic structure , instead of hcp , composed of groups of three strongly interacting @xmath9 molecules @xcite . all these candidates were found to be insulating , and the reason for this was claimed to be the opening of a hybridization gap at the fermi level @xcite . diffusion monte carlo ( dmc ) calculations by natoli et al . @xcite have essentially confirmed results by kaxiras et al . @xcite , also at the quantitative level for the angles between different molecular units . they considered neither nagara s @xmath7 nor tse s orthorombic structure . this dmc calculation also finds an insulating behaviour and , interestingly , zero - point - motion effects due to the protons appear to be quite structure independent . as a consequence , zero - point - motion would turn out to be irrelevant as regards the 150 gpa transition , contrarily to the claims of surh et al . @xcite . well - converged local density functional ( lda ) calculations supplemented with the zero - point - motion energy contribution taken from frozen - phonon calculations @xcite , suggest that the molecular - hcp phase goes over directly to an atomic squeezed - hexagonal phase , by breaking the intramolecular bonds , at a pressure of 380 gpa . however , this calculation did not take into account non - hexagonal structures . hence , also from the theoretical point of view , it is not clear up to now whether metallization happens already in the molecular phase , or whether it is a consequence of dissociation , thus occuring simultaneously with the molecular - atomic transition @xcite . a rhombohedral structure was also postulated to supersede the squeezed - hexagonal phase at higher pressures ( above 400 gpa ) @xcite , while the transition to a body - centered - cubic ( bcc ) phase was located at around 1100 gpa @xcite . in general , static methods suffer from the drawback of having to carry out the investigation by _ guessing _ different structures . in this respect , the molecular dynamics results by tse and klug seem to be the most reliable . however , there is still some doubt because it is not obvious that their 96-atom supercell with @xmath10-point sampling and fixed cell volume and shape calculation is a sufficiently good description . a recently proposed method that combines state - of - the - art electronic structure calculations with variable cell shape molecular dynamics simulations @xcite , thus avoiding every undesired bias , is currently being used to investigate the low - temperature phase of h as a funcion of pressure @xcite . it is important to remark that , in general , energy differences between different structures are quite small , so that different levels of approximation often lead to different ground - state structures . in this respect , fully quantum - mechanical dmc calculations are very indicative . ceperley and alder @xcite located the molecular - to - atomic phase transiton at a pressure of about 300 gpa , by studying only cubic structures . further refinements of these dmc calculations have shown how dramatic the effect of improving the description can be @xcite . eventually , it appears that the atomic ground state structure in the vicinity of the molecular - atomic transition is the diamond structure , but the energy difference with respect to others is very small and , in particular , with respect to @xmath11-sn and hexagonal diamond , it is well within the error bars . the diamond structure was considered as a candidate for the ground state in one of the previous dft calculations , but found to be less favorable than a distorted hexagonal phase @xcite . the monoatomic bcc structure is unequivocally identified as the ground state in the very high density limit . there , the electronic kinetic energy is largely dominant over the remaining contributions , such that the electrons behave essentially as free fermions ; i.e. the electronic subsystem decouples from the protons and becomes a rigid , homogeneous electron gas , which acts only as a neutralizing background for the protons . this is well - known as the one - component plasma ( ocp ) model , whose classical version was shown to crystallize into the bcc structure . this follows from a simple calculation of the madelung energy , although the energy differences relative to other structures turn out to be rather small . in the classical ocp , which was thoroughly studied during the past two decades @xcite , the bcc structure turns out to be the stable one also at finite temperature and up to the melting point . the quantum version of the ocp at finite temperature has not yet been studied in detail . a numerical study using the path integral monte carlo technique ( pimc ) is currently under development @xcite . at higher temperatures the protons begin to behave as classical particles . the degeneracy temperature for the protons can be estimated to be of the order of @xmath12 , by comparing the mean interprotonic distance and the thermal de broglie wavelength . the electron degeneracy temperature ( @xmath13 ) lies well above ( by a factor @xmath14 ) such that , at temperatures lower than @xmath15 , electrons can be considered to follow the protons adiabatically , always staying in the the ground state compatible with the current protonic configuration ( fermi temperatures are about twice these numbers , e.g. @xmath16 ) . this latter approximation allowed hohl et al . @xcite to examine the hot , dense _ molecular _ phase by means of ab initio molecular dynamics ( aimd ) simulations , and also to make some progress regarding the low - temperature structure . the aimd study of the _ atomic _ phase is the subject of this and of a previous publication @xcite . the effect of excited electronic states in the same regime has very recently been explored by a novel ab initio molecular dynamics scheme using mermin s density functional @xcite , and the fully - quantum pimc method was applied to simulate the very high - temperature fluid phase @xcite , where the number of excited electronic states involved becomes too large to be treated efficiently with mermin s functional . the low - temperature regime , where protons in turn become degenerate , requires different simulation techniques which are currently being developed @xcite . high - temperature high - pressure measurements on hydrogen have been very recently reported in shock wave experiments @xcite . pressures between 1 and 2 mbar and temperatures of some thousand @xmath17 are now feasible and rather controllable , and further expansion of this range is in sight . in such experiments , nellis et al . observed the metallization of fluid molecular hydrogen at @xmath18 mbar and @xmath19 , a temperature which is significantly lower than that predicted for the plasma phase transition by chabrier et al . @xcite on the basis of a multicomponent thermodynamic theory that employs an approximate equation of state . although more accurate pimc simulations have essentially confirmed the predictions of the theory , including the existence of a first order phase transition at high pressures and a somewhat lower value for the transition temeprature @xcite , the discrepancy with respect to shock wave experiments and aimd results @xcite still holds . the latter would rather indicate a continuous transition in which metallization and dissociation of h@xmath5 molecules are closely related phenomena . a possible scenario would be the existence of two phase transitions : a continuous one , at relatively low temperatures , where molecular hydrogen already metallizes and begins to dissociate , and a discontinuous one at higher temperatures where a massive dissociation occurs with a concomitant jump in the electrical conductivity . hydrogen at finite temperatures and high densites consitutes , hence , a strongly coupled proton - electron plasma which is of great astrophysical interest since it is , in particular , the major constituent in jovian planet interiors . actually , the latter are basically h plasmas with a few percent admixture of he , and the statistical properties of the mixture ( mixing / demixing transition ) are of great importance to account for experimental observations and to establish models for the evolution of these planets @xcite . nevertheless , the answer to many relevant questions in astrophysics requires a full understanding of the statistical properties of the pure main constituents . hydrogen is present in both atomic and molecular fluid phases . these are separated by a boundary located at some distance from the center of the planet . the characteristics of this boundary , e.g. precise location , width , etc . depend on the density and temperature profile , i.e. on the equation of state . the matter of dissociation and metallization is particularly relevant because the large magnetic field measured in jupiter would include a non neglgible component generated in the outer molecular fluid phase , provided that temperature is above the metallization threshold . to quantify this effect the important quantities are the dissociated fraction and the electrical conductivity , which can be obtained from simulations in the molecular phase . the hydrogen plasma is perhaps the most fundamental , simple many - body system , with _ all _ the interactions ( proton - proton , proton - electron and electron - electron ) described exactly by the _ coulomb potential . in spite of this , the phase diagram appears to be surprisingly rich . the purpose of this article is to present a detailed study of the very high pressure physical properties and phase diagram of this simple and fascinating system . we are going to deal always with atomic phases , the molecular phases appearing at lower densities than those studied here . in section ii we briefly describe some details of the simulations , and we address some important technical issues such as the interplay between pseudopotentials and basis set expansions . section iii is devoted to the analysis of the validity of linear response theory in its ability to describe proton - proton interactions in terms of an effective pair potential of the screened coulomb form . in section iv we address a crucial problem that arises in the ab initio simulation of metallic systems , in particular liquid metals , i.e. size effects and fermi surface sampling . in section v we enter directly into the properties of the hydrogen plasma by describing the solid phases and the melting at very high pressure . section vi is concerned with the fluid phase , which we characterize as atomic - like . diffusion and vibrational properties are presented in section vii , while section viii is devoted to a thorough study of the collective dynamics , as a approach to the metal - insulator transition coming from the metallic side . the behavior of the longitudinal collective mode is studied as a function of density and temperature , and proposed as a probe for the metal - insulator transition at finite temeprature . finally , we conclude in section ix . in the remaining part of this section we define some useful parameters and ratios . the plasma is made up of @xmath20 electrons and as many ions ( protons ) per unit volume , such that the usual dimensionless density parameter is @xmath21 , where @xmath22 is the mean interionic distance ( ion - sphere radius ) and @xmath23 is the bohr radius . adopting atomic units throughout , the fermi momentum is @xmath24 , and the thomas - fermi screening length is @xmath25 . the dimensionless coulomb coupling constant associated with the classical ions is @xmath26 . note that @xmath27 , and that the electron degeneracy parameter is @xmath28 . typical densities inside jovian planets are between 1 and 10 @xmath29 and temperatures are of the order of 100 to 10000 @xmath17 , i.e. @xmath30 ranging from 0.5 to 1.5 , and @xmath31 @xcite . note that within this range @xmath32 and for the present calculations , whose aim is precisely to address some aspects of astrophysical plasmas , we can resort to the adiabatic approximation by assuming that the electrons are always in their instantaneous ground state for any given ionic configuration . higher temperatures require the relaxation of this hypothesis . our simulations were carried out using the standard car - parrinello ( cp ) aimd scheme @xcite . the ions were considered as classical point - like particles , whereas the electrons were described by density functional theory ( dft ) within the local density approximation ( lda ) @xcite . the electronic density was constructed via kohn - sham single - electron orbitals , expanded in a plane wave basis set up to a specified energy cutoff @xmath33 ( see below ) . we have extensively studied a system consisting of 54 h atoms contained in a simple cubic simulation box under periodic boundary conditions ( pbc ) , and we have also performed simulations for larger supercells , containing 128 and 162 h atoms , in order to analyse finite - size effects . the plane wave expansion was carried out around the @xmath10-point of the supercell s brillouin zone . finite - size and brillouin zone sampling effects will be analysed later , in a forthcoming section . we used the exchange - correlation functional deduced from the results of quantum monte carlo calculations on the uniform electron gas @xcite , and the bare coulomb potential for the ion - electron interaction . the singularity of the ion - electron coulomb attraction implies that there is never absolute convergence of the electronic quantities as the energy cutoff for plane waves is increased . in particular , the cusp condition is never satisfied , the density reaching the origin with zero slope . this is because the fourier expansion of @xmath34 is also a long - range function ( @xmath35 ) , meaning that there always exists a spatial scale small enough to require a representation involving large-@xmath36 components . to satisfy the cusp condition , a different ( localized ) basis set is needed instead of plane waves , e.g. slater - type orbitals @xcite . a finite cutoff @xmath37 translates into the following effective potential : @xmath38 this description is variational in the number of plane waves ( i.e. in the energy cutoff ) and , even if the energy never fully converges , other quantities , like the electronic distribution around the protons , are affected only in a tiny region around the nuclei , the remainder being essentially converged at a finite cutoff . this is shown in fig . 1 for protons fixed at the ideal bcc lattice sites . the choice of the cutoff strongly depends on the average density : the lower @xmath30 , the higher the cutoff . it is the number of plane waves which has to be kept approximately constant in order to achieve similar levels of convergence . in order to have the electronic density correctly described from less than one third of the nearest - neighbor ionic distance onward , we decided to vary the cutoff between 230 ry for @xmath39 and 60 ry for @xmath1 and 1.2 . it can be seen in fig . 1 ( @xmath39 ) that differences in the proton - electron correlation function become noticeable at distances smaller than @xmath40 , while the nearest neighbor distance in the bcc lattice is @xmath41 . it should be pointed out that a cutoff of 60 ry is sufficient to achieve convergence in the properties of the h@xmath5 molecule , like the proton - proton distance ( 1.5 a.u . = 0.79 ) and the vibrational frequency ( 4160 @xmath42 ) . the difference with respect to the experimental quantities ( 0.74 and 4400 @xmath42 ) may be ascribed exclusively to the lda . isothermal ionic dynamics ( particularly in the fluid phase ) was achieved by using a nos - hoover thermostat . the time step for integration of the cp equations of motion was chosen typically between 0.25 and 1.5 atomic units ( depending on the temperature and density ) , i.e. roughly 10@xmath43 s , compared to the ionic plasma period @xmath44 a.u . in some cases another nos thermostat was necessary to keep the fictitious kinetic energy of the electronic degrees of freedom at low values . most runs extended over 0.4 to 0.6 psec , which amounts to more than 50 plasma oscillations . before this , we allowed for thermal equilibration during an initial period of 0.2 psec in the solid phase , and 0.8 to 1.2 psec in the fluid phase . the densities @xmath39 , @xmath1 and @xmath45 were explored in detail for temperatures in the range @xmath46 , which correspond to the strong coupling regime ( @xmath47 ) , and cover both fluid and solid phases of the ionic component . in particular , @xmath48 and @xmath49 are typical conditions inside jovian planets @xcite . the simulations at @xmath45 were carried out only in the fluid phase because the stable solid has no longer the bcc symmetry , and thus it is not compatible with the choice of 54 h atoms in a simple cubic supercell . in the very high density limit ( @xmath50 ) the electrons are barely polarized by the ionic charge distribution , due to their rapidly increasing fermi energy , and the hydrogen plasma practically reduces to two decoupled components : a classical ionic `` one - component - plasma '' ( ocp ) , and a degenerate , rigid electron `` jellium '' . both systems have been extensively studied by classical @xcite and quantum @xcite computer simulations . for finite , but small @xmath30 ( @xmath51 ) , the coupling between the two components may be treated perturbatively within linear screening theory @xcite . however , due to the strength of the unscreened coulomb interaction between protons and electrons , such a perturbative approach is expected to break down rapidly as @xmath30 increases . in this section we investigate the high density limit and the validity of the description of the hydrogen plasma in terms of a linear response picture ( lrt ) for the electronic component . to this end we compare the proton - electron pair correlation function taking into account the full lda response : @xmath52 and the one obtained by replacing the electronic charge density by its linear response expression in terms of the ionic charge density : @xmath53 where @xmath54 is the static electron density response function . in eq . ( 2 ) , @xmath55 denotes the @xmath56 spherical bessel function , @xmath57 is a fourier component of the microscopic ionic density , and @xmath58 is a fourier component of the electronic density corresponding to the instantaneous ionic configuration @xmath59 . the latter represents , within the adiabatic approximation , a ground - state expectation value . for the density response we adopt the random phase approximation ( rpa ) corrected by the long wavelength limit of the local field function @xmath60 @xcite . @xmath61 } \label{pot}\ ] ] with @xmath62 the lindhard function and @xmath63 the thomas - fermi wavenumber . for the local field correction we use the following expression , which is consistent with the lda and slater s local exchange : @xmath64~\frac{k^2}{k_f^2}\ ] ] where the correlation energy density @xmath65 is that proposed by perdew and zunger @xcite , and @xmath66 . at typical densities studied in this work the exchange contribution to @xmath67 turns out to be largely dominant over correlation , and both of them represent a small correction relative to the rpa susceptibility . in fact , this is reasonable because the rpa ( or lindhard approximation ) is known to be good in the high density limit . the proton - electron distribution function reads accordingly : @xmath68}\right)}\end{aligned}\ ] ] in fig . 2 we compare the proton - electron radial distribution functions in the ideal bcc structure as obtained from the lda and from lrt , for @xmath39 and @xmath1 . as expected , @xmath69 is considerably more structured at the lower density . the lda and lrt distribution functions are very close at @xmath39 , but significant differences are clearly apparent at @xmath1 , reaching a region well beyond the first ionic shell and signalling the break - down of the linear screening regime . when the ions are at finite temperature the differences are significantly enhanced , particularly at short distances . also the location of the first minimum turns out to be shifted outwards by lrt in the fluid phase ( unlike in the solid phase ) , thus indicating that the lrt description of the electronic charge distribution worsens for increasing temperature . this is shown in fig . the dashed curves have been obtained by averaging the adiabatic electronic charge distribution around the protons along the aimd trajectory , assuming that the configurations generated in this way are also representative of the lrt . in fact , the magnitude of the differences observed in fig . 3 suggests that the lrt trajectories will differ significantly from the lda ones , implying that this averaged electronic distribution might be meaningless within lrt . in other words , the fully - consistent lrt is likely to be worse than the rasults presented here . following a novel procedure @xcite we have fitted , for @xmath1 , an effective proton - proton pair potential to our set of aimd configurations generated at several temperatures spanning both solid and fluid phases . in fig . 4 we compare this potential to the one obtained within lrt , i.e. by fourier transforming @xmath70/k^2 $ ] , where @xmath54 is given by expression ( [ pot ] ) . the nearest neighbor distance in the bcc structure for @xmath1 is 1.76 a.u . , and the position of the first peak in the proton - proton radial distribution function decreases down to values of the order of 1.5 a.u . in the fluid phase and upon heating ( see below ) . this means that nearest pairs of protons will spend most of the time at distances of this order . looking carefully at fig . 4 it can be seen that , at those distances , the two potentials differ by more than 10% , the lda one being steeper . on the other side , the lrt pair potential appears to be more long - ranged than the lda one . this implies that the lda pair potential has a characteristic screening length shorter than its lrt counterpart , i.e. screening is more efficient than linear at @xmath1 . the departure from the linear screening regime , also in the form of many - body effects beyond the pair potential approach ( e.g. three - body terms or embedding functions ) , becomes more pronounced as @xmath30 is increased , leading to ground state - atomic phases other than bcc , fcc or hcp ( e.g. hexagonal and diamond ) and , eventually , to recombination and to the different h@xmath5 molecular phases . the reason for this early departure from the lr regime has to be traced back to the unusual strength of the bare coulomb interaction , arising from the absence of core electrons . in fact , other alkali metals can be reasonably well described by lrt at much lower electronic densities ( @xmath71 ) @xcite . interestingly , the range of validity deduced here for the lrt is much wider than expected in previous theoretical work based on perturbative expansions @xcite , where @xmath72 was identified as the upper limit . in the solid phase , the combination of @xmath10-point sampling and a finite system size is not expected to provide a very accurate description of the electronic component , because all the periodicities beyond the size of the supercell are not properly included . this is particularly important in metallic systems , where no point in the brillouin zone can be taken as representative of the band structure of the bulk solid . the reason is that occupied states ( contributing to the electronic density ) and empty states ( which do not contribute ) coexist in the same band , corresponding to different k - points . a particular choice ( e.g. the @xmath10-point and its refolded images ) will sample the conduction band in some specific points , but the character of the fermi surface could be misinterpreted if empty states are taken as if they were occupied and viceversa . for quasi - spherical fermi surfaces this is unlikely to happen , but for transition metals or semimetals ( like graphite ) this effect is crucial , and a very fine sampling of the brillouin zone is needed to obtain the right physics . in the case of simple metals the effect of quantization of the electronic states in the simulation box is more important . the brillouin zone of the unit cell is sampled with a finite number of points , which arise from the refolding of the @xmath10-point of the supercell . these points reflect the symmetries of the system in the sense that , if the @xmath10-point refolds onto some point @xmath73 , then all the points in the star of @xmath73 must arise from some other refolding of @xmath10 . otherwise , the symmetry is broken and spurious forces appear that drive the system away from the symmetric ground state . depending on the case at hand , the distortion can be rather large , especially if the symmetry of the supercell is very different from that of the unit cell . if the correct symmetry is used for the supercell then , since all the points in the star are equivalent , the eigenvalues associated with them are degenerate , thus giving rise to the formation of electronic shells . if a better sampling of the brillouin zone is performed , or the size of the system is increased , new shells appear that eventually give rise to a continuous energy band . but for finite systems and restriction to the @xmath10-point , the electronic density of states consists of a set of discrete peaks ( the shells ) . a problem arises with the highest occupied shell because in finite - size metallic systems it is partially occupied , unless a very fortuitous situation occurs . the immediate consequence is that it is not possible to fullfil the symmetry requirements by occupying all the points in the star with an _ integer _ number of electrons . then , unless _ fractional _ occupation is introduced , the symmetry is broken even when the symmetry of the unit cell is mantained for the supercell . this is analogous to the jahn - teller effect in molecular systems , where a lower - energy state can be obtained by reducing the symmetry and breaking the electronic degeneracy . however , in the present case this effect is unphysical . in a fluid phase these effects are normally expected to become less important because of the breakdown of the discrete translational invariance , as reflected in the very existence of a finite brillouin zone via bloch s theorem . the unit cell becomes of infinite size and the brillouin zone reduces to a single point , i.e. the @xmath10-point . however , for practical reasons , computer simulations are bound to represent the infinite ( fluid ) system with a limited number of particles ( typically of the order of some hundreds in ab initio calculations ) , while repeating periodically the supercell ( via pbc ) . in fact , this introduces a spurious brillouin zone , which is associated with periodicities that are absent in the infinite fluid . the properties of the fluid are assumed to be recovered from the finite sample in terms of pbc combined with statistical averages , which in the case of our aimd simulations are computed as time averages . this is a reasonable justification for a purely classical system and also for liquid semiconductors but , for metallic systems , as soon as electronic states are introduced , the quantization of these states in the ( unphysical ) simulation box acquires a crucial role . the problem of partial occupation of the star persists in the fluid , in the sense that spurious forces appear that modify in a non - trivial way the structural and dynamical properties of the system . a qualitative picture of the consequences of these observations can be obtained by comparing the proton - proton radial distribution function for systems of different sizes . a system of 54 h atoms ( of valence charge equal to 1 ) is really a fortuitous case of compatibility of a closed electronic shell ( the whole star is fully occupied ) with a bcc atomic arrangement in a simple cubic supercell . in the case of 128 h atoms only 57 states are doubly occupied , and the next shell of 24 degenerate states has to accomodate 14 electrons ( i.e. only 30% of the shell ) . on the contrary , 162 h atoms is not compatible with a bcc arrangement in a simple cubic supercell , but it closes the former electronic shell thus amounting to 81 doubly occupied states . therefore , in the fluid phase it should behave essentially as the 54-atom supercell but better converged in system size ( or k - points ) . in fig . 5(a ) we show the curves corresponding to these 3 different system sizes at @xmath1 and @xmath74 . in fact , up to the boundary of the 54-atom cell ( @xmath75 ) there is hardly any relevant difference between the @xmath76 corresponding to 54 and 162-atom supercells . the same kind of picture holds also at higher temperatures , meaning that , as regards static thermodynamic properties , 54 h atoms already give a very reasonable picture . very different is the situation with open shell systems like the 128 h atoms one ( short - dashed line in fig . it is clear that this @xmath76 has little to do with those of 54 and 162 atoms . the first peak is significantly lower and broader , and also its position is shifted downwards . the first valley is much shallower and , in practice , the pair distribution becomes structureless beyond it , while with 162 atoms the atomic shell structure is still visible at least up to the second valley . the 128-atom distribution resembles , in fact , the one that would have been obtained with closed shells at a higher temperature value . the correct way to account for open shell structures at finite temperature is to doubly occupy the lowest - lying electronic states at every time step of the aimd simulation , while keeping empty the rest of that shell . the curves in fig . 5(a ) have been obtained by applying the standard procedure of considering explicitly , in the description of the electronic component , a number of electronic states equal to one half of the number of electrons , i.e. including those strictly necessary . the justification for this approach is that typical temperatures are much lower than the fermi temperature . atomic motion leads to a degeneracy lifting of the order of a few percent of an ev , implying that the distance between electronic states is usually much larger than the width of the fermi - dirac distribution ; this latter is of the order of @xmath77 , i.e. about 0.1 ev at @xmath74 . this leads to a symmetry breaking in the sampling of the fermi surface , which is expected to be recovered in terms of statistical averaging . however , following the approach above , in practical simulation times we have not noticed a convergence to the closed - shell picture . in fig . 6 we show the fermi - dirac distribution function for three different temperatures , namely 1000 @xmath78 , 5000 @xmath17 , and 10000 @xmath17 , compared to the electronic density of states averaged along an aimd run at @xmath74 , and to a snapshot of the instantaneous kohn - sham eigenvalues . it is clear that the occupation numbers fall from 2 to 0 in an energy scale narrower than the thermal splitting of the eigenvalues . the drawback of this kind of approach is that for open shells the ordering of the states within the highest ( partially ) occupied shell changes continuously during the md evolution and , in particular , occupied states become empty and viceversa . the standard cp approach is not able to take into account this phenomenon , and this is one reason for the well - known failure in metallic systems , reflected in the energy transfer between electronic and ionic degrees of freedom ( cooling the ions and heating the electronic orbitals ) . the double nos thermostat proposed by blchl and parrinello @xcite helps in fixing the temperature of each of the components , but still does not take into account level crossing effects . to our knowledge , there are three methods capable of solving this problem : one is to abandon the cp lagrangian strategy in favor of a self - consistent minimization at each time step @xcite , a second one is to abandon the description of the electronic component in terms of single - particle orbitals to work directly with the electronic density which is , by definition , constructed with the lowest occupied states @xcite , and finally a third one consisting of a rigorous lagrangian formulation which incorporates the occupation numbers ( actually the conjugate variables , i.e. the kohn - sham eigenvalues ) as dynamical variables @xcite . however , this latter procedure exhibits the odd feature that the fictitious kinetic energy of the electronic orbitals still increases at the expense of the ionic component , due to the appearance of low - energy excitations introduced by the dynamics of the eigenvalues . an approximate solution can still be found within the lagrangian approach by evolving explicitly the whole open shell , but initially occupying with an integer number of electrons only the lowest states of the this shell . this will approximately take into account level crossing in terms of mixing of states within the shell , during the time evolution . the very same mechanism that leads to energy transfer between ions and electronic orbitals , i.e. the vanishing energy gap @xcite , is responsible for mixing occupied and empty orbitals , thus allowing for initially empty states ( if explicitly included ) to become occupied and viceversa . statistical averaging completes the task by generating a more uniform sampling of the fermi surface . 5(b ) shows how the results obtained with 128 h atoms can reproduce approximately those obtained with 54 and 162 h atoms . it has to be pointed out that this procedure , although not rigorously justified , has the nice feature that the fictitious kinetic energy of the electronic orbitals is practically constant , behaving exactly as in systems with a gap , i.e. there is no energy exchange between ionic and electronic degrees of freedom . in the ocp ( @xmath50 ) limit the ionic bcc lattice is known to be the stable crystalline structure up to melting ( which occurs at @xmath79 @xcite ) . we have studied the stability of the bcc structure at low temperatures and finite @xmath30 by performing canonical md simulations ; initial conditions were constructed by giving the ions a small random displacement ( about 3 % of the nearest neighbor distance @xmath80 ) from their alleged equilibrium positions ( the bcc lattice sites ) . the bcc structure was found to be dynamically stable against such displacements at least up to @xmath39 . at @xmath1 the bcc structure was found to be unstable for @xmath81 , where a close - packed structure appears to be favored . this is consistent with a description in terms of lrt . in fact , at high densities the effective lrt potential behaves essentially like a yukawa potential @xmath82 while friedel oscillations are practically negligible . the phase diagram of a classical system of particles interacting via the above yukawa potential ( [ yukawa ] ) has been extensively studied by computer simulations and lattice dynamics @xcite . these calculations point to a bcc - fcc phase transition when the density - dependent screening wavenumber increases , i.e. when the effective interaction becomes of shorter range . at @xmath83 the bcc structure is found to be stable up to @xmath84 , beyond which the fcc phase becomes the stable structure . the @xmath30 at coexistence shifts to higher values at finite temperatures , such that the system goes through a structural fcc - bcc phase transition as the temperature increases along an isochore . the situation here is reminiscent of the behavior of alkali metals . na exhibits an hcp ground state , while li goes from bcc to fcc and eventually to hcp at very low t. the heavier alkalis k , rb and cs undergo a structural transition to fcc upon cooling below @xmath85 . in all these cases the entropic contribution of the bcc structure , arising from the valley in the phonon dispersion along the ( 110 ) direction , wins over at finite temperature and stabilizes this phase . this is exactly what we observe in our simulations for h , where at @xmath1 the bcc structure appears to be stable for @xmath86 . still , finite energy cutoff , finite system size , and coarse brillouin zone sampling may have a large influence on the stability of the ground state structure . the study of this part of the phase diagram deserves special attention because also zero - point motion effects on the protons have been shown to influence the stability of different structures at @xmath83 @xcite . disregarding the problem of zero - point energy ( zpe ) , and only as a check of the present calculations ( which do not include zpe ) , we have performed total energy full - potential linear muffin tin orbitals ( fp - lmto ) calculations for solid , monoatomic h in the bcc , fcc and hcp structures , at @xmath39 and @xmath1 . the energy differences turned out to be very small , but at @xmath1 the fcc and hcp structures are significantly lower in energy than the bcc . moreover , the energy differences are enhanced if a coarse sampling of the brillouin zone is performed . these calculations also identify the hcp structure as the lowest energy one , but the difference with respect to the fcc is within the accuracy of the calculations . the reason for this can be found in fig . 4 . the differences between fcc and hcp structures begin only at the level of third nearest neighbors , a region where the effective potential shows negligibly small friedel oscillations . between @xmath1 and 1.2 a phase transition occurs that takes the system from hcp to a simple - hexagonal phase with a compressed @xmath87 ratio ( squeezed - hexagonal ) . this is compatible with static total energy calculations by barbee et al . @xcite , and it is an additional confirmation of the breakdown of lrt , because pair potentials ( like the lrt one ) are not able to stabilize anisotropic structures like the simple - hexagonal . a more detailed study of the low - temperature atomic phases of hydrogen is currently under way @xcite . next , the melting of the ionic crystal was investigated by gradually increasing the temperature and monitoring the time - dependent mean square displacement of the ions @xmath88 . in the crystalline phase , @xmath89 goes over to @xmath90 for sufficiently long times , where @xmath91 denotes the static mean square displacement ( @xmath92 are the equilibrium positions of the ions ) . in the fluid phase diffusion sets in , so that @xmath93 at long times , with @xmath94 the ionic diffusion constant . at @xmath39 diffusion was found to set in at @xmath95 , which may be identified with the limit of mechanical stability of the ( overheated ) metastable crystal . the thermodynamic transition occurs at lower temperature ( higher @xmath10 ) ; its location may be estimated by assuming that the lindemann ratio @xmath96 at melting is the same as for the ocp , i.e. @xmath97 @xcite . this leads to @xmath98 compared to @xmath99 , indicating a strong influence of electron screening on the melting transition . this confirms recent predictions based on free energy comparisons , obtained by means of an approximate density functional theory @xcite . we have also studied the melting transition at @xmath1 using the same procedure ; the melting temperature drops sharply from @xmath100 to @xmath101 ( @xmath102 ) . this indicates the possible proximity of a triple point bcc - hcp(fcc)-liquid , analogous to that found for yukawa potentials . however , the nonlinearity of the screening at these values of @xmath30 is likely to bring the triple point from @xmath103 @xcite to @xmath3 . moreover , the hcp structure goes over to a simple hexagonal one at @xmath30 somewhere between 1 and 1.2 , and this has to be a consequence of the appearance of anisotropic forces , beyond the level of pair - wise additivity . the very low value of the melting temperature might also be related to the appearance of these forces . interestingly , at @xmath1 the fermi temperature of the ionic component is @xmath104 , a value close to the melting point ( @xmath105 ) . therefore , the influence of quantum effects for the protons on the melting transition can not be ignored , and will probably also play an important role in the above structural phase transition . in particular , they might destabilize the hexagonal phase in favor of some more isotropic configuration which has a higher energy within a framework of classical protons . turning to the fluid phase , a quantitative measure of ion - electron correlations is provided by the sphericalized average of the ion - electron pair correlation function @xmath69 , as computed from eq . the effect of temperature on @xmath69 is illustrated in fig 7 , for @xmath39 and @xmath1 . the distribution function is seen to be remarkably insensitive to @xmath106 over the whole range of temperatures , covering the solid and fluid phases , as already noticed at lower @xmath10 ( higher @xmath106 ) by dharma - wardana and perrot in the framework of an approximate static dft - hnc calculation @xcite . the observed weak temperature dependence implies that the main effect of ionic thermal motion ( electrons are always at @xmath83 ) , is to enhance the localization of the electronic charge close to the protons . this behavior is to be contrasted with the predictions of higher - level theories that go beyond the born - oppenheimer approximation by including excited electronic states . both , fully - quantum pimc @xcite and mermin functional @xcite simulations imply that ion - electron correlations become weaker as temperature increases . this is to be intuitively expected , but for temperatures much higher than the ones studied here . excited electronic states are much more insensitive to ionic polarization effects , because they correspond to larger kinetic energies . in this way , for temperatures larger than a threshold value that can be estimated around @xmath107 , i.e. @xmath108 for @xmath1 , the above localization effect due to ionic disorder starts to be compensated by the effect of electronic excitations , so that eventually the opposite trend will take over . the ion - electron pair correlation functions for _ all _ temperatures are seen to intersect at a well - defined ( reduced ) distance from the proton site , irrespective of thermal ionic disorder and almost independently of density . we locate this value at @xmath109 , and notice that the ratio @xmath110 is related to the ratio of the location of the nodes corresponding to the first two spherical bessel functions ( @xmath55 and @xmath111 ) . in fact , the electronic problem can be modelled , in a very crude approximation , as that of a particle in a spherical well ; the corresponding radial solutions are precisely the spherical bessel functions . temperature effects can be mimicked by increasing the relative population of excited states with respect to @xmath55 . however , the location of the node of @xmath111 ( the leading excitation ) relative to @xmath55 ( the ground state ) does not depend on temperature . since the location of the nodes is defined in units of the radius of the well , and this is identified with @xmath80 , which is proportional to @xmath30 , the crossing should not depend significantly on density . the first maximum of @xmath69 is clearly associated with the location of the first coordination shell ( the first maximum in @xmath112 see below ) , which is quite natural since the electronic density peaks at the proton sites . it is interesting to notice that differences in the electronic screening properties between @xmath1 and @xmath45 are significative only in the vicinity of the protons , up to @xmath113 . the proton - proton pair correlation function for @xmath1 is illustrated in fig . 8(a ) , as a function of temperature . it can be observed that the first peak remains clamped at the nearest neighbor distance ( @xmath114 ) for temperatures below the mechanical stability limit @xmath115 , while in the fluid phase it shifts continuously to shorter distances . the same plot shows that the location of the first minimum is quite insensitive to temperature . this , together with the fact that equivalent results are found at @xmath39 , defines quite univocally a first coordination shell of radius @xmath116 ( with @xmath22 the ion - sphere radius ) . the integrated number of particles is shown in fig . 8(b ) as a function of temperature . the main result is that the first coordination shell contains 14 atoms on average , implying that the short - range structure of the liquid is quite reminiscent of that of the solid , since the first coordination shell of the bcc structure ( containing first and second nearest neighbors ) also contains 14 atoms . simulations performed with 162 h atoms show that this is a genuine feature and not an artifact of the small system size . a second coordination shell is also well - defined in the fluid phase provided that the temperature is low enough , i.e. @xmath117 ( @xmath118 ) , as can be observed in fig . however , the fluid becomes structureless beyond the first coordination shell at temperatures of the order of 5000 k ( at @xmath1 ) . summarizing , the fluid phase of the h plasma at moderately high temperatures and very high densities ( typical of the inner h shell of jovian planets ) behaves like an atomic liquid with a well - defined first coordination shell . the influence of electron screening is also apparent when comparing the ion - ion and charge - charge static structure factors @xmath119 and @xmath120 . while at @xmath39 , the two are nearly indistinguishable , the amplitudes of their main peaks differ significantly for @xmath1 ( by roughly 9 % ) , and @xmath45 ( 12% ) . due to the discrete sampling of the electronic density in reciprocal space , the curves are noisy and will not be reproduced here . the strong polarization of the electronic component leads to a damping of the local charge fluctuations , and hence to a reduction of @xmath120 . again , the presence of a well - defined first peak and valley in @xmath121 is an indication that the fluid is well structured in this region of the parameter space . our aimd simulations give direct access to the ionic dynamics by analysing the time evolution of atomic coordinates and velocities . diffusion coefficients have been calculated using the asymptotic relation @xmath93 ; i.e. by measuring the slope of the mean square displacement of the atoms as a function of time . our results are displayed in table i for @xmath39,@xmath1 and @xmath45 , in reduced plasma units @xmath122 , where @xmath123 is the bare ionic plasma frequency . the results are shown in a log - log plot in fig . the relationship between @xmath124 and @xmath125 follows quite accurately a power - law of the type @xmath126 . we have fitted our data to such an expression , obtaining for @xmath1 the following values : @xmath127 and @xmath128 , and for @xmath39 the values : @xmath129 and @xmath130 . it is interesting to note that the diffusion coefficient follows the same relationship as in the ocp model . the ocp values for the parameters , fitted to classical md simulations @xcite are @xmath131 and @xmath132 . the value of the exponent seems to be unaffected by electronic screening , at least within the accuracy of our calculations . the prefactor , however , is clearly enhanced from its ocp value , and this can be readily understood in terms of the response of the electronic component to the motion of the protons . a rigid uniform electronic background ( as in the ocp ) does not have any influence on the dynamics of the protons . a polarizable background weakens the proton - proton interaction thus increasing the mobility of the ions . in fact , the results obtained at @xmath39 are quite close to the ocp values , represented by the dashed line in fig . 10 . the difference becomes much larger at @xmath1 , where @xmath124 differs from its ocp values by a factor of 3 ; this contrasts with the thomas - fermi md results of zrah et al . @xcite , who observed a much milder effect ( a factor 1.4 at @xmath133 and @xmath1 ) . therefore , the diffusion coefficient appears to be very sensitive to the treatment of electron screening . the case of @xmath45 is slightly different , because the exponent appears to be larger than the ocp value . however , it can be seen that the error bars are also compatible with an ocp - like power - law , represented by a line parallel to that of the other densities . the diffusion coefficient is also related to the ion velocity autocorrelation function ( vacf ) , which , in the ocp limit , exhibits a striking oscillatory behavior due to a strong coupling of the single - particle motion to the collective ionic plasma oscillations @xcite . such oscillations were recently shown to persist at finite @xmath30 , by md simulations using the approximate thomas - fermi kinetic energy functional instead of the kohn - sham version @xcite . the present ab initio calculations qualitatively confirm this behavior . in fig . 11 we present the vacf for a typical simulation in the solid phase ( @xmath134 ) and then for three different temperatures in the fluid phase . the latter were computed in the supercell containing 162 atoms . finite size effects are not very significant as regards the general features of the vacf . however , a small frequency shift is observed , and decorrelation happens faster in the larger sample . it is interesting to note that the fastest oscillation , i.e. the one associated with the ion plasma oscillations , is essentially temperature - independent . the power spectra of the ionic vacf are plotted in fig . 12 for @xmath1 , at several temperatures . the spectra exhibit a high - frequency peak ( or shoulder at the highest temperature ) at a frequency which amounts to 55 % of the bare ion plasma frequency ( @xmath135 ) . the power spectra for @xmath39 are similar , with the difference that the high - frequency peak occurs now at a value which is 70 % of the bare plasma frequency @xcite . as expected , electron screening shifts the vibrational spectra to lower frequencies . temperature , however , does not affect the position of the high - frequency peak , which remains the only well - defined feature at high temperatures , while the rest of the spectrum merges into a structureless continuum . these spectra were obtained from 54 h atom simulations . the 162 h atom sample yields a frequency 5 % lower than the 54 h atom one . the oscillations in the vacf point to a long - lived longitudinal collective mode , related to the ionic plasmon mode of the ocp @xcite . we have computed the charge autocorrelation function : @xmath136 where the fourier components of the microscopic charge density are : @xmath137 with @xmath138 and @xmath139 the aimd - generated time - dependent densities . in keeping with the born - oppenheimer approximation , @xmath140 is a fourier component of the expectation value of the electronic density for the instantaneous ion configuration . in practice , we computed the average of @xmath141 over the shell of equal - modulus @xmath142-vectors @xmath143 from this , we computed the dynamical structure factor @xmath144 by fourier transforming @xmath145 . in the @xmath146 ( ocp ) limit , where the electrons form a uniform ( non polarizable ) background , @xmath144 , which there reduces to the dynamical structure factor of the bare ions , is simply the k - dependent spectrum of the ionic plasma oscillations @xcite . in the long wavelength limit the mode is undamped , and its characteristic frequency is the ion plasma frequency @xmath135 . adiabatic electron polarization transforms this ionic plasmon ( or optic ) mode into an acoustic mode for any finite value of @xmath30 @xcite . this mode is to be identified with the familiar low - frequency ion - acoustic mode . only if the system were treated as a fully dynamical ion - electron plasma , would the high - frequency plasma oscillation mode appear , which is related to the fast electronic motions . this mode is obviously not accessible by adiabatic md simulations . the conjectured scenario is confirmed by the results of our aimd simulations . the dynamical structure factor @xmath144 was computed for the smallest wavenumber k compatible with the pbc ( @xmath147 for the 54 atom system ) and for selected larger wavenumbers ( @xmath148 ) . the resulting @xmath144 for @xmath39 , @xmath1 and @xmath45 are shown in fig . 13 , for the smallest available wavenumber , and at a temperature just above melting . the sharp peaks are characteristic of the long - lived ( weakly damped ) mode anticipated above . the peaks shift to lower frequencies as @xmath30 increases due to enhanced electron screening , and in accord with the behavior of the plasmon - like peak in the spectrum of the vacf ( the single - particle excitation coupled to the collective plasmon mode ) . the k - dependence of the spectrum @xmath144 is illustrated in fig . 14 , for @xmath1 ; the resulting dispersion curve is shown in fig . 15 . a striking feature is the nearly constant width of the resonance peaks for @xmath149 , pointing to a nearly k - independent damping mechanism . the damping increases dramatically at larger wavenumbers , while the dispersion curve bends over ; the behaviour is reminiscent of that observed for a classical fluid of atoms interacting via an effective yukawa potential @xcite . the bending over may be regarded as a remnant of the negative dispersion of the plasmon mode observed in the strongly - coupled ocp @xcite . from the initial slope of the dispersion relation , we estimate a sound velocity of @xmath150 km / s at @xmath1 , which is is consistent with the extrapolation of very recent results by alavi et al . to the ultra - high density regime @xcite . sound velocities are relevant to the determination of global free oscillations of jovian planets , which have been recently measured for jupiter @xcite . the remarkable feature is that this collective mode is much sharper than the usual sound mode observed in metals at comparable wavenumbers . moreover , the peaks do not shift significantly with temperature , although they broaden . however , particularly for low values of @xmath151 , the signature of the collective mode can be detectable up to quite high temperatures ( well above 3000 @xmath17 ) . this is seen in fig . 16 , where we plot the dynamical structure factor for @xmath152 ( i.e. just before the bending - down point ) in the 162-atom sample at @xmath1 and for three different temperatures . it is interesting to note that this is precisely the temperature range where a transition is expected to occur to the molecular ( h@xmath5 ) fluid phase at lower densities @xcite . the observed collective behavior may be regarded as characteristic of the metallic phase of hydrogen , and is expected to change dramatically at the transition towards the molecular phase , which begins to show up at low temperatures at @xmath153 . thus , an analysis of @xmath144 may provide an efficient diagnostic to locate the plasma phase transition at finite temperature . the main conclusions to be drawn from the present aimd simulations of the hydrogen plasma in the high - density ( @xmath154 ) regime may be summarized as follows : \a ) due to the significant spacing between the quantized electronic states in the vicinity of the fermi surface , the n - dependence of the statistical averages must be treated with great care , in order to extract meaningful results . \c ) the bcc structure , which is the stable low temperature solid phase at least up to @xmath39 , becomes unstable at lower densities , where hcp and simple - hexagonal phases appear . more work is needed to determine the full low - temperature phase diagram , also including zero - point - motion effects . \d ) the melting temperature drops sharply with decreasing density , due to the enhanced efficiency of electron screening of the effective interaction between ions . new interesting physics is likely to arise in the region of @xmath3 and @xmath4 , where the existence of a bcc - hcp(fcc)-liquid triple point is argued , in a region where quantum effects in the protons are non - negligible . \e ) the fluid metallic phase behaves very much like a simple atomic liquid from a structural point of view , but the longitudinal collective dynamics of the ions retain a strong plasma - like character at intermediate wavenumbers . this reflects itself in unusually sharp peaks in the charge - fluctuation spectrum , which are gradually shifted to lower frequencies with decreasing density , as a result of electron screening . the damping , however , appears to be surprisingly insensitive to density , but is significantly enhanced by temperature . the acoustic character of the longitudinal mode is recovered at sufficiently small wavenumbers ( @xmath155 ) , in qualitative agreement with a simple linear screening picture . a strong damping of the mode at intermediate wavenumbers should be a clear - cut signature of the plasma - to - molecular phase transition , which is expected to start at @xmath83 around @xmath156 , and to move to finite temperatures of the order of a few thousand @xmath17 at lower densities ( @xmath157 ) @xcite . \f ) the single - particle motion of the ions couples to the longitudinal collective mode , and reflects itself in a striking oscillatory behaviour of the velocity autocorrelation function , which is reminiscent of the behavior of the ocp , despite the action of strong electron screening . the resulting ionic self - diffusion constant is strongly enhanced at lower densities , for identical values of the plasma coupling constant @xmath10 , but follows a power law similar to that observed in the ocp . one of us ( jk ) would like to thank furio ercolessi for facilitating his code to fit the two - body potential , and ruben weht for helping with the fp - lmto calculations . we acknowledge helpful discussions with ali alavi , detlef hohl , hong xu , gilles zerah , stephane bernard , carlo pierleoni , pietro ballone , erio tosatti , giorgio pastore and sandro scandolo . e. wigner and h. b. huntington , j. chem . phys . * 3 * , 764 ( 1935 ) ; h. k. mao and r. j. hemley , science * 244 * , 1462 ( 1989 ) . see h. k. mao and r. j. hemley , rev . phys . * 66 * , 671 ( 1994 ) and references therein . see also i. f. silvera in _ simple molecular systems at very high pressure _ , eds . a. polian , p. loubeyre and n. bocarra ( plenum , ny , 1989 ) . l. cui , n. h. chen and i. f. silvera , phys . . lett . * 74 * , 4011 ( 1995 ) . a. ruoff , private communication . n. w. ashcroft , physics world , july 1995 , p. 43 . m. hanfland , r. j. hemley and h. k. mao , phys . lett . * 70*,3760 ( 1993 ) . r. j. hemley and h. k. mao , phys . * 61 * , 857 ( 1988 ) . e. kaxiras , j. broughton and r. j. hemley , phys . . lett . * 67 * , 1138 ( 1991 ) ; e. kaxiras and j. broughton , europhys . * 17 * , 151 ( 1992 ) . h. chacham and s. g. louie , phys . lett . * 66 * , 64 ( 1991 ) . m. p. surh , t. w. barbee iii and c. mailhiot , phys . * 70 * , 4090 ( 1993 ) . h. nagara and t. nakamura , phys . * 68 * , 2468 ( 1992 ) . j. s. tse and d. d. klug , nature ( in press ) . i. i. mazin and r. e. cohen , phys . b * 52 * , r8597 ( 1995 ) . v. natoli , r. m. martin and d. m. ceperley , phys . lett . * 70 * , 1952 ( 1995 ) . t. w. barbee iii , a. garcia , m. l. cohen and j. l. martins , phys . lett . * 62 * , 1150 ( 1989 ) . t. w. barbee iii and m. l. cohen , phys . b * 44 * , 11563 ( 1991 ) . p. focher , g. l. chiarotti , m. bernasconi , e. tosatti and m. parrinello , europhys . lett . * 36 * , 345 ( 1994 ) . j. kohanoff , s. scandolo , g. l. chiarotti and e. tosatti , unpublished . d. m. ceperley and b. j. alder , phys . rev . b * 36 * , 2092 ( 1987 ) . v. natoli , r. m. martin and d. m. ceperley , phys . lett . * 70 * , 1952 ( 1993 ) . m. baus and j .- p . hansen , phys . reports * 59 * , 1 ( 1980 ) ; s. ichimaru , h. iyetomi and s. tanaka , _ ibid _ * 149 * , 91 ( 1987 ) . d. m. ceperley , private communication . d. hohl , v. natoli , d. m. ceperley and r. m. martin , phys . lett . * 71 * , 541 ( 1993 ) . j. kohanoff and j .- p . hansen , phys . lett . * 74 * , 626 ( 1995 ) . a. alavi , j. kohanoff , m. parrinello and d. frenkel , phys . lett . * 73 * , 2599 ( 1994 ) . c. pierleoni , d. m. ceperley , b. bernu and w. r. magro , phys . * 73 * , 2145 ( 1994 ) . j. kohanoff , c. pierleoni and d. m. ceperley ( unpublished ) . w. j. nellis , m. ross and n. c. holmes , science * 269 * , 1249 ( 1995 ) , s. t. weir , a. c. mitchell , and w. j. nellis , preprint . g. chabrier , d. saumon , w. b. hubbard , and j. i. lunine , astrophys . j. * 391 * , 817 ( 1992 ) , and references therein . w. r. magro , d. m. ceperley , c. pierleoni , and b. bernu , phys . rev . lett * 76 * , 1240 ( 1996 ) . a. alavi , m. parrinello and d. frenkel , science * 269 * , 1252 ( 1995 ) . d. saumon , w. b. hubbard , g. chabrier , and h. m. van horn , astrophys . j. * 391 * , 827 ( 1992 ) . r. car and m. parrinello , phy . lett . * 55 * , 2471 ( 1985 ) ; for a recent review see g. galli and m. parrinello in _ computer simulations in materials science _ , , m. meyer and v. pontikis ( eds . ) ( kluwer , dodrecht , 1991 ) . for a comprehensive treatise see , e.g. r. g. parr and w. yang , _ density functional theory of atoms and molecules _ , o.u.p . h. xu , j .- hansen and d. chandler , europhys . * 26 * , 419 ( 1994 ) . d. m. ceperley and b. j. alder , phys . * 45 * , 566 ( 1980 ) ; d. m. ceperley , phys . b * 18 * , 3126 ( 1978 ) . z. lin and j. harris , j. phys . : condens . matter * 4 * , 1055 ( 1992 ) . w. b. hubbard and m. s. marley in _ strongly coupled plasma physics _ , nato asi series b * 154 * , eds . d. rogers and h. e. dewitt . s. galam and j .- p . hansen , phys . a * 14 * , 816 ( 1976 ) . j. p. perdew and a. zunger , phys . b * 23 * , 5048 ( 1981 ) . f. ercolessi and g. b. adams , europhys * 26 * , 583 ( 1994 ) . j. hafner , _ from hamiltonians to phase diagrams _ , springer series in solid - state sciences * 70 * ( springer , berlin , 1987 ) . p. blchl and m. parrinello , phys . b * 45 * , 9413 ( 1992 ) . t. a. arias , m. c. payne and j. d. joannopoulos , phys . lett . * 69 * , 1077 ( 1992 ) . grumbach , d. hohl , r. m. martin , and r. car , j. phys . : condens . matter * 6 * , 1999 ( 1994 ) . g. pastore , e. smargiassi and f. buda , phys . a * 44 * , 6334 ( 1991 ) . m. o. robbins , k. kremer and g. s. grest , j. chem . phys . * 88 * , 3286 ( 1988 ) . m. w. c. dharma - wardana and f. perrot , phys . a * 26 * , 2096 ( 1982 ) . c. pierleoni , private communication . a. alavi , private communication . j. p. hansen , i. r. mcdonald and e. l. pollock , phys . a * 11 * , 1025 ( 1975 ) . g. zrah , j. clrouin and e. l. pollock , phys . lett . * 69 * , 446 ( 1992 ) . f. postogna and m. p. tosi , nuovo cimento * 55 b * , 399 ( 1980 ) . j. l. barrat , j. p. hansen and h. totsuji , j. phys . c * 21 * , 4511 ( 1988 ) . b. mosser et al . , astron . astrophys . * 267 * , 604 ( 1993 ) .

we study the finite - temperature behaviour of two - dimensional heisenberg antiferromagnets with very weak easy - axis and easy - plane exchange anisotropies . by means of quantum monte carlo simulations , based on the continuous - time loop and worm algorithm , we obtain a rich set of data that allows us to draw conclusions about both the existence and the type of finite - temperature transition expected in the considered models . we observe that the essential features of the ising universality class , as well as those of the berezinskii - kosterlitz - thouless ( bkt ) one , are preserved even for anisotropies as small as 10 times the exchange integral ; such outcome , being referred to the most quantum case , rules out the possibility for quantum fluctuations to destroy the long or quasi - long range order , whose onset is responsible for the ising or bkt transition , no matter how small the anisotropy . besides this general issue , we use our results to extract , out of the isotropic component , the features which are peculiar to weakly anisotropic models , with particular attention for the temperature region immediately above the transition . by this analysis we aim to give a handy tool for understanding the experimental data relative to those real compounds whose anisotropies are too weak for a qualitative description to accomplish the goal of singling out the genuinely two - dimensional critical behaviour .

cond-mat0209316

in the last few decades the heisenberg antiferromagnet ( hafm ) on the square lattice has been thoroughly studied by means of several theoretical , numerical and experimental techniques @xcite . such research hands us a picture where three classes of substantially different models appear , the isotropic , the easy - axis , and the easy - plane antiferromagnets . the reference models for the two latter classes are the 2d - ising and 2d - xy antiferromagnets , defining their respective universality class and known to display a finite - temperature phase transition . it is also known that quantum effects , despite being enhanced by the reduced dimensionality , do not substantially change the essential features of the above models , not even in the extreme @xmath0 case , though evidently renormalizing their thermodynamic properties . this results from renormalization group @xcite and semiclassical approaches @xcite , as well as from quantum monte carlo ( qmc ) simulations @xcite . in the above picture , however , there still is a grey area , where our knowledge is not detailed enough to allow a precise reading of the experimental data : this is the area of very weak anisotropies and strong quantum effects , which is of particular interest as most of the real layered compounds whose magnetic behaviour is properly modeled by the @xmath0 hafm on the square lattice are characterized by anisotropies as small as @xmath2 times the exchange integral . these compounds exhibit a phase transition at a finite temperature @xmath3 which is often too large for the interlayer coupling to be the unique player , while the idea of a two - dimensional anisotropic criticality as the trigger of the transition appears well sound @xcite : such idea is corroborated by the measured values of some critical exponents @xcite . the experimental observation tells us that three - dimensional long - range order is present below @xmath3 and that well above @xmath3 no trace of anisotropic behaviour is left ; it is slightly above @xmath3 that one hence expects evidences of genuine 2d anisotropic behaviour to be detectable . in order to let these experimental evidences surface out of the sea of the isotropic thermodynamics , precise numerical data for the @xmath0 nearly isotropic hafm are needed : it is the purpose of this work to fulfill such need . in particular , we address the problem of the existence and characterization of the phase transition , making use of continuous - time qmc cluster algorithms , which are well suited in the neighbourhood of a critical point , since they do not suffer from critical slowing - down . the general resulting picture is that a phase transition is induced by an arbitrary amount of anisotropy , and that several distinctive features of the expected universality class can be traced out . the structure of the paper is as follows : in section [ s.2dxxz ] the general properties of the anisotropic models are described , extracting the most significant difference with respect to the isotropic model . in section [ s.defqmcfse ] the qmc methods are presented , the thermodynamic quantities under investigation are defined , and the finite - size scaling ( fss ) theory used in our analysis is briefly recalled . the results for nearly isotropic models , as from both fss analysis and thermodynamic behaviour , are presented and discussed in sections [ s.eaxxz ] and [ s.epxxz ] for the easy - axis and easy - plane case , respectively . in section [ s.phasediag ] the critical - temperature vs anisotropy phase - diagram is discussed . eventually , conclusions are drawn in section [ s.conclusions ] . the xxz model is defined by the hamiltonian @xmath4~ , \label{e.xxzmodel } \end{aligned}\ ] ] where @xmath5 runs over the sites of an @xmath6 square lattice , @xmath7 connects each site to its four nearest neighbours , @xmath8 is the antiferromagnetic exchange integral , @xmath9 and @xmath10 are the easy - axis ( ea ) and easy - plane ( ep ) anisotropy parameters , respectively , hereafter getting positive values smaller than unity . as @xmath11 sets the energy scale , the dimensionless temperature @xmath12 will be used in the following . the spin operators @xmath13 ( @xmath14 ) obey the _ su_(2 ) commutation relations @xmath15= i\varepsilon^{\alpha\beta\gamma}\delta_{\bm{ij } } \hat{s}^\gamma_{\bm{i}}$ ] and are such that @xmath16 . besides the isotropic model @xmath17 , eq . defines the ea ( @xmath18 , @xmath19 ) and ep ( @xmath20 , @xmath21 ) magnets , whose respective reference models are the ising model ( @xmath22 ) and the xy ( also known as xx0 ) one ( @xmath23 ) . what is known about these models can be summarized as follows : _ i _ ) the isotropic model has no finite - temperature transition ; its ground state is ordered for any @xmath24 and a critical region of divergent correlations is clearly observed at very low temperatures ; _ ii _ ) the ea models exhibit an ising - like transition at a critical temperature @xmath25 which is an increasing function of both @xmath9 and @xmath24 ; for @xmath26 , @xmath25 is finite for all anisotropies ; _ iii _ ) the ep models exhibit a transition of the berezinskii - kosterlitz - thouless ( bkt ) type , at a critical temperature @xmath27 which is an increasing function of both @xmath9 and @xmath24 ; for @xmath26 , @xmath27 is finite for all anisotropies . some of the above statements are rigorously proved , others come from the combination of theoretical , numerical and experimental results . more precisely , for what concerns ground state properties , an antiferromagnetically ordered ground state has been exactly proved @xcite for @xmath0 only in the case of strong anisotropy , @xmath28 and @xmath29 . for weaker anisotropies clear numerical evidence exists of an ordered ground state down to the isotropic limit , as coming from qmc and exact diagonalization studies @xcite . as for the finite - temperature properties , according to renormalization group analysis @xcite , the critical temperature vanishes in the isotropic limit as @xmath30^{-1}$ ] . in the ea case the existence of the transition for arbitrary @xmath9 has been rigorously proved only in the classical limit @xmath31 , while in the quantum case the proof is restricted to a spin - dependent finite amount of anisotropy @xcite . in the ep case the situation is even less clear , since also in the classical limit a rigorous proof @xcite of the existence of a transition is limited to the case @xmath32 , while no rigorous proof exists in the quantum case . in the @xmath0 case , evidences of a phase transition for anisotropies as small as @xmath33 in the ea case @xcite and @xmath34 in the ep case @xcite were suggested by previous qmc approaches ; however , these computations employed local algorithms which can not easily access the critical region and a rigorous fss analysis could not be performed . the situation is still unclear also because recent works based on real - space renormalization - group @xcite predict the existence of a critical value of the anisotropy in the ea case ( @xmath35 ) below which the transition would be destroyed by quantum fluctuations . in this work we hence consider @xmath0 and four nearly isotropic systems , two ea , @xmath33 and @xmath36 , and two ep ones , @xmath34 and @xmath37 . as usually done in the existing literature on qmc , in this section and in appendices [ a.worm ] and [ a.helicity ] we employ the notation @xmath38 and @xmath39 . the qmc method for the @xmath0 xxz model is based on the trotter - suzuki decomposition of the partition function , which can be approximated by the following expression @xcite : @xmath40 where @xmath41 represents the amplitude of propagation of a pair of nearest - neighbour spins from a configuration @xmath42 to @xmath43 in the ( imaginary- ) time step @xmath44 , @xmath45 being the trotter number and @xmath46 ( @xmath47 ) the basis set diagonalizing the @xmath48 operator . the two bond configurations define a space - time _ plaquette _ configuration @xmath49 , so that @xmath41 can be seen also as the weight of a given plaquette configuration @xmath50 . the index @xmath51 runs over all plaquettes on the space - time lattice , and the index @xmath52 runs over all configurations of the system . at each time step plaquettes are defined on different groups of bonds @xmath53 , so that all bonds involved in the propagation at the same time step do not share any spin ; moreover , each plaquette shares its corner spins with two plaquettes on the previous and two on the subsequent time step . the expression becomes exact in the limit @xmath54 . in the case of the xxz model , only the 6 plaquette configurations shown in fig . [ plaquettes ] have non - zero weights @xmath41 , whose expansion to first non trivial order in @xmath55 are @xcite : @xmath56 plaquettes 1 , 2 , 3 , and 4 propagate the state of the spin pair unchanged , while plaquettes 5 and 6 introduce an exchange of state for the interacting spins , hereafter denoted as _ any standard monte carlo ( mc ) method can be applied for numerically evaluating the partition function ; in principle one could do single local metropolis moves which update clusters of plaquettes sharing corner spins in order to move to configurations with non - zero weight @xcite . this approach is however inconvenient , being affected by the well - known critical slowing - down . moreover , in order to control the trotter - approximation error , it is necessary to repeat the simulation for increasing trotter number . in recent years efficient cluster algorithms were proposed to overcome the problem of critical slowing down in qmc simulations . in the context of quantum spin systems , the _ loop - cluster algorithm _ and the _ worm algorithm _ have revealed very effective ; moreover , they can be also formulated in continuous imaginary time , thus removing completely the trotter - approximation error . one can completely eliminate the critical slowing - down by introducing the so - called _ loop - cluster algorithm _ @xcite , which is the quantum analog of the swendsen - wang @xcite and wolff @xcite cluster algorithms introduced for classical spin systems . within the multi - cluster approach @xcite ( analogue to the swendsen - wang scheme ) the loop algorithm consists of probabilistically assigning to each plaquette a _ break - up decomposition _ ( or graph ) @xmath57 , i.e. , a way of grouping its spins in subgroups , so that the grouped spins can be flipped all at once bringing the plaquette into a configuration with non - vanishing weight ; in the case of the xxz model the above condition allows only grouping of spins into pairs ( _ non - freezing _ breakups ) or all together ( _ freezing _ breakup ) , as shown in fig . [ breakups ] . assigning a breakup @xmath57 to each plaquette , one univocally defines a breakup decomposition @xmath58 of the whole configuration @xmath59 into loops . different breakup decompositions @xmath57 are assigned to each plaquette configuration @xmath50 according to weights @xmath60 obeying the general sum rule @xmath61 . since the xxz model is symmetric under time reversal , it is straightforward to require that plaquettes connected by time - reversal must have the same weights , i.e. , @xmath62 , @xmath63 and @xmath64 . these weights must be positive and obey the detailed balance condition . the presence of frozen plaquettes leads to the generation of very big loops , whose flip is expected to be less and less effective in generating successive uncorrelated configurations ; therefore , a general strategy to be followed is that of minimizing the probability of frozen plaquettes to appear . if one sets @xmath65 for each p , a solution with positive breakup weights is obtained only in the case @xmath66 , i.e. , in the ep and isotropic case . when @xmath67 , i.e. , in the ea case , one must allow for freezing of at least a single plaquette configuration to achieve positiveness . in particular , in the antiferromagnetic case , it is @xmath68 that must be different from zero in order to have positive weights . by further imposing independence of the breakup decomposition from the plaquette configuration , i.e. , @xmath69 one gets that , within the same loop decomposition @xmath70 , all of the configurations @xmath71 obtained from the starting one @xmath72 by flipping some of the loops , are reached with the same probability , i.e. , each loop is flipped independently of all the other ones . the complete set of non - zero breakup weights used in our calculation thus reads @xcite ( to first order in @xmath55 ) : * easy - plane and isotropic case ( @xmath73 ) : @xmath74 * easy - axis case ( @xmath75 ) : @xmath76 the fact that freezing is required in the ea case can be physically interpreted as a reflection of the enforced antiferromagnetic correlations of the @xmath77-component of the spins : as a matter of fact , freezing of plaquettes of type 3 and 4 implies preservation of nearest - neighbour antiferromagnetic correlation along the mc flow , i.e. , antiferromagnetic configurations are more `` resistant '' to mc fluctuations , with respect to ferromagnetic ones . in the limit of continuous imaginary time @xcite , @xmath78 , plaquettes with no _ kink _ , i.e. , of types 1 ( 2 ) and 3 ( 4 ) , acquire unitary weight , while plaquettes with a kink , i.e. , of type 5 ( 6 ) , get an infinitesimal weight , still keeping a finite weight per unit time @xmath79 ; therefore kink - bearing plaquettes must be regarded as poissonian events in the imaginary - time evolution of each pair of interacting spins . at the same time the breakup decomposition creating no kink in imaginary time evolution , acquires a unitary weight , while all the other breakups acquire infinitesimal weight , still keeping a finite weight per unit time @xmath80 . in the case of plaquette 5 ( 6 ) , since the breakup weights have to be normalized to the plaquette weights , they become finite probabilities @xmath81 in the ep case , and @xmath82 , @xmath83 in the ea case . the algorithm then proceeds as follows : * distribute breakups @xmath84 , @xmath85 and @xmath86 on the continuous segments (= infinite sequences of infinitesimal plaquettes 1(2 ) and 3(4 ) ) along the imaginary - time evolution of each pair of interacting spins , according to the poisson distribution having as parameter @xmath87 ; for each kink in the propagation , choose a breakup with probabilities @xmath88 . * reconstruct the loops defined by the decomposition of each infinitesimal plaquette ; * decide whether to flip each loop independently with probability one half . the above procedure ( multi - cluster update ) represents a single mc step in our code . we have generally performed @xmath89 mc steps for thermalization for each value of the temperature , and @xmath90 mc steps for evaluation of thermodynamic observables . the algorithm is very efficient in both the ea and the ep case , with autocorrelation times which always remain around unity for all the lattice sizes @xmath91 we considered , i.e. , @xmath9216 , 32 , 64 , 128 , 200 . the autocorrelation time @xmath93 has been estimated by the _ blocking technique_@xcite as : @xmath94 where @xmath95 is the variance of the time - series @xmath96 ( @xmath97 ) produced for the variable @xmath98 , while @xmath99 is the variance of the block variable @xmath100 ( @xmath101 ) , with @xmath102 for the estimate to be sensible . the introduction of freezing breakups has the effect of making a loop branch into subloops ; as the flip of large branched clusters is less effective in generating successive uncorrelated configurations , the inclusion of freezing is generally thought to lower the efficiency of the loop algorithm @xcite , although no direct evidence of such conclusion exists . nevertheless , for the ea anisotropies we consider , no significant loss of efficiency ( i.e. , increase in correlation time ) is observed . moreover , in the ising limit , the above algorithm reproduces the swendsen - wang algorithm for the ising model , which is known not to suffer from slowing down ; therefore efficiency is possibly lost only for intermediate values of the anisotropy , which is not the case we are interested in . we have implemented improved estimators @xcite for all the quantities of interest . a separate , more careful analysis is needed in the case of off - diagonal observables , whose most general bilinear example may be @xmath103 . in absence of freezing , the improved estimator simply reads @xcite : @xmath104 , if @xmath105 and @xmath106 belong to the same loop , @xmath107 otherwise . this result can be understood by observing that the quantity we are considering may be seen as the ratio between the partition function of the system with density operator @xmath108 and the partition function of the original model considered , with density operator @xmath109 . the final outcome for the estimator given above thus follows from the observation that in absence of freezing , i.e. , in the isotropic or ep class , according to eqs . the weight assigned to a loop decomposition is actually independent of the configuration onto which it is defined , since each kind of breakup has the same weight on every plaquette configuration where its weight is non - zero ; therefore the same loop decomposition is reached with equal probability in the original and in the modified system , with the additional constraint , in the case of the modified system , that the two points @xmath105 and @xmath106 must belong to the same loop . when freezing is present , the above argument breaks down , since only one plaquette configuration admits freezing and the constraint of having @xmath105 and @xmath106 on the same loop is no longer sufficient to have a non - zero contribution to the estimator . in principle it is possible to define the estimator for off - diagonal observables even in case of freezing ; however , not only its implementation is highly non trivial from the point of view of programming , but its evaluation would also consume a considerable amount of computational time . as a matter of fact , for each pair of points @xmath106 belonging to the same branch of a loop , one should look for freezing breakups by checking all the loop segments between the two points , an operation that , being the loops structure involved , becomes very costly at low temperatures . we have therefore renounced to implement such estimators in case of freezing . to have a complete picture of the thermodynamics of the system in the ea case we have then resorted to a different ( and generally less efficient ) qmc scheme , within which the calculation of the off - diagonal observables in the ea case is relatively straightforward , i.e. , to the so - called worm algorithm . the _ worm _ algorithm represents an alternative way to overcome the problem of critical slowing down in qmc simulations . the original idea of the algorithm can be found in ref . , but we here formulate the algorithm in a different way , so that it appears as a direct generalization of the loop algorithm ; our formulation is more directly related with the so - called `` operator - loop update '' introduced in the framework of the stochastic series expansion @xcite . the worm algorithm starts by choosing a point at random in space - time , inserting two discontinuities in the local imaginary - time evolution , and then keeping one fixed ( the `` tail '' of the worm ) while letting the other ( the `` head '' of the worm ) freely travel through the lattice . the single - worm update ends when the head happens to `` eat '' the tail ( the worm closes ) , so that the isolated discontinuities disappear and the system is led to a new configuration having non - zero weight . all the segments of imaginary - time evolution touched by the worm s head have to be flipped , i.e. , the worm s head performs a real - time update of the system . its motion conventionally goes forward ( backward ) in imaginary time while updating segments with up ( down ) spins , and it is ruled by detailed balance condition , to be locally satisfied on each ( infinitesimal ) plaquette it touches . in the context of the worm algorithm one has to abandon the concept of plaquette breakup , and the decision about how a plaquette is passed through by the worm must be taken from time to time . one of the distinctive feature of the algorithm is that the worm can pass through the same plaquette many times , each time in a different way , depending on the local configuration left at its previous passage . general detailed balance conditions for the single plaquette update when the worm s head passes through it , flipping two spins , read @xmath110 where we have already introduced the time- and space - reversal symmetries , so that here `` 1 '' means 1 or 2 , `` 3 '' means 3 or 4 , and `` 5 '' means 5 or 6 , depending on the way the worm s head travels through the plaquette . moreover , the transition probabilities must satisfy the sum rules @xmath111 moving to the continuous - time limit , we express the transition probabilities from a plaquette with a non - vanishing weight to another plaquette , as @xmath112 where @xmath113 and @xmath114 have the meaning of ( positive ) transition probabilities per unit imaginary time , while @xmath115 and @xmath116 are ( negative ) corrections to the transition probabilities among plaquettes taking non - vanishing weights ; these corrections arise from the poissonian occurrence of kinks in imaginary - time evolution . with the above parameterization of the transition probabilities , the first two of the sum rules take the form @xmath117 while the third remains unchanged , as @xmath118 and @xmath119 keep their meaning of dimensionless probabilities for the different ways the worm s head can pass through a kink in imaginary - time evolution . the set of detailed balance equations in the continuous - time limit takes the form @xmath120 together with the sum rules , they give as unique solution the following set of transition probabilities : @xmath121 it is immediate to see that this solution is equivalent to the set of breakup weights , which means that , at this level , the worm algorithm is nothing but the wolff - type ( single - cluster ) version of the loop algorithm . however , as seen in the loop algorithm , in the ea case transition probabilities are not always positive , and some other transition mechanism must be invoked to overcome this problem . as seen before , the remedy in the case of the loop algorithm was to allow for branching of the loops ; if one hence allows for branching also in the worm algorithm , the single - cluster version of the loop algorithm for the ea case is obtained . a different strategy can be adopted in the case of the worm algorithm by introducing a new type of motion , named _ bouncing _ , where the worm s head , when attempting to update a plaquette , is bounced off and hence forced to locally trace back its route . from the physical point of view , the existence of a bounce mechanism protects some plaquettes from being updated , possibly those plaquettes containing local spin configurations which give a relevant contribution to the thermodynamics of the system ; such local configurations are preserved along the mc flow and the effect of bouncing is seen to be very similar to that of freezing in the context of the loop algorithm . in the case of the ea antiferromagnet , the most relevant local configurations are those containing antiferromagnetic correlations of the @xmath77-components , i.e. , in terms of plaquettes , p=3,4 . therefore , we allow for bounce on these plaquette configurations , introducing a finite bounce probability @xmath122 which has to be accounted for in the sum rule @xmath123 the detailed balance condition for the bounce probability is trivial , reading @xmath124 . eqs . and form an underdimensioned set and @xmath125 can be hence chosen arbitrarily , with the only constraint of positive transition probabilities . as in the case of freezing , it is highly convenient to minimize the bounce probability : when the worm s head bounces , part of its update operations are lost as it locally traces back its way , so that the efficiency in updating the configuration , keeping the number of elementary update operations fixed , is lowered . the following solution for the transition probabilities , minimizing the bounce probability , is found @xmath126 the worm algorithm with the bounce process is a _ pure - quantum _ cluster algorithm : in the ising model , which is a substantially classical statistical model , the algorithm loses its cluster nature , since only bounce processes survive , thus confining the worm on a single site . in this case the worm algorithm reproduces the single - flip metropolis algorithm , as shown in appendix [ a.worm ] . as in the case of the loop algorithm , each of our simulations consists of @xmath89 mc steps for thermalization and of @xmath127 mc steps for evaluation of thermodynamic observables . during thermalization , the number of worms to be produced at each step is adjusted so that the total length of the worms in the imaginary - time direction roughly equals the size of the ( d+1)-dimensional lattice , @xmath128 ; this number is then kept fixed during the measurement phase . in this way , autocorrelation times of the order of unity are achieved for all values of the ea anisotropy considered . at variance with the loop algorithm , the efficiency is here expected to drastically decrease as the anisotropy increases , given that , as the model moves toward the ising limit , the cluster algorithm transforms into the local metropolis algorithm ; however , the case of strong anisotropy is not of our interest here . as mentioned in the previous section , the estimator for bilinear off - diagonal quantities like @xmath129 can be thought of as a partition function for a modified model , in which two spin discontinuities are inserted in the system configuration at the points @xmath105 , @xmath106 . now it becomes clear that configurations giving a non - zero contribution to such a partition function are generated during the worm update whenever the discontinuities associated to the head and tail of the worm coincide with the above points , both in the ea and in the ep case . therefore the off - diagonal observables are measured on - the - fly during the motion of worm s head @xcite , and each worm update produces a statistics for the estimators which grows linearly with the length of the worm . as in the case of the loop algorithm , this kind of estimators give to the worm geometry a physical meaning : the further the worm head travels away from its tail , the larger will be the off - diagonal correlations . on the other hand , improved estimators are not defined for diagonal quantities in the ea case ; to this respect , worm and loop algorithms are seen to be exactly complementary . as final remarks , we mention that the worm algorithm retains its full efficiency also in presence of a uniform magnetic field applied to the spins , while the loop algorithm is known to exponentially slow down as the field is increased and/or the temperature is lowered @xcite . details of the worm algorithm in a finite field are not relevant to the present work and will be given elsewhere . independently of us , syljusen and sandvik @xcite have recently developed a very similar ( _ directed loop _ ) algorithm , in the framework of both stochastic series expansion and path - integral monte carlo . we briefly report here the definition of the relevant thermodynamic quantities measured in our qmc study , together with their respective estimators . the mc average of the estimator will be hereafter denoted as @xmath130 . the internal energy @xmath131 is estimated as the mc average of @xmath132 where @xmath133 takes the value @xmath134 if at imaginary time @xmath135 there is an infinitesimal plaquette configuration of type 1 ( 2 ) , @xmath136 if of type 3 ( 4 ) , @xmath137 if of type 5 or ( 6 ) . this corresponds to the continuous - time limit of the energy estimator as defined in ref . . the specific heat @xmath138 is estimated from energy fluctuations as @xmath139 where @xmath140 is the number of kinks present in each generated configuration . the variance of the specific heat has been estimated via binning analysis of the time series related to the energy estimator and the kink number . the staggered magnetization @xmath141 is estimated as the mc average of @xmath142 the spin - spin correlation function is @xmath143 where @xmath144 in the equal - time correlator ( et ) and @xmath145 in the time - averaged ( ta ) one . in both cases the numerical calculation of the correlation function takes advantage of the existence of the improved estimator defined in the previous subsection . the generalized susceptibility is @xmath146 the time - averaged susceptibility corresponds to the thermodynamic definition ( second derivative of the free energy ) while the equal - time one corresponds to @xmath147 , where @xmath148 is the static structure factor as measured , e.g. , in neutron scattering experiments . from the general definition above follow those of the uniform susceptibility @xmath149 and of the staggered one @xmath150 susceptibilities and correlation functions have been measured both along the @xmath77-axis ( @xmath151 , @xmath152 ) and in the @xmath153-plane ( @xmath154 , @xmath155 ) ; in the ea case , the latter have been evaluated by means of the worm algorithm . in what follows , we will show and comment our data relative to the uniform ta susceptibility and to the staggered et susceptibility , being such quantities the more relevant ones from the experimental point of view . the correlation length @xmath156 is defined _ via _ the long - distance exponential decay of the staggered correlation function , @xmath157 . a direct estimate @xmath158 of the correlation length may be hence found by fitting the long - distance behaviour of @xmath159 with a model - dependent function , as discussed in the following sections . such procedure , however , is strongly dependent on the quality and stability of the fit , and does not always lead to a univocally defined result in case of a finite - size system in presence of a phase transition , i.e. , of a diverging correlation length . an alternative strategy , which we have also used , is offered by the so - called second moment definition @xcite @xmath160 which can be directly extracted by the simulation data , supplemented by a binning analysis of susceptibility time series in order to estimate the variance . another relevant observable , in the ep case , is the helicity modulus @xmath161 , which is a measure of the response of the system to the application of a twist @xmath162 in the boundary condition along a given direction : @xmath163_{\phi=0}~ , \label{thermohelix}\ ] ] where @xmath164 . in appendix [ a.helicity ] we show that , starting from the above definition as explicitly written in terms of spin operators , the estimator of the helicity modulus of the @xmath0 xxz ep model reads @xmath165 where @xmath166 , @xmath167 being the total winding number of spin paths ( paths traced by a fixed spin configuration , up or down ) in the 1(2 ) lattice direction . remarkably , this estimator is directly related with that of the superfluid density of bosonic systems @xcite . an efficient improved version of the estimator has been introduced by harada and kawashima @xcite in the context of the loop algorithm , and is the one employed in this work . a fss analysis @xcite can give strong indications on the existence of a phase transition at some temperature @xmath168 , possibly leading to a full characterization of its universality class . the simplest evidence that a transition occurs is found when , for increasing lattice size , the order parameter scales to a finite value below a certain temperature , indicating that a non - zero order parameter develops in the thermodynamic limit . in the case of second - order phase transitions , the ansatz @xcite for the scaling behaviour of a generic finite - size thermodynamic quantity @xmath169 in the neighbourhood of the critical point reads @xmath170~ , \label{e.fss.at}\ ] ] where @xmath171 is the critical exponent of @xmath172 , i.e. , @xmath173 , @xmath174 is the exponent for the correlation length , while @xmath175 is the universal scaling function . at the critical point eq . implies @xmath176 . in the case of @xmath177 this means linear scaling at criticality , without any assumption on the universality class ; therefore , looking for the temperature at which a properly defined@xcite @xmath178 scales linearly with the system size gives an unbiased estimate of the critical temperature . . implies that the scaling plot of @xmath179 vs @xmath180 , with a proper estimate of @xmath168 , shows the data for different lattice sizes to collapse onto the universal curve @xmath181 . in the case of a bkt transition , in which no order parameter is given , the presence of topological order at finite temperature is shown when the helicity modulus scales to a finite value below a certain temperature . the use of the scaling ansatz to locate the critical temperature can be generalized to the case of a bkt transition , though most of the critical exponents are not defined . however the kosterlitz - thouless theory predicts @xmath182 at the critical point , so that a scaling behaviour of the susceptibility as @xmath183 is a good signature of the critical temperature . moreover , kosterlitz s renormalization group equations @xcite provide the critical scaling law for the helicity modulus in the form @xcite @xmath184 where @xmath185 is a constant this relation has been widely used to locate the bkt critical temperature of the classical 2d planar - rotator model @xcite and of the @xmath0 quantum xy model @xcite . we end this section with a general remark . it is observed that the smaller the anisotropy , the bigger the lattice sizes required to enter the asymptotic scaling regime , where fss holds . this is essentially due to the fact that the critical region is shifted to lower temperature : the correlation length of the isotropic model , acting as a lower bound for that of the nearly isotropic ones , increases exponentially upon lowering the temperature , and therefore , keeping the lattice size fixed , the ratio @xmath186 , that drives the onset of asymptotic scaling near the transition , gets smaller . the values of the anisotropy here considered are @xmath187 ( also used in ref . ) and @xmath188 . they are comparable with the characteristic anisotropies of real compounds ; yet , for such small anisotropy there is no universal consensus on the existence of a transition @xcite . from previous works @xcite the transition is expected in the temperature range @xmath189 in both systems . at higher temperature the behaviour gets closer to that of the isotropic model , which has been extensively investigated by means of qmc in recent years @xcite ; we have extended our analysis up to @xmath190 in order to identify those deviations from the isotropic behaviour that can be experimentally detected above the critical region . in our approach , evidences for the existence of an ising - like transition follow from a detailed fss analysis of the data ; subsequently , we analyze the temperature dependence of some relevant thermodynamic quantities , emphasizing the signatures of the ea nature . our analysis proceeds in three steps : we give evidence for a transition to occur , then the transition temperature is located , and eventually the ising critical scaling is tested . after the discussion made in section [ ss.fss ] the fss analysis for @xmath188 is expected to be more delicate than for @xmath187 . indeed , for the lattice sizes used ( @xmath191 ) some quantities show to have well entered the asymptotic scaling regime , while others do not . anyway , clear ( though not complete ) evidence for the ising universality class is given also for @xmath188 ; larger lattices would be required to reach a full characterization . let us first consider the order parameter , i.e. , the staggered magnetization given in eq . . in fig [ ea - szscale ] @xmath192 for @xmath188 is seen to scale to a finite value if @xmath193 , so that the magnetization in the thermodynamic limit becomes finite ; the same behaviour is _ a fortiori _ observed in the case @xmath187 . we then invoke the scaling ansatz for the longitudinal correlation length @xmath194 . the scaling plot of @xmath195 , as specifically defined in section [ ss.ea-xi ] below , is shown in fig . [ ea - xiscale ] for @xmath188 and gives @xmath196 . a similar analysis yields @xmath197 . hitherto , no assumptions were made about the universality class . in order to identify it , we consider the so called binder s fourth cumulant @xcite , shown in fig . [ ea - u4scale ] and defined by @xmath198 which assumes the universal critical value @xmath199 at @xmath25 in the 2d ising model on the square lattice @xcite , and increases ( decreases ) with @xmath91 , below ( above ) @xmath25 . for @xmath187 , we verify such behaviour and obtain @xmath200 , consistently with the above unbiased estimate from the scaling of @xmath194 . the scaling ansatz for the staggered magnetization , @xmath201 at @xmath202 , constitutes a further way of checking the 2d ising behaviour , since the critical exponents @xmath203 and @xmath204 are involved . the data reported in fig . [ ea - szscalemu.990 ] give @xmath205 . in the case of the weakest anisotropy @xmath188 , both the binder s fourth cumulant and the staggered magnetization have not yet well entered the asymptotic scaling region for the lattice sizes considered , and @xmath25 can not be reliably estimated by this technique . a further test of the universality class involves the longitudinal staggered susceptibility @xmath206 , eq . : in this case the scaling ansatz gives @xmath207 , with 2d - ising critical exponents @xmath208 and @xmath204 , as shown in fig . [ ea - chiscalemu.999 ] for the case @xmath209 . the estimated critical temperatures result @xmath210 and @xmath211 , in full agreement with the above unbiased estimates . to summarize , in the case @xmath187 we find consistency for the 2d - ising critical exponent ratios @xmath212 and @xmath213 , thus fully verifying the universality class . for @xmath188 the evidence , though limited to the matching of the estimates of @xmath25 obtained in figs . [ ea - xiscale ] and [ ea - chiscalemu.999 ] , is quite convincing . as a check that the magnetization and the staggered susceptibility have actually reached the asymptotic scaling regime with the considered lattice sizes , we have constructed their scaling plots after eq . , which are reported in figs . [ ea - szscplotmu.990 ] and [ ea - chiscplotmu.999 ] . data collapse for different lattice sizes is verified for the staggered susceptibility in the case @xmath188 for @xmath214 , taking @xmath215 , and _ a fortiori _ in the case @xmath187 ; the staggered magnetization is instead seen to have entered the asymptotic scaling regime for @xmath214 only in the case @xmath187 . from the above analysis a strong indication for the existence of an ising phase transition is therefore given for both the considered anisotropies . estimates of the critical temperature @xmath216 from the different criteria described in this section are summarized in table [ t.ea.tc ] ; all estimates are consistent , and amongst them we choose those realizing the best data collapse onto the universal scaling function in the scaling plots of staggered susceptibility and magnetization shown in fig . [ ea - szscplotmu.990 ] and [ ea - chiscplotmu.999 ] : the resulting values are @xmath217 and @xmath218 . such values will be indicated with a thin arrow in the following figures . .[t.ea.tc ] 2d ising transition temperature @xmath216 as obtained by fss analysis and fit of critical behaviours . [ cols= " < , < , < " , ] [ t.tbkteta ] although the identification of the bkt universality class is less complete than in the ising case , substantial consistence between the different estimates of the critical temperature obtained by different predictions of the kosterlitz - thouless theory proves that the two anisotropic models display a bkt critical regime . the estimates for the critical temperature @xmath219 given in this section are summarized in table [ t.ep.tc ] for the two anisotropies considered . putting together these estimates we choose as reference values @xmath220 and @xmath221 . such values will be indicated by a thin arrow in the figures of the following sections . the specific heat does not show large systematic deviations from the isotropic case within the resolution reached by the simulations for both anisotropies considered . only a small temperature region , well above the estimated transition temperature , displays an anomaly in the form of a tiny peak , as shown in fig . [ ep - cv ] ; such peak is possibly reminiscent of the rounded peak shown by the specific heat of the quantum @xmath0 xy model above its bkt transition @xcite . we must however mention that , at variance with our results , previous qmc data @xcite , also reported in fig . [ ep - cv ] significantly deviate from the isotropic model . according to the generally low sensitivity shown by the specific heat to weak anisotropies , as seen for instance in the ea case , we find this result a bit unlikely . as in the ea case , the uniform susceptibility reported in fig . [ ep - chi0 ] shows strong evidences of the anisotropy . moving down from the high - temperature region , where the isotropic behaviour is reproduced , the in - plane @xmath222 and out - of - plane @xmath223 uniform susceptibilities separate from each other and from the isotropic data . the in - plane component decreases faster than in the isotropic case . at variance with the ea case , however , @xmath222 is not expected to vanish at @xmath224 , due to the continuous rotational symmetry of the ground state in the @xmath153 plane . indeed , in a semiclassical picture , such symmetry allows the staggered magnetization to align along the in - plane axis perpendicular to the field , making possible the canted spin configuration with a finite ferromagnetic magnetization parallel to -and linear in- the field , so that @xmath222 stays finite . the out - of - plane susceptibility is instead enhanced with respect to the isotropic case , showing a minimum well above the transition . such minimum marks the onset of a completely different behaviour with respect to the isotropic model , entirely due to the presence of the small anisotropy . according to kosterlitz theory @xcite , in presence of a bkt transition the correlation length @xmath225 is expected to diverge exponentially at finite temperature as @xmath226~. \label{chixicritical}\ ] ] as for the @xmath227 model , this behaviour is in fact observed in ref . , where it is used to estimate the critical temperature . we use the estimates @xmath228 obtained by fitting the in - plane correlation function to eqs . and . discarding the values affected by finite - size saturation and thus considering only those satisfying @xmath229 , we also observe the predicted behaviour : in particular , singling out the bkt critical region by successively dropping points at high temperature until a stable fit is obtained , we obtain @xmath230 , @xmath231 , and the estimate @xmath232 , which agrees with the value found via fss analysis . furthermore , near criticality it is expected that the staggered in - plane susceptibility is related to the in - plane correlation length by the relation @xcite @xmath233 where @xmath234 is a nonuniversal constant and @xmath182 . by plotting @xmath235 together with @xmath236 , as done in fig . [ f.ep-bktxichi ] , one observes that this prediction also holds for the weakly - anisotropic quantum model ; remarkably , the curve @xmath236 collapses onto the @xmath225 on a wide range of temperature so that @xmath237 : this property is not shared by , e.g. , the classical planar rotator model or by the 2d quantum xy model . a closer look to the validity of the scaling relation can be obtained by plotting the ratio @xmath238 which converges to the value @xmath239 when @xmath240 ; this is clearly shown by the data plotted in the inset of fig . [ f.ep-bktxichi ] . in the @xmath241 case , neither @xmath242 nor @xmath243 exhibit the expected bkt critical behaviour for the considered lattice sizes . the correlation length obtained by fitting the correlator @xmath244 to the function is also not of much help . this suggests the in - plane correlation length to behave as in the isotropic model up to relatively large values ( @xmath245 ) , and the same holds for the staggered susceptibility . such findings closely resemble those of neutron scattering experiments on very weakly anisotropic layered @xmath0 compounds , such as sr@xmath246cuo@xmath246cl@xmath246 @xcite , la@xmath246cuo@xmath247 @xcite and pr@xmath246cuo@xmath247 @xcite , that do not show signature of the existing anisotropy in the correlation length and static structure factor data . both the out - of - plane staggered susceptibility and correlation length have a non - critical behaviour , with a maximum well above the transition , at a temperature which roughly coincides with that of the minimum of the out - of - plane uniform susceptibility , marking the onset of an anisotropy - dominated regime . both maxima are clearly decoupled from the transition temperature , at variance with the maximum of the transverse staggered susceptibility and correlation length observed in the ea case . to this respect we mention a definite disagreement with ref . , where out - of - plane correlation length is conjectured to diverge exponentially when @xmath248 . we show such conjecture to be wrong , as @xmath194 is clearly seen to saturate to a finite value . the detailed analysis presented above for the ea and ep models separately , is now composed to form the phase diagram @xmath249 vs @xmath250 in fig . [ eaep - phdiagr+ding ] , where our best estimates for the critical temperatures relative to the four models considered are shown , together with data from refs . and . critical temperatures are seen to be strongly reduced with respect to the classical values , as given for instance in ref . : however , the diagram clearly suggests the critical temperatures to stay finite for any finite anisotropy , both in the ea and in the ep case , thus leading to the conclusion that quantum fluctuations can not destroy the transition . we can actually see that the above conclusion is the consequence of a more general finding . if one numerically analyses @xmath251 finds that a logarithmic dependence is well consistent with our data , as shown in fig . [ eaep - phdiagr+ding ] . such dependence , already predicted by renormalization group techniques @xcite , is rederived in appendix [ a.sketchy ] on the basis of a fully classical argument . it is found that @xmath252 and @xmath253 where @xmath254 and @xmath255 are constants , while @xmath256 is the spin stiffness of the classical isotropic model , entering the above expressions via the exponential divergence of its correlation length . the dominant effect of quantum fluctuations on such divergence is embodied in the spin stiffness renormalization ; therefore , if the ordering process we are here observing is the same as in the classical case , we expect @xmath257 , where the value @xmath258 has been taken for the renormalized @xmath0 isotropic spin stiffness@xcite . from the logarithmic fits of the quantum data we indeed find @xmath259 and @xmath260 as prefactors of the logarithm , which are remarkably near to the predicted value , despite the simplicity of the argument that led to it . for @xmath227 and @xmath187 , where a direct comparison is possible , the critical temperatures are not fully consistent with the values given in refs . and . we notice that the latter were estimated as free parameters of fitting functions for the critical behaviour of the susceptibility and correlation length ; the precision of this approach is hindered by the fact that the critical regime of both quantities was not always properly entered in the simulations of refs . and , mainly due to technical limitations which are nowadays overcome . we therefore propose our data as more precise estimates . in this paper we have presented an extensive numerical study of thermodynamic and critical properties of weakly anisotropic two - dimensional quantum antiferromagnets described by the 2d @xmath0 xxz model with both ea and ep anisotropy . use has been made of the continuous - time qmc method based on the loop algorithm , implemented here for the first time also in the ea case , and on the worm algorithm , here reformulated as a variant of the loop algorithm . the general outcome of the numerical simulations is that the thermodynamics of 2d quantum antiferromagnets is extremely sensitive to the presence of anisotropies of magnitude comparable to those of real compounds , i.e. , as small as @xmath2 times the dominant isotropic coupling . in the models studied we see a finite temperature transition to persist with clear signatures of ising and bkt critical behaviour , in the ea and ep case , respectively ; in the more anisotropic case ( @xmath261 ) full consistency with the expected universality class is found . the most striking evidences of the presence of the exchange anisotropy are observed in the thermodynamic behaviour of correlation lengths and susceptibilities . moreover , the dependence of the critical temperature on the anisotropy is found to be quantitatively consistent with the prediction relative to the classical case , with properly renormalized parameters . this tells us that quantum effects can neither destroy the transition , nor change the ordering mechanism responsible for the transition to occur and that our quantum models , despite having @xmath0 and very weak anisotropies , do actually behave as renormalized classical ones . given the results of ref . for @xmath26 , we can say that this conclusion generally holds for quantum heisenberg antiferromagnets on the square lattice . as for the thermodynamic behaviour of the specific observables considered , we find all the non - diverging quantities to be highly sensitive to the anisotropy , while critical quantities show deviations , with respect to the isotropic case , which are generally harder to detect . this is due to the fact that , in order to discriminate between @xmath262 isotropic and finite-@xmath263 anisotropic divergences one must come very close to the critical point of the anisotropic model , which is a non - trivial issue both numerically ( due to severe finite - size effects ) and experimentally ( due to finite experimental resolution and intralayer coupling ) . as for the ep case , we underline that the considered values of anisotropy compare to that of several real compounds . on the other hand , we have clearly shown , for instance in figs . [ ep - chi0 ] , [ ep - xi ] and [ ep - chi ] , that in the ep case traces of 2d anisotropic behaviour are detectable above the transition , due to the fact that some quantities display either a minimum or a maximum in a temperature region well apart from @xmath27 , where the in - plane correlation length has not diverged yet , and experimental observation should hence be more feasible . we therefore think that our results could constitute a sound basis for a possible experimental observation of genuinely 2d ep behaviour in real magnets . the numerical calculations which led to the present work have been performed on the parallel cluster at cilea ( milano , italy ) , on the beowulf cluster cineca ( bologna , italy ) through infm grant no.1114174655 , on the beowulf cluster at inoa ( firenze , italy ) and on the pc s of the institt fr theoretische physik at the university of leipzig ( germany ) . we thank all these institutions for their generous support . we thank t. sauer and w. janke for participation to the early stage of this project . t.r . acknowledges fruitful conversations with b.b . beard , n. kawashima , p. carretta and m. laurati . t.r . further acknowledges the institt fr theoretische physik at leipzig university for hospitality during the early steps of this project , and crui for financial support via the vigoni programme . this work has been partially supported by the cofin2000-murst fund we show that the ising limit of the worm algorithm is nothing but the standard single - flip metropolis algorithm , considering the simple case of an ising antiferromagnetic spin dimer . independent transitions via single spin flip are only two : ( _ i _ ) @xmath264 and ( _ ii _ ) its reversal . within the worm algorithm , the transition ( _ i _ ) occurs with probability 1 , since the spin configuration @xmath265 admits no bounce ; this equals the metropolis result , since a net energy gain @xmath266 is obtained in the process ( _ i _ ) . starting from the configuration @xmath267 a bounce process is instead possible with a probability per unit time @xmath268 ; if the worm experiences even a single bounce , it will u - turn finding on its way back only plaquettes of type 1(2 ) , on which it can not bounce , until it reaches its tail leaving the configuration unchanged . therefore the transition occurs with probability @xmath269 , where @xmath270 is the probability of experiencing at least one bounce . since the number of bounces follows a poisson distribution with parameter @xmath271 , we have that @xmath272^k}{k ! } ~e^{-\beta \pi(3,b ) } = e^{-\beta j^z/2}\ ] ] which is exactly the metropolis transition probability for a process involving an energy increase @xmath266 . in this section we derive the qmc estimator for the helicity modulus starting from its thermodynamic definition . the derivation is a finite - temperature generalization of the one given by sandvik in ref . , in the context of stochastic series expansion , to estimate the spin stiffness , i.e. , at zero temperature . we start from the `` twisted '' xxz hamiltonian , with the twist applied along the 1-direction of the lattice , as : @xmath273 \\ \nonumber & + & { \hat{\cal h}}_2 \label{twist}\end{aligned}\ ] ] where @xmath274 and @xmath275 is the term containing only bonds along the 2-direction , which remains unchanged . we expand the twisted hamiltonian to second order in @xmath276 as : @xmath277 where @xmath278 is the @xmath279-component of the spin current operator , and @xmath280 carefully deriving the free energy with respect to the twist , i.e. , taking care of the non - commutativity between the @xmath281 , @xmath282 and @xmath283 operator , one obtains for the helicity modulus , averaged over the @xmath279- and @xmath284direction of the applied twist , the expression : @xmath285 where @xmath286 and @xmath287 . such expression stands as the direct quantum analogue of the classical expression as given in , _ e.g. _ , ref . for the plane rotator model ; in the limit of zero temperature it reproduces the expression given by ref . . the qmc estimator can be obtained directly starting from in the case @xmath0 . in the continuous - time limit , the estimator for the bond exchange operators @xmath288 takes the form : @xmath289 where @xmath290 is the number of ( + ) -kinks ( of the type ) on the @xmath291 bond , and analogously for @xmath292 . therefore the estimator for the xy - energy takes the form : @xmath293 where @xmath294 is the total number of ( @xmath295)-kinks present in each configuration . the current - current correlator present in can be decomposed into bond - pair contributions as follows : @xmath296 \big[\hat{t}^{+}_{{\bm{j}}{\bm d_1}}(\tau ) -\hat{t}^{-}_{{\bm{j}}{\bm{d_1}}}(\tau)\big]\ ] ] taking into account the @xmath297 constraint @xmath298 , one obtains that @xmath299 \big[\hat{t}^{+}_{{\bm{j}}{\bm{d_1}}}(\tau ) -\hat{t}^{-}_{{\bm{j}}{\bm{d_1}}}(\tau)\big]\big\rangle= \nonumber \\ & & ~~ = \frac{1}{\beta } ~ \bigg[~\big\langle \big(n^{+}_{{\bm{i}}{\bm{d_1}}}- n^{-}_{{\bm{i}}{\bm{d_1}}}\big ) \big(n^{+}_{{\bm{j}}{\bm{d_1}}}-n^{-}_{{\bm{j}}{\bm{d_1}}}\big ) \big\rangle_{_{\rm mc } } \nonumber \\ & & \hspace{28mm}-~\delta_{{\bm{i}}{\bm{j}}}~ \big\langle\big(n^{+}_{{\bm{i}}{\bm{d_1}}}+ n^{-}_{{\bm{i}}{\bm{d_1}}}\big ) \big\rangle_{_{\rm mc } } ~\bigg ] \label{correstimator}\end{aligned}\ ] ] putting together and with one obtains : @xmath300 where @xmath301 is the total number of @xmath295-kinks on 1(2)-bonds . since a ( + ) -kink and a ( @xmath302)-kink affect a spin path crossing the kink by shifting it of a lattice spacing in opposite directions , we have that the spin - path winding number can be expressed as @xmath303 and this reduces the estimator to the expression . we give here a sketchy description of the ordering mechanism in slightly anisotropic 2d magnets , referring to the classical limit where the antiferro- and ferromagnetic cases are thermodynamically equivalent . _ ea case_. we rewrite the hamiltonian in the classical limit as @xmath304 where @xmath305 is a unitary classical vector and @xmath306 is the classical exchange constant . in the above form the hamiltonian is written as the isotropic heisenberg hamiltonian plus , as long as @xmath307 , a small ising perturbation @xmath308 . the isotropic heisenberg model has an exponentially divergent correlation length as @xmath248 @xcite , @xmath309 , where @xmath256 is the spin stiffness of the classical model . at very high temperatures , i.e. , @xmath310 , the spins are fully uncorrelated , so that the anisotropy has little effect . when @xmath311 correlations set on and clusters of almost aligned spins form on the length scale @xmath177 . very roughly , one can imagine the @xmath312 spins of each cluster @xmath313 to be fully aligned , so that the anisotropy term can be written as @xmath314 where @xmath315 is the polar angle of the spin orientation on each cluster and border terms of order @xmath177 are neglected . hence , the anisotropy term creates an effective potential for the orientation of each correlated cluster that has two minima in @xmath316 and @xmath317 ( up and down ) separated by an energy barrier @xmath318 . when @xmath177 increases upon lowering @xmath263 , the barrier becomes comparable to the thermal energy , so that the orientation of each cluster is confined to one side of the potential barrier : the system becomes ising - like and a finite magnetization appears . using the isotropic behaviour of @xmath177 , this happens when @xmath319 solving this equation gives the critical temperature as in eq . . the above simplified picture accounts for the logarithmic dependence of @xmath320 upon the ( small ) anisotropy @xmath321 that was earlier obtained via more sophisticated approaches @xcite ; the logarithm appears to follow from the exponential correlation length in the isotropic model . _ ep case_. the above argument can be essentially rephrased , this time taking as perturbation of the isotropic hamiltonian the term @xmath322 which , in presence of clusters on the scale @xmath177 , becomes @xmath323 the anisotropy potential has the minimum at @xmath324 , i.e. , for a cluster orientation on the @xmath153-plane , and the well depth is @xmath325 . as in the ea case , the anisotropy becomes relevant once @xmath263 is comparable to @xmath326 : the out - of - plane fluctuations are suppressed making the system effectively planar , so that vortex excitations appear and the bkt transition can take place . this roughly happens when eq . holds . a. cuccoli , v. tognetti , p. verrucchi , and r. vaia , phys . b * 51 * , 12840 ( 1995 ) ; l. capriotti , a. cuccoli , v. tognetti , p. verrucchi , and r. vaia , physica d * 119 * , 68 ( 1998 ) ; a. cuccoli , t. roscilde , v. tognetti , p. verrucchi , and r. vaia , phys . rev . b * 62 * , 3771 ( 2000 ) . j. d. reger and a. p. young , phys . b * 37 * , 5978 ( 1988 ) ; t. barnes , d. kotchan and e. s. swanson , phys . b * 39 * , 4357 ( 1989 ) ; h .- q . lin , j. s. flynn and d. d. betts , phys . b * 64 * , 214411 ( 1999 ) . h. g. evertz , g. lana and m. marcu , phys . lett . * 70 * , 875 ( 1993 ) ; h. g. evertz , _ the loop algorithm _ , in _ numerical methods for lattice quantum many - body problems _ , edited by d. j. scalapino , ( perseus book , frontiers in physics , reading , 2001 ) . the weigth of plaquette configurations of type 1 ( 2 ) differ from those given in eq.(36 ) of ref . as we did not multiply by the arbitrary factor @xmath327 ; this also affects the break - up weights given in eqs . and . n. v. prokofev , b. v. svistunov , and i. s. tupitsyn , pisma zh . . fiz . * 64 * , 853 ( 1996 ) [ jetp lett . * 64 * , 911 ( 1996 ) ] ; phys . a * 238 * , 253 ( 1998 ) ; zh . fiz . * 114 * , 570 ( 1998 ) [ j. exp . * 87 * , 310 ( 1998 ) ] .

we consider the time evolution of quantum states by many - body schrdinger dynamics and study the rate of convergence of their reduced density matrices in the mean field limit . if the prepared state at initial time is of coherent or factorized type and the number of particles is large enough then it is known that is the correct rate of convergence at any time . we show in the simple case of bounded pair potentials that the previous rate of convergence holds in more general situations with possibly correlated prepared states . in particular , it turns out that the coherent structure at initial time is unessential and the important fact is rather the speed of convergence of all reduced density matrices of the prepared states . we illustrate our result with several numerical simulations and examples of multi - partite entangled quantum states borrowed from quantum information . : 81s30 , 81s05 , 81t10 , 35q55 , 81p40 + : mean field limit , reduced density matrices , wigner measures , entangled quantum states .

1411.6284

the mean field theory provides in principle a fair approximation of time evolved quantum states by many - body schrdinger dynamics in the mean field scaling ; namely when the number of particles is large and the pair interaction potential is proportionally weak . during the last decade , a strong activity around the mean - field problem has occurred within the community of mathematical physics . this in particular have led to a rigorous justification of the mean field approximation for singular potentials including coulomb interaction as well as the derivation of the gross - pitaevskii equation from many - body quantum dynamics ( see for instance @xcite and also @xcite for older results ) . more recently , emphasis has been placed on the speed of convergence of the mean - field approximation . this seems to be motivated by providing useful quantitative bounds and understanding higher order corrections ( see @xcite ) . the aim of our article , is to give at the level of a simple model more insight on the aforementioned problem . actually , the rate of convergence is essentially understood in the case of coherent or factorized type states with a particular structure . so , we can ask the following natural questions : * what should we expect if we start from another prepared state which is more correlated ? * is the specific coherent structure of the known examples important ? * can we determine the optimal rate of convergence in some examples ? * does the rate of convergence improves under the effect of the quantum dynamics ? we will show that the rate of convergence at a given time depends essentially on the rate of convergence of all reduced density matrices of the prepared state at time @xmath2 . in fact , we are able to give a general condition on the prepared state that guaranties a given speed of convergence at any time . the assumption we require at time zero , which is rather easy to check in initial states , is true at any time if it holds at @xmath2 . this allows in particular to consider the question of improvement of the convergence over time while the question of optimality will be addressed through numerical analysis . consider for instance the many - body schrdinger hamiltonian of an @xmath0-boson system [ schrod ] _ n=_i=1^n-_x_i+ _ 1i < jn v(x_i - x_j ) , where @xmath3 and @xmath4 is a real bounded potential satisfying @xmath5 . the self - adjoint operator @xmath6 acts on the space @xmath7 of symmetric square integrable functions . a function @xmath8 is symmetric if @xmath9 for any permutation @xmath10 on the symmetric group @xmath11 . suppose that the system is in a prepared quantum state @xmath12 at initial time @xmath2 ( that is @xmath12 is a non - negative trace class operator with @xmath13=1 $ ] ) . so , under the action of the schrdinger dynamics the system at time @xmath14 evolves into the state @xmath15 the mean field approximation at the dynamical level is usually understood as the following picture : if the system is in an uncorrelated state @xmath16 , with @xmath17 , at initial time @xmath2 then it will evolve into a state close in some sense to an uncorrelated one @xmath18 when @xmath0 is large and @xmath19 is the solution of the nonlinear hartree equation @xmath20{l } i\partial_t \varphi_t=-\delta \varphi_t+(v*\left|\varphi_t\right|^{2})\varphi_t\,,\\ \varphi_{t=0}=\varphi\ , . \end{array } \right.\end{aligned}\ ] ] the above convergence is neither a strong nor a weak one but rather in the sense of reduced density matrices . more precisely , the convergence is understood as _ n= _ t^p , a _ t^p_l^2(^dp ) , for any bounded ( or compact ) operator @xmath21 on @xmath22 and any @xmath23 ( @xmath24 is kept fixed while @xmath25 ) . + in some sense , the mean field approximation says essentially that the measurements [ correl ] , np , for any observable @xmath21 on @xmath26 converge , when @xmath0 goes to infinity while @xmath24 is kept fixed , to some classical or one particle quantities to be determined . hence , the main quantities to be analyzed are the reduced density matrices of the time evolved states @xmath27 . recall that for each @xmath28 , the @xmath24-reduced density matrix of @xmath27 is the unique non - negative trace class operator @xmath29 on @xmath30 satisfying [ eq.1 ] = , for any bounded operator @xmath21 on @xmath26 . therefore , the point is to determine for each @xmath28 the limit and the rate of convergence of these quantities when the number of particles @xmath0 goes to infinity . it turns out that the limit at @xmath2 may not exist and actually there is a difference between requiring convergence in for all bounded operators @xmath21 on @xmath26 , or convergence for compact operators only , since the weak and weak-@xmath31 topologies differ on the space of trace - class operators . however , one can characterize all the limit points of @xmath32 with respect to the weak-@xmath31 topology in the space of trace - class operators ( which is the dual space of compact operators ) and also describe their structure . indeed , at time @xmath2 , we can show that there exists always a subsequence @xmath33 such that for each @xmath34 , @xmath35 , the reduced density matrices @xmath36 converge to non - negative trace - class operators @xmath37 in the weak-@xmath31 topology . moreover , there exists a borel probability measure @xmath38 on @xmath39 such that @xmath40 in this way we have characterized all the possible limit points via subsequences of the reduced density matrices @xmath32 and identified their structure . more details are given in subsection [ sub.wigner ] while here we summarize the main result in the proposition below . we will use often the notation @xmath41 , @xmath42 , to refer to the schatten classes with @xmath43 denoting their norms . [ definetti ] let @xmath44 be a sequence of density matrices with @xmath45 for each @xmath46 . suppose that for any @xmath23 and each compact operator @xmath47 the sequence @xmath48)_{n\in{{\mathbb{n}}}^*}$ ] converges . then there exists a unique borel probability measure @xmath49 on @xmath39 invariant with respect to the unitary group @xmath50 and such that for any @xmath23 and any @xmath47 , _ ^(p)=_l^2(^d ) |z^pz^p| d_0(z ) . moreover , the measure @xmath49 is concentrated on the unit ball @xmath51 of @xmath39 centered at the origin and of radius one ( i.e. : @xmath52 . actually , the measure @xmath49 is the unique wigner measure of the sequence @xmath53 ( see subsection [ sub.wigner ] for definition and details ) . once this is understood we can consider the problem of rate of convergence for more general correlated states . [ main - th ] let @xmath54 be a sequence of positive numbers with @xmath55 and such that @xmath56 is bounded . let @xmath44 and @xmath57 be two sequences of density matrices with @xmath45 and @xmath58 for each @xmath59 . assume that there exist @xmath60 , @xmath61 and @xmath62 such that for all @xmath59 with @xmath63 : [ init - ineq ] ^(p)_n-_^(p)_1 c_0 . then for any @xmath64 there exists @xmath65 such that for all @xmath66 $ ] and all @xmath67 with @xmath63 , [ main - ineq ] ^(p)_n(t)-_^(p)(t)_1 c_t , where _ ^(p)(t)=_l^2(^d ) |z^pz^p| d_t(z ) , with @xmath68 is the push - forward of the initial measure @xmath49 ( given in proposition [ definetti ] ) by the well defined and continuous hartree flow @xmath69 on @xmath39 of the equation ( given in subsection [ sec.classical-quantum ] ) . * our result holds true in a more general framework . we can replace @xmath39 by any separable hilbert space @xmath70 , @xmath71 by any self - adjoint operator @xmath72 , and @xmath4 by any two - particle bounded interaction ( see subsection [ sec.classical-quantum ] ) . so from now on we will consider this setting , which has the advantage of covering several situations : e.g. either finite or infinite dimensional systems , as well as semi or non relativistic ones . * the assumption implies that we can apply proposition [ definetti ] and hence obtain the existence of the initial measure @xmath49 at @xmath2 . * the condition @xmath61 in the main assumption of theorem [ main - th ] can be replaced by @xmath73 at the cost of slightly changing the conclusion , by replacing @xmath74 in by @xmath75 . * we can apply theorem [ main - th ] backward in time . so , if the estimates hold true at a given time @xmath14 , then should also hold at time @xmath2 . this answers the question of improvement of the rate of convergence under the action of the quantum evolution . indeed , if we suppose that inequalities hold with a faster rate of convergence @xmath76 , @xmath77 , then the `` initial '' estimate should also hold with @xmath76 instead of @xmath78 by backward evolution . * the proof of theorem [ main - th ] allows to start with a rate of convergence @xmath78 faster than @xmath1 at time @xmath2 . however , we ca nt recover a better convergence at time @xmath79 . this is why we have restricted @xmath78 to be of order @xmath0 or less . however , this feature do not seem to be an artefact of the proof : numerical simulations on product states indicate a @xmath1 order of convergence even when at time @xmath2 the reduced density matrices coincide with their limit . the mathematical analysis of the mean field limit is quite rich and indeed there are several approaches and techniques applicable to this problem . for example coherent states analysis @xcite , bbgky hierarchy method @xcite , egorov type theorem @xcite , wigner measures approach @xcite or deviation estimates @xcite . hence the combination of these different techniques may lead to interesting results . the proof of our main theorem [ main - th ] relies on two ingredients : an egorov type theorem proved in @xcite and a wigner measures characterization of the limit points of reduced density matrices studied in @xcite . so the first step is to use second quantization formalism and wick observables , then the result in @xcite provides the asymptotics of time - evolved wick observables as [ int - expa ] e^it _ n b^wick e^-it _ n _ |l^2_s(^dn)=b(t)^wick _ |l^2_s(^dn)+r(n ) , where @xmath80 in some specific sense and where @xmath81 is an infinite sum of wick operators with time - dependent kernels or symbols ( see subsections [ sub.wick ] and [ sub.expansion ] ) . the mean field expansion gives actually the convergence of the correlation functions ( [ eq.1 ] ) . so that , if we use the idea of wigner measures extended to this framework in @xcite , we can obtain the rate of convergence for the quantities ( [ eq.1 ] ) . once this is proved , one can get the announced trace norm estimates for the difference between reduced density matrices . the article is organized as follows . the second quantization formalism and wick symbolic calculs is recalled in subsection [ sub.wick ] . the mean field expansion is explained in subsection [ sub.expansion ] while the the quantum and classical dynamics are introduced in subsection [ sec.classical-quantum ] . in section [ sec : rdm ] , we analyse the relationship between reduced density matrices ( rdm ) and wigner measures and provide the proof of proposition [ definetti ] . our main result is proved in section [ sec.rate ] with some preliminary lemmas . examples and numerical simulations are discussed in the last section [ sec.ens ] . the mean field theory is concerned with quantum dynamical systems which preserve the number of particles and can be worked out in the setting of multi - particles schrdinger operators . nevertheless , it is advantageous to use the more general setting of second quantization , as the reader will notice throughout the following sections . actually , the hamiltonian can be reformulated as @xmath82 with @xmath83 and @xmath84 a hamiltonian on the symmetric fock space over @xmath85 given by @xmath86 where @xmath87 are the usual creation - annihilation operator - valued distributions , i.e. : @xmath88=\delta(x - y)\,,\;\ ; [ a^*(x),a^*(y)]=0=[a(x),a(y)]\,.\ ] ] our investigation of the mean field approximation for the quantum dynamics is made through the analysis of the hamiltonian . the strategy relies on a specific schwinger - dyson expansion of the time dependent correlation functions elaborated in @xcite combined to some tools ( wigner measures ) from semiclassical analysis extended to infinite dimensional setting in @xcite . the schwinger - dyson expansion , called here mean field expansion , is explained in subsection [ sub.expansion ] and leads naturally to the consideration of several multiple commutators which we need to normal order using wick s theorem . so , for reader convenience we recall some basic results on normal ordering and wick operators written in more systematic and in some sense more efficient way : it makes possible the use of a symbolic calculus , for an algebra of wick operators , similar to the pseudo - differential calculus in finite dimension ( for other presentations of second quantization see @xcite ) . from now on we will wok in a general setting . let @xmath89 be a separable hilbert space . the symmetric fock space over @xmath89 is the hilbert space @xmath90 where @xmath91 denotes the @xmath0-fold symmetric tensor product . the orthogonal projection of @xmath92 onto the closed subspace @xmath93 is denoted by @xmath94 . the dense subspace of finite particle vectors is denoted by @xmath95 so , the creation and annihilation operators @xmath96 and @xmath97 , parameterized by @xmath98 , are then defined by : @xmath99 they extend to closed operators , they are adjoint and satisfy the canonical commutation relations ( ccr ) : @xmath100=\hbarr\la f , g\ra\ , 1 , \;\;\;[a^*(f),a^*(g)]=[a(f),a(g)]=0\,,\quad\forall f , g\in \z.\end{aligned}\ ] ] the weyl operators are @xmath101}\,,\quad f\in\z\,,\ ] ] and they satisfy the weyl commutation relations @xmath102 for any ( possibly unbounded ) operator @xmath103 we define @xmath104 as @xmath105 where @xmath106 denotes the @xmath0-fold algebraic symmetric tensor product of @xmath107 . + any wick operator preserving the number of particles could be written in the case of @xmath108 as a quadratic form using the integral formula b^wick=^k _ ^2kd _ i=1^k a^*(x_i ) b(x_1,,x_k;y_1,,y_k ) _ j=1^k a(y_j ) dx_1dx_k dy_1dy_k , with @xmath109 denotes the distribution kernel of the operator @xmath110 on @xmath111 . for general hilbert spaces , this formula can be generalized as follows . [ def.symbol ] for any @xmath112 , define @xmath113 to be the space of homogeneous complex - valued polynomials on @xmath70 such that @xmath114 if and only if there exists a ( unique ) bounded operator @xmath115 such that for all @xmath116 : [ eq.id ] b(z)=z^q , z^p . we will often use the identification between homogeneous polynomials @xmath114 and their associated operators @xmath115 according to . the algebraic sum @xmath117 is clearly an algebra of polynomials . these spaces @xmath118 and @xmath113 play a role similar , in some sense , to classes of symbols in pseudo - differential calculus . for this reason we sometimes call the polynomials @xmath119 symbols ( see for instance @xcite ) . the subspace of @xmath113 made of polynomials @xmath120 such that @xmath121 is a compact operator is denoted by @xmath122 and @xmath123 a _ wick operator _ with symbol @xmath124 is a linear operator @xmath125 with domain @xmath126 defined as @xmath127 where @xmath121 denotes the operator associated to the symbol @xmath120 according to . remark that for simplicity we have used the notation @xmath125 without stressing the dependence on the scaling parameter @xmath128 . an interesting feature of the above quantization is that it maps the algebra of symbols or polynomials @xmath118 into an algebra of operators in the fock space . in particular , the composition of two given wick operators @xmath129 and @xmath130 is again a wick operator @xmath131 . more interesting is that its symbol @xmath132 belongs to @xmath118 and is given by an explicit formula like in pseudo - differential calculus of finite dimension . this comparison goes much further in fact , and the commutator of wick operators is a sum of quantized multiple poisson brackets with @xmath128 playing the role of a semi - classical parameter . let us introduce the precise meaning of the multiple poisson brackets . remark that all polynomials in @xmath113 admit frchet differentials and therefore they all have directional derivatives . remark also that we do nt need a particular conjugation on the hilbert space @xmath70 in order to define the derivatives @xmath133 and @xmath134 . in fact , for @xmath114 we define @xmath135=\bar\partial_r b(z+ r u)_{|r=0 } , & & \partial_{z } b(z)[u]= \partial_{r } b(z+ r u)_{|r=0}\,,\end{aligned}\ ] ] where @xmath136 are the usual derivatives over @xmath137 . moreover , @xmath138 naturally belongs to @xmath139 ( i.e. : @xmath140-linear symmetric functionals ) while @xmath141 is identified via the scalar product with an element of @xmath142 , for any fixed @xmath143 . for @xmath144 , @xmath145 and @xmath146 , set @xmath147\ ; \in\p_{p_1+p_2-k , q_1+q_2-k}\quad.\ ] ] the multiple _ poisson brackets _ are defined by @xmath148 [ wick - calc ] let @xmath149 et @xmath150 . for all @xmath151 , the polynomial @xmath152 belongs to @xmath153 with the following formulas holding true on @xmath126 : @xmath154^{wick}.\ ] ] @xmath155=\sum_{k=1}^{\max(\min(p_{1},q_{2}),\min{(p_{2},q_{1})})}\frac{\varepsilon^{k}}{k ! } \big[\{b_{1},b_{2 } \}^{(k ) } \big]^{wick}.\ ] ] consider a polynomial @xmath156 such that @xmath157 is bounded and symmetric . in all the sequel we consider the many - body quantum hamiltonian of bosons to be the operator defined by @xmath158 where @xmath159 is a given self - adjoint operator on @xmath70 with domain @xmath160 . by standard perturbation theory , and thanks to the conservation of the number of particles , it is easy to prove that @xmath161 is essentially self - adjoint on @xmath162 . we denote respectively the time evolution of the perturbed and the free quantum system by u(t)=e^-i h _ u_0(t)=e^-i ( h_0 ) . it is known that in the mean field limit we obtain the hartree equation , when the many - body schrdinger hamiltonian is considered . in our abstract setting the limit equation has the energy functional @xmath163 which is actually the wick symbol of the quantum hamiltonian . so , the associated nonlinear field equation reads @xmath164 with @xmath165 is the vector field given by @xmath166 . in order to solve this equation we write it in the integral form @xmath167 since @xmath168 is a bounded operator then a standard fixed point argument implies that ( [ hartree.int ] ) admits a unique continuous local solution for each initial condition @xmath169 . thanks to the conservation of the hilbert norm on @xmath70 we see that any local solution extends to a global continuous one . therefore , we have a well defined global continuous flow on @xmath70 which we denote by @xmath170 . in other terms @xmath171 is a @xmath172-map satisfying @xmath173 and @xmath174 solves ( [ hartree.int ] ) for any @xmath169 . moreover , if @xmath175 is the solution of @xmath176 and @xmath177 is the polynomial in @xmath178 given as @xmath179 , then the curve @xmath180 solves the differential equation @xmath181 hence , a simple computation yields for any @xmath114 the identity b(w_t)&= & i _ zq_t(w_t ) [ _ |z b(w_t)]+ _ z b(w_t)[-i _ |zq_t(w_t ) ] + & = & i \{q_t , b}(w_t ) , where the brackets are defined according to . so , we obtain the following duhamel formula for all @xmath182 : [ class - integ - form ] b(z_t)=b_t(w_0)+i_0^t\{q_t_1,b_t}(w_t_1 ) dt_1 , with @xmath183 a ( mild ) solution of the nonlinear field equation and @xmath180 . the main point is to study the time evolution of wick operators with respect to the small mean field parameter @xmath128 which is essentially the inverse of the number of particles . this was done in @xcite and in fact we can prove in some sense that [ formal_wick ] e^i h _ a^wick e^-i h_=a(t)^wick+r ( ) , with @xmath184 when @xmath185 ( see @xcite and also ( * ? ? ? * thm . 5.5 ) ) and where @xmath186 is an infinite sum of wick operators with time - dependent symbols related to the hartree dynamics . in order to prove , we use an iterated integral formula ( the so - called dyson - schwinger expansion ) with a specific use of wick calculus ( proposition [ wick - calc ] ) in order to expand commutators of wick operators with respect to the @xmath128 parameter . we will work in the interaction representation . hence , the following notation is useful @xmath187 for @xmath119 and @xmath188 ( remark that @xmath189 belongs to @xmath118 ) . we also know that multiple commutators in the schwinger - dyson expansion lead to wick operators with multiple poisson brackets symbols . for this reason we make the following definition . for @xmath190 and @xmath191 , we associate to any @xmath192 the polynomial : @xmath193 for simplicity the dependence of @xmath194 on the symbol @xmath120 is not made explicit and sometimes we will write @xmath195 for shortness . the above polynomials @xmath196 satisfy the following iteration formula . for @xmath197 and @xmath191 , [ iter - com ] & = & c^(m)_0(t_m,,t_1,t)^wick + & & + ( \{q_t_m , c^(m-1)_0(t_m-1,,t_1,t)}^(2))^wick . this is a straightforward consequence of the definition of @xmath196 and the composition formula in proposition [ wick - calc ] . we consider a sequence @xmath53 of density matrices such that @xmath198 . for shortness , we denote @xmath199 and for simplicity write @xmath200 for the wick operator with symbol @xmath201 with @xmath202 . [ prop-1 ] let @xmath53 be a sequence of density matrices such that @xmath198 for each @xmath46 . then for any @xmath59 such that @xmath203 , @xmath202 , @xmath204 and @xmath205 : [ formula - m ] tr[_n ( t ) a^wick ] & = & _ k=0^m-1 i^k _ 0^t dt_1_0^t_k-1 dt_k + + _ k=1^m i^k & & _ 0^t dt_1_0^t_k-1 dt_k + + i^m & & _ 0^t dt_1_0^t_m-1 dt_m , with @xmath206 given in and the multiple poisson bracket defined in . the expansion is obtained by iteration . let @xmath207 then @xmath208 u_0(t)^ * u(t)_{|\vee^n\z}\,.\ ] ] a simple integration yields [ eq.iter.dh ] u(t)^ * u_0(t ) b^wick u_0(t)^ * u(t)_|^n= b^wick _ |^n+ _ 0^t u(t)^ * u_0(t ) [ q_t^wick , b^wick ] u_0(t)^ * u(t)_|^n . taking @xmath209 in the above formula , gives u(t)^ * a^wicku ( t)_|^n & = & u_0(t)^ * a^wicku_0(t)_|^n + & & + _ 0^t dt_1 u(t_1)^ * u_0(t_1 ) u_0(t_1)^ * u(t_1)_|^n . hence using lemma [ iter - com ] , we get u(t)^ * a^wicku ( t)_|^n & = & u_0(t)^ * a^wicku_0(t)_|^n + & & + i _ 0^t dt_1 _ ( ) _ |^n + & & + i _ 0^t dt_1 u(t_1)^ * u_0(t_1 ) ( \{q_t_1,a_t}^(2))^wick u_0(t_1)^ * u(t_1)_|^n . remark that the first two terms in the right hand side are of order @xmath210 while the last one is of order @xmath211 . by using again to expand the term ( t ) above with @xmath212 we obtain , after taking the trace with @xmath12 , the formula for @xmath213 . so , iterating this process @xmath214 times and following the same scheme of splitting commutators into two parts one of order @xmath210 and the second of order @xmath211 , we get u(t)^ * a^wicku ( t)_|^n & = & + & & _ k=0^m-1 i^k _ 0^t dt_1_0^t_k-1 dt_k c^(k)_0(t_k,,t_1,t)^wick + & & + i^m _ 0^t dt_1_0^t_m-1 dt_m u(t_m)^ * u_0(t_m ) ^wick u_0(t_m)^ * u(t_m ) + & & + _ k=1^m-1 i^k _ 0^t dt_1_0^t_k-1 dt_k u(t_k)^ * u_0(t_k ) ( \{q_t_k ; c^(k-1)_0(t_k-1,,t_1,t)}^(2))^wick u_0(t_k)^ * u(t_k ) . hence , by taking the trace with @xmath12 we prove the proposition . the next step is to let @xmath215 in the formula . but to do this we need to prove some estimates which guarantee the absolute convergence of these series . [ tech.lem.1 ] for any @xmath207 the symbols @xmath216 and @xmath217 with the following inequalities holding true : + ( i ) @xmath218 ( ii ) for any @xmath190 , @xmath219 here @xmath220 and @xmath221 are respectively the operators associated to the polynomials @xmath222 and @xmath196 according to definition [ def.symbol ] . see ( * ? ? ? * lemma 5.8 , 5.9 ) . [ prop-2 ] let @xmath53 be a sequence of density matrices such that @xmath198 for each @xmath46 . then for any @xmath59 such that @xmath203 , @xmath202 and @xmath223 and @xmath224 : [ form - inf ] tr[_n ( t ) a^wick ] & = & _ k=0^ i^k _ 0^t dt_1_0^t_k-1 dt_k + + _ k=1^ i^k & & _ 0^t dt_1_0^t_k-1 dt_k , with @xmath206 are given in , the multiple poisson bracket defined in . proposition [ prop-1 ] says that @xmath225 = \sum_{k=0}^{m-1 } c_k + \frac{\hbarr}{2 } \sum_{k=1}^{m } b_k + r_m\,,\ ] ] where @xmath226 and @xmath227 are short notations for the terms appearing in . applying lemma [ tech.lem.1 ] and using the fact that we integrate @xmath140-times , we get for @xmath228 @xmath229 therefore , using the bound @xmath230 , we see that for times @xmath223 the two series @xmath231 and @xmath232 are absolutely convergent and @xmath233 when @xmath215 . [ prop-3 ] let @xmath53 be a sequence of density matrices such that @xmath198 for each @xmath46 . then for any @xmath61 there exists @xmath60 such that for any @xmath59 , @xmath203 , @xmath202 , @xmath234 and @xmath224 : [ est - inf ] |tr[_n ( t ) a^wick ] -_k=0^ i^k _ 0^t dt_1_0^t_k-1 dt_k | c_0 ||a|| . this follows by proposition [ prop-2 ] and estimate . in fact , we see that the left hand side of is bounded by _ k=1^(k+p)^2(8|t|||q||)^k ||a|| _ k=1^(k^2+p^2)(8|t|||q||)^k ||a|| ( 3+p^2 ) ||a|| . taking @xmath235 for @xmath61 , we obtain the inequality . in this section , we explain the notion of wigner measures and its relationship with reduced density matrices . most of the results we need are proved in @xcite , but for reader convenience we briefly recall them since they play an essential role in the proof of our main result . the main observation is that reduced density matrices of a given sequence of normal states have limit points with respect to the weak-@xmath31 topology when @xmath236 and these limit points have a very particular structure . actually , a non - commutative de finetti theorem @xcite due to strmer ( motivated by classification of @xmath237-algebras and type factors ) provides in some sense the structure of these limiting states . this is more apparent in the work of hudson and moody @xcite where the authors focus on normal states which are also used in our setting . actually , it turns out that with wigner measures we can characterize the structure of the limit points more easily , without appealing to @xmath237-algebras formalism , and using probability measures in more natural sets . moreover , some compactness defect phenomena can be easily understood with the latter tool ( see @xcite ) . more recently , the authors lewin , nam and rougerie in @xcite gave an alternative proof of the non - commutative de finetti theorem ( see also @xcite for application of this type of result ) . in finite dimension , wigner ( or semi - classical ) measures are well - known tools in the analysis of pdes with particular scaling ( see for instance @xcite ) . this idea was extended to the infinite dimensional case in @xcite and adapted to the framework of the mean field problem . actually , the borel probability measures @xmath49 appearing in proposition [ definetti ] is what we call wigner measures of the sequence of density matrices @xmath53 . this concept is more general and one can deal with arbitrary families of normal states ( or even trace class operators ) on the fock space . the main advantage is that we can identify these measures @xmath49 by means of simpler quantities involving the weyl operators ( see theorem [ th.wig-measure ] ) according to the formula : @xmath238= \int_{\z } e^{i\sqrt{2 } { \rm re}\langle\xi , z\rangle } \ , d\mu_0(z)\ , , \quad\forall\xi\in\z\,,\ ] ] where @xmath239 refers to the weyl operator on the fock space @xmath240 with @xmath83 . therefore the mean field problem becomes a propagation problem of wigner measures along the nonlinear flow of the ( hartree ) equation . to enlighten the discussion let us consider a concret example . let @xmath241 with @xmath242 and @xmath243 . it is easy to see that the @xmath24-reduced density matrices of @xmath244 are @xmath245 and one can compute explicitly the wigner measure of the sequence @xmath246 according to proposition [ definetti ] : @xmath247=\la \varphi^{\otimes p},b\varphi^{\otimes p}\ra= \int_{\z } \la z^{\otimes p } , b z^{\otimes p}\ra \ ; d\mu_0(z ) , \quad \mbox { with } \quad \mu_0=\frac{1}{2\pi } \int_0^{2\pi } \delta_{e^{i\theta}\varphi } \ , d\theta\,,\ ] ] where @xmath248 denotes the dirac measure on @xmath70 at the point @xmath249 . so theorem [ main - th ] gives in particular the convergence of the evolved reduced density matrices and in our example it yields @xmath250= \int_{\z } \la z^{\otimes p } , b z^{\otimes p}\ra \ ; d\mu_t(z),\quad \mbox { with } \quad \mu_t=(\phi_t)_\sharp\mu_0=\frac{1}{2\pi } \int_0^{2\pi } \delta_{e^{i\theta}\varphi_t } \ , d\theta\,,\ ] ] where @xmath19 is the solution of the nonlinear field ( hartree ) equation with initial condition @xmath251 . so , working with wigner measures allows to understand the superposition of states that may interact in the mean field limit ( see @xcite ) ; and hence it provides a general and flexible point of view . we recall below the result that gives the construction of wigner measures . it is a slight adaptation of ( * ? ? ? * theorem 6.2 ) including ( * ? ? ? * lemma 2.14 ) . [ th.wig-measure ] let @xmath53 be a sequence of density matrices such that @xmath198 for each @xmath46 . then there exists a subsequence @xmath252 and a borel probability measure @xmath38 on @xmath70 , called a wigner measure , such that for any @xmath253 , @xmath254= \int_{\z } e^{i\sqrt{2 } { \rm re}\langle\xi , z\rangle } \ , d\mu(z)\,,\ ] ] with @xmath239 referring to the weyl operator on the fock space @xmath240 with the scaling @xmath255 . moreover , the probability measure @xmath38 is @xmath50 invariant and it is concentrated on the unit ball @xmath256 of the hilbert space @xmath70 ( i.e. : @xmath52 . the @xmath50-invariance of the measure @xmath38 is a straightforward consequence of the fact that @xmath257 for each @xmath46 . so , the above theorem says that the set of wigner measures of a sequence @xmath53 is never empty and we denote it by @xmath258 in practice and without loss of generality , one can assume in the analysis of the mean field problem that the set @xmath259 only contains a single measure . in this subsection we give the proof of proposition [ definetti ] which can be considered as a non - commutative de finetti theorem . moreover , the convergence extends to wick quantized symbols with compact kernels belonging to @xmath260 and hence this proves the weak- * convergence of reduced density matrices . this result is proved in ( * ? ? ? * corollary 6.14 ) and a slight adaptation of it is recalled below . [ pr.polycomp ] let @xmath44 be a sequence of density matrices such that @xmath198 for each @xmath46 and assume that @xmath261 . then the convergence @xmath262 = \tr[\varrho_{\infty}^{(p ) } a]\ , , \quad \mbox { with } \quad \varrho_{\infty}^{(p)}=\int_{\z } |z^{\otimes p}\rangle\langle z^{\otimes p}| \ , d\mu_0(z)\,,\ ] ] holds for any @xmath23 and any @xmath263 . * proof of proposition [ definetti ] * : + suppose that for each @xmath23 and each compact operator @xmath263 the sequence @xmath48)_{n\in{{\mathbb{n}}}^*}$ ] converges . then there exist trace - class operators @xmath264 , @xmath23 , such that @xmath265 = \tr[\varrho_{\infty}^{(p ) } a]\,,\quad \forall a\in\l^{\infty}(\z)\,.\ ] ] let @xmath38 be any wigner measure in @xmath266 . then by proposition [ pr.polycomp ] , up to extraction of subsequences , we see that @xmath267 so , this provides the existence of a borel probability measure @xmath38 on @xmath70 with the appropriate properties . the uniqueness follows by ( * ? ? ? * proposition 6.15 ) . the convergence in proposition [ pr.polycomp ] is with respect to the weak- * topology on @xmath268 which is the topological dual of @xmath269 and the statement ( [ eq.convpolww ] ) does not hold in general for all @xmath270 , @xmath23 . counterexamples exhibiting this phenomenon of dimensional defect of compactness are given in @xcite ( we call it in this way because of the similarity with finite dimension , although the source of defect here is the fact the phase - space is of infinite dimension and so bounded sets are not relatively compact in the norm topology ) . actually , the extension of ( [ eq.convpolww ] ) to all bounded operators @xmath270 and @xmath23 depends on the sequence @xmath44 and it turns out to be an important point in the mean field problem : we need this information when we take the limit @xmath236 in the mean field expansion . let @xmath44 be a sequence of density matrices such that @xmath198 for each @xmath46 . the reduced density matrices @xmath271 weakly converges to @xmath272 if @xmath273=\tr[\varrho_{\infty}^{(p ) } \,a]\ , , \quad \;\forall a\in \l(\vee^p \z)\,.\end{aligned}\ ] ] the following proposition provides a strong relationship between the wigner measures of a sequence of density matrices and the convergence of their reduced density matrices in the @xmath274-norm topology . [ pr.wignergp0 ] let @xmath44 be a sequence of density matrices with @xmath257 for each @xmath46 . suppose that the reduced density matrices @xmath275 weakly converge to @xmath272 for each @xmath23 according to . then there exists a unique @xmath50-invariant borel probability measure @xmath49 on @xmath70 such that for any @xmath23 , _ n ||^(p)_n-_^(p)||_1=0 , _ ^(p)= _ measure of @xmath53 and it is concentrated on the unit sphere @xmath276 of @xmath70 centred at the origin and of radius one ( i.e. : @xmath277 . the assumption on @xmath53 imply that @xmath278 are non - negative trace class operators with @xmath279=1 $ ] and @xmath280 converges to @xmath281 with respect to the weak topology in @xmath282 . but since @xmath280 and @xmath283 are non negative trace - class operators with @xmath284=1=\tr\left[\varrho^{(p)}_{\infty}\right]$ ] , the @xmath274-norm convergence follows according to @xcite . in a more general framework , it is said that @xmath285 has the kadec - klee property ( kk * ) in the weak- * topology ( see @xcite and references therein ) . the ( kk * ) property on a dual banach space means that the weak- * and norm convergence coincide on the unit sphere . + thanks to the proof of proposition [ definetti ] , we know that @xmath49 is the unique wigner measure of the sequence @xmath53 . so , the measure @xmath49 is @xmath50-invariant and it is concentrated on the unit ball of @xmath70 according to theorem [ th.wig-measure ] . now , using the fact that @xmath286=1 $ ] , we get @xmath287 this easily yields that the measure is actually concentrated on the unit sphere . [ pr.eqppi ] let @xmath44 be a sequence of density matrices with @xmath257 for each @xmath46 and such that @xmath288 . the two following conditions are equivalent : @xmath289 suppose that the wigner measure @xmath49 is concentrated on the unit sphere , then by proposition [ pr.polycomp ] we see that @xmath275 weak- * converges to @xmath283 which is a non - negative trace - class operator with @xmath290=1 $ ] . so , again by the kadec - klee property ( kk * ) of @xmath291 we obtain the @xmath274-norm convergence for each @xmath23 . in this section we give the proof of our main result ( theorem [ main - th ] ) . we start by proving an elementary estimate in subsection [ sub.est ] and then prove the result in subsection [ sub.proof ] . instead of estimating the quantities @xmath292 in the trace norm , we will work essentially with @xmath293-\int_\z \langle z^{\otimes p } , a\ , z^{\otimes p}\rangle \,d\mu_0\right|\,.\ ] ] in that way , we can use the mean field expansion based on wick calculus . the two quantities are comparable and this is given by the lemma below . [ lem.1 ] let @xmath53 be a sequence of density matrices such that @xmath294 for each @xmath46 and satisfying the assumptions of theorem [ main - th ] . then for any @xmath295 : | ||^(p)_n-_^(p)||_1- _ a0 @xmath53 and @xmath296 for any @xmath202 and @xmath224 : @xmath297=\frac{n \cdots ( n - p+1)}{n^p } \,\tr[\varrho_n^{(p ) } a ] \quad \mbox { and } \quad \tr[\varrho_\infty^{(p)}a]=\int_{\z } \langle z^{\otimes p } , a z^{\otimes p}\rangle\ , d\mu_0(z)\,.\ ] ] hence , we get | -_z^p , a z^pd_0| & & |1-| ||a||+ |-| + & & ||a|| , and also | | & & |1-| ||a||+ |-_z^p , a z^pd_0| . the inequality 1-=1-_j=1^p-1 ( 1-j n ) 1-(1-)^p-1 , gives the sought estimate . recall that @xmath298 in theorem [ main - th ] with @xmath49 is the unique wigner measure of the sequence @xmath53 provided by proposition [ definetti ] . [ tech.lem.2 ] for any @xmath299 such that @xmath300 : @xmath301 we know already that the measure @xmath49 is concentrated in the ball of radius @xmath302 centred at the origin according to proposition [ definetti ] . hence , we deduce the inequality @xmath303 the right hand side of is absolutely convergent whenever @xmath304 thanks to the estimate ( ii ) of lemma [ tech.lem.1 ] : _ k=0^ _ 0^tdt_1_0^t_k-1dt_k |_0 ( c^(k)_0(t_k, ,t_1,t))|2^p-1 ||a|| _ k=0^ ( 8 |t| ||q||)^k . recall that according to the classical solution @xmath305 verifies for any @xmath207 , b(z_t)=b_t(w_0)+i_0^t\{q_t_1,b_t}(w_t_1 ) dt_1 , w_t = e^i th_0 z_t . iterating this formula , with @xmath306 , and using the absolute convergence checked above gives for all @xmath307 , @xmath308 integrating with respect to @xmath49 and using the fact that @xmath309 yields . * proof of theorem [ main - th ] : * + for reader convenience , we recall the assumptions of theorem [ main - th ] . let @xmath54 be a sequence of positive numbers with @xmath310 and such that @xmath56 is bounded . consider @xmath44 and @xmath57 to be two sequences of density matrices with @xmath45 and @xmath58 for each @xmath59 . assume that there exist @xmath60 , @xmath61 , @xmath62 such that for all @xmath59 with @xmath63 , [ init - ineq - bis ] ^(p)_n-_^(p)_1 c_0 . we first prove the estimate for short times and than extend it to arbitrary times . so , suppose that @xmath311 with @xmath61 the constant provided by the main assumption . thanks to lemma [ lem.1 ] it is enough to estimate the quantity @xmath312-\mu_t\big(\langle z^{\otimes p } , a\ , z^{\otimes p}\rangle\big)|$ ] for any bounded operator @xmath202 . so , the estimate in proposition [ prop-3 ] yields [ eq.fes ] |tr[_n ( t ) a^wick ] -_t(z^p , a z^p ) | & & c_0 ||a|| + ( t ) , for some @xmath60 and ( t)&= & |_k=0^ i^k _ 0^t dt_1_0^t_k-1 dt_k -_t(z^p , a z^p)| + & = & |_k=0^ i^k _ 0^t dt_1_0^t_k-1 dt_k -_0(c_0^(k)(t_k,,t_1,t))| . the last equality is a consequence of lemma [ tech.lem.2 ] . using now the main assumption and the fact that @xmath313 are polynomials in @xmath314 , we get the inequality ( we can assume that @xmath315 without loss of generality ) [ t1 ] ( t ) & & _ k=0 ^ -p _ 0^t dt_1_0^t_k-1 dt_k |tr-_0(c_0^(k)(t_k,,t_1,t))| + [ t2 ] & & + ^n - p_k =- p+1 _ 0^t dt_1_0^t_k-1 dt_k [ t3 ] & & + _ k = n - p+1^ _ 0^t dt_1_0^t_k-1 dt_k |_0(c_0^(k)(t_k,,t_1,t))| . thanks to the estimate ( ii ) of lemma [ tech.lem.1 ] , the right hand side of - is bounded by & & _ k=0 ^ -p ( 8 t c ||q||)^k ||a|| ||a|| + & & 2^n - p_k =- p+1 ( 8 t ||q||)^k ||a||2||a|| + & & _ k = n - p+1^(8 t ||q||)^k ||a||=||a|| . since @xmath311 , we easily get the bounds , with @xmath316 which depends only on @xmath74 and the sequence @xmath54 . collecting theses estimates , we conclude that ( t ) & & ||a|| . so , by lemma [ lem.1 ] there exists @xmath317 such that ^(p)_n(t)-_^(p)(t)_1 c_1 , uniformly in time whenever @xmath318 $ ] . now , iterate the same reasoning as much as needed to cover a time interval @xmath319 $ ] with @xmath64 arbitrary . then one gets the existence of @xmath65 such that for all @xmath66 $ ] , ^(p)_n(t)-_^(p)(t)_1 c_t . in order to illustrate the main result of this article , it is useful to consider some examples and numerical simulations of states with an increasing degree of correlation . the notion of correlation is quite important in quantum information theory and it is related to the so - called quantum entanglement . so , there are several interesting examples of states in the latter field which are also useful for our purpose ( bell state , cat state , w state , ghz state , ... ) . this is the most known example in mean field theory . it appears in the literature under the name of chaos , factorized , product or also hermite states . it emphasizes the fact that , in the mean field limit , states @xmath320 that are prepared uncorrelated will evolve into states which are close to be uncorrelated , namely @xmath321 where @xmath19 is a solution of the hartree equation with initial condition @xmath251 . it is easy to see that the factorized states @xmath322 satisfy the assumption of theorem [ main - th ] . in fact , the @xmath24-reduced density matrices of @xmath12 coincide with the limit [ prod ] _ n^(p)=|^p^p|=_^(p ) . this means that in this example the rate of convergence at initial time @xmath2 is arbitrary fast . remember that according to theorem [ main - th ] the @xmath24-particle reduced density matrix @xmath323 is _ |z^pz^p| d_t(z)=|_t^p_t^p| _ t= _ 0 ^ 2 _ e^i_t d , where @xmath19 is a solution of the nonlinear field equation . a numerical simulation performed on a discrete model ( figure [ fig : plot1 ] ) , shows that for the first marginal _ ^(1)_n(t)-_^(1)(t)_1=-(1+)(n ) + o(1 ) ; with a deviation @xmath324 well within the expected computational inaccuracy . this is in very good agreement with our mathematical prevision , and indicates that the estimate in theorem [ main - th ] is not far from being optimal , in some sense . the w state is a multi - partite @xmath0-qubit entangled quantum state @xmath325 where @xmath326 denotes a one particle excited state and @xmath327 denotes the one particle ground state of two mode system . more generally , if @xmath70 is a hilbert space and @xmath328 are two normalized orthogonal vectors in @xmath70 then [ def.wst ] |w=(|^n-1 + |^n-2 + ... + |^n-1 ) . let @xmath53 denotes a sequence of w states as in . then for all @xmath59 such that @xmath329 : @xmath330 a simple computation yields for any @xmath202 : w , a1^(n - p ) w= ^p , a ^p+ w_p , a w_p , where @xmath331 . so that the @xmath24-reduced density matrices of @xmath12 is _ n^(p)&= & |^p^p|+ |w_pw_p| + & = & _ ^(p)+ |w_pw_p| , hence the estimate follows since @xmath332 is a normalized vector . the ghz ( greenberger - horne - zeilinger ) state is a multipartite entangled quantum state . in a two - mode system it is given by the formula @xmath333 so , it can be generalized as follows [ def.ghz ] |= , where @xmath251 and @xmath334 are two normalized ( orthogonal ) vectors in a given hilbert space @xmath70 . so , the @xmath0-partite ghz states are superposition of uncorrelated states and it is again easy to check that their @xmath24-reduced density matrices coincide with their limit as in . hence , theorem [ main - th ] provides a rate of convergence for this example too with @xmath335 rate . let @xmath336 be two normalized orthogonal vectors . the _ twin states _ are rank one projectors @xmath244 given by [ twin - vec ] _ _ n _ 1^n_1_2^n_2 , with @xmath337 and @xmath338 . this sequence of states have a unique wigner measure @xmath38 computed in @xcite . so , after identification of the hilbert space @xmath70 as @xmath339 , with @xmath340 the orthogonal subspace to @xmath341 , the measure @xmath38 reads [ twin - mes ] = ^s^1_^s^1__0^ ^s^1_= _ 0 ^ 2 _ e^i d , j=1,2 . remark that in this example the measure @xmath342 is quite correlated because of the nonlinear effect of the flow and the situation differs significantly from the simple picture of uncorrelated states ( here @xmath69 is the flow of the nonlinear field equation ) . [ twin ] let @xmath53 be a sequence of twin states with @xmath38 its wigner measure given in . then for any @xmath59 such that @xmath343 : _ n^(p)-_^(p)_1 2^p _ ^(p)=_|z^pz^p| d. let @xmath344 be the vector given by . a simple computation yields _ n , a1^(n - p ) _ n&= & _ , ( n ) t__1^n_1_2^n_2 , a1^(n - p ) t__1^n_1_2^n_2 , where @xmath345 denotes the operator on @xmath346 defined for any @xmath347 by @xmath348 for @xmath349 , @xmath350 , we denote @xmath351 so , there is a correspondence between permutations @xmath352 and maps @xmath353 according to [ sigi ] t__1^n_1_2^n_2= _ i(1)_i(n)=:(i ) . since the cardinal of the set of @xmath354 such that holds for the same @xmath355 is equal to @xmath356 , we see that _ n , a1^(n - p ) _ n&= & ( ) ^2 _ i , j_n^(n_1 ) ( i ) , a1^(n - p ) ( j ) . in the above sum if @xmath357 on the set @xmath358 then the scalar product is null because of the orthogonality condition on the vectors @xmath359 . so this simplifies the sum and actually we can decompose it according to the number of occurrence of @xmath360 in the first @xmath24 vectors constituting @xmath361 , i.e. : _ n^(n_1)=_k=0^p \ { i_n^(n_1 ) , i^-1(\{1 } ) \{1,,p}=k}=:_k=0^p _ n , k^(n_1 ) . hence , _ a1^(n - p ) _ n= _ i_n , k^(n_1 ) _ j_n^(n_1 ) , j = i_| p+1,,n ( i ) , a1^(n - p ) ( j ) . if we fix the first @xmath24 values of @xmath362 and @xmath363 and variate the @xmath364 others then the scalar product @xmath365 will not change as long as @xmath366 . actually , there is @xmath367 configurations for each choice of @xmath368 such that @xmath369 . hence , we get _ n , a1^(n - p ) _ n= _ k=0^p c_n - p^n_1-k _ j_p^(k ) ( i ) , a ( j ) . observe that for all @xmath370 and @xmath371 : @xmath372 so , we see that the limit of the @xmath24-reduced density matrices is @xmath373 where @xmath38 is the wigner measure of the sequence @xmath374 given in . in particular , the orthogonality of the family @xmath375 gives 1=||_^(p)||_1=_k=0^p _ k^2 therefore a simple estimate yields _ n^(p)-_^(p)_1 & & _ k=1,,p | 1 - 2^p| _ k=0^p ||_k||^2 + & & _ k=1,,p | 1 - 2^p| . so , the result follows once we prove _ k=1,,p | 1 - 2^p| 2^p . in fact , for any @xmath376 such that @xmath377 the following simple estimates hold true [ est - el1 ] 0&1-_i=1^r ( 1-a_i)&r _ 1,,r a_i + [ est - el2 ] 0&_i=1^r ( 1+a_i)-1&2^r-1 r _ 1,,r a_i . by writing 2^p=_t_1 ^t_2 we can see that the product @xmath378 while the last one is @xmath379 with @xmath380 ( some of the @xmath381 are null ) . hence applying - , we obtain & & 2^p-1 p_1, ,p _ j + p _ 1, ,p _ i + & & 2^p-1 p+p . again in this example , a numerical simulation indicates a @xmath1 order of convergence for the first reduced density matrix ( figure [ fig : plot2 ] ) . finally , we bring to reader s attention the fact that any rate of convergence is actually possible . in fact take the following example @xmath382 with @xmath54 such that @xmath383 , @xmath384 and @xmath385 are two normalized orthogonal vectors . so , it is easy to see that @xmath386 therefore , for each @xmath23 , the following equality is satisfied : @xmath387 the research of the second and third authors has been supported respectively by the centre henri lebesgue anr-11-labx-0020 - 01 and anr-11-is01 - 0003 lodiquas . l. tartar . h - measures , a new approach for studying homogenization . oscillations and concentration effects in partial differential equations , proceedings of the royal society edinburgh , 115-a ( 1990 ) , 193230 .

we report on the results from our deep _ chandra _ observation ( 120 ks ) of the supernova remnant ( snr ) n49 and soft-ray repeater ( sgr ) 052666 in the large magellanic cloud . we firmly establish the detection of an ejecta `` bullet '' beyond the southwestern boundary of n49 . the x - ray spectrum of the bullet is distinguished from that of the main snr shell , showing significantly enhanced si and s abundances . we also detect an ejecta feature in the eastern shell , which shows metal overabundances similar to those of the bullet . if n49 was produced by a core - collapse explosion of a massive star , the detected si - rich ejecta may represent explosive o - burning or incomplete si - burning products from deep interior of the supernova . on the other hand , the observed si / s abundance ratio in the ejecta may favor type ia origin for n49 . we refine the sedov age of n49 , 4800 yr , with the explosion energy 1.8 10 erg . our blackbody ( bb ) + power law ( pl ) model for the quiescent x - ray emission from sgr 052666 indicates that the pl photon index ( 2.5 ) is identical to that of psr 1e1048.15937 , the well - known candidate transition object between anomalous x - ray pulsars and sgrs . alternatively , the two - component bb model implies x - ray emission from a small ( 1 km ) hot spot(s ) ( 1 kev ) in addition to emission from the neutron star s cooler surface ( 10 km , 0.4 kev ) . there is a considerable discrepancy in the estimated column toward 052666 between bb+pl and bb+bb model fits . discriminating these spectral models would be crucial to test the long - debated physical association between n49 and 052666 .

1201.5056

the supernova remnant ( snr ) n49 is the third brightest x - ray snr in the large magellanic cloud ( lmc ) . the blast wave is interacting with dense clumpy interstellar clouds on the snr s eastern side @xcite , producing bright emission in optical , infrared , ultraviolet , and x - ray bands @xcite , which revealed the shock structures in n49 down to sub - pc scales over wide ranges of the velocity ( @xmath8 @xmath9 10@xmath10 km s@xmath11 ) , temperature ( @xmath12 @xmath9 10@xmath13 k ) , and the interacting density ( @xmath14 @xmath9 110@xmath15 @xmath16 ) . the blast wave in the western side of the snr appears to be propagating into a lower density medium than in the east @xcite . early studies estimated overall low metal abundances in n49 , indicating that metal - rich ejecta might have been substantially intermixed with the interstellar medium ( ism ) of the lmc @xcite . the origin of n49 has been elusive ( thermonuclear vs. core - collapse explosion ) . the interstellar environment of n49 , indicated by the presence of nearby molecular clouds @xcite and young stellar clusters @xcite and its location within an ob association @xcite generally support a core - collapse origin for n49 from a massive progenitor star . one model suggested that n49 is interacting with a dense shell of a strmgren sphere created by a massive b - type progenitor @xcite . the positional coincidence of a soft @xmath0-ray repeater ( sgr ) 052666 within the boundary of n49 @xcite may also suggest a core - collapse origin from a massive star , if this sgr is the compact remnant of the sn that created n49 . however , the physical association between snr n49 and sgr 052666 is uncertain @xcite . based on the _ asca _ data , hughes et al . ( 1998 ) found no evidence for surrounding ism that had been modified by stellar winds from the massive progenitor . nucleosynthesis studies of metal - rich ejecta in snrs using spatially - resolved x - ray spectroscopy are useful to constrain the progenitor s mass ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? although bright x - ray features in n49 are apparently dominated by emission from the shock - cloud interaction , previous _ observations have shown marginal evidence for enhanced emission lines and overabundant metal elements from a few sub - regions in the snr @xcite . for instance , a promising ejecta candidate was a small protrusion beyond the snr s southwestern boundary . this emission feature was suggested to be a candidate ejecta bullet , but metal overabundances were not conclusive because of poor photon statistics @xcite . also , because n49 was detected significantly off - axis ( by @xmath96@xmath175 ) in the previous _ chandra _ data ( obsid 1041 ) used by park et al . ( 2003a ) , the point spread function ( psf ) was substantially distorted , and the true morphology of this feature ( point - like vs. extended ) was uncertain . we note that there were three other on - axis _ _ observations of n49@xcite . these previous observations were not useful to study the ejecta bullet candidate because of either the use of a ccd subarray or a very short exposure . thus , a firm detection of metal - rich ejecta in n49 remains elusive . here we report on initial results from our new _ observation of n49 . data present @xmath94 times deeper exposure ( for the entire n49 ) than the data used by park et al . ( 2003a ) , with an on - axis pointing to n49 . with these new data , we clearly detect some metal - rich ejecta features in n49 . combining the new data with archival data , we also perform a detailed spectral analysis of sgr 052666 . in this paper , we report on the results from our spectral analysis of sgr 052666 and the detected metal - rich ejecta features in n49 . our studies of complex shock - cloud interaction regions of n49 and the timing analysis of sgr 052666 will be presented in a subsequent paper . we describe the observations and the data reduction in section [ sec : obs ] . the image and spectral analyses are presented in section [ sec : result ] , and a discussion in section [ sec : disc ] . a summary will be presented in section [ sec : sum ] . we observed n49 with the advanced ccd imaging spectrometer ( acis ) on board _ chandra x - ray observatory _ on 2009 july 18 september 19 ( obsids 10123 , 10806 , 10807 , and 10808 ) during ao10 . we chose the acis - s3 chip to utilize the best sensitivity and energy resolution of the detector in the soft x - ray band . the pointing ( @xmath18@xmath19 = 05@xmath20 25@xmath21 58@xmath22.8 , @xmath23@xmath19 = @xmath2466@xmath25 05@xmath26 00@xmath270 ) was roughly toward the geometrical center of n49 . sgr 052666 and the ejecta bullet candidate are detected within @xmath930@xmath28 and @xmath940@xmath28 of the aim point , respectively . we used a 1/4 subarray of the acis - s3 to ensure low photon pileup ( @xmath29 5% ) of sgr 052666 , and to allow us to detect the 8 s pulsations from the sgr , while still obtaining full coverage of the entire snr n49 . we processed the raw event files following the standard data reduction methods using ciao 4.2 and _ chandra _ caldb 4.3.0 , which includes correction for the charge transfer inefficiency ( cti ) . we applied the standard data screening by status and grade ( _ asca _ grades 02346 ) . we performed this data reduction on individual observations . the overall light curve from the entire s3 chip for each obsid showed no evidence of flaring particle background . then , we merged four data sets into a single event file by re - projecting them onto obsid 10807 s tangential plane . after the data reduction , the total effective exposure is @xmath9108 ks . we also used two data sets ( obsids 747 and 1957 ) available in the _ chandra _ archive6@xmath175 , obsid 1041 ) and a short exposure ( @xmath97 ks , obsid 2515 ) . ] . the pointing of these archival data was toward the position of sgr 052666 ( @xmath18@xmath19 = 05@xmath20 26@xmath21 00@xmath22.7 , @xmath23@xmath19 = @xmath2466@xmath25 04@xmath26 35@xmath270 ) , and thus they provide a high resolution imaging of n49 as well as the sgr . although these archival data can not be used to study south - southwestern regions ( including the ejecta bullet candidate ) of n49 due to the use of 1/8 subarray of the acis - s3 , they covered east - northeastern parts of the snr with a decent exposure ( a total of @xmath988 ks by combining obsids 747 and 1957 ) . thus , these data are useful to study eastern regions of n49 and sgr 052666 . we re - processed these two archival data sets in the same way as we did for our new observation data . combining all these observations , our data present the deepest coverage with the high resolution x - ray imaging spectroscopy for snr n49 ( @xmath9108 ks in the south - west regions and @xmath9196 ks in the north - east regions ) and sgr 052666 ( @xmath9196 ks ) . observations and archival data used in this work are summarized in table [ tbl : tab1 ] . an x - ray color image of n49 shows complex asymmetric x - ray emission features ( figure [ fig : fig1]a ) . these features have been partially revealed by previous _ data @xcite . our new deep _ image reveals all those fine x - ray structures throughout the entire snr with substantially improved photon statistics and resolution ( figure [ fig : fig1]a ) . the outer boundary of n49 appears as a thin shell radiating in soft x - rays ( reddish in figure [ fig : fig1]a ) that represents the blast wave sweeping through the general ambient medium of the lmc . a notable exception is the candidate ejecta bullet extending beyond the snr shell in the southwest , which is distinctively blue in color ( figure [ fig : fig1]c ) . bright x - ray filaments in the eastern parts of the snr are regions where the shock is interacting with dense clumpy clouds ( * ? ? ? * and references therein ) . these shock - cloud interaction regions show a wide range in x - ray colors at angular scales down to several arcseconds . the western parts of n49 are relatively faint and x - ray emission is emphasized in green to blue while showing some red filamentary features as well ( figure [ fig : fig1]a ) . this general color variation across the snr appears to be related to the temperature variation caused by the overall interstellar density distribution , in which dense clouds are interacting with the shock mostly in the eastern half of the snr ( e.g. , * ? ? ? the spectrally - hard , blue protrusion extending beyond the main snr shell in the southwestern boundary was proposed to be a candidate ejecta bullet based on the previous low - resolution _ chandra _ data @xcite . if it is indeed an ejecta bullet , this feature is reminiscent of `` shrapnel '' found in the vela snr @xcite , but nearly two orders of magnitude more distant . data clearly resolve this feature . it is an extended feature with a head ( @xmath94@xmath28 in radius ) and a tail that connects the head to the main snr shell ( the `` head '' and `` tail '' regions in figure [ fig : fig1]a , respectively ) . the blue color of the head region is likely due to enhanced line emission in the hard x - ray band : e.g. , the si ( he@xmath18 + ly@xmath18 ) line _ equivalent width _ ( ew ) map shows a strong enhancement in the head region ( figure [ fig : fig2]a ) , which suggests a metal - rich ejecta for the origin of this feature . in fact , our spectral analysis of the head region shows highly overabundant si and s ( see section [ subsubsec : ejecta ] ) . the tail region appears to be composed of a central hard ( blue ) emission region surrounded by a soft ( red ) , conical region ( figure [ fig : fig1]c ) . it is likely the turbulent region behind the bullet , in which the metal - rich ejecta and shocked ism are intermixed , possibly enclosed by the sides of a bow - shock produced by the bullet s supersonic motion . the bow shock , however , is not prominent ahead of the bullet , possibly due to unfavorable viewing geometry . the axis bisecting the opening angle ( @xmath987@xmath25 ) is roughly pointing back to the geometric center of the snr , but not exactly toward sgr 052666 . we noticed that there is a small region in the eastern part of the snr ( the `` east '' region in figure [ fig : fig1]a ) , which shows similar characteristics to those of the head region : i.e. , this region stands out with a distinctively blue color in the bright eastern half of the snr , and it is coincident with a strongly enhanced si line ew ( figure [ fig : fig2]a ) . the east region corresponds to an inter - cloud region ( figure [ fig : fig2]b ) in which a possibility of metal overabundance was suggested by our previous work @xcite . our spectral analysis of this region indeed reveals its metal - rich ejecta nature ( see section [ subsubsec : ejecta ] ) . based on results from our image analysis , we extracted spectra from several regions that appear to characteristically represent distinctive features : i.e. , the swept - up interstellar medium ( the `` shell '' and `` sgr_bg '' regions ) , shocked metal - rich ejecta ( the `` head '' and `` east '' regions ) , and the connecting region between the ejecta bullet and the main snr shell ( the `` tail '' region ) . these regions are marked in figures [ fig : fig1]a & [ fig : fig1]b . we also extracted the spectrum from the quiescent x - ray counterpart of sgr 052666 ( the `` sgr '' region in figure [ fig : fig1]b ) . we present the spectral analysis of these swept - up medium , ejecta features , and 052666 in sections [ subsubsec : ism ] , [ subsubsec : ejecta ] , and [ subsubsec : sgr ] , respectively . the brightest x - ray emission in the eastern parts of n49 originates from the shock interaction with clumpy clouds with a density gradient spanning a few orders of magnitudes @xcite . x - ray emission from such a clumpy medium would represent a complex mixture of gas temperatures and ionization states . such a plasma condition is likely localized in strong shock - cloud interaction regions . while the western regions do not show evidence of shock interaction with dense clumpy clouds , they also show some significant x - ray color variation ( figure [ fig : fig1]a ) . for those regions , it is not straightforward to characterize the x - ray emission spectrum of the general ambient medium shocked by the blast wave . thus , to study the nature of the general swept - up ism and the snr s overall dynamics , we extracted the x - ray spectrum from a region in the outermost rim of the snr in the southern boundary ( the `` shell '' region in figure [ fig : fig1]a ) in which no complex spectral and/or spatial substructures are seen . we also used a small annular region surrounding sgr 052666 ( the `` sgr_bg '' region in figure [ fig : fig1]b ) to characterize background emission for sgr 052666 . since the sgr_bg region is in the _ transition _ area between the bright eastern and faint western parts of the snr , it may also effectively show an average plasma condition of the complex shock structures across the snr . for our spectral analysis , we re - binned the observed spectra to contain a minimum of 20 counts per energy channel . we fit the x - ray spectrum with a non - equilibrium ionization ( nei ) state plane - shock model ( vpshock model in conjunction with the nei version 2.0 in the xspec software , * ? ? ? * ) that is based on atomdb @xcite . we used an augmented version of this atomic database to include inner - shell processes ( e.g. , lines from li - like ions ) and updates of the fe l - shell lines , whose effects are substantial in under - ionized plasma , but are not incorporated in the standard xspec nei version 2.0 . we fixed the galactic column at @xmath30 = 6 @xmath3 10@xmath31 @xmath32 toward n49 @xcite . we fit the foreground column by the lmc ( @xmath33 ) assuming the lmc abundances available in the literature ( he = 0.89 , c = 0.30 , n = 0.12 , o = 0.26 , ne = 0.33 , na = 0.30 , mg = 0.32 , al = 0.30 , si = 0.30 , s = 0.31 , cl = 0.31 , ar = 0.54 , ca = 0.34 , cr = 0.61 , fe = 0.36 , co = 0.30 , and ni = 0.62 , * ? ? ? * ; * ? ? ? hereafter , elemental abundances are with respect to solar @xcite . the southern boundary of n49 was not covered by obsids 747 and 1957 because of the use of a 1/8 subarray . thus , we used our new _ chandra _ data to extract the spectrum from the shell region . the background was subtracted using the spectrum extracted from two circular source - free regions ( with a radius of 8@xmath28 ) beyond the south - western boundary of the snr . the shell region spectrum contains significant photon statistics of @xmath94600 counts , and can be well fitted by the nei plane - shock model ( @xmath34/n = 49.1/68 with the electron temperature @xmath35 = 0.57 kev and the ionization timescale @xmath36 = 6.35 @xmath3 10@xmath37 @xmath16 s , figure [ fig : fig3]a ) . the foreground column by the lmc is estimated to be @xmath33 @xmath9 0.9 @xmath3 10@xmath38 @xmath32 . the best - fit metal abundances are generally consistent with the lmc values . results from this spectral model fit are summarized in tables [ tbl : tab2 ] & [ tbl : tab3 ] . the sgr_bg region was observed by obsids 747 and 1957 as well as our new _ chandra _ observations , and thus , we used all of those data to extract the spectrum from this region . the total photon statistics for this region is @xmath913600 counts . the background was subtracted using the spectrum extracted from a circular source - free region ( with a radius of 8@xmath28 ) beyond the northern boundary of the snr . we initially fit these three data sets simultaneously with all fitted parameters tied with each other . although the overall fit may be acceptable ( @xmath34/n = 1.28 ) , there appears to be a small systematic offset in the normalization at @xmath39 @xmath29 1 kev between data taken in 2000/2001 ( obsids 747 and 1957 ) and in 2009 ( obsids 10123 + 10806 + 10807 + 10808 ) . considering the fact that the time separation between 2000/2001 and 2009 observations is substantial and that the effect appears to be emphasized in the soft band ( @xmath39 @xmath29 1 kev ) , this small discrepancy is likely related to the calibration effect by the time - dependent quantum efficiency degradation in the acis . since the quantum efficiency degradation affects the spectral modeling as if there is an `` extra '' foreground absorption , we attempted the model fit with @xmath40 untied among three spectra . the best - fit @xmath40 values are the same between 2000 and 2001 data , while it is @xmath925% larger for the 2009 data . thus , we repeated the spectral model fit with @xmath40 tied only between 2000 and 2001 data , and with that for 2009 data allowed to vary freely . the fit improved somewhat ( @xmath34/n = 1.19 ) . the best - fit columns are @xmath40 = 1.39@xmath41 @xmath3 10@xmath38 @xmath32 and 1.76@xmath42 @xmath3 10@xmath38 @xmath32 for the 2000/2001 and 2009 data , respectively . while showing a marginal calibration effect , these values are consistent within statistical uncertainties ( hereafter , errors are with a 90% confidence level ) . in the following discussion , we assume the average value of @xmath40 = 1.58@xmath43 @xmath3 10@xmath38 @xmath32 . the best - fit spectral parameters for the sgr_bg region are @xmath35 @xmath9 0.56 kev and @xmath36 @xmath9 9.65 @xmath3 10@xmath37 @xmath16 s , all of which are consistent with those for the shell region within uncertainties . the best - fit metal abundances are also generally consistent with those from the shell region . these model parameters are summarized in tables [ tbl : tab2 ] & [ tbl : tab3 ] . data clearly resolve the extended nature of the ejecta fragment bullet ( the head region ) and the trailing tail - like hot gas ( the tail region , figure [ fig : fig1]c ) . because the bullet region was not covered by the 2000/2001 data , we used only 2009 data for the spectral analysis of this feature . we extracted @xmath91800 counts from the head region . in contrast to those for the swept - up medium , the x - ray spectrum of the head region shows remarkably enhanced emission lines from highly ionized si and s ( figure [ fig : fig4]a ) , which is consistent with the corresponding bright si line ew image ( figure [ fig : fig2]a ) . we fitted this spectrum with an nei plane - shock model following the method described in section [ subsubsec : ism ] . the background was subtracted using the same spectrum as we used for the shell region . we initially fixed metal abundances at the best - fit values for the shell region ( table [ tbl : tab3 ] ) , which resulted in a statistically unacceptable fit ( @xmath34/n = 3.3 ) primarily because of the strongly enhanced si and s lines . varying si and s abundances provides an acceptable fit ( @xmath34/n = 67.0/59 , figure [ fig : fig4]a ) . the best - fit lmc column ( @xmath33 @xmath9 1.3 @xmath3 10@xmath38 @xmath32 ) is consistent with those estimated from the shell and sgr_bg regions . the electron temperature ( @xmath35 @xmath9 1 kev ) and the ionization timescale ( @xmath36 @xmath44 2 @xmath3 10@xmath45 @xmath16 s ) are significantly higher than those from swept - up medium . the si ( @xmath92.3 ) and s ( @xmath93.2 ) abundances are an order of magnitude larger than the lmc values , firmly establishing the metal - rich ejecta nature of the bullet . varying other elemental abundances ( o , ne , mg , and fe ) somewhat improves the fit ( @xmath34/n = 49.3/55 , but the corresponding f - probability is marginal [ @xmath90.002 ] ) , and the best - fit electron temperature is very high ( @xmath35 @xmath9 4 kev ) . in addition to the overabundant si and s , the best - fit abundance appears to be moderately enhanced for o ( @xmath90.7 ) , ne ( @xmath92.1 ) , and mg ( @xmath91.0 ) . however , these abundances are not well constrained ( with a factor of @xmath462 uncertainties ) , and thus the overabundance for these elements is not convincing based on the current data . we conclude that the main differences in the fitted parameters resulting from varying metal abundances other than si and s ( i.e. , the high @xmath35 @xmath9 4 kev and possibly enhanced abundances for o , ne , and mg ) are not compelling . otherwise , it does not affect our discussion of the ejecta features ( see section [ subsec : disc_ejecta ] ) . thus , we hereafter discuss the nature of the bullet based on our spectral model fit with only si and s abundances varied . the results from the head region model fit are summarized in tables [ tbl : tab2 ] & [ tbl : tab3 ] . the tail region is faint , and we extracted @xmath91500 counts in this region . we fit this spectrum with an nei plane shock model @xmath34/n = 44.9/44 , figure [ fig : fig4]c ) . the tail region is fitted by the x - ray spectrum from a hot gas ( @xmath35 @xmath9 2 kev ) in a significantly under - ionized state ( @xmath36 @xmath9 7 @xmath3 10@xmath47 @xmath16 s ) . the best - fit elemental abundances are consistent with those estimated from the shell and sgr_bg regions . we note that the tail region shows substructures of the spectrally - hard ( blue ) _ interior _ surrounded by soft ( red ) outer layer ( figure [ fig : fig1]c ) . it suggests that the tail region may be a mixture of metal - rich ejecta and shocked swept - up medium . thus , we attempted an alternative spectral model , assuming two shock components , a soft component for the swept - up ism and a hard component for the metal - rich ejecta . the spectral parameters , except for the normalization , and metal abundances for the ejecta component were fixed at the best - fit values estimated from the head region . the metal abundances for the soft component were fixed at the values estimated from the shell region . the electron temperature , ionization timescale , and the normalization parameters for the soft component were varied freely . the observed tail spectrum can be equally fitted by this two - component model ( @xmath34/n = 52.4/49 ) . in this model fit , the metal - rich ejecta component contributes @xmath915% of the total observed x - ray flux in the 0.510 kev band . these one- and two - component models are not statistically distinguished with the current data . results from these tail region spectral model fits are summarized in tables [ tbl : tab2 ] & [ tbl : tab3 ] . the east region shows a distinctively blue color ( figure [ fig : fig1]a ) with a strong enhancement in the si line ew image ( figure [ fig : fig2]a ) , which resembles the spectral properties of the head region of the bullet . this region corresponds to a hot inter - cloud region @xcite . this region may also be a candidate ejecta feature , probably isolated from the complex emission from shocked dense clumpy clouds ( figure [ fig : fig2]b ) . since this region was covered by both the 2000/2001 and 2009 data , we used all of these data to extract the x - ray spectrum . we extracted the x - ray spectrum from a small circular region with a radius of 2@xmath28 , in which the _ chandra _ data allowed us total photon statistics of @xmath93600 counts . the background was subtracted using the spectrum extracted from two circular source - free regions ( with a radius of 8@xmath28 ) beyond the eastern boundary of the snr . we fit the observed spectrum with the nei plane shock model . because enhanced si - k and s - k lines are evident , we varied si and s abundances while fixing other abundances at average values estimated from the shell and sgr_bg regions . the fitted parameters are @xmath33 @xmath9 0.8 @xmath3 10@xmath38 @xmath32 , @xmath35 @xmath9 1.05 kev , and @xmath36 @xmath44 52.4 @xmath3 10@xmath37 @xmath16 s. the estimated si and s abundances are high ( si @xmath9 1.8 and s @xmath9 1.4 ) . while the fit may be acceptable ( @xmath34/n = 173.4/134 ) , there appear to be some systematic residuals at @xmath39 @xmath29 1 kev . based on the results from our spectral analysis of the ejecta ( the head region ) and the swept - up ism ( the shell and sgr_bg regions ) regions , we consider that the east region emission is a superposition of the spectrally - hard metal - rich ejecta and the projected soft swept - up ism components . to account for the contribution of soft x - ray emission from the shocked ism in the observed spectrum of the east region , we added an nei model component for which we fixed the electron temperature , ionization timescale , and metal abundances at the average values estimated from the shell and sgr_bg regions . only the normalization parameter was varied for this component . this model fit significantly improves ( @xmath34/n = 144.4/133 ) without affecting the best - fit parameters for the hard , ejecta component . the only noticeable change is a larger value for the best - fit @xmath33 @xmath9 3 @xmath3 10@xmath38 @xmath32 . this larger estimate for @xmath48 appears to be consistent with the presence of dense molecular clouds toward eastern parts of the snr . the assumed soft component from the swept - up ism contributes @xmath920% of the total observed x - ray flux in the 0.55 kev band . the results from this fit are summarized in table [ tbl : tab2 ] & [ tbl : tab3 ] . the x - ray spectrum of sgr 052666 is extracted from a circular region of 2@xmath28 radius ( figure [ fig : fig1]b ) . the background spectrum is characterized by emission from an annular region ( 2@xmath28 in thickness ) around the source ( the sgr_bg region , figure [ fig : fig1]b ) to average the complex diffuse emission from snr n49 projected against the sgr . since archival _ chandra _ data ( obsids 747 and 1957 ) were pointed at the sgr with a decent exposure ( table [ tbl : tab1 ] ) , we used these archival data as well as our new data for the spectral analysis of sgr 052666 . we fit these three spectra simultaneously . just like our spectral analysis of n49 , we assumed that the sgr spectrum is absorbed by both galactic ( fixed at @xmath30 = 6 @xmath3 10@xmath31 @xmath32 ) and lmc columns ( @xmath33 ) . initially , we fit the sgr spectrum with a single power law ( pl ) in which we tied all fitted parameters between the three sets of the spectrum . the best - fit model from this fit is not acceptable ( @xmath49@xmath50/@xmath14 @xmath9 1.8 ) . the best - fit model shows significant residuals at @xmath39 @xmath46 0.8 kev , because it underestimates the overall flux level for the 2000/2001 data while it overestimates the flux level for the 2009 data . thus , we untied the normalization parameter between the 2000/2001 and 2009 data , and then repeated the model fit . the fit improved ( @xmath49@xmath50/@xmath14 @xmath9 1.5 ) with a @xmath915% lower normalization for the 2009 data than that for the 2000/2001 data . although statistically improved by removing the normalization offset between the 2000/2001 and 2009 data , we do not accept this fit because systematic residuals are evident over the entire bandpass with a relatively high @xmath49@xmath50/@xmath14 @xmath9 1.5 . alternatively , we attempted a single blackbody ( bb ) model fit . the best - fit models are statistically unacceptable ( @xmath49@xmath50/@xmath14 @xmath9 2.2 and 2.5 with the untied and tied normalization parameter between the 2000/2001 and 2009 data , respectively ) . kulkarni et al . ( 2003 ) showed that a pl+bb model was needed to adequately describe the quiescent x - ray emission from sgr 052666 . on the other hand , park et al . ( 2003a ) argued that there could have been a considerable contamination from the surrounding soft thermal x - ray emission in the sgr spectrum presented by kulkarni et al . ( 2003 ) , and that the need for a bb component was not conclusive . our sgr spectrum with @xmath94 times higher photon statistics than that used by park et al . ( 2003a ) indicates that a pl model can not describe the observed sgr spectrum , supporting the conclusion by kulkarni et al . thus , we fit the sgr spectrum with a pl+bb model . we repeated the general steps as we did with a single pl model fits : i.e. , we first tied all fitted parameters between the 2000/2001 and 2009 data sets , and then untied them one at a time ( i.e. , bb temperature [ @xmath51 , pl photon index [ @xmath52 , and normalizations for the bb and pl components ) . the best - fit values for @xmath53 and @xmath54 are consistent within uncertainties between the 2000/2001 and 2009 data . the normalization for the bb and/or pl components are fitted to be @xmath91030% lower for the 2009 data than they are for the 2000/2001 data . the fit improvement due to this normalization change is statistically significant , and the fit is acceptable ( @xmath49@xmath50/@xmath14 = 577.4/582 ) . the normalization change appears to be more effective for the pl component than for the bb , probably because the pl component covers a broader range of the observed spectrum than the bb component ( the pl component contributes @xmath970% of the total observed flux in the 0.510 kev band ) . while the x - ray spectra of sgrs and anomalous x - ray pulsars ( axps ) are usually fitted with a bb+pl model ( e.g. , * ? ? ? * ) , double bb models ( bb+bb ) have also been suggested to fit some axps ( e.g. , * ? ? ? a recent work showed that a bb+bb model fit can successfully reproduce the observed _ chandra _ spectrum ( taken in 2000 and 2001 ) of sgr 052666 @xcite . thus , we attempted to fit our sgr spectrum with a bb+bb model . we tied bb temperatures between 2000/2001 and 2009 data since we found no evidence of change in the temperatures . just like our bb+pl modeling , we varied normalizations freely between 2000/2001 and 2009 . the fit is statistically good ( @xmath49@xmath50/@xmath14 = 616.4/582 ) . although statistical uncertainties are relatively large , a @xmath915% decrease in the best - fit value of the normalization parameters are suggested for both bb components , which is similar to the results from the bb+pl model fit . both of the bb+pl and bb+bb model fits suggest that the overall x - ray flux of sgr 052666 in the 0.510 kev band has been reduced by @xmath915% in 2009 compared with that in 2000/2001 . we note that we used a 1/4 subarray of the acis - s3 in our 2009 observations , while a 1/8 subarray was used in 2000 and 2001 . based on the event grade distribution , we found that the photon pileup effect is similarly small in both data sets ( @xmath92% in 2000 and 2001 vs. @xmath94% in 2009 ) . nonetheless , assuming the x - ray flux observed in 2000/2001 , we estimated the pileup effect on the x - ray flux measurements in 2009 data using the portable interactive multi - mission simulator ( pimms ) . our pimms simulations predict that the pileup effect in 2009 data would reduce the 0.510 kev band x - ray flux in 2009 by @xmath295% compared with that measured in 2000/2001 . we also tested the pileup effect by applying an acis pileup model @xcite for our spectral model fits . this model indicated that the 0.510 kev flux estimates with the 2009 data are probably affected by @xmath295% due to the use of 1/4-subarray . thus , although the flux change between 2000/2001 and 2009 is partially caused by pileup effect , the pileup is not responsible for the entire flux change . the apparent flux change is unlikely due to uncertainties in the detector calibration for the time - dependent quantum efficiency degradation , because there is no evidence for discrepancy in the fitted normalization parameters that is emphasized in the soft band ( @xmath39 @xmath29 1 kev ) between 2000/2001 and 2009 data . thus , while follow - up _ chandra _ observations are required to draw a firm conclusion on the nature of the long - term x - ray light curve of 052666 , we tentatively conclude that the discrepancy in the observed x - ray flux of sgr 052666 between 2000/2001 and 2009 is probably real rather than an artifact due to the photon pile - up and/or an inaccurate calibration of the time - dependent quantum efficiency degradation between two epochs . the sgr spectrum and the best - fit bb+pl model are presented in figure [ fig : fig4]d . results from the best - fit bb+pl and bb+bb model fits are summarized in table [ tbl : tab4 ] and [ tbl : tab5 ] , respectively . based on the volume emission measure ( @xmath55 ) estimated from the spectral fit of the shell region , we calculate the post - shock electron density ( @xmath56 ) behind the blast wave . we assumed an emission volume of @xmath57 @xmath9 9.4 @xmath3 10@xmath58 cm@xmath15 for a 3@xmath28 @xmath3 18@xmath28 region ( corresponding to @xmath90.7 pc @xmath3 4.4 pc , hereafter we assume @xmath60 = 50 kpc for the lmc ) with a @xmath91 pc path - length along the line of sight . we also assumed @xmath56 = 1.2 @xmath61 for a mean charge state with normal composition . the best - fit @xmath55 (= 6.75 @xmath3 10@xmath62 @xmath16 ) implies @xmath56 @xmath9 9.3@xmath63 @xmath16 and @xmath61 @xmath9 7.7@xmath63 @xmath16 , where @xmath64 is the volume filling factor of the x - ray emitting gas . the pre - shock h density is then @xmath65 @xmath9 1.9@xmath63 @xmath16 for a strong adiabatic shock where @xmath61 = 4@xmath65 . assuming an adiabatic shock in electron - ion temperature equipartition , the gas temperature is related with the shock velocity ( @xmath8 ) as @xmath12 = 3@xmath66@xmath67/16@xmath68 where @xmath68 is the boltzmann constant and @xmath66 @xmath9 0.6 @xmath69 is the mean molecular weight with the proton mass @xmath69 = 1.67 @xmath3 10@xmath70 g. for the gas temperature of @xmath35 = 0.57 kev , the shock velocity @xmath8 @xmath9 700 km s@xmath11 is estimated . thus , for the radius of @xmath98.5 pc ( assuming the angular distance of @xmath935@xmath28 between the shell region and the geometric center of n49 ) , we estimate the sedov age of the snr , @xmath71 @xmath9 4800 yr . the estimated @xmath71 and @xmath65 imply an explosion energy of @xmath72 @xmath9 1.8@xmath63 @xmath3 10@xmath4 erg . the estimated sedov age is @xmath930% lower than our previous estimate ( @xmath96600 yr , * ? ? ? * ) for which we assumed @xmath65 = 0.9 @xmath16 based on optical data @xcite and the canonical value of the sn explosion energy ( @xmath72 = 1 @xmath3 10@xmath4 erg ) . also , the previous estimate was based on a more complex spectral modeling of multi - temperature plasma in the eastern regions of the snr for which an accurate estimate of the distance from the snr center was not feasible . thus , we conclude that our new estimate of the snr age is more reliable , while being in plausible agreement with the previous estimate by park et al . ( 2003a ) . the presence of the ejecta bullet beyond the southwestern shell of n49 is conclusively established by our new _ chandra _ observation . the on - axis acis image clearly resolves the bullet into a head and a tail extending back to the main shell of the snr . the head region itself is an extended feature ( @xmath94@xmath28 in radius ) with an enhanced intensity toward the inferred direction of motion ( figure [ fig : fig1]c ) . the x - ray color of the head is distinctively blue in contrast to reddish color of the overall snr shell . the foreground column ( @xmath33 @xmath9 1.3 @xmath3 10@xmath38 @xmath32 ) toward the head is consistent with those to n49 s main shell and the bullet s tail region , which supports that the blue color is intrinsic for the head rather than being caused by a significantly larger foreground absorption ( in which case the head would probably be a background source ) . in fact , the observed x - ray spectrum ( figure [ fig : fig4]a ) shows that the hardness of the head is due to highly enhanced line emission from he- and h - like si and he - like s ions . the estimated abundances for si ( @xmath92.3 ) and s ( @xmath93.2 ) are an order of magnitude larger than the lmc values . we note that park et al . ( 2003a ) suggested a possibility of the enhanced abundance for o in addition to si and s in the bullet . however , the overabundance for o is not confirmed by our new data . nonetheless , the highly overabundant si and s firmly establish the identification of the bullet as metal - rich stellar fragment ejected from the deep interior of the progenitor star . the east region shows nearly identical spectral characteristics to the head region : i.e. , a distinctively blue color compared with surrounding regions and strongly enhanced si and s lines . just like the head region , significantly enhanced si and s abundances are estimated , but overabundances for other metal species are not required to describe the observed spectrum . the foreground absorption for the east region appears to be larger than other regions of n49 , and it is probably caused by nearby dense molecular clouds which are interacting with the snr in the eastern regions ( e.g. , * ? ? ? * ; * ? ? ? if the _ excess _ column ( @xmath73@xmath48 @xmath9 2 @xmath3 10@xmath38 @xmath32 compared with the average @xmath33 @xmath9 1.2 @xmath3 10@xmath38 @xmath32 estimated for other regions of n49 , table [ tbl : tab2 ] ) originates from these nearby molecular clouds , the corresponding average cloud density of @xmath61 @xmath9 90 @xmath16 is implied for the overall cloud size of @xmath74 @xmath97 pc ( the size estimated by banas et al . this average density is in plausible agreement with the pre - shock density range of the clumpy clouds ( @xmath65 @xmath9 20940 @xmath16 , vancura et al . 1992 ) with which n49 is interacting , while it is significantly larger than the value we estimated for the shell region ( @xmath65 @xmath9 1.9 @xmath16 ) where the shock is propagating into the low density ambient medium . based on the si and s abundances and volume emission measures of these metal - rich ejecta features , we estimate the si to s ejecta mass ratio . the observed spectrum of the head and east regions shows that he- and h - like ionization states dominate for si , while s ions are primarily in he - like state . thus , for simplicity , we assumed a `` pure '' ejecta case with electron to ion density ratios of @xmath75 @xmath76 12.5 @xmath77 and @xmath78 @xmath76 14 @xmath79 . for dominant isotopes of @xmath80si and @xmath81s , the measured si and s abundances imply the ejecta mass ratio @xmath82/@xmath83 @xmath9 1.2 and 1.6 ( @xmath84/@xmath85)@xmath86 , where @xmath84 and @xmath85 are the volume of si- and s - rich hot ejecta gas , respectively . assuming @xmath84 @xmath76 @xmath85 , the estimated mass ratios are @xmath82/@xmath83 @xmath91.2 for the head region and @xmath91.6 for the east region . the mixture of h in metal - rich ejecta may not significantly affect our mass ratio estimates as long as the ism mixture rate is similar between the si and s ejecta . we compare these @xmath82/@xmath83 with standard sn nucleosynthesis models . our estimated @xmath82/@xmath83 appears to be smaller than those for core - collapse sn models in which @xmath82/@xmath83 typically ranges @xmath924 depending on the progenitor s mass ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? the si to s mass ratio for type ia sn models ( @xmath82/@xmath83 @xmath9 1.5 1.8 , e.g. , nomoto et al . [ 1997b ] ; iwamoto et al . [ 1999 ] ) are generally smaller than core - collapse cases , which are closer to our estimates for n49 . the lack of evidence for o - rich ejecta in the head and east regions is generally suggestive of a type ia origin as well . if we take @xmath90.3 ( table [ tbl : tab3 ] ) as an _ upper limit _ for the o abundance of the sn nucleosynthesis in n49 , @xmath87/@xmath82 @xmath88 1.4 can be inferred for the head region . this limit appears to be consistent with type ia models ( @xmath87/@xmath89 @xmath29 1 , e.g. , nomoto et al . [ 1997b ] ; iwamoto et al . [ 1999 ] ) , while being smaller than those for core - collapse models ( @xmath87/@xmath82 @xmath44 2 , e.g. , nomoto et al . [ 1997a ] ; limongi & chieffi [ 2003 ] ) . on the other hand , the head and east regions do not show evidence for overabundant fe , which is usually considered as an iconic feature for type ia snrs . the lack of fe - rich ejecta is problematic for a type ia interpretation , and it would rather support a core - collapse origin . also , the suggested type ia origin for n49 is inconsistent with its interstellar environment with recent star - forming regions and nearby molecular clouds , which rather suggests a massive progenitor for n49 ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? if n49 is a remnant of a core - collapse explosion from a massive progenitor rather than a type ia , the si - rich nature of the head and east regions may be generally considered to be explosive o - burning or incomplete si - burning products from deep inside of the core - collapse sn . in fact , si - rich ejecta knots have been detected in core - collapse snrs : e.g. , shrapnel a in vela snr @xcite and those found in cassiopeia a @xcite . while the possibility of type ia origin for n49 is intriguing , we caution that the utility of our @xmath82/@xmath83 estimate to identify the sn type is limited , because it is based on small localized ejecta features whereas sn nucleosynthesis model calculations are for the integrated ejecta material from the entire sn . thus , based on the current data , our discussion on the origin of n49 is far from conclusive . a more extensive ejecta search and comprehensive nucleosynthesis study are required to reveal the true origin of n49 . we point out that the correct identification of n49 s origin is particularly important because of its astrophysical implications in the context of the snr s environment . for instance , if n49 is the remnant of a type ia sn , the long - standing argument for its physical association with sgr 052666 is unambiguously ruled out . a type ia origin for n49 may also suggest an intriguing case for a prompt population ia sn from a relatively young white dwarf progenitor in analogy to the scenario suggested for snr 010472.3 in the small magellanic cloud @xcite . observation firmly establishes the previously suggested two - component x - ray spectrum for sgr 052666 . the pl index ( @xmath54 @xmath9 2.5 ) of its x - ray spectrum is intermediate between those for other sgrs ( @xmath54 @xmath9 12 ) and axps ( @xmath54 @xmath9 34 ) , which suggests that sgr 052666 may be a transition object between the sgr and axp states @xcite . in fact , our estimated pl index ( @xmath54 @xmath9 2.5 ) is identical to that of axp 1e 1048.15937 , which is considered to be a clear case of a sgr - axp transition object @xcite . the implied size of the bb radiation ( @xmath90 @xmath9 56 km ) is smaller than the canonical size of neutron stars . the small @xmath90 suggests the existence of restricted hot spots on the surface of the neutron star . alternatively , a bb+bb model can equally describe the observed spectrum of 052666 the two - component bb model was preferred for some axps , because the second bb component is physically more reasonable than the steep pl component to explain the low flux limit at longer wavelengths ( e.g. , * ? ? ? recently , nakagawa et al . ( 2009 ) showed that the x - ray spectrum of the quiescent emission from 052666 can be fitted by a two - component bb model . our result is consistent with that by nakagawa et al . the hard bb component indicates a small emission area ( @xmath74 @xmath9 1 km ) with a hot temperature of @xmath35 @xmath9 1 kev . the soft bb component indicates an area corresponding to the entire surface of the neutron star ( @xmath74 @xmath9 10 km ) with a lower temperature of @xmath35 @xmath9 0.4 kev . these overall characteristics are reminiscent of the peculiar types of neutron stars found at the center of several young snrs ( e.g. , * ? ? ? * ; * ? ? ? * ) , but the estimated bb temperatures of 052666 are significantly higher than those estimated in others . the overall x - ray flux in the 0.510 kev band in 2009 is @xmath915% lower than it was in 20002001 . we show a long - term x - ray light curve of sgr 052666 in figure [ fig : fig5 ] . in figure [ fig : fig5 ] , we plot the mean flux of 2000 and 2001 data ( the middle data point ) because the observed flux is indistinguishable between the two epochs . we added the fluxes estimated by the _ rosat _ hri data in this light curve . for the _ rosat _ fluxes , we converted the observed hri count rate ( 1.51@xmath910.13 @xmath3 10@xmath92 counts s@xmath11 , * ? ? ? * ) into the 0.510 kev band energy flux using pimms . in this calculation , we assumed two cases of bb+pl and bb+bb models with the best - fit parameters listed in tables [ tbl : tab4 ] and [ tbl : tab5 ] , respectively . we also assumed the fractional contributions in the observed 0.12.4 kev band hri count rate from each of the model components , based on the results summarized in table [ tbl : tab4 ] and [ tbl : tab5 ] . the calculated _ rosat _ hri fluxes are @xmath93 @xmath9 1.34 ( bb+pl ) and @xmath91.17 ( bb+bb ) @xmath3 10@xmath94 erg @xmath32 s@xmath11 . while the x - ray flux appears to have decreased by @xmath930% since 1992 based on the bb+pl case , the overall flux decrease is less certain for the bb+bb case . if it is real , the suggested x - ray flux change for 052666 would not be surprising , because long - term variabilities by a factor of up to @xmath910 in several years have been detected in some axps ( e.g. , * ? ? ? * ; * ? ? ? follow - up _ chandra _ observations are essential to reveal the true nature of the long - term light curve of sgr 052666 . we note that the foreground column to 052666 shows a significant discrepancy by a factor of @xmath93 between values estimated by two different spectral modelings ( bb+pl vs. bb+bb ) . the column estimated by the bb+bb model fit ( @xmath40 @xmath9 1.7 @xmath3 10@xmath38 @xmath32 ) is generally in agreement with those measured for n49 ( table [ tbl : tab2 ] ) , and particularly , it is fully consistent with the column toward the sgr_bg region ( @xmath40 @xmath9 1.6 @xmath3 10@xmath38 @xmath32 ) . these consistent columns between n49 and 052666 are supportive of the long - suggested physical association between them . on the other hand , our bb+pl model fit of 052666 spectrum shows a substantially larger foreground column for sgr 052666 ( @xmath48 @xmath9 5.4 @xmath3 10@xmath38 @xmath32 ) than that for sgr_bg region . if this is the case , the large difference in the foreground absorption between snr n49 and sgr 052666 brings into question their physical association . if we assumed an average interstellar density of @xmath65 @xmath9 12 @xmath16 ( see section [ subsec : disc_ejecta ] ) near n49 , sgr 052666 may be @xmath95001000 pc beyond n49 . thus , this model - dependent difference in @xmath40 for 052666 deserves full attention for further studies with follow - up observations . an h i survey of the lmc shows that n49 and 052666 are projected toward the dense boundary between two supergiant shells @xcite . the estimated h i column density toward this region appears to be @xmath40 @xmath9 5 @xmath3 10@xmath38 @xmath32 . it is difficult to discriminate our model - dependent @xmath40 measurements toward 052666 based on this relatively large column estimated by h i data with a poor angular resolution ( @xmath91@xmath26 which is comparable with the angular size of n49 ) . we note that , if 052666 and n49 are associated , the projected angular separation ( @xmath922@xmath28 ) of 052666 from the geometric center of n49 requires a large kick - velocity ( i.e. , an average transverse velocity of @xmath8 @xmath9 1100 km s@xmath11 for the snr age of @xmath94800 yr ) . at least , the proper measurement of the foreground column toward 052666 is directly related with two important astrophysical issues : ( 1 ) the origin of the quiescent x - ray emission from sgr 052666 , and ( 2 ) the physical relationship between snr n49 and sgr 052666 . it may also provide a useful piece of puzzle to reveal the origin of snr n49 ( thermonuclear vs. core - collapse ) . using our deep _ chandra _ observation , we detect metal - rich ejecta features in snr n49 . these ejecta features include a `` bullet '' that is most likely a stellar fragment travelling beyond the southwestern boundary of the snr . we also find an ejecta feature in the eastern part of the snr , nearly in the opposite side of the snr to the bullet . both of these ejecta features show highly enhanced si and s abundances by an order of magnitude above the lmc values . we do not find compelling evidence for overabundant o and/or fe in these ejecta features . if n49 is a remnant of a core - collapse explosion of a massive progenitor , the si - rich nature of ejecta may be considered to be nucleosynthesis products of explosive o - burning or incomplete si - burning from deep inside of the sn . on the other hand , the estimated si- and s - rich ejecta mass ratio appears to favor a type ia origin for n49 . if n49 was the remnant of a type ia sn , the suggested physical association between n49 and sgr 052666 would be unambiguously ruled out . however , we note that our sn ejecta study is limited because we used only some small localized ejecta features rather than the integrated ejecta composition from the entire snr to represent the true sn nucleosynthesis . follow - up studies to reveal the comprehensive ejecta composition in n49 are required to unveil the true origin of this snr . our new _ chandra _ observation allows us to detect the blast wave forming the swept - up shell in the southern boundary with significant photon statistics . since this part of the shell is not affected by spectral complications due to the snr s interaction with dense clumpy clouds , it provides a useful opportunity to reveal the dynamics of the snr more accurately than previous works . we estimate the sedov age of @xmath71 @xmath9 4800 yr and the explosion energy @xmath72 @xmath9 1.8 @xmath3 10@xmath4 erg for n49 . our spectral analysis of the quiescent x - ray emission from sgr 052666 using the deep exposure clearly reveals the presence of a bb emission ( @xmath53 @xmath9 0.44 kev ) in addition to a pl component . the implied bb emitting area is relatively small ( @xmath74 @xmath9 56 km ) compared to the canonical size of neutron stars . the estimated pl photon index ( @xmath54 @xmath9 2.5 ) is identical to that of axp 1e 1048.15937 , the well - known candidate transition object between axps and sgrs . alternatively , the observed x - ray spectrum of 052666 can be equally fitted by a two - component bb model . this model indicates that x - ray emission originates from a small ( @xmath74 @xmath9 1 km ) , hot ( @xmath35 @xmath9 1 kev ) spot(s ) in addition to a cooler ( @xmath35 @xmath9 0.4 kev ) surface of the neutron star ( @xmath74 @xmath9 10 km ) . we find marginal evidence for a slow decay in the observed x - ray flux of 052666 ( @xmath92030% for the last @xmath917 yr ) . continuous x - ray monitoring of 052666 is needed to clarify the nature of its long - term light curve . we find a considerable difference in the foreground column by a factor of @xmath93 between two modelings ( bb+pl vs. bb+bb ) of the x - ray spectrum of 052666 . while such a model - dependent discrepancy has been noticed for several other axps , the 052666 case is particularly intriguing because discriminating these competing models may be able to provide some critical clues on the nature of 052666 and n49 : e.g. , their physical association , the origin of x - ray emission of 052666 , and the origin of n49 . this work has been supported in part by the sao under the _ chandra _ grants go9 - 0072a to the university of texas at arlington , and go9 - 0072x to the pennsylvania state university . j.p.h . has been supported in part by the _ grant go0 - 11090x . cccc 10123 & 2009 - 7 - 18 & 26.8 & acis - s3 ( 1/4 subarray ) + 10806 & 2009 - 9 - 19 & 26.5 & acis - s3 ( 1/4 subarray ) + 10807 & 2009 - 9 - 16 & 26.0 & acis - s3 ( 1/4 subarray ) + 10808 & 2009 - 7 - 31 & 28.7 & acis - s3 ( 1/4 subarray ) + 747 & 2000 - 1 - 4 & 39.9 & acis - s3 ( 1/8 subarray ) + 1957 & 2001 - 8 - 31 & 48.4 & acis - s3 ( 1/8 subarray ) + cccccc shell & 0.89@xmath95 & 0.57@xmath96 & 6.35@xmath97 & 6.75@xmath98 & 49.1/68 + sgr_bg & 1.58@xmath43 & 0.56@xmath910.03 & 9.65@xmath99 & 15.45@xmath100 & 258.9/218 + tail & 1.01@xmath101 & 2.02@xmath102 & 0.68@xmath103 & 0.50@xmath104 & 44.9/44 + tail & 1.68@xmath105 & 0.71@xmath106 & 4.02@xmath107 & 1.61@xmath108 & 52.4/49 + head & 1.32@xmath109 & 1.04@xmath110 & @xmath4420.3 & 2.25@xmath111 & 67.0/59 + east & 3.11@xmath112 & 1.09@xmath113 & @xmath4467.0 & 2.68@xmath114 & 144.4/133 + ccccccc shell & 0.32@xmath115 & 0.27@xmath116 & 0.25@xmath910.10 & 0.36@xmath117 & 0.84@xmath118 & 0.21@xmath119 + sgr_bg & 0.24@xmath120 & 0.17@xmath121 & 0.15@xmath122 & 0.24@xmath123 & 0.59@xmath124 & 0.13@xmath125 + tail & 0.28@xmath126 & @xmath881.54 ( 0.58 ) & 0.37@xmath127 & 0.88@xmath128 & @xmath882.30 ( 0.49 ) & 0.28@xmath129 + head & 0.32 & 0.27 & 0.25 & 2.28@xmath130 & 3.22@xmath131 & 0.21 + east & 0.27 & 0.21 & 0.20 & 1.92@xmath132 & 1.39@xmath133 & 0.17 + lccc @xmath35 ( kev ) & 0.44@xmath910.02 & - & - + @xmath54 & - & 2.50@xmath134 & - + @xmath135 ( km ) & 6.0@xmath136 & - & - + @xmath137 ( km ) & 5.5@xmath910.6 & - & - + @xmath138 ( 10@xmath94 erg @xmath32 s@xmath11 ) & 0.40@xmath139 & 0.78@xmath910.13 & 1.18@xmath140 + @xmath141 ( 10@xmath94 erg @xmath32 s@xmath11 ) & 0.33@xmath142 & 0.69@xmath910.11 & 1.01@xmath143 + @xmath144 ( 10@xmath145 erg s@xmath11 ) & 1.68@xmath146 & 3.84@xmath910.65 & 5.52@xmath147 + @xmath148 ( 10@xmath145 erg s@xmath11 ) & 1.37@xmath149 & 3.36@xmath910.54 & 4.74@xmath150 + @xmath33 ( 10@xmath38 @xmath32 ) & - & - & 5.44@xmath151 + @xmath30 ( 10@xmath38 @xmath32 ) & - & - & 0.6 ( fixed ) + @xmath34/n & - & - & 577.4/582 + lccc @xmath35 ( kev ) & 0.39@xmath910.01 & 1.01@xmath120 & - + @xmath135 ( km ) & 9.7@xmath152 & 1.0@xmath910.2 & - + @xmath137 ( km ) & 9.0@xmath910.5 & 0.9@xmath910.2 & - + @xmath138 ( 10@xmath94 erg @xmath32 s@xmath11 ) & 0.73@xmath153 & 0.40@xmath154 & 1.13@xmath155 + @xmath141 ( 10@xmath94 erg @xmath32 s@xmath11 ) & 0.62@xmath156 & 0.35@xmath157 & 0.96@xmath155 + @xmath144 ( 10@xmath145 erg s@xmath11 ) & 0.90@xmath120 & 0.42@xmath158 & 3.96@xmath159 + @xmath148 ( 10@xmath145 erg s@xmath11 ) & 0.77@xmath160 & 0.36@xmath161 & 3.39@xmath162 + @xmath33 ( 10@xmath38 @xmath32 ) & - & - & 1.70@xmath163 + @xmath30 ( 10@xmath38 @xmath32 ) & - & - & 0.6 ( fixed ) + @xmath34/n & - & - & 616.4/582 +

direct submillimeter imaging has recently revealed the 850 m background to be mostly composed of a population of distant ultraluminous infrared galaxies , but identifying the optical / near - infrared ( nir ) counterparts to these sources has proved difficult due to the poor submillimeter spatial resolution . however , the proportionality of both centimeter and submillimeter data to the star formation rate suggests that high resolution radio continuum maps with subarcsecond positional accuracy can be exploited to locate submillimeter sources . in this paper we present results from a targeted scuba survey of micro - jansky radio sources in the flanking fields of the hubble deep field . the sources were selected from the uniform ( 8 jy at 1 ) 1.4 ghz vla image of richards ( 1999b ) . even with relatively shallow scuba observations ( a detection limit of 6 mjy at 850 m ) , we were successful at making submillimeter detections of optical / nir - faint ( and ) radio sources , and our counts closely match the bright counts from submillimeter surveys . an important corollary is that a large fraction of the bright ( mjy ) submillimeter sources in untargeted submillimeter surveys have extremely faint optical / nir counterparts and hence are inaccessible to optical imaging and spectroscopy . however , redshift estimates can be made from the ratio of the submillimeter flux to the radio flux across the 100 ghz break in the spectral energy distribution . this procedure , which we refer to as _ millimetric _ redshift estimation , places the bright submillimeter population at where it forms the high redshift tail of the faint radio population . the star formation rate density ( sfrd ) due to ultraluminous infrared galaxies increases by more than two orders of magnitude from to . the sfrd at high redshift inferred from our mjy submillimeter observations is comparable to that observed in the ultraviolet / optical .

astro-ph0001096

recent detections of distant dusty galaxies with the scuba camera ( the submillimeter common user bolometer array ; holland et al . 1999 ) on the 15 m james clerk maxwell telescope constitute a substantial fraction of the cosmic fir background detected by the _ firas _ and _ dirbe _ experiments on the _ cobe _ satellite ( puget et al . 1996 ; guiderdoni et al . 1997 ; schlegel , finkbeiner , & davis 1998 ; fixsen et al . 1998 ; hauser et al . 1998 ; lagache et al . 1999 ) . since the observed fir background is comparable to the total unobscured emission at ultraviolet / optical wavelengths , a full determination of the global star formation history of the universe requires a comprehensive understanding of this dust - enshrouded galaxy population . the 850 @xmath0 m scuba surveys to date have reported galaxy number counts that are in general agreement ( smail , ivison & blain 1997 ; barger et al . 1998 ; hughes et al . 1998 ; blain et al . 1999a ; eales et al . 1999 ; barger , cowie , & sanders 1999 ) ; the cumulative surface density above 2 mjy is about @xmath9 . the discrete sources have bolometric luminosities that are characteristically @xmath10 if they lie at @xmath11 . moreover , the sources for which measurements exist at multiple wavelengths ( e.g. ivison et al . 1998 ) show thermal spectral energy distributions ( seds ) . thus , the scuba sources are inferred to be the distant analogs of the local ultraluminous infrared galaxy ( ulig ; sanders & mirabel 1996 ) population . an essential observational goal is to determine the redshift distribution of the submillimeter population in order to trace the extent and evolution of obscured emission in the distant universe . however , identifying the optical / nir counterparts to the submillimeter sources is difficult due to the uncertainty in the scuba positions . barger et al . ( 1999b ) presented a spectroscopic survey of possible optical counterparts to a flux - limited sample of galaxies selected from the 850 @xmath0 m survey of massive lensing clusters by smail et al . candidate optical counterparts in the scuba error - boxes were identified using moderately deep ground - based and _ hst _ exposures ( @xmath12 and @xmath13 , respectively ) . one - quarter of the sources could be reliably identified , and those had redshifts in the range @xmath8 . a lower limit of 20 per cent of the full sample showed signs of agn activity . however , for the majority of the submillimeter sources there were either no optical counterparts or the optical associations were not secure . such sources could either be at very high redshift or be so highly obscured that they emit their energy almost entirely in the submillimeter . high resolution radio continuum maps with subarcsecond positional accuracy and resolution offer new opportunities for locating submillimeter sources and determining their physical properties . the unique advantage of centimeter and fir observations is that galaxies and the intergalactic medium are transparent at these wavelengths , so observed flux densities are proportional to intrinsic luminosities . in galaxies without a powerful agn , the radio luminosity is dominated by diffuse synchrotron emission from relativistic electrons accelerated in supernovae remnants from stars more massive than @xmath14 . these massive stars live @xmath15 yr ; the relativistic electrons probably live @xmath16 yr ( condon 1992 ) . thus , radio observations probe very recent star formation . fir observations of starburst galaxies are also a direct measure of massive star formation . as summarized by condon ( 1992 ) , radio continuum emission and thermal dust emission are empirically observed to be tightly correlated due to both being linearly related to the massive star formation rate . if the fir - radio correlation applies to high redshift , as is plausibly the case ( though at the very highest redshifts compton cooling of the relativistic electrons by the microwave background may suppress the radio emission ) , then very sensitive radio observations can be used to pinpoint distant submillimeter sources . in this paper we investigate the feasibility of using radio data to identify and characterize the bright submillimeter source population . richards ( 1999b ) recently obtained an extremely deep very large array ( vla ) 1.4 ghz image centered on the hubble deep field ( hdf ) . richards et al . ( 1999 ) matched ground - based optical data from barger et al . ( 1999a ) to the radio image and found that @xmath17 per cent of the galaxies in the sample could not be identified to optical magnitude limits of @xmath18 . in other respects , such as radio size and spectral index , the optically - faint objects were not any different from the remaining population . richards et al . ( 1999 ) proposed four possible scenarios to explain this population , including @xmath19 obscured starbursts ( beyond @xmath20 the sensitivity to star forming galaxies cuts off due to the flux density limits of the radio data ) , extreme redshift ( @xmath21 ) agn , @xmath22 obscured agn , or one - sided radio jets . in the first phase of our program we observed with lris on the keck ii 10 m telescope a complete subsample of the radio sources . our primary objective was to determine the redshifts of the optical / nir - faint ( @xmath23 ) radio sources . although we were able to spectroscopically identify nearly all the objects in our subsample to @xmath24 ( all had @xmath25 ) , we were unable to obtain redshifts for the fainter objects . either these sources are distant ( @xmath26 ) with spectral features lying outside the optical wavelength range , or the visibility of remarkable features is strongly affected by dust in the galaxies . if the optical / nir - faint radio sources are highly dust obscured systems , then it is possible that they will be detectable in the submillimeter . in the second phase of our program we observed with scuba 15 of the 22 optical / nir - faint radio sources in the central @xmath27 square arcminute region of the radio map ; another 4 were observed by hughes et al . ( 1998 ; hereafter h98 ) in the hdf - proper . the jiggle map mode enabled simultaneous observations of a large fraction ( 31/48 ) of the optical / nir - bright radio sources . even with relatively shallow scuba observations ( a @xmath2 detection limit of 6 mjy at 850 @xmath0 m ) , we detected 5 of the optical / nir - faint radio sources ; a sixth with submillimeter flux @xmath28 mjy was detected in the deep hdf - proper submillimeter map of h98 . in contrast , none of the optical / nir - bright radio sources were detected . we additionally detected two @xmath5 mjy sources that did not have radio counterparts . thus , our targeted scuba survey of optical / nir - faint radio sources turned up @xmath29 per cent of the bright submillimeter sources in our surveyed areas . in the final phase of our program , we explored the feasibility of obtaining redshift estimates from the submillimeter - to - radio flux ratios , as recently suggested by carilli & yun ( 1999 ) . we find that the redshifted arp 220 sed reasonably describes both redshifted local ulig data and known high redshift submillimeter sources and hence can be used as a rough redshift estimator . we estimate that all of our bright submillimeter sources fall in the redshift range @xmath6 , consistent with the redshifts for the lensed submillimeter sources of barger et al . ( 1999b ) . once we have redshifts for the distant submillimeter source population , we can determine the global evolution of star formation in dust - obscured galaxies . previous studies of the star formation rate density ( sfrd ) have primarily used rest - frame ultraviolet data ( e.g. madau et al . however , the ultraviolet emission from a galaxy is heavily affected by the presence of even small amounts of dust , and the extinction corrections are highly uncertain ; for example , corrections for dust obscuration at @xmath30 range from factors of @xmath31 ( pettini et al . 1997 ) to factors of @xmath32 ( meurer et al . 1997 ) , though the more recent estimate of meurer , heckman , & calzetti ( 1999 ) lowers the latter to a factor of @xmath33 . only direct measurements of the reradiated light at submillimeter wavelengths can securely address the sfrd at high redshifts . the true sfrd will receive contributions from both the ultraviolet / optical and the submillimeter . we find that the submillimeter contribution to the sfrd in the @xmath6 range from our @xmath5 mjy sources is comparable to the observed ultraviolet / optical sfrd contribution . the ultraviolet lyman - break galaxies are on the average undetected in the submillimeter at a @xmath34 level of @xmath35 mjy , and thus the lyman - break galaxy population is largely distinct from the bright submillimeter population ( chapman et al . 1999 ) . if we assume that fainter submillimeter sources have the same redshift distribution and properties as the @xmath5 mjy sample , then the sfrd of the entire population contributing to the submillimeter background is about an order of magnitude higher than the observed ultraviolet / optical sfrd . the contribution from uligs to the sfrd increases by more than two orders of magnitude from @xmath7 to @xmath6 , which supports a scenario in which the distant submillimeter sources are the progenitors of massive spheroidal systems . this rapid evolution in the ulig population sampled by scuba can be reproduced in models ( e.g. , blain et al . 1999b , c ) . 2 we present our radio sample and optical / nir imaging , along with our new scuba and lris observations . in 3 we determine our radio - selected submillimeter source counts . in 4 we introduce the predicted high redshift submillimeter - radio flux correlation and obtain millimetric redshift estimates for our sources . in 5 we use the complementary information from radio and submillimeter fluxes to gain insights into the characteristics of our radio - selected submillimeter source population . in 6 we compare the rest - frame colors of the radio sources that have spectroscopic redshifts with the rest - frame colors of our submillimeter sources and find that the latter are likely to fall in the extremely red object category . in 7 we calculate the luminosities , number densities , and sfrds of our submillimeter sources . we compare the submillimeter contribution to the sfrd with contributions from radio and ultraviolet / optical wavebands over a range of redshifts . in 8 we summarize our main conclusions . we take @xmath36 km s@xmath37 mpc@xmath37 and consider both @xmath38 , @xmath39 and @xmath40 , @xmath41 , which should cover the full range of possible cosmologies . the present study is based on deep radio maps centered on the hdf that were observed with the vla at 1.4 ghz ( richards 1999b ) and 8.5 ghz ( richards et al . the primary 1.4 ghz image covers a @xmath42 diameter region with an effective resolution of @xmath43 and a @xmath44 completeness limit of 40 @xmath0jy . the 8.5 ghz images have an effective resolution of @xmath45 and a @xmath44 completeness limit of 8 @xmath0jy over a radius of @xmath46 from the hdf center , rising to 40 @xmath0jy at @xmath47 . the 1.4 ghz hdf map was trimmed to match the 79.4 arcmin@xmath48 nir and optical imaging of the field obtained by barger et al . ( 1999a ) . the absolute radio positions are known to @xmath49 rms in the hdf ; the alignment of the optical data to the radio data left residual astrometric uncertainties of @xmath50 ( richards et al . there are 70 sources in our final radio sample . [ tab1 ] table 1 gives the 1.4 ghz radio catalog for the central 79.4 arcmin@xmath48 region discussed above . the first five columns are catalog number , ra(2000 ) , dec(2000 ) , 1.4 ghz radio flux , and radio flux uncertainty ( richards 1999b ) . the remaining columns , to be discussed in the following subsections , are @xmath51 , @xmath52 , @xmath53 , @xmath54 , @xmath55 , and @xmath56 magnitudes , redshift , submillimeter flux , submillimeter flux uncertainty , and @xmath57 and 15 @xmath0 m isocam fluxes ( aussel et al . the last column of table 1 gives the radio spectral index @xmath58 ( @xmath59 ) or limit for the sources where 8.4 ghz data are available and can be used to determine the origin of the radio emission ( richards 1999b ) . inverted spectrum sources ( @xmath60 ) invariably have self - absorbed synchrotron emission associated with an agn . flat spectrum sources ( @xmath61 ) can be produced by agn activity or by optically thin bremsstrahlung radiation from star formation at higher radio frequencies ( @xmath62 ghz ) . steep spectrum sources ( @xmath63 ) consist of diffuse synchrotron emission often associated with either radio jets or star formation in galaxies . in this paper we use the radio spectral index as a crude discriminator between agn and star formation activity . we arbitrarily classify any source with an available spectral index that has @xmath64 ( there are six such sources in our sample ) to be primarily powered by agn activity . wide - field and deep @xmath51 observations of the hdf and flanking fields were obtained using the university of hawaii quick infrared camera ( quirc ; hodapp et al . 1996 ) on the 2.2 m university of hawaii ( uh ) telescope and the 3.6 m canada - france - hawaii telescope ( cfht ) . the @xmath51 ( @xmath65 @xmath0 m ) filter is described in wainscoat & cowie ( in preparation ) ; the empirical relation between @xmath51 and @xmath66 is @xmath67 , which simplifies to @xmath68 , assuming the median @xmath69 galaxy color ( barger et al . in table 1 we use a @xmath70 limit of @xmath71 for our wide - field image and a @xmath70 limit of @xmath72 for our deep image in the area around the hdf proper . deep johnson @xmath53 and kron - cousins @xmath52-band observations were made with the cfht over a much larger area using the uh8k ccd mosaic camera built by metzger , luppino , and miyazaki . details of the above @xmath51 and optical observations can be found in barger et al . ( 1999a ) . in table 1 we use @xmath70 limits of @xmath73 and @xmath74 . in february 1998 we observed four of the sources in our radio sample with the near - infrared camera ( nirc ; matthews & soifer 1994 ) on the keck i 10 m telescope . these sources previously had only @xmath51 limits but were detected in the nirc observations . conditions were photometric with seeing @xmath75 fwhm . nirc has a @xmath76 insb array with @xmath77 pixels , giving a @xmath78 field of view . we imaged at 2.1 @xmath0 m ( @xmath79 ) with total exposure times for each object of @xmath80 s. the data were obtained in sets of 120 s exposures , and the center of the field was moved in a @xmath81 grid pattern with @xmath82 on a side . the centers of successive grids were moved by @xmath83 between each set . a fifth source was subsequently detected with nirc with a longer exposure under non - photometric conditions . the data were processed using median sky flats generated from the dithered images and calibrated onto the @xmath51 images using other galaxies in the field . we used the low - resolution imaging spectrometer ( lris ; oke et al . 1995 ) on the keck ii 10 m telescope in march 1997 and february 1998 to obtain @xmath55-band and kron - cousins @xmath54-band images , respectively , of a strip region @xmath84 in size that crosses the hdf . the total exposure times were 1680 s and 1600 s for @xmath55 and @xmath54 , respectively , and the seeing was @xmath85 in @xmath54 and @xmath86 in @xmath55 . the @xmath70 limits are @xmath87 and @xmath88 . details of the observations can be found in cowie & hu ( 1998 ) and cowie , songaila , & barger ( 1999 ) . finally , a deep @xmath89 ) image of an area @xmath90 arcmin@xmath48 centered on the hdf - proper was obtained using the orbit ccd on the uh telescope . the @xmath70 limit for the @xmath56 image is @xmath91 . details of the @xmath56 observations can be found in wilson et al . ( in preparation ) . all photometric magnitudes were measured in @xmath82 diameter apertures and then corrected to @xmath92 diameter ( near total ) magnitudes following the procedures of cowie et al . ( 1994 ) . serjeant et al . ( 1997 ) obtained deep isocam observations of the hdf and flanking fields at 6.75 and 15 @xmath0 m . the aussel et al . ( 1999 ) reductions of the data produced a main source list of 49 objects ( @xmath93 ) and a supplementary list of an additional 51 objects ( @xmath2 for 15 @xmath0 m and @xmath44 for 6.75 @xmath0 m ) . despite the large point spread function ( @xmath94 at 15 @xmath0 m ) , in most cases the optical or nir identification of the isocam source was straightforward . the redshifts for sources in the sample are in the range @xmath95 to @xmath96 ( median @xmath97 ) . we used lris during a two night run on keck ii in march 1999 to obtain spectroscopic observations of a sub - sample of the 1.4 ghz sources in a strip region centered on the hdf . although the hdf and flanking field region has been intensely studied by a number of groups ( see cohen et al . 1999 for a summary and references ) , most of the sources in our sample had not been previously observed due to their faint optical fluxes . we used @xmath98 wide slits and the 400 lines mm@xmath37 grating blazed at 8500 , which gives a wavelength resolution of @xmath99 and a wavelength coverage of @xmath100 . the wavelength range for each object depends on the exact location of the slit in the mask but is generally between @xmath101 and 10000 . three of the slit masks were constructed at a position angle of @xmath102 , and the remaining three were nearly identical versions constructed at a position angle of @xmath103 . this procedure enabled us to obtain sufficient wavelength coverage for all the objects in our sample , including those that fell close to the edges of the masks . the observations were 1.5 hr per slit mask , broken into three sets of 30 minute exposures . three hdf slit masks were observed per night . some of the objects were in all of the slit masks and hence were observed for 9 hrs ( see table [ tab2 ] ) . conditions were photometric with seeing @xmath104 fwhm both nights . the objects were stepped along the slit by @xmath105 in each direction , and the sky backgrounds were removed using the median of the images to avoid the difficult and time - consuming problems of flat - fielding lris data . details of the spectroscopic reduction procedures can be found in cowie et al . ( 1996 ) . ccccc 16 & 23.39 & 25.30 & & 6 + 17 & 18.75 & 22.49 & 1.013 & 9 + 23 & 16.36 & 19.09 & 0.456 & + 24 & 19.10 & 22.29 & 1.219 & + 25 & 19.25 & 22.71 & & 3 + 26 & 22.29 & 25.30 & & 6 + 28 & 21.22 & 25.30 & & 6 + 30 & 21.23 & 25.30 & & 3 + 31 & 18.06 & 20.86 & 0.857 & + 32 & 17.85 & 21.04 & 1.013 & + 33 & 21.65 & 24.72 & & 3 + 35 & 18.36 & 20.76 & 0.961 & + 36 & 22.60 & 25.30 & & 3 + 37 & 21.00 & 23.20 & & 3 + 38 & 18.74 & 21.09 & 0.475 & + 40 & 17.36 & 19.88 & 0.410 & + 41 & 21.50 & 25.30 & & + 42 & 16.52 & 18.70 & 0.321 & + 43 & 19.31 & 21.88 & 1.275 & + 44 & 21.50 & 25.30 & & 6 + 46 & 22.60 & 25.30 & & 9 + 47 & 19.55 & 22.64 & 0.474 & + 48 & 18.52 & 20.92 & 0.761 & + 50 & 20.14 & 24.81 & 0.884 & + 51 & 19.21 & 21.89 & 1.243 & + 52 & 18.18 & 20.55 & 0.902 & + 53 & 20.29 & 24.70 & & 3 + 54 & 20.76 & 25.13 & & 3 + 56 & 17.98 & 19.87 & 0.422 & + 59 & 19.51 & 22.66 & & 3 + 60 & 20.08 & 22.90 & & 6 + 61 & 20.38 & 22.90 & & + 62 & 17.36 & 19.80 & 0.558 & + 64 & 17.59 & 19.90 & 0.411 & 7.5 + 65 & 20.90 & 25.30 & & 3 + 66 & 19.39 & 22.14 & 1.019 & 3 + 69 & 21.50 & 25.30 & & 1.5 + in table [ tab2 ] we list all the objects in the radio sample that fall in the lris strip region . the columns are radio catalog number from table 1 , @xmath51 mag , @xmath52 mag , redshift , and exposure time for any object targeted in our march 1999 spectroscopic survey . over an area @xmath106 arcmin@xmath48 , 19 of the 37 radio sources now have secure redshift identifications ; all are at @xmath25 . figure [ figzhkandzi]a , b shows @xmath51 versus redshift and @xmath52 versus redshift . spectroscopic redshifts are in general relatively straightforward to obtain for radio objects with @xmath24 . we note that waddington et al . ( 1999 ) claim a redshift identification of @xmath107 for object 30 based on a ly@xmath108 detection . however , the position of their detection is @xmath109 away from the radio source position and counterpart optical galaxy , and so an association with the radio source is not secure . the radio source is detected in the mid - ir , which would suggest @xmath110 , consistent with the redshift estimated using the millimetric redshift technique described in [ secze ] . however , we did not see any strong [ oii ] 3727 feature in our spectrum from @xmath111 , which would suggest @xmath26 . thus , the redshift of this object remains uncertain . in all of our slit - masks ( equivalent to a 9 hr integration ) we also included the position of the brightest scuba source , hdf850.1 , from the deep 850 @xmath0 m map of the hdf - proper by h98 . we centered our slit across the position of the 1.3 mm detection of hdf850.1 reported by downes et al . ( 1999 ) , which also coincides with the 8.5 ghz supplemental radio source 3651 + 1226 ( richards et al . we oriented the slit such that it fell across both the arc - like feature 3 - 593.0 that is favored by downes et al . ( 1999 ) as the optical counterpart to hdf850.1 and the nearby red galaxy 3 - 586.0 . however , we were unable to determine a secure redshift for either source from our spectroscopic data . our scuba jiggle map observations were flexibly scheduled in mostly excellent observing conditions during two runs in april and june 1999 for a total of five observing shifts . the maps were dithered to prevent any regions of the sky from repeatedly falling on bad bolometers . the chop throw was fixed at a position angle of @xmath102 so that the negative beams would appear 45 arcsec on either side east - west of the positive beam . the data were reduced using beam weighted extraction routines that included both the positive and negative portions of the chopped images , thereby increasing the effective exposure times . regular `` skydips '' ( lightfoot et al . 1998 ) were obtained to measure the zenith atmospheric opacities at 450 and 850 @xmath0 m , and the 225 ghz sky opacity was monitored at all times to check for sky stability . the median 850 @xmath0 m optical depth for all nights together was 0.265 . pointing checks were performed every hour during the observations on the blazars 0954 + 685 , 1418 + 546 , 0923 + 392 , or 1308 + 326 . the data were calibrated using 30 arcsec diameter aperture measurements of the positive beam in beam maps of the primary calibration source , mars , and one of three secondary calibration sources , crl618 , irc+10216 , or oh231.8 . the data were reduced in a standard and consistent way using the dedicated scuba user reduction facility ( surf ; jenness & lightfoot 1998 ) . due to the variation in the density of bolometer samples across the maps , there is a rapid increase in the noise levels at the very edges . we have clipped the low exposure edges from our images . we present our scuba maps in fig . [ figmaps ] . we also re - reduced the archival hdf - proper scuba data of h98 in order to make a consistent analysis with the present data . we could only make use of the maps that were taken with a fixed ra chop ( 90 per cent of the data sample ) , which included 34 hours with a 30 arcsec chop throw and 28 hours with a 45 arcsec chop throw . we combined these data separately to form two independent maps . in table 1 we quote submillimeter fluxes determined from the weighted average of measurements made in each map . the surf reduction routines arbitrarily normalize all the data maps in a reduction sequence to the central pixel of the first map ; thus , the noise levels in a combined image are determined relative to the quality of the central pixel in the first map . in order to determine the absolute noise levels of our maps , we first eliminated the @xmath112 real sources in each field by subtracting an appropriately normalized version of the beam profile . we then iteratively adjusted the noise normalization until the dispersion of the signal - to - noise ratio measured at random positions became @xmath113 . our noise estimate includes both fainter sources and correlated noise . we centered on the positions of the radio sources and measured the submillimeter fluxes in both the positive and negative beams . the extracted fluxes were calibrated to the 30 arcsec diameter aperture fluxes of the brightest sources . the brighter submillimeter sources were subtracted from the maps before the fainter sources were extracted . in cases where the positions of two radio sources are very close , we may be slightly overestimating the submillimeter flux allocated to the bright source with this procedure and underestimating that allocated to the faint source , but since the total submillimeter flux should be reasonable , we will not be making any gross errors in our later estimates of the total star formation rate . from the total sample of 70 radio - selected galaxies in the @xmath27 square arcminute central region of the radio map , we take the 22 with @xmath23 to be our optical / nir - faint radio sample , for which there are now submillimeter observations of 19 . even though our scuba observations were relatively shallow ( the 850 @xmath0 m @xmath2 limit was 6 mjy ) , we detected 5 of the 15 optical / nir - faint radio sources that we observed ; a sixth significant source ( @xmath114 mjy ) was detected in the hdf - proper scuba map of h98 ( hdf850.2 in their notation ) . the submillimeter flux we measure for the brightest source ( hdf850.1 ) in the hdf - proper map is @xmath115 mjy , which is slightly lower than the value of @xmath116 found by h98 but is consistent within the statistical and systematic errors . the jiggle map observing mode enabled simultaneous observations of a large fraction ( 31/48 ) of the optical / nir - bright radio sources , none of which were detected . there are two @xmath5 mjy submillimeter sources in our jiggle maps that were not in the radio sample . in fig . [ figcolor]a , b we show the 1.4 ghz fluxes of the radio sources in our sample versus their @xmath51 and @xmath52 magnitudes , indicating those with measured redshifts and those with submillimeter detections . we find that the radio selection technique is effective in locating the majority of bright submillimeter sources . we document this in fig . [ figcounts1 ] where we compare the combined 850 @xmath0 m source counts from blank field submillimeter surveys ( h98 ; eales et al . 1999 ; barger , cowie , & sanders 1999 ) with our new radio - selected 850 @xmath0 m source counts . a source of flux strength @xmath117 contributes @xmath118 to the counts per unit area , where @xmath119 is the area over which there is @xmath120 sensitivity to @xmath117 . the cumulative counts , @xmath121 , are given by the sum of the inverse areas of all sources brighter than @xmath122 . however , for our radio - targeted search , the effective area ( 65 square arcminutes ) is greater than if we had done an untargeted search with the same number of pointings and is given by the fraction ( number @xmath23 sources observed)/(total number @xmath23 sources in sample ) times the radio survey area ( 79.4 square arcminutes ) . because of the correspondence between the optical / nir - faint radio population and the bright submillimeter population , we do not need to survey the entire radio field in order to observe the bright submillimeter population . it is interesting to speculate whether the correspondence between optical / nir - faint radio sources and bright submillimeter sources also holds at fainter submillimeter flux levels . a necessary requirement for this to be the case is that the surface density of faint radio sources be comparable to or exceed the surface density of submillimeter sources . extrapolating a power - law parameterization of the 1.4 ghz counts over the range @xmath123 @xmath0jy ( richards 1999b ) down to a 1 @xmath0jy cut - off gives @xmath124jy ) @xmath125 arcmin@xmath126 . the empirically estimated surface density of submillimeter sources ( barger , cowie , & sanders 1999 ) is @xmath127 arcmin@xmath126 . a possible counter - argument to a close correspondence between radio and submillimeter sources at faint submillimeter flux levels comes from the study by smail et al . ( 1999b ) who compared their lensed submillimeter survey data with radio data . they found that the carilli & yun ( 1999 ) radio - submillimeter spectral indices differed for their bright and faint submillimeter subsamples . they suggested that the difference could arise if the faint submillimeter sources lie at higher redshifts or have different fractions of radio - loud agn or have different dust temperatures . however , with the low statistics , strong conclusions can not be drawn at this time . optical spectroscopic surveys of submillimeter galaxies are difficult because of the very different behaviors of the @xmath66-corrections in the optical and submillimeter and the poor submillimeter resolution . spectroscopic surveys to date ( barger et al . 1999b ; lilly et al . 1999 ) give potentially conflicting results and are very limited in size . photometric redshift estimates of submillimeter sources have been made from candidate optical counterparts ( h98 ; lilly et al . 1999 ) , but these are questionable both because the counterpart identification is not secure and because the seds of the submillimeter sources may be unlike those of the optical sources used in making the redshift estimates due to dust extinction and possible agn contributions . consequently , a new approach is required if the positions of the submillimeter sources are to be reliably determined and their nature and redshift distribution understood . the remarkably tight local correlation between the global fir and nonthermal radio luminosities provides a promising alternative method for identifying and studying individual submillimeter sources , provided that the correlation holds to high redshift . in [ secsmmcounts ] we found that targeting with scuba optical / nir - faint 1.4 ghz sources is an efficient technique for identifying the majority of the bright submillimeter source population . our results indicate that a large fraction of bright sources in submillimeter surveys have extremely faint optical / nir counterparts and hence are inaccessible to optical spectroscopy ( see also smail et al . this conclusion is consistent with results from the barger et al . ( 1999b ) spectroscopic survey of lensed submillimeter sources discussed in the introduction . although we are unable to obtain spectroscopic redshifts for the optical / nir - faint radio - selected submillimeter sources , we can use the submillimeter - to - radio flux ratios to obtain millimetric redshift estimates . figure [ figarp ] illustrates how the slope of a dusty galaxy s sed changes abruptly at frequencies higher than 100 ghz . below 30 ghz synchrotron radio emission is dominant , free - free emission is largest in the range @xmath128 ghz , and thermal dust emission dominates above 200 ghz . because of the opposing spectral slopes of the blackbody spectrum in the submillimeter and the synchrotron spectrum in the radio , the submillimeter - to - radio flux ratio rises extremely rapidly with increasing redshift ( see fig . [ figulig ] ) . carilli & yun ( 1999 ; hereafter cy99 ) have therefore suggested using the submillimeter - to - radio flux ratio as a redshift estimator . the dust emission from local luminous infrared galaxies is well described by optically thin single - temperature modified blackbodies with extinction coefficient @xmath129 , where @xmath130 to 2 . the dust temperatures derived lie between 30 and 60 k. arp 220 , with its high bolometric luminosity produced almost entirely from starburst activity ( downes & solomon 1998 ) , is an appropriate prototype for high redshift submillimeter sources , as we will justify subsequently . over the interval @xmath131 ghz ( @xmath132 m ) arp 220 s sed is well represented by a modified blackbody with @xmath133 and a dust temperature @xmath134 k ; the luminosity of this dominant cooler component is @xmath135 ( klaas et al . 1997 ) . in terms of the blackbody distribution @xmath136 the submillimeter flux for an arp 220-like galaxy at redshift @xmath137 is given by @xmath138 ^ 2 \label{smmexact}\ ] ] where @xmath139 is the observed submillimeter frequency , @xmath140 is the dust temperature , and @xmath141 is the luminosity distance . we allow for an overall strength factor , @xmath142 , relative to arp 220 , for which @xmath143 . the above normalization constant is fixed by a @xmath144 fit to the arp 220 flux measurements . in the rayleigh - jeans long - wavelength limit , the submillimeter flux is approximately given by @xmath145 ^ 2 \label{smmlong}\ ] ] which is normalized to the measured 375 ghz ( 800@xmath0 m ) flux value for arp 220 from rigopoulou , lawrence , & rowan - robinson ( 1996 ) ; here @xmath146 . the approximation in eq . [ smmlong ] is reasonably accurate for @xmath147 ; at higher @xmath137 it extrapolates above the result of eq . [ smmexact ] . the radio flux is given by @xmath148 ^ 2 \label{radio}\ ] ] which is normalized to the measured 8.4 ghz flux for arp 220 from condon et al . ( 1991 ) . since both @xmath149 and @xmath150 depend linearly on the star formation rate , the scale factor @xmath142 is the same . in local star forming galaxies there is an observed spectral flattening at 1.4 ghz due to free - free absorption ( condon 1992 ) . since 1.4 ghz measurements at higher redshifts sample higher frequencies that are not sensitive to these absorption effects , we adopt the standard value @xmath151 from condon ( 1992 ) to extrapolate local 8.4 ghz fluxes to 1.4 ghz fluxes , in order to be consistent with high redshift observations . the ratio of eqs . [ smmlong ] and [ radio ] with @xmath152k gives @xmath153 a relation of this type was obtained by cy99 based on the sed of m82 . in their relation , the factor of 1.1 in eq . [ ratio ] is replaced by 0.25 and the power - law index @xmath154 is replaced by 4.3 . however , m82 , with more than an order of magnitude lower luminosity than arp 220 ( @xmath155 ; hughes , dunlop , & rawlings 1997 ) , is not a ulig , and so we would argue that our result is more appropriate for the distant submillimeter sources . in addition to their analytic result , cy99 give empirical results for the flux ratio versus redshift for both arp 220 and m82 . other authors have used the span of the four cy99 curves in their fig . 1 to estimate redshift ranges ; however , the use of m82 again introduces an inappropriate uncertainty in obtaining redshifts . inverting eq . [ ratio ] , the value of @xmath137 can be related to the flux ratio by @xmath156 which holds for @xmath147 . for higher redshifts the ratio of eqs . [ smmexact ] and [ radio ] must be used . the flux ratio method for estimating redshifts has the advantage that the luminosity distance drops out , and thus there is no dependence on @xmath157 or on the cosmology . we make the assumption that unevolved local uligs are representative of the distant ulig population . we show in the following that the arp 220 sed provides a good representation of the submillimeter - to - radio flux ratios of the ensemble of local uligs placed at appropriate redshifts . we use a compilation of submillimeter measurements at 3000 , 857 , 667 , and 375 ghz from rigopoulou et al . ( 1996 ) to infer 353 ghz fluxes at redshifts of 7.5 , 1.44 , 0.9 , and 0.06 . fluxes at 1.4 ghz were extrapolated from the 8.4 ghz measured values of condon et al . ( 1991 ) using a standard synchrotron emission spectrum with @xmath158 , as justified earlier . we plot the mean 353 ghz to 1.4 ghz flux ratios versus redshift as solid diamonds in fig . [ figulig]a with ranges from the minimum to maximum flux ratios . the number of objects in the bins are , in order of decreasing redshift , 25 , 2 , 4 , and 7 . the data are to be compared with the redshifted arp 220 flux ratio calculated from eqs . [ smmexact ] and [ radio ] and plotted as a solid line in the figure . the spread in measured ratios is about a multiplicative factor of two relative to arp 220 , as indicated by the dashed curves . thus , the redshifted arp 220 flux ratio is likely to approximate the flux ratios of uligs at high redshifts to this accuracy . the above conclusion is strengthened by fig . [ figulig]b where we have plotted as solid ( star forming galaxies ) and open ( agn ) squares available submillimeter observations with corresponding radio and redshift information . the star forming galaxies fall neatly on the curves . the average submillimeter - to - radio flux ratio of the galaxies in our survey which have spectroscopic redshifts and have been observed in the submillimeter are shown in two redshift bins as the large filled diamonds with @xmath34 uncertainties . the lowest point is consistent with a @xmath2 null detection , but in the higher redshift bin there is a strong positive detection consistent with the arp 220 ratio , showing that these objects have cool dust emission and obey the fir - radio relation at these redshifts . the short horizontal lines show the redshift ranges for our six significant radio - selected submillimeter detections ; all are in the @xmath6 redshift range . the top portion of table [ tab3 ] gives the fluxes and the millimetric redshifts and ranges for these six sources , as well as their bolometric luminosities for both @xmath39 and @xmath41 ( see [ secfluxz ] ) . crrrrrrr 33 & 11.9 & 0.124 & 1.0 & 1.5 & 2.1 & 5.0 & 9.1 + 49 & 10.7 & 0.324 & 1.6 & 2.3 & 3.1 & 4.5 & 9.0 + 7 & 7.8 & 0.151 & 1.3 & 1.8 & 2.5 & 3.3 & 6.3 + 11 & 7.5 & 0.148 & 1.2 & 1.8 & 2.4 & 3.2 & 6.0 + 13 & 7.1 & 0.071 & 1.8 & 2.4 & 3.3 & 3.0 & 6.0 + 46 ( hdf850.2 ) & 2.4 & 0.046 & 1.3 & 1.8 & 2.5 & 1.0 & 1.9 + + & 19.0 & @xmath159 & @xmath160 & @xmath161 & @xmath162 & @xmath163 & @xmath164 + & 8.8 & @xmath159 & @xmath165 & @xmath166 & @xmath167 & @xmath168 & @xmath169 + ( hdf850.1 ) & 5.4 & @xmath159 & @xmath170 & @xmath171 & @xmath172 & @xmath173 & @xmath174 + the ulig sources plotted in fig . [ figulig ] have a range of emissivity indices , @xmath175 , and dust temperatures in the vicinity of 50 k. it is therefore quite remarkable that all the sources follow a constrained envelope around the arp 220 submillimeter - to - radio flux ratio . the implication of this empirical correlation is that for practical purposes a temperature - redshift degeneracy ( blain 1999 ) does not present a serious problem to millimetric redshift estimation . a recent empirical analysis by carilli & yun ( 2000 ) , which appeared subsequent to the submission of our paper , finds conclusions that are consistent with our analysis above . in our analysis we assume a flat universe , @xmath176 , where @xmath177 is the matter density and @xmath178 the vacuum density . the absolute flux values in eqs . [ smmexact ] and [ radio ] depend on the cosmology through the luminosity distance ( carroll , press , & turner 1992 ) @xmath179^{1/2}\ ] ] the basic physics underlying the joint use of radio and submillimeter observations is contained in figs . [ figfalloff ] and [ figsvss ] . figure [ figfalloff ] shows predicted arp 220 radio and submillimeter fluxes versus redshift for the @xmath38 , @xmath39 cosmology ( @xmath180 km s@xmath37 mpc@xmath37 ) with selected overall strengths @xmath181 , 1 , 3 , and 6 relative to arp 220 . @xmath182 is flat with redshift for @xmath11 due to the negative @xmath66-correction , whereas @xmath183 falls sharply with increasing redshift due to the @xmath184 dependence . these behaviors are characteristic of other cosmologies as well . thus , radio surveys detect a high proportion of low redshift sources whereas submillimeter surveys offer comparable sensitivity to both moderate ( @xmath185 ) and high redshifts . in fig . [ figsvss ] we illustrate the dependence on cosmology of the absolute fluxes in the @xmath183 versus @xmath182 plane . figure [ figsvss]a shows predicted arp 220 fluxes for the @xmath39 cosmology , and fig . [ figsvss]b shows the same for the @xmath41 cosmology favored by recent distant supernovae type ia observations ( perlmutter et al . 1999 ; riess et al . redshift labels are given on the @xmath143 , 3 , and 6 curves . in fig . [ figsvss]a the bright submillimeter sources fall near the @xmath186 contour , indicating that these sources have luminosities several times the luminosity of the arp 220 prototype . for the @xmath41 cosmology we would infer somewhat higher luminosities from the curves , although the redshift estimates remain the same . from fig . [ figsvss ] , we see that any submillimeter sources above our detection threshold of 6 mjy would not be detectable in the radio if @xmath187 . of the significant submillimeter sources detected in our survey , two had no radio counterparts and one , hdf850.1 , had only a supplemental 8.5 ghz radio detection . based on the millimetric redshift estimator , these sources are potentially at very high redshift . the bottom three lines of table [ tab3 ] give the fluxes and the millimetric redshifts and ranges , along with the bolometric luminosities for @xmath39 and @xmath41 , for the three significant scuba detections without corresponding 1.4 ghz detections ; here the numerical entries are based on the 0.040 mjy ( @xmath44 ) radio flux limit , which corresponds to the lowest possible redshift and @xmath188 . since we have covered only one - third of our @xmath27 square arcminute area with our targeted scuba observations , there could in principle be more high redshift submillimeter sources with no radio detections . however , our present radio counts already saturate the bright submillimeter counts distribution from the combined blank field submillimeter surveys ( see fig . [ figcounts1 ] ) and hence argue against the probability of finding many such additional sources . in contrast , sources of arp 220 or sub - arp 220 strength could be seen in the radio but not in the submillimeter with our present 6 mjy threshold if @xmath189 . figure [ figfalloff]b shows that we have many such candidates . our targeted scuba observations of optical / nir - faint radio sources are therefore selecting the high redshift end of the faint radio source population . in fig . [ figrest ] we plot rest - frame ab(2800)ab(8140 ) color versus redshift for the radio sample with spectroscopic identifications . the colors of field galaxies in the hdf and ssa22 fields are indicated with tiny solid circles to illustrate how the radio galaxies ( solid triangles for sources with @xmath64 but otherwise solid circles ) generally have very red ab colors . these colors range from @xmath190 to @xmath191 . there is no evidence for a color trend with redshift . the rest - frame ab colors of the submillimeter galaxies with @xmath51 detections ( solid pentagons ) , although mostly lower limits , are not inconsistent with the colors of the lower - redshift radio sources . at @xmath192 , ab(2800)ab(8140 ) roughly corresponds to ab(@xmath52)ab(@xmath51 ) , and thus submillimeter sources with similar colors to the radio sources would have observed vega - based @xmath193 colors in the range @xmath194 , which would mostly place them in the extremely red object ( ero ) category ( @xmath195 or @xmath196 ; e.g. , mccracken et al . 1999 and references therein ) . this result is consistent with the recent detections of bright ( @xmath197 ) eros as submillimeter sources ( cimatti et al . 1998 ; dey et al . 1999 ; smail et al . . in general , however , the submillimeter sources are so optically faint that identifying them as eros is difficult . the luminosities of our submillimeter sources were obtained by scaling the arp 220 luminosity at redshift @xmath137 by the relative source strengths , @xmath142 ; these are given in table [ tab3 ] . the fir luminosity provides a direct measure of the current star formation rate ( sfr ) ; with present detection capabilities , the radio provides greater sensitivity to sfrs at low redshifts ( @xmath198 ) while the submillimeter is superior at high redshifts . in this sense radio and submillimeter data complement each other in getting the sfr over the full range of redshifts . this relative capability is illustrated in fig . [ figradiolum ] where the strength factors , @xmath142 , as determined from the radio luminosities relative to redshifted arp 220 luminosities , are plotted versus redshift in the @xmath39 cosmology . here the dotted curve represents the radio threshold of 40 @xmath0jy , and the solid line at a value of two represents our 6 mjy submillimeter threshold of roughly two times the luminosity of arp 220 for the @xmath39 cosmology ( see fig . [ figsvss ] ) . these two threshold curves cross at @xmath199 . since the bulk of the radio detections are at @xmath25 and have strength factors @xmath200 , their non - detection in the submillimeter is as expected ; our @xmath201 sources are detected in both the radio and submillimeter . at redshifts @xmath202 the submillimeter may provide the only access to distant uligs . figure [ figradiolum ] also provides a means to establish the completeness of the ulig number distribution versus redshift . our distant ulig sample is complete over the redshift range @xmath6 ( except possibly for the 4 out of 22 optically - faint radio sources that were not observed ) because these @xmath6 sources are above both radio and submillimeter detection thresholds . at @xmath203 there is the risk that we may lose sources that fall below the rising 1.4 ghz threshold relative to the arp 220 luminosity at redshift @xmath137 . the space density of our five significant @xmath204 mjy submillimeter sources in the redshift range @xmath6 is @xmath205 in units @xmath206 . in comparison , the space density of local uligs ( @xmath207 ; kim & sanders 1998 ) with bolometric luminosities above @xmath208 is only @xmath209 for @xmath39 and @xmath210 for @xmath41 , about a factor of 200 and 80 lower than our high redshift results . the discrepancy in space density is still larger if we compare with local uligs that are comparably luminous to the @xmath5 mjy population . thus , enormous evolution in the ulig population must take place between @xmath201 and the present . it is interesting to note a possible empirical similarity between the redshift distribution of quasars and that of submillimeter sources , which might be expected if quasars are a successor stage to merger processes forming submillimeter sources ( sanders & mirabel 1996 ) . the relative space density of a complete radio - selected quasar sample ( shaver et al . 1998 ) , which should be unaffected by dust obscuration , is approximately described by a gaussian as a function of redshift , given by @xmath211 the peaking of this distribution at @xmath199 and the rapid fall - off at low redshift seems consistent with our measured submillimeter redshift distribution . recent rosat x - ray studies have found a very high redshift tail ( @xmath212 ) to the quasar distribution that may be above the gaussian expectation ( hasinger 1998 ) . since the rosat sources were not seen in the optical , it is plausible that they are dust obscured . similarly , we have found a few submillimeter sources that plausibly lie at @xmath203 . the observed ultraviolet light can most simply be translated to a metal production rate where , as first pointed out in cowie ( 1988 ) , there is an extremely tight relation since both metals and ultraviolet light are produced by the same massive stars . however , it is usual ( particularly in the fir ) to translate the metal production rate to a total stellar mass production rate , which may be more intuitively interpreted . unfortunately , this extrapolation requires knowledge of the shape of the initial mass function ( imf ) , which is quite uncertain . in the following we shall assume a salpeter imf ( @xmath213 , @xmath214 ) . the sfr is related to the rest - frame 2800 luminosity per unit frequency by @xmath215 with @xmath216 in ergs s@xmath37 hz@xmath37 and @xmath217 in m@xmath218 . the cowie ( 1988 ) estimate of @xmath219 from nucleosynthesis arguments and galaxy sed modelling is @xmath220 for the above imf limits . later songaila , cowie , & lilly ( 1989 ) suggested that @xmath219 should be about a factor 1.7 lower based on stellar synthesis . madau et al . ( 1996 ) obtained a value of @xmath221 based on the evolutionary models of bruzual & charlot ( 1993 ) . the relatively invariant estimates of @xmath219 are a direct consequence of the nucleosynthesis arguments underlying the modelling ; the primary uncertainty in @xmath217 lies in the stellar imf . we adopt the madau et al . value for consistency with other recent papers that follow this normalization . in terms of the solar luminosity , the sfr relation is then @xmath222 where @xmath223 . the conversion of the bolometric luminosity to a total stellar mass production rate is an even more invariant prediction of the nucleosynthesis arguments since knowledge of the exact shape of the ultraviolet sed of the galaxies is not needed . the bolometric luminosity in units of the solar luminosity is @xmath224 where @xmath225 is the bolometric correction . for a flat @xmath226 over the wavelength range 912 to 22000 , @xmath227 . rowan - robinson et al . ( 1997 ) computed a bolometric correction of 3.5 from the bruzual & charlot ( 1993 ) models , which we adopt . then the sfr is related to @xmath188 by @xmath228 @xmath217 would be increased by a factor of 3.3 if we had instead assumed a miller - scalo imf over the same mass range . equations [ mdot2 ] and [ mdot3 ] therefore provide a self - consistent description of the mass production rates seen in the optical and submillimeter . light which directly escapes the galaxy can be mapped with the ultraviolet light , while reprocessed light , which escapes the galaxy in the submillimeter , can be calibrated with eq . [ mdot3 ] . we now use the arp 220 bolometric luminosity @xmath229 ( klaas et al . 1997 ) and the 8.4 ghz radio flux of 0.148 jy ( condon et al . 1991 ) to calibrate the sfrs for our radio sources . we find @xmath230 ; thus @xmath231 where the normalization factor is in units of ergs s@xmath37 hz@xmath37 . from the equations of condon ( 1992 ) translated to the same assumptions about the imf , we would obtain a normalization of approximately @xmath232 or about a factor of 3.5 smaller than our result . our normalization translated with the @xmath233 spectral shape to 1.4 ghz is @xmath234 using eq . [ radiosfr ] , we can convert our radio luminosities into sfrds . we know that 95 per cent of the micro - jansky radio sources in the hdf region have been resolved at @xmath235 resolution with the multi - element radio linked interferometer ( merlin ) at 1.4 ghz . the median angular size for the radio emission is @xmath236 ( richards 1999a ) , which suggests radio emission on galactic or sub - galactic size scales . thus , the radio emission in most of these systems may originate primarily from star formation , although we can not exclude contributions to the radio flux densities from embedded agn . we divide our radio sources with spectroscopic redshifts into two bins , @xmath237 and @xmath238 , excluding sources with radio spectral indices @xmath239 that might be agn . of the 38 presumed star forming objects with @xmath240 in our sample , 23 have spectroscopic redshifts . most of the remaining 15 sources have not been observed and thus may be expected to follow the same redshift distribution as the identified sources . we therefore calculate the differential comoving volume over a survey area @xmath241 the volume is integrated from @xmath242 , the lower redshift limit of the bin , to @xmath243 , either the radio luminosity limit of the survey or the upper redshift limit of the bin , whichever is smaller . the comoving volume element in mpc@xmath244 is @xmath245\ ] ] where @xmath246 is the area in arcmin@xmath48 and @xmath141 is in mpc . we find that for the @xmath247 bin , the sfrd in units @xmath248 is @xmath249 and for the @xmath238 bin , @xmath250 where the uncertainties are poissonian based on the number of sources . these represent lower limits to the sfrd since we have not attempted to correct for the contributions of faint sources below our flux limits . the sfrd from our submillimeter sources can be estimated under the assumption that star formation dominates agn contributions . we can calculate the contribution of the @xmath204 mjy submillimeter sources at @xmath6 to the sfrd using eq . [ mdot3 ] ; we find @xmath251 unlike the number distribution versus redshift , the sfrd needs large corrections for completeness due to the fact that we are detecting only relatively bright sources here ( at about six times the luminosity at which the submillimeter background is primarily resolved ) , and the distribution @xmath252 increases rapidly as @xmath122 decreases . to estimate the completeness correction , we assume that the dependences of @xmath253 on @xmath122 and @xmath137 factorize , @xmath254 . this is a plausible assumption in the submillimeter where the fluxes are nearly independent of redshift ; nonetheless , in view of the smail et al . ( 1999b ) study , this assumption remains to be confirmed . we can determine the completeness correction for the sfrd using the empirical number distribution at 850 @xmath0 m versus @xmath122 @xmath255 that describes the measured submillimeter counts above 2 mjy ( barger , cowie , & sanders 1999 ) . we are making the assumption that the flux to @xmath256 conversion based on arp 220 applies in the low submillimeter flux region ; the justification is that even the dominant @xmath113 mjy population are near - ulig sources . the completeness correction over all submillimeter fluxes is therefore the measured @xmath257 m extragalactic background light ( ebl ) divided by the @xmath257 m light above 6 mjy . the @xmath257 m ebl measurement of @xmath258 mjy deg@xmath126 from puget et al . ( 1996 ) and @xmath259 mjy deg@xmath126 from fixsen et al . ( 1998 ) imply correction factors in the submillimeter of 11 and 15 , respectively . thus , the estimated total submillimeter contribution to the sfrd in units of @xmath248 is @xmath260 where we have used the factor of 11 completeness correction . more speculatively , we can determine the sfrd from higher redshift sources using our two @xmath5 mjy submillimeter sources without radio counterparts . in this case we use the volume from @xmath261 and the actual area surveyed in the submillimeter . we find @xmath262 after including the factor of 11 completeness correction , this becomes @xmath263 h98 used photometric redshift estimates to infer that four of their five @xmath264 mjy sources were in the redshift range @xmath265 . while these source identifications were problematic , it appears likely from the present work that the redshifts do lie in this rough redshift range . using our parameters and a scaled arp 220 sed , we find that the sfrd for their sources with @xmath39 is @xmath266 . if we make a completeness correction to include the contribution below 2 mjy , we obtain sfrd=@xmath267 , in good agreement with our result in eq . [ eqtruesfrdsmma ] . the presence of a substantial fraction of agn - dominated ulig sources would reduce the above sfrds . in a recent near - infrared spectroscopic study of 64 local uligs , veilleux , sanders , & kim ( 1999 ) found agn characteristics in @xmath268 per cent of the sample , which increased to @xmath269 per cent for the sample with @xmath270 . thus , our @xmath5 mjy contributions to the sfrd may need to be reduced by a factor @xmath271 . however , the lower agn fraction in fainter uligs seen locally suggests that agn contamination may be less of an issue for the extrapolated sfrd of the whole submillimeter population . the determination of the sfrd from optical observations has been a subject of intense investigation . observations first indicated a rather rapid rise in the sfrd from @xmath272 followed by a sharp decline at higher redshifts with the peak sfrd being @xmath273 ( madau et al . a recent modification in the inferred optical sfrd at low redshifts was made by cowie , songaila , & barger ( 1999 ) , whose data indicated a more gradual rise in the sfrd than had previously been found by lilly et al . ( 1996 ) . it was realized that dust obscuration effects could result in factors of 3 to 5 ( pettini et al . 1997 ; meurer , heckman , & calzetti 1999 ) increases in the sfrd at high redshift . with these rather uncertain dust corrections taken into account , it has been argued that the sfrd flattens at a constant @xmath274 in the @xmath39 cosmology for @xmath275 ( steidel et al . 1999 ) . in fig . [ figsfrdvsz ] we compare the star formation history in the optical ( without extinction corrections ) with that which we obtain in the submillimeter both before ( eqs . [ eqsfrdsmma ] and [ eqhisfrdsmma ] ; solid triangles ) and after ( eqs . [ eqtruesfrdsmma ] and [ eqhitruesfrdsmma ] ; solid circles ) correcting for incompleteness . we also include our newly determined radio sfrd limits ( eqs . [ eqlowradioa ] and [ eqhiradioa ] ) on the figure as solid squares . the submillimeter contribution to the sfrd inferred from our @xmath5 mjy observations is comparable to the ultraviolet / optical contribution to the sfrd . the two wavelength regimes are likely sampling different stages in galaxy formation . the submillimeter detects the formation of massive spheroids while the ultraviolet / optical detects the formation of smaller disk or bulge systems . the approximate equality of the optical and submillimeter backgrounds supports this hypothesis ; the metal density in present - day disks is roughly comparable to that in the spheroidal components of galaxies , so comparable amounts of light are expected to be produced in their formation ( cowie 1988 ) . the completeness - corrected submillimeter sfrd , shown by the solid circles in fig . [ figsfrdvsz ] , is based on the assumption that fainter submillimeter sources have the same redshift distribution and properties as the @xmath5 mjy sample . hence the sfrd from the entire population contributing to the submillimeter background is about an order of magnitude higher than the extinction - uncorrected ultraviolet / optical sfrd . since the submillimeter measures reradiated optical light , the observed ultraviolet / optical sfrd contribution should be added to the submillimeter contribution . we also plot in fig . [ figsfrdvsz ] the sfrd determined from the sum of the sfrd of the local ulig data ( @xmath207 ; @xmath276 ) from kim & sanders ( 1998 ) and the local near - ulig data ( @xmath277 ; @xmath278 ) from sanders et al . ( in preparation ) . the latter @xmath188 selection is imposed in order to be able to compare with our completeness - corrected submillimeter data points , which should include all sources with luminosities @xmath279 . we note that the radio data in fig . [ figsfrdvsz ] can not be straightforwardly compared to the ulig data because the radio sources may not satisfy the same luminosity criteria . the completeness - corrected sfrd in fig . [ figsfrdvsz ] shows a very rapid evolution , @xmath280 , in the sfrd of uligs from @xmath7 to @xmath8 . fast evolution is not surprising if the distant submillimeter sources are associated with major merger events giving rise to the formation of massive spheroidal systems ( smail et al . 1998 ; eales et al . 1999 ; lilly et al . 1999 ; trentham et al . 1999 ) . barger , cowie , & sanders ( 1999 ) showed that the volume density of submillimeter sources at high redshift ( @xmath281 mpc@xmath282 for @xmath283 ) is comparable to the volume density of present - day elliptical galaxies ( @xmath284 mpc@xmath282 ) . we have carried out an observational program designed to establish the overlap between the optical / nir - faint radio population and the submillimeter population and to exploit the complementary information that the radio and submillimeter wavelengths provide to gain insights into the evolution of the galaxy populations . our major conclusions are as follows @xmath285 we have found that submillimeter sources at bright flux levels ( @xmath286 mjy ) are associated with optical / nir - faint radio sources ( @xmath287jy ) . this association provides a powerful means to conduct submillimeter surveys by preselecting potential bright submillimeter sources through high - resolution deep radio maps . @xmath285 we have shown that the redshifted submillimeter - to - radio flux ratio of an unevolved arp 220 sed reproduces the ensemble of local ulig data placed at appropriate redshifts , and thus arp 220 provides a prototype for ulig sources at high redshift . from the arp 220 model we have derived a _ millimetric _ redshift estimator to determine the redshifts of the submillimeter sources using the ratio of the submillimeter to radio fluxes . our estimator is consistent with another recent study by carilli & yun ( 2000 ) . the flux strengths relative to the redshifted arp 220 sed indicate that the @xmath6 bright ( @xmath5 mjy ) submillimeter sources are @xmath288 the luminosity of arp 220 for @xmath283 and @xmath289 the luminosity of arp 220 for @xmath290-dominated models . @xmath285 through radio versus submillimeter flux plots of the arp 220 predictions with redshift , we have shown that our present survey is sensitive only to sources with strengths comparable to or greater than arp 220 and that our @xmath291jy radio threshold precludes the detection of very high redshift sources ( @xmath292 ) if they are radio sources ; however , we serendipitously observed two bright submillimeter sources without radio detections that may lie at extreme redshifts . an alternate possibility is that they could be colder galaxies than arp 220 located at @xmath293 . it remains to be established whether the very faint radio population down to @xmath294jy have submillimeter counterparts . @xmath285 at @xmath295 the ultraviolet / optical contribution to the sfrd dominates . at @xmath296 the submillimeter contribution from @xmath5 mjy sources is comparable to the observed ultraviolet / optical contribution . the ultraviolet / optical and bright submillimeter are likely sampling two different stages in galaxy formation . the bright submillimeter detects the formation of massive spheroids while the ultraviolet / optical detects the formation of smaller disk or bulge systems . joint radio / submillimeter surveys are a powerful way to explore dust - obscured galaxies in the distant universe since the radio and submillimeter approaches are complementary and allow us to infer redshifts and sfrds for galaxy sources that we now conclusively see are inaccessible in shorter wavelength observations . 0.5 cm we thank dave sanders for valuable discussions , nicolas biver for expert observing assistance , and an anonymous referee for helpful comments about the manuscript . support for this work was provided by nasa through hubble fellowship grants hf-01117.01 - 99a and hf-01123.01 - 99a awarded by the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , for nasa under contract nas 5 - 26555 . shaver , p.a . , hook , i.m . , jackson , c.a . , wall , j.v . , kellermann , k.i . 1998 , in highly redshifted radio lines , asp conf . series vol . 156 , eds . carilli , s.j.e . radford , k.m . menten , g.i . langston , p. 163

a transcendental entire function is called _ geometrically finite _ if the intersection of the set of singular values with the fatou set is compact and the intersection of the postsingular set with the julia set is finite . ( in particular , this includes all entire functions with finite postsingular set . ) if is geometrically finite , then is either empty or consists of the basins of attraction of finitely many attracting or parabolic cycles . let be a repelling or parabolic periodic point of such a map . we show that , if has finite order , then there exists an injective curve consisting of escaping points of that connects to . ( this curve is called a _ dynamic ray_. ) in fact , the assumption of finite order can be weakened considerably ; for example , it is sufficient to assume that can be written as a finite composition of finite - order functions .

0707.4135

in polynomial dynamics , _ dynamic rays _ , which foliate the set of escaping points ( points which tend to @xmath6 under iteration ) , were introduced by douady and hubbard as a tool in their famous work on the mandelbrot set @xcite . since then , dynamic rays , and their landing properties in particular , have been an essential ingredient in the success of polynomial dynamics . one of the fundamental results in this area , which goes back to douady , states that each repelling or parabolic ( pre)periodic point of a polynomial with connected julia set is the landing point of at least one ( pre)periodic dynamic ray ( * ? ? ? * theorem 18.11 ) . in the polynomial case , dynamic rays arise naturally as preimages of straight rays under the bttcher isomorphism at @xmath6 . when @xmath0 is a transcendental entire function , @xmath6 is no longer a superattracting fixed point , but rather an essential singularity . hence the _ escaping set _ @xmath7 is no longer open . nevertheless , it has long been known @xcite that for certain classes of entire functions there exist curves in the escaping set which can be seen as analogs of dynamic rays . we shall refer to these curves ( which are also often known as _ hairs _ ) themselves as dynamic rays of @xmath0 to stress the analogy to the polynomial case . ( see definition [ defn_ray ] for a formal definition of dynamic rays , and also below for the special case of _ periodic _ rays that is of main interest to us . ) recently rottenfuer , rckert , rempe and schleicher @xcite proved that , for an entire function that is a composition of finite - order entire functions with bounded sets of singular values , the escaping set consists of dynamic rays and ( some of ) their endpoints . this provides us with a large class of functions where we can study the topology of julia sets by looking at landing properties of dynamic rays . this approach to the study of julia sets and escaping sets has been used with great success in certain families of entire transcendental maps like the exponential family @xmath8 or the cosine family @xmath9 . to state the main question of interest to us , let us begin by formally defining what we mean by a _ periodic ( dynamic ) ray_. [ defn_periodic_rays ] let @xmath0 be an entire function . then a _ periodic ray _ of @xmath0 is an injective curve @xmath10 such that @xmath11 and such that there is an @xmath12 with @xmath13 for all @xmath14 . the minimal such @xmath15 is called the _ period _ of @xmath16 . as usual , we say that @xmath16 _ lands _ at @xmath5 if @xmath17 . a natural question suggested by douady s theorem and the results of @xcite is the following . suppose that @xmath0 is a finite - order entire function whose singular set is bounded , and suppose also that @xmath18 . is every repelling or parabolic periodic point of @xmath0 the landing point of a periodic ray of @xmath0 ? even for exponential and cosine maps , this question is still open . ( for a partial result on exponential maps , compare @xcite . ) however , schleicher and zimmer @xcite obtained a positive answer for exponential maps satisfying certain dynamical assumptions . in this paper , we generalize this statement , under similar conditions , to a much larger class of entire functions . for an entire transcendental function @xmath0 , @xmath1 denotes the set of singular values , @xmath3 the postsingular set , and @xmath2 and @xmath4 the fatou and julia sets of @xmath0 , respectively . ( for definitions , see section [ notations_section ] . ) [ def_gf ] a map @xmath0 is called _ geometrically finite _ if @xmath19 is compact and @xmath20 is finite . we can now state our main result . recall that @xmath0 has _ finite order _ if @xmath21 as @xmath22 . [ thm1 ] let @xmath0 be geometrically finite and assume that @xmath0 has finite order . then , for any repelling or parabolic periodic point @xmath23 of @xmath0 , there is a periodic ray landing at @xmath23 . in fact , the condition that @xmath0 is geometrically finite can be weakened ; what we need is that every iterate of @xmath0 has an `` admissible expansion domain '' at every repelling or parabolic fixed point ( see definition [ dfn_adm_exp_dom ] ) . furthermore , the assumption of finite order is not essential ; we require that `` all periodic rays exist '' . ( see theorem [ maintheorem ] for the precise statement . ) by @xcite , this condition is satisfied by all finite compositions of finite - order functions with bounded sets of singular values . our result implies in particular that each singular value in the julia set of a map @xmath0 to which the theorem applies is the landing point of some periodic dynamic ray . using these rays , one can define a dynamically natural partition of the julia set , as done for exponential and cosine maps in @xcite , which is useful for studying the topological dynamics of @xmath0 in combinatorial terms . we remark that , under more restrictive function - theoretic conditions , our result can be considerably strengthened . indeed , if @xmath24 is a cosine map for which both critical values are strictly preperiodic , then schleicher showed in @xcite that _ every _ point @xmath25 is either on a dynamic ray or the landing point of a dynamic ray . in @xcite we generalize this result to any finite - order _ subhyperbolic _ function @xmath0 such that @xmath4 contains no asymptotic values and the degree of critical points in @xmath4 is uniformly bounded . ( a subhyperbolic entire function is a geometrically finite map without parabolic cycles . ) furthermore , the julia set of such a function can be described as a `` pinched cantor bouquet '' in the sense of @xcite ( see also ( * ? ? ? * definition 1.2 ) ) . here the assumption that there are no asymptotic values in the julia set is essential : indeed , there is _ no _ exponential map @xmath26 for which the asymptotic value belongs to the julia set and every point of @xmath4 is either on a dynamic ray or the landing point of a dynamic ray @xcite . in the case of geometrically finite exponential maps , our theorem is due to schleicher and zimmer @xcite ( although in the case where the singular value is preperiodic some of the details are only sketched ) . this was extended to cosine maps with preperiodic critical values in @xcite . the general strategy of our proof , which we will now describe , follows the same idea as these papers . by passing to a suitable iterate , we can assume that the considered repelling or parabolic periodic point @xmath5 is a fixed point . the idea is to start with _ any _ given curve connecting @xmath5 to infinity , and pull back this curve using iterates of the map @xmath0 . using hyperbolic contraction arguments , we prove that this procedure yields only finitely many different curves up to homotopy . this then allows us to associate a combinatorial object ( a `` periodic external address '' ) to these curves . when we know from the existence theorems of @xcite that there exists a periodic ray corresponding to this address , it then easily follows that this ray lands at @xmath5 . since the `` ad - hoc '' method that was used to obtain hyperbolic contraction estimates in @xcite appears to be difficult to adapt to our more general setting , we develop a rather natural construction using hyperbolic geometry . our argument gives a more streamlined proof even in the established cases . we would like to emphasize that the case when @xmath3 is finite requires less technical constructions than the general setting . this is why we consider this special case separately in the proofs of some results required for the proof of theorem [ thm1 ] ( see section [ sec_3 ] ) . [ ackref ] i would like to thank especially my supervisor , lasse rempe , for his great help and support . i would also like to thank adam epstein , freddie exall and mary rees for helpful and interesting discussions . the complex plane , riemann sphere and the punctured plane are denoted by @xmath27 , @xmath28 and @xmath29 , respectively . we write @xmath30 for the unit disk and @xmath31 for the unit circle ; @xmath32 is the punctured disk . the euclidean distance between two sets @xmath33 is denoted by @xmath34 . the closure @xmath35 and the boundary @xmath36 of a set @xmath37 is always understood to be taken relative to the complex plane . let @xmath38 be an entire transcendental function . then the set @xmath1 of _ singular values _ of @xmath0 equals the closure of the set of all finite critical and asymptotic values of @xmath0 . if @xmath39 is an open set with @xmath40 , then the map @xmath41 is a covering map . denote by @xmath42 the @xmath15-th iterate of @xmath0 . then the _ postsingular set _ of @xmath0 is defined by @xmath43 . [ lem_u_hyp ] let @xmath0 be an entire transcendental function . then @xmath44 . assume first that @xmath0 has only one singular value , at @xmath45 say , since otherwise the claim already follows . for simplicity , assume @xmath46 . a well - known result that follows from covering space theory says that @xmath0 is of the form @xmath47 for some @xmath48 . in this case the asymptotic value @xmath46 is also an omitted value , so @xmath3 contains at least two points . the _ fatou set _ @xmath2 of @xmath0 is the set of all points @xmath25 so that @xmath49 forms a normal family in the sense of montel in a neighbourhood of @xmath23 . its complement @xmath50 is called the _ julia set _ of @xmath0 . recall from the introduction that the _ escaping set _ @xmath51 is the set of all those points @xmath25 such that @xmath52 when @xmath53 . we note that @xmath54 for all @xmath12 . to avoid confusion , we call a point @xmath25 _ preperiodic _ if some image @xmath55 , @xmath12 , of @xmath23 is periodic but not @xmath23 itself . we call a point _ eventually periodic _ if it is either preperiodic or periodic . let @xmath23 be a periodic point of @xmath0 of ( minimal ) period @xmath15 . we call @xmath56 the _ multiplier _ of @xmath23 . a periodic point @xmath23 is called _ attracting _ if @xmath57 , _ indifferent _ if @xmath58 and _ repelling _ if @xmath59 . an attracting periodic point @xmath23 is called _ superattracting _ if @xmath60 . we will denote the union of all attracting periodic points of @xmath0 by @xmath61 . since the multiplier of an indifferent periodic point is of the from @xmath62 with @xmath63 , we can distinguish between _ rationally _ and _ irrationally indifferent _ periodic points , according to whether @xmath14 is rational or not . a rationally indifferent periodic point is also called _ parabolic_. we denote the union of all parabolic cycles of @xmath0 by @xmath64 . an irrationally indifferent periodic point in the julia set is called a _ cremer point_. the set of all points whose orbits converge to an attracting periodic cycle is called the _ attracting basin _ of this cycle . likewise , the set of points whose orbits converge nontrivially to a parabolic cycle is called the _ parabolic basin _ of that cycle . every component of @xmath2 is either an _ eventually periodic domain _ or a _ wandering domain_. the only possible periodic fatou domains are _ immediate attracting basins _ , _ immediate parabolic basins _ , _ siegel disks _ and _ baker domains _ ( for a detailed explanation see ( * ? ? ? * theorem @xmath65 ) ) . if @xmath0 belongs to the _ eremenko - lyubich class _ @xmath66 then @xmath0 has no baker domains , each component of @xmath2 is simply - connected and @xmath67 ( ( * ? ? ? * proposition 3 , theorem 1 ) ) . for further background on holomorphic dynamics we refer the reader to @xcite and @xcite . a domain ( i.e. , open connected set ) @xmath68 is called _ hyperbolic _ if @xmath69 contains at least two points . we denote the density of the hyperbolic metric on @xmath70 ( i.e. , the unique complete conformal metric of constant curvature @xmath71 ) by @xmath72 . to each curve @xmath73 we assign the _ hyperbolic length _ @xmath74 of @xmath75 . for any two points @xmath76 the _ hyperbolic distance _ @xmath77 is the smallest hyperbolic length of a curve connecting @xmath23 and @xmath45 in @xmath70 . pick s theorem ( * ? ? ? * theorem 2.11 ) states that , any holomorphic map @xmath78 between two hyperbolic domains does not increase the respective hyperbolic metrics . in fact , it is a local isometry if and only if @xmath0 is a covering map ; otherwise @xmath0 is a strict contraction . in particular , if @xmath79 , then @xmath80 for all @xmath81 . we will use the following standard estimates on the hyperbolic metric of a hyperbolic domain @xmath68 ( * ? ? ? * corollary @xmath82 ) : @xmath83 where the inequality on the left - hand side only holds if @xmath70 is simply - connected . by the uniformization theorem ( * ? ? ? * theorem 2.1 ) , any hyperbolic domain @xmath70 is conformally isomorphic to a quotient of the form @xmath84 , where @xmath85 is a fuchsian group acting on @xmath30 . let @xmath86 be a universal covering map . a _ geodesic _ on @xmath70 is the image of a geodesic in @xmath30 , i.e. an arc of a circle orthogonal to @xmath87 . if @xmath16 is a geodesic connecting two points @xmath76 , then @xmath16 has minimal hyperbolic length among curves connecting @xmath23 and @xmath45 in the same homotopy class ( understood , as usual , relative @xmath88 ) . it will often be important to know that we can replace any curve by a geodesic in the same homotopy class . [ roy ] suppose that @xmath89\to\overline{u}$ ] is a curve with @xmath90 . then there exists a unique geodesic @xmath16 of @xmath70 that is homotopic to @xmath75 . suppose that @xmath45 is an isolated point of @xmath88 ; such a point is called a _ puncture_. by ( * ? ? ? * proposition 3.8.9 ) , there exists a covering map @xmath91 such that @xmath92 extends to a continuous map @xmath93 , sending @xmath94 to @xmath45 , and such that , for sufficiently small @xmath95 , the restriction @xmath96 is one - to - one . if @xmath97 has this property , then the simple closed curve @xmath98 is called a _ horocycle at @xmath45 _ ; the component @xmath99 of @xmath100 whose boundary is @xmath101 is called a _ horosphere at @xmath45_. the following result states that geodesics on hyperbolic domains will stay away from the punctures . [ jorg ] let @xmath70 be a hyperbolic domain and let @xmath45 be a puncture of @xmath70 . there exists @xmath95 such that each simple geodesic entering the horosphere @xmath99 ends at the point @xmath45 . next , we want to review the constructions of tracts and external addresses for a function @xmath0 with bounded postsingular set , following @xcite . let @xmath102 be any jordan domain containing @xmath3 and define @xmath103 and @xmath104 . then each component @xmath105 of @xmath39 is a simply - connected unbounded domain whose boundary @xmath106 is a jordan arc tending to @xmath6 at both ends . such a component @xmath105 is called a _ tract _ of the function @xmath0 ; the restriction @xmath107 is a universal covering ( * ? ? ? * theorem 1.1 ) . there can be only finitely many tracts having non - empty intersection with @xmath102 . next , we choose a curve @xmath108 not intersecting any tract @xmath105 , such that @xmath109 connects @xmath110 to @xmath6 . the preimage @xmath111 induces a partition of each tract , cutting it into countably many components called _ fundamental domains_. every such domain is mapped conformally to @xmath112 under @xmath0 . . ] if @xmath25 with @xmath113 for all @xmath114 , then the _ external address _ of @xmath23 is the sequence @xmath115 of fundamental domains defined by @xmath116 . note that the fact that @xmath117 for any tract @xmath105 of @xmath0 guarantees that @xmath55 does indeed belong to a ( unique ) fundamental domain . if @xmath118 then there exists an integer @xmath119 such that @xmath120 for all @xmath121 . let @xmath122 denote the one - sided shift operator , i.e. @xmath123 . we say that @xmath124 is _ periodic _ if @xmath125 for some @xmath12 and _ preperiodic _ if some image @xmath126 of @xmath124 , @xmath12 , is periodic but not @xmath124 itself . [ defn_ray ] a _ ray tail _ of @xmath0 is an injective curve @xmath127 ( where @xmath128 ) such that for each @xmath129 , @xmath130 and such that , as @xmath131 , @xmath132 uniformly in @xmath14 . a _ dynamic ray _ of @xmath0 is then a maximal injective curve @xmath133 such that @xmath134 is a ray tail for every @xmath128 . if @xmath16 is a dynamic ray , then there exists @xmath128 such that , for each @xmath114 , the curve @xmath135 is contained in a fundamental domain @xmath136 . the sequence @xmath137 is called the _ external address of @xmath16_. it is not difficult to check that any periodic ray , as defined in definition [ defn_periodic_rays ] , is indeed a dynamic ray in the sense of definition [ defn_ray ] with a periodic external address . conversely , any dynamic ray @xmath16 with @xmath138 is a periodic ray in the sense of definition [ defn_periodic_rays ] , after a suitable reparametrization . it is less obvious that any dynamic ray with a periodic external address is also itself periodic . however , if there were two dynamic rays , say @xmath139 and @xmath140 , with the same external address , it would follow from the proof of ( * ? ? ? * corollary 3.4 ) that @xmath139 is a subset of @xmath140 or the other way around ( the stated reference implies only that @xmath139 and @xmath140 intersect but the proof shows more , namely that one of the two rays would have to be a subset of the other one ) . it would then follow from ( * ? ? ? * lemma 3.3 ) that @xmath139 equals @xmath140 . set @xmath141 recall that by definition [ def_gf ] , @xmath0 is called geometrically finite if @xmath142 is compact and @xmath143 is finite . note that every such map belongs to the class @xmath144 . furthermore , since @xmath145 and @xmath146 , it follows that every iterate of a geometrically finite map is again geometrically finite . following mcmullen @xcite , a rational function @xmath147 is called geometrically finite if @xmath148 is finite . using classical results on dynamics of rational maps one can easily deduce that the fatou set of such a map is the union of finitely many attracting and parabolic basins ( * ? ? ? * chapter @xmath65 ) . the following statement shows that the same holds for a geometrically finite entire transcendental map . [ poss_f_comp ] let @xmath0 be a geometrically finite entire transcendental function . then the fatou set of @xmath0 is either empty or consists of finitely many attracting and parabolic basins . furthermore , every periodic cycle in the julia set is repelling or parabolic . first note that @xmath0 can not have wandering domains . indeed , if @xmath149 was a wandering domain , then all limits of orbits of points in @xmath149 would belong to @xmath143 ( * ? ? ? * theorem ) , and hence the iterates in @xmath149 would converge locally uniformly to a single periodic orbit in @xmath143 . this orbit clearly can not be repelling or parabolic . by a result of perez - marco @xcite , this orbit also can not be irrationally indifferent . hence @xmath0 has no wandering domains . additionally , if @xmath0 had a siegel disk , then its boundary would be contained in @xmath3 ( * ? ? ? * theorem 7 ) . this is again impossible because @xmath143 is finite , so @xmath0 has no siegel disks . thus the fatou set is the union of attracting and parabolic basins . the set of attracting and parabolic basins forms an open cover of the compact set @xmath19 . hence there exist finitely many attracting and parabolic basins that cover this set . on the other hand , every attracting or parabolic basin must contain at least one point of @xmath1 ( * ? ? ? * theorem 7 ) . this proves the first claim . furthermore , if @xmath5 was a cremer point of @xmath0 , then there would be a sequence @xmath150 of points in @xmath3 converging nontrivially to @xmath5 ( * ? ? ? * corollary 14.4 ) . since @xmath143 is finite and @xmath151 consists of finitely many parabolic cycles , this is also impossible . in particular , if @xmath0 is geometrically finite , then @xmath3 is bounded . recall that @xmath0 is called _ subhyperbolic _ if it is geometrically finite and has no parabolic cycles . note that an entire function is subhyperbolic if and only if @xmath152 is compact and @xmath143 is finite . an entire function is called _ postsingularly finite _ if @xmath3 is finite . the fatou set of such a map is either empty or the union of finitely many superattracting basins . clearly , every postsingularly finite map is subhyperbolic , and in particular geometrically finite . to prove our main theorem we will use a hyperbolic domain @xmath70 such that our function @xmath0 is expanding with respect to the hyperbolic metric of @xmath70 , and such that @xmath70 has simple topology . our requirements are formalized in the following definition . [ dfn_adm_exp_dom ] let @xmath0 be an entire transcendental function and let @xmath5 be a fixed point of @xmath0 which is either repelling or parabolic . a domain @xmath153 is called an _ admissible expansion domain _ of @xmath0 at @xmath5 , if the following properties hold : * @xmath154 and @xmath6 is an isolated boundary point of @xmath70 . * @xmath155 . * @xmath70 is finitely - connected . furthermore , @xmath156 . * the point @xmath157 is accessible from @xmath70 . if @xmath158 is the component of @xmath69 containing @xmath5 , then @xmath159 has finitely many components . note that every map for which there is an admissible expansion domain at some repelling or parabolic fixed point must have a bounded postsingular set . furthermore , it follows from definition [ dfn_adm_exp_dom ] and ( * ? ? ? * theorem 7 ) that , if @xmath160 is a puncture of @xmath70 , then @xmath5 is a repelling fixed point of @xmath0 . we will show later that , if @xmath3 is finite , then we can choose @xmath69 to be finite as well . also note that if @xmath70 is an admissible expansion domain of @xmath0 at @xmath5 , then @xmath70 is also an admissible expansion domain of @xmath42 at @xmath5 . let @xmath70 be an admissible expansion domain of @xmath0 . observe that by lemma [ lem_u_hyp ] , @xmath70 is hyperbolic . by definition , there exists at least one point @xmath161 . note that @xmath45 has has infinitely many preimages under @xmath0 ; otherwise , @xmath45 would be a ( picard ) exceptional value but every such point is also an asymptotic value of @xmath0 ( * ? ? ? * chapter 5 , theorem 1.1 ) . throughout section [ subs_l ] and [ subsec_33 ] we will assume that @xmath0 is an entire transcendental map and @xmath5 is a repelling or parabolic fixed point of @xmath0 with an admissible expansion domain , denoted by @xmath70 . as mentioned in the introduction , we will prove that theorem [ thm1 ] holds when we replace the condition that @xmath0 is geometrically finite by the property that every iterate of @xmath0 has an admissible expansion domain at every repelling or parabolic fixed point . it will become clear that the existence of admissible expansion domains is what is essential for our idea to work . we will show in section [ subs_aed_gf ] that _ every _ iterate of a geometrically finite map has an admissible expansion domain at _ any _ of its repelling or parabolic fixed points ; we will also give an example of a map which is not geometrically finite but to which our methods still apply . we note that we do not require that @xmath0 has finite order . indeed , we will prove a general combinatorial statement ( theorem [ thm_finite ] ) , requiring only our standing assumption . theorem [ thm1 ] will then be deduced by applying this result to an iterate of the original geometrically finite function of finite order . [ def_spider ] a _ leg _ is an injective curve @xmath89\rightarrow u\cup\lbrace z_0 , \infty\rbrace$ ] such that * @xmath163 , * @xmath164 and @xmath165 . two legs @xmath166 and @xmath167 are called _ equivalent _ ( @xmath168 ) if they are homotopic in @xmath70 relative to the set of endpoints @xmath169 . for a leg @xmath75 we will denote its equivalence class by @xmath170 $ ] . by assumption , @xmath5 is not a critical point of @xmath0 , so every leg ending at @xmath5 has a unique preimage curve ending at @xmath5 and this is again a leg . the map which assigns such a pullback to each leg @xmath75 will be called the _ leg map _ and denoted by @xmath162 . as usual , we will denote the @xmath15-th iterate of @xmath162 by @xmath171 . it follows from the homotopy lifting property that if @xmath168 , then this also holds for their images , i.e. @xmath172 . hence , _ the leg map @xmath162 descends to a map on the set of equivalence classes of legs_. we will often replace pieces of arbitrary legs by pieces of geodesics in their homotopy classes , which is possible by proposition [ roy ] . we will call a leg that is a geodesic with respect to the hyperbolic metric on @xmath70 a _ geodesic leg_. we are now able to formulate the main result of section [ sec_3 ] . [ thm_finite ] let @xmath75 be a leg . then there exist integers @xmath173 and @xmath15 such that @xmath174 . the proof of this theorem will be given at the end of section [ subsec_33 ] . from now on , let us denote the density of the hyperbolic metric on the admissible expansion domain @xmath70 of @xmath0 at @xmath5 by @xmath175 . we define @xmath176 note that @xmath177 does not have to be connected . every component @xmath178 of @xmath177 is again a hyperbolic domain with corresponding density map @xmath179 . if @xmath81 , then @xmath23 lies in a unique component @xmath178 and for simplicity , we will denote the density of the hyperbolic metric at @xmath23 by @xmath180 . [ prop_y ] let @xmath181 be a compact connected set with finitely many boundary components . then there exists a constant @xmath182 such that @xmath183 for any compact subset of @xmath184 there is certainly such a constant @xmath185 , since @xmath186 has only finitely many components . hence we have to consider only sufficiently large points @xmath187 . let @xmath188 be a non - exceptional value of @xmath0 . then @xmath45 has infinitely many preimages under @xmath0 and all but finitely many of them are contained in @xmath189 . there exists a sequence @xmath190 and a constant @xmath191 such that @xmath192 and @xmath193 holds for all @xmath194 . a sketch of a proof can be found in ( * ? * proof of lemma 5.1 ) . for completeness we will elaborate the arguments given in @xcite . let @xmath195 be a jordan curve , such that the bounded component of @xmath196 contains @xmath1 but not @xmath45 , and let @xmath197 denote the unbounded component of @xmath196 . then @xmath198 is a countable union of tracts @xmath199 and @xmath200 is a universal covering for every @xmath201 . let us pick a tract @xmath202 . since @xmath203 , there is an infinite sequence @xmath204 of preimages of @xmath45 in @xmath202 , such that the distance @xmath205 measured in the hyperbolic metric of @xmath202 is constant . we can assume w.l.o.g . that @xmath206 , and so by equation [ hyp_est ] we obtain @xmath207 let @xmath208 . it follows that @xmath209 where @xmath210 \rightarrow t_0 $ ] is any rectifiable curve that connects @xmath211 and @xmath212 . hence @xmath213 the claim now follows with @xmath214 . recall that @xmath70 contains a punctured disk at @xmath6 , hence @xmath215 on the other hand , @xmath216 , and it follows from ( * ? ? ? * proposition 2.1 ) that @xmath217 hence @xmath218 as @xmath22 and the statement follows . [ lindisk ] assume that @xmath5 is an isolated boundary point of @xmath70 . then for every horosphere @xmath219 there exists a horosphere @xmath220 such that @xmath221 and @xmath222 . furthermore , @xmath223 can be replaced by any @xmath224 . first recall that @xmath5 is neccessarily a repelling fixed point of @xmath0 . the first statement is obvious since @xmath225 holds for any horosphere @xmath219 and since @xmath5 is a fixed point of @xmath0 . there exist a covering map @xmath91 and a constant @xmath226 such that @xmath92 maps @xmath227 one - to - one to the horosphere @xmath228 at @xmath5 . by the riemann removable singularity theorem , the embedding @xmath229 can be continued holomorphically to @xmath94 . for any @xmath230 let @xmath231 , where @xmath232 , and denote by @xmath233 and @xmath234 its inner and outer radius , respectively . clearly , @xmath235 as @xmath236 as well as @xmath237 by composing @xmath0 with a linear transformation , we can assume that @xmath238 , so the power series of the function @xmath0 has the form @xmath239 in a neighbourhood of @xmath94 , where @xmath240 is the multiplier of @xmath94 . let @xmath241 . then @xmath242 for every sufficiently small @xmath223 . hence every point @xmath243 is mapped outside the circle at @xmath94 with radius @xmath234 and the statement follows . recall that our goal in section [ sec_3 ] is to prove theorem [ thm_finite ] which states that the iteration of the leg map produces only finitely many equivalence classes of legs . since equivalence classes of legs arise , roughly speaking , by winding around components of @xmath88 , we want to find a compact subset @xmath244 of @xmath70 so that producing additional homotopy implies increase of length of leg - pieces contained in @xmath244 . by choosing @xmath244 so that proposition [ prop_y ] applies , we can later use uniform contraction arguments to control the lengths of the considered pieces of legs . note that if @xmath75 is _ any _ leg , it is fairly impossible to make useful statements about the location of its iterated images @xmath245 related to an _ arbitrary _ compact set @xmath244 . only by constructing @xmath244 carefully using hyperbolic geometry and working with geodesic legs rather than arbitrary legs , we obtain additional tools that enable us to control the lengths of geodesic leg - pieces . [ thm_y ] there exists a compact path - connected set @xmath181 with finitely many boundary components such that : * if @xmath16 is a geodesic leg , then @xmath246 is non - empty and connected . furthermore , if @xmath247 and @xmath248 are distinct components of @xmath249 , then @xmath247 and @xmath248 are contained in two distinct components of @xmath250 . * let @xmath251 and @xmath252 denote the components of @xmath253 that contain @xmath5 and @xmath6 , respectively , as boundary points , and for a leg @xmath75 , denote by @xmath254 the hyperbolic length in @xmath70 of the longest subpiece of @xmath75 connecting the boundaries of @xmath251 and @xmath252 in @xmath70 . then there exists a constant @xmath255 , such that if @xmath16 is a geodesic leg and @xmath256 $ ] is another leg , then @xmath257 . * there exists a constant @xmath258 , such that if @xmath16 is a geodesic leg , then there exists a leg @xmath259 $ ] with @xmath260 . let @xmath261 be the punctures of @xmath70 including @xmath6 , so let us assume that @xmath262 . for every @xmath263 choose a sufficiently small horosphere @xmath264 which satisfies the conclusion of proposition [ jorg ] , and such that @xmath265 whenever @xmath266 . recall from section [ subs_tracts ] that @xmath267 is a countable union of tracts @xmath199 , and any compact subset of @xmath70 can intersect only finitely many tracts . if @xmath5 is one of the punctures , say @xmath268 , we also require that @xmath269 satisfies the conclusion of proposition [ lindisk ] . define @xmath270 case i : _ @xmath69 is finite _ let @xmath271 . clearly , @xmath244 is a compact path - connected set with finitely many boundary components . @xmath272 : if @xmath16 is a geodesic leg , then @xmath16 does not intersect @xmath273 for all @xmath274 , while it intersects @xmath275 and @xmath276 exactly once ( * ? ? ? * proposition 3.3.9 ) , so in particular @xmath246 is non - empty and connected . since every puncture of @xmath70 belongs to a unique component of @xmath250 , statement @xmath272 follows . @xmath277 : observe that among all curves in a given homotopy class which connect the two horocycles @xmath275 and @xmath276 , the unique geodesic realizes the smallest distance , hence the claim follows with @xmath278 . @xmath279 : if @xmath16 is any geodesic leg , then , by @xmath272 , @xmath16 intersects @xmath280 and @xmath281 exactly once , while it does not enter any other horosphere . also recall that by proposition [ lindisk ] , the inverse branch of @xmath0 that maps @xmath5 to itself maps @xmath251 into itself . hence the only components of @xmath282 that might have non - empty intersection with @xmath283 are domains that arise as intersection of @xmath284 and a tract @xmath105 ( a component of @xmath267 ) , that eventually contains @xmath285 . observe that @xmath285 intersects @xmath106 in exactly one point , say @xmath286 , while it is possible that @xmath285 has more than one intersection point with @xmath276 lying in @xmath105 ( see figure [ bild5](a ) ) . let @xmath287 be the last intersection point of @xmath285 and @xmath276 . if @xmath288 , then the longest subpiece of @xmath285 connecting @xmath280 and @xmath276 is itself a subpiece of @xmath289 and the claim follows with @xmath290 and @xmath291 . otherwise , let @xmath292 denote the subpiece of @xmath285 connecting @xmath211 and @xmath212 . clearly , @xmath293 . since @xmath294 is a geodesic in @xmath295 , it follows that @xmath296 is a geodesic in @xmath105 , hence @xmath292 is a subpiece of a geodesic in @xmath105 connecting @xmath211 to @xmath6 . recall that @xmath106 is an analytic curve , hence @xmath297 depends continuously on the point @xmath211 . furthermore , the set of those points @xmath298 for which the geodesic in @xmath105 from @xmath45 to @xmath6 intersects @xmath276 is a compact subset of @xmath70 . together with the fact that only finitely many tracts intersect the set @xmath244 , this implies that there is a finite number @xmath299 such that @xmath300 . hence @xmath301 . ( 200,150 ) of a geodesic leg @xmath16 : @xmath285 intersects @xmath106 exactly once , while it can have more than one intersection point with @xmath276 , ( b ) : the point @xmath5 is a non - isolated boundary point contained in the component @xmath158 of @xmath69 : the separating curves @xmath302 intersect only in @xmath5.,title="fig : " ] ( 200,150 ) of a geodesic leg @xmath16 : @xmath285 intersects @xmath106 exactly once , while it can have more than one intersection point with @xmath276 , ( b ) : the point @xmath5 is a non - isolated boundary point contained in the component @xmath158 of @xmath69 : the separating curves @xmath302 intersect only in @xmath5.,title="fig : " ] case ii : _ @xmath69 is infinite _ it follows from definition [ dfn_adm_exp_dom]@xmath303 that @xmath69 is a finite union of compact sets . let @xmath304 be those components of @xmath69 that are not punctures of @xmath70 . now let @xmath305 be a component such that @xmath306 . by the plane separation theorem ( * ? * chapter vi , theorem 3.1 ) there exists a simple closed curve @xmath307 , entirely contained in @xmath70 , which separates @xmath305 from any other component of @xmath249 . by ( * proposition 3.3.8 , proposition 3.3.9 ) , there is a unique geodesic @xmath308 which is a simple closed curve homotopic to @xmath307 . denote by @xmath309 the component of @xmath310 whose boundary consists of @xmath311 . let @xmath312 where the union is taken over all components @xmath305 of @xmath69 such that @xmath313 . by ( * ? ? ? * proposition 3.3.9 ) any two geodesics @xmath314 are disjoint and by our initial choice , any two horocycles @xmath315 or geodesics @xmath308 are disjoint as well . hence the obtained set @xmath316 is a subset of @xmath70 with finitely many boundary components , each of which is either a horocycle @xmath315 or a simple closed geodesic @xmath308 . case iia : _ @xmath5 is a puncture of @xmath70 _ define @xmath317 . by construction , @xmath244 is a compact and path - connected set with finitely many boundary components . @xmath272 : let @xmath16 be a geodesic leg . since the boundary of @xmath244 consists of geodesics and horocycles and since @xmath16 intersects only the horocycles at @xmath5 and @xmath6 , it follows that @xmath246 is connected . furthermore , it follows from the previous construction that every component of @xmath186 surrounds exactly one component of @xmath88 , hence @xmath272 follows . @xmath277-@xmath279 : these statements follow by exactly the same arguments as in case i. case iib : _ @xmath5 is not a puncture of @xmath70 _ let @xmath158 be the component of @xmath69 that contains @xmath5 . by definition [ dfn_adm_exp_dom]@xmath318 , @xmath319 has finitely many components , say @xmath320 . it follows from the plane separation theorem that for every @xmath321 there is a simple closed curve @xmath322 that separates @xmath323 from every component of @xmath249 other than @xmath158 , as well as from every @xmath324 , where @xmath266 . furthermore , @xmath325 ( see figure [ bild5](b ) ) . each @xmath302 is homotopic relative the start- and endpoint @xmath5 to a unique geodesic @xmath326 in @xmath70 , and any two such geodesics @xmath326 and @xmath327 intersect only in @xmath5 . for every @xmath328 let @xmath329 be the component of @xmath330 bounded by @xmath331 , @xmath326 and @xmath332 ( see figure [ bild6](a ) ) , and let @xmath333 it follows that @xmath334 and that @xmath5 is accessible through exactly @xmath335 sectors , each of which lies between two geodesics @xmath326 and @xmath336 ( modulo @xmath335 ) . let @xmath337 , @xmath328 , be a collection of simple geodesic arcs , each of which connects a point in @xmath326 to a point in @xmath336 , such that the domain @xmath338 bounded by @xmath337 , @xmath332 , @xmath326 and @xmath336 is a simply - connected subdomain of @xmath70 ( see figure [ bild6](b ) ) . define @xmath339 @xmath272 : the statement follows by exactly the same arguments as in case iia . @xmath277 : let @xmath251 denote the unique component of @xmath340 that contains @xmath5 and set @xmath341 . now , @xmath246 does not necessarily realize the shortest distance between @xmath280 and @xmath276 in its homotopy class ; still , if @xmath75 is a leg in @xmath342 $ ] , then there exists a component @xmath343 of @xmath344 which is homotopic to @xmath246 relative @xmath345 and we obtain @xmath346 . @xmath279 : recall that there is a unique @xmath347 such that @xmath16 itersects @xmath337 ; their intersection point , say @xmath348 , is unique and the piece of @xmath16 connecting @xmath5 and @xmath348 is entirely contained in @xmath338 . let @xmath349 be the component of @xmath350 such that @xmath285 intersects @xmath351 and let @xmath352 be their unique intersection point . observe that the piece of @xmath285 that connects @xmath5 and @xmath352 is entirely contained in @xmath349 . if @xmath353 , then the situation is reduced to the previous case and we can choose @xmath290 . otherwise , we replace the subpiece of @xmath285 that connects @xmath5 and @xmath352 by the unique homotopic geodesic @xmath354 of the hyperbolic domain @xmath349 connecting @xmath5 and @xmath352 . if @xmath355 , then we are only interested in the longest piece of @xmath354 connecting two points in @xmath280 , and otherwise in the longest piece of @xmath354 that connects @xmath280 and @xmath352 . again , by continuity and compactness arguments ( as in the case of intersections with tracts ) , it follows that the length in @xmath70 of every such piece is globally bounded . the claim now follows from the fact that there are only finitely many domains @xmath349 . ( 200,150 ) is bounded by @xmath356 , ( b ) : geodesics @xmath357 and @xmath358 , one for each sector through which @xmath5 is accessible.,title="fig : " ] ( 200,150 ) is bounded by @xmath356 , ( b ) : geodesics @xmath357 and @xmath358 , one for each sector through which @xmath5 is accessible.,title="fig : " ] let @xmath244 be a compact set that satisfies the conclusions of theorem [ thm_y ] . by proposition [ prop_y ] there exists a constant @xmath182 such that @xmath359 holds for all @xmath360 . hence if @xmath361 is any rectifiable curve , then @xmath362 . let @xmath16 be a geodesic leg . by theorem [ thm_y]@xmath279 there exists a universal constant @xmath363 and a leg @xmath364 $ ] such that @xmath260 . together with the uniform contraction this yields the estimate @xmath365 let @xmath366 $ ] be a geodesic leg . by theorem [ thm_y]@xmath272 , @xmath367 is connected , hence @xmath368 . it then follows from theorem [ thm_y]@xmath277 that there exists a universal constant @xmath369 such that @xmath370 . altogether , we obtain @xmath371 by proceeding inductively it follows that if @xmath372 $ ] is the geodesic leg , then @xmath373 so in particular , if @xmath374 and @xmath375 , then @xmath376 . recall that by theorem [ thm_y ] every component of @xmath250 contains exactly one component of @xmath249 , hence there can be only finitely many geodesic legs with globally bounded length . so for all @xmath129 , the geodesic legs @xmath377 belong to only finitely many equivalence classes . we will now show that geometrically finite maps have admissible expansion domains . such maps provide us with many examples to which our main result applies . still , there are functions that admit expansion domains but are not geometrically finite ; an example will be given at the end of this section . [ prop_u ] let @xmath0 be a geometrically finite map and let @xmath5 be an arbitrary but fixed repelling or parabolic fixed point of @xmath0 . then @xmath0 has an admissible expansion domain @xmath70 at @xmath5 . furthermore , if @xmath0 is postsingularly finite , then @xmath70 can be chosen such that @xmath69 is finite . @xmath3 is finite _ recall that @xmath0 can not have any parabolic cycles . if @xmath378 , then define @xmath379 . otherwise , since every function in class @xmath144 has infinitely many repelling fixed points @xcite , there is a repelling fixed point @xmath45 of @xmath0 that belongs to @xmath380 . in this case define @xmath381 . in both cases there is a point in @xmath382 that has preimages arbitrarily close to @xmath6 . the set @xmath70 is open , connected and finitely - connected , @xmath6 is an isolated boundary point and @xmath69 is a finite union of points . since @xmath383 , it follows that @xmath384 . but the set @xmath385 is not compact , hence it follows that @xmath386 and @xmath155 . furthermore , @xmath5 is an isolated boundary point and hence accessible from @xmath70 . case ii : _ @xmath3 is infinite _ recall that in this case @xmath387 . there exists a full set @xmath388 whose boundary has only finitely many components such that @xmath389 and @xmath390 . denote by @xmath391 the set of those singular values of @xmath0 that are attracted by a cycle in @xmath61 . for every point in @xmath61 we choose a linearising neighbourhood ; let @xmath392 be their union . so @xmath392 is a finite union of jordan domains and it satisfies @xmath393 and @xmath394 . since @xmath391 is a compact subset of the fatou set , we have that @xmath395 . for each point @xmath396 we pick a euclidean disk @xmath397 of radius less than @xmath398 centered at @xmath23 . the union of these disks forms an open cover of @xmath391 and since the set is compact , there is a subcover consisting of finitely many disks , say @xmath399 . let @xmath400 . each of the disks @xmath401 is mapped into the set @xmath392 after finitely many iterations , hence there is an integer @xmath402 such that @xmath403 . define @xmath404 to be the union of the sets @xmath35 and @xmath405 . denote by @xmath406 the set of those singular values of @xmath0 that are attracted by a cycle in @xmath64 and let @xmath407 . for simplicity let us assume that @xmath408 is a fixed point of multiplicity @xmath409 , where @xmath410 ; the periodic case is analogous . by the parabolic flower theorem ( * ? ? ? * theorem 10.7 ) there are @xmath15 ( attracting ) petals of arbitrarily small diameter attached to @xmath408 . let @xmath411 be such a collection of petals at @xmath408 . then there is a positive number @xmath223 so that if @xmath412 holds for all @xmath15 , then @xmath23 is contained in some @xmath413 ( * ? ? ? * chapter @xmath414 , @xmath415 ) . denote by @xmath416 the set of all points in @xmath406 that converge to @xmath408 under iteration . for each point @xmath417 we choose a euclidean disk @xmath397 centered at @xmath23 with sufficiently small radius such that its closure is contained in the fatou set . as before , the union of these disks is a cover of @xmath416 and it has a subcover consisting of finitely many disks , say @xmath418 . let @xmath419 . there is an integer @xmath420 such that @xmath421 is contained in the @xmath223-neighbourhood of @xmath408 for all @xmath422 . we now define the set @xmath423 to be the union of @xmath424 and @xmath425 , and the set @xmath426 to be the union of all sets @xmath423 , where @xmath407 . observe that @xmath427 is not necessarily connected . define @xmath388 to be the union of the set @xmath428 and the bounded components of its complement . since @xmath429 , it follows that @xmath430 . let @xmath431 be a component of the interior of @xmath388 . then @xmath431 and @xmath432 are bounded and so by the open mapping theorem , @xmath433 . hence @xmath434 and by montel s theorem , @xmath435 is a normal family . it follows that the interior of @xmath388 is contained in the fatou set . furthermore , it follows from our construction that @xmath436 , @xmath390 and that @xmath388 has only finitely many boundary components , yielding the statement of the claim . we define @xmath437 since @xmath70 is the complement of a full set , it follows that @xmath70 is a domain with @xmath6 as an isolated boundary point . also , @xmath156 since @xmath438 . furthermore , @xmath439 , hence @xmath155 . since @xmath440 has finitely many components and since @xmath143 is a finite set , it follows that @xmath70 is finitely - connected . if @xmath5 is repelling , then it is a puncture of @xmath70 and in particular an accessible boundary point . if @xmath5 is parabolic , then it belongs to a non - trivial component @xmath158 of @xmath388 , which is disjoint from the repulsion vectors at @xmath5 , hence @xmath5 is accessible ( * ? ? ? * lemma @xmath441 ) . by construction , @xmath158 is a union of finitely many petals at @xmath5 , hence @xmath442 has finitely many components . since every iterate of a geometrically finite map is again geometrically finite , we immediately obtain the following statement . [ cor_u ] let @xmath0 be a geometrically finite map and let @xmath42 be an iterate of @xmath0 . then for every repelling or parabolic fixed point @xmath443 of @xmath42 , there is an admissible expansion domain of @xmath42 at @xmath443 . the map @xmath444 was introduced in @xcite as an example of an entire transcendental function that has a completely invariant fatou component @xmath177 , which contains an indirect singularity in its boundary . we will state here some of the properites of @xmath0 ; for more details see @xcite . the map @xmath0 has infinitely many critical values , all of which are contained in a closed interval @xmath445\subset [ 0,\infty)$ ] , and which accumulate at the asymptotic value @xmath94 . furthermore , @xmath94 is a parabolic fixed point of multiplier @xmath414 and @xmath446 is contained in its basin of attraction @xmath177 . so in particular , @xmath19 is not a compact set and hence @xmath0 is not geometrically finite . since @xmath0 maps @xmath447 into itself and since every singular value of @xmath0 converges to @xmath94 , there is a compact intervall @xmath448 $ ] that contains @xmath3 and is mapped by @xmath0 into itself . it follows that if @xmath5 is any repelling or parabolic fixed point of @xmath0 , then the domain @xmath449)$ ] is an admissible expansion domain of @xmath0 at @xmath5 . moreover , if @xmath443 is a repelling or parabolic fixed point of an iterate @xmath42 , then @xmath450)$ ] is an admissible expansion domain of @xmath42 at @xmath443 . more generally , let @xmath451 . there exists a real number @xmath452 such that for all @xmath453 the map @xmath454 has an attracting fixed point @xmath455 whose basin of attraction @xmath456 contains @xmath446 . since for every such @xmath109 the map @xmath454 has a repelling fixed point at @xmath94 , it follows that @xmath457 and so again , @xmath454 is not a geometrically finite map . without remarkable differences to the previous case , we can construct admissible expansion domains for every iterate @xmath458 at any of its repelling or parabolic fixed points . this section is devoted to the proof of our main result ; together with corollary [ cor_u ] and the results from @xcite it will imply theorem [ thm1 ] stated at the beginning of the article ( see also corollary [ cor_fin_comp ] ) . [ maintheorem ] let @xmath0 be an entire transcendental function , @xmath5 a repelling or parabolic fixed point of @xmath0 and assume that there is an admissible expansion domain @xmath70 of @xmath0 at @xmath5 . if for any periodic external address @xmath124 there exists a periodic ray of @xmath0 with address @xmath124 , then there is a periodic ray of @xmath0 landing at @xmath5 . let us pick a horosphere @xmath459 in the admissible expansion domain @xmath70 . recall that the preimage of @xmath459 under @xmath0 is a countable union of tracts , which we will denote by @xmath199 . moreover , each tract @xmath199 can be split into fundamental domains , depending on the choice of a curve that connects @xmath276 to @xmath6 without intersecting any of the tracts . let us fix such a selection of fundamental domains @xmath460 . it is necessary to give an idea of how to define an external address , which is respected by homotopies , for a leg @xmath75 . so let @xmath75 be any leg . note that @xmath75 does not even need to intersect a tract . on the other hand @xmath461 is eventually contained in a tract and @xmath462 is eventually contained in a fundamental domain . the application of this procedure to any other leg in @xmath170 $ ] leads to the same fundamental domain . let @xmath16 be a geodesic leg . by theorem [ thm_finite ] the equivalence class @xmath342 $ ] is eventually periodic . since @xmath70 is also an admissible expansion domain of every @xmath42 , we can assume , by passing to a suitable iterate , that @xmath342 $ ] is actually fixed . we can also assume that @xmath16 is eventually contained in a fundamental domain , say @xmath463 , and so are its images , since by the previous discussion this is true for all @xmath464 with @xmath465 . hence we can assign to @xmath16 the fixed external address @xmath466 by assumption , there exists a periodic ray @xmath467 with address @xmath124 , hence there is a constant @xmath468 such that @xmath469 for all @xmath470 . there is also a constant @xmath471 so that @xmath472 for all @xmath473 . let @xmath474 . we homotope @xmath16 to a leg @xmath475 $ ] by keeping @xmath476}$ ] fixed , such that @xmath477 and @xmath478 holds for all @xmath479 . note that this is always possible since every fundamental domain is a simply connected subset of @xmath70 . now , the tails @xmath480 and @xmath481 are both entirely contained in the same fundamental domain @xmath463 , hence we can replace @xmath480 by the ray tail @xmath482 , without changing the equivalence class . when we apply @xmath162 to the tails of @xmath16 and @xmath483 , then the resulting curves approach @xmath6 through the same fundamental domain @xmath463 , and the same holds for the following iterates . hence , after replacing a tail of @xmath16 by a tail of the dynamic ray and applying @xmath162 we again obtain a leg eventually contained in @xmath463 . now , we want to show that in the limit , the iteration of @xmath162 on such a leg yields a dynamic ray that lands at @xmath5 . let @xmath484 be a point on @xmath485 close to @xmath5 . the sequence of iterated images of @xmath486 under the leg map ( more precisely , under the corresponding inverse branch of @xmath0 ) will converge to the point @xmath5 . on the other hand , it follows from pick s theorem that the hyperbolic length of @xmath487 decreases . hence the sequence @xmath488 also converges to @xmath5 , which means that @xmath483 lands at @xmath5 . let @xmath5 be an arbitrary but fixed repelling or parabolic periodic point of @xmath0 of period @xmath15 . then @xmath5 is a repelling or parabolic fixed point of @xmath42 and by corollary [ cor_u ] @xmath42 has an admissible expansion domain at @xmath5 . note that @xmath42 also can be written as a finite composition of maps in class @xmath144 with finite order . it follows from ( * ? ? ? * theorem 2.4 ) and ( * ? ? ? * proposition 4.5 , theorem 4.2 ) that for every periodic external address there is a corresponding periodic ray of @xmath42 . theorem [ maintheorem ] finally implies that @xmath5 is the landing point of a periodic ray of @xmath42 . since every periodic ray of @xmath42 is also a periodic ray of @xmath0 and since @xmath5 was an arbitrary repelling or parabolic periodic point of @xmath0 , we obtain the first claim . let us assume that all singular values of @xmath0 belong to @xmath4 ; the case when @xmath387 uses similar ideas . recall that @xmath492 . by corollary [ cor_fin_comp ] every singular value @xmath204 of @xmath0 is the landing point of an eventually periodic ray , say @xmath493 . let @xmath494 then @xmath495 is a simply connected domain and @xmath496 is a covering map . we call @xmath497 a _ dynamical partition _ of @xmath27 . the components @xmath498 of @xmath495 are simply - connected domains , called _ itinerary domains _ , and the restriction @xmath499 is a conformal map for any @xmath201 . such a partition is very useful for studying dynamics of @xmath0 in combinatorial terms ; for instance , since no singular value of @xmath0 escapes to @xmath6 , every dynamic ray of @xmath0 is contained in a unique itinerary domain . 99 , ` the geometry of julia sets ' , _ trans . _ 338 ( 1993 ) , no . 2 , 897918 . , ` trees and hairs for some hyperbolic entire maps of finite order ' , _ math . z. _ 253 ( 2007 ) , 3359 . , ` iteration of meromorphic functions ' , _ bull . _ 29 , no . 2 ( 1993 ) , 151188 , arxiv : math.ds/9310226 . , ` on the limit functions of iterates in wandering domains ' , _ ann . ser a i math . _ 18 ( 1993 ) , 369375 . , ` a question of eremenko and lyubich concerning completely invariant domains and indirect singularities ' , _ proc . _ 130 ( 2002 ) , 32313236 . , ` dynamics of entire functions near the essential singularity ' , _ ergodic theory dynamical systems _ 6 ( 1986 ) , 489503 . , ` a dynamical approximation to the exponential map by polynomials ' , _ preprint _ ( 1986 ) . , ` etude dynamique des plynomes complexes ' , _ publications mathmatique dorsay , 84 , universit de paris - sud _ ( 1984 ) . , ` dynamics of analytic transformations ' , _ leningrad math . j. _ 1 , no . 3 ( 1990 ) , 563633 . , ` dynamical properties of some classes of entire functions ' , _ ann . fourier , grenoble _ 42 , no . 4 ( 1992 ) , 9891020 . , ` value distribution of meromorphic functions ' ( english translation ) , originally appeared in _ nauka , moscow _ ( 1970 ) . , ` teichmller theory and applications to geometry , topology and dynamics volume i : teichmller theory . 1 ' , _ matrix editions , ithaca , new york _ ( 2006 ) . , ` simple geodesics on riemann surfaces ' , _ proc . _ 86 , no . 1 ( 1982 ) , 120122 . , ` on the fixpoints , multipliers and value distribution of certain classes of meromorphic functions ' , _ ann . acad . _ 23 ( 1998 ) , 135150 . , ` hausdorff dimension and conformal dynamics ii : geometrically finite rational maps ' , _ comm . _ 75 , no . 1 ( 2000 ) , 535593 . , ` orbifolds of subhyperbolic transcendental maps ' , _ preprint _ ( 2008 ) . , ` dynamics in one complex variable ' , _ annals of mathematics studies , princeton university press , princeton , nj , vol . 160 , edition 3 _ ( 2006 ) . , ` sur une question de dulac et fatou ' , _ c. r. acad . sr i math . _ 321 ( 1995 ) , 10451048 . , ` fixed points and circle maps ' , _ acta . _ 179 , no . 2 ( 1997 ) , 243294 . , ` siegel disks and periodic rays of entire functions ' , _ journal fr reine und angewandte mathematik _ 624 ( 2008 ) , 81102 , arxiv : math.ds/0408041 . , ` topological dynamics of exponential maps on their escaping sets ' , _ ergodic theory dynamical systems _ 26 , no . 6 ( 2006 ) , 19391975 , arxiv : math.ds/0309107 . , ` rigidity of escaping dynamics for transcendental functions ' , to appear in _ acta math . _ , arxiv : math.ds/0605058 . , ` on nonlanding dynamic rays of exponential maps ' , _ ann . _ 32 ( 2007 ) , 353369 , arxiv : math.ds/0511588 . , ` on a question of eremenko concerning escaping components of entire functions ' , _ bull . london math . _ 39 ( 2007 ) , 661666 , arxiv : math.ds/0610453 . , ` on the dynamical fine structure of entire transcendental functions ' , _ phd thesis _ ( 2005 ) . , ` dynamic rays of bounded type entire functions ' , _ preprint _ ( 2005 ) , arxiv : math.ds/0704.3213 . , ` the conformal rigidity of certain subdomains on a riemann surface ' , _ trans . _ 76 , no . 1 ( 1954 ) , , ` periodic points and dynamic rays of exponential maps ' , _ ann . acad . _ 28 ( 2003 ) , 327354 . , ` the dynamical fine structure of iterated cosine maps and a dimension paradox ' , _ duke math . journal _ 136 , no . 2 ( 2007 ) , 343356 , arxiv : math.ds/0406255 . , ` analytic topology ' , _ american mathematical society colloquium publications _ , xxviii ( 1971 ) .

the wishart distribution , defined on the open convex positive definite cone , plays a central role in multivariate analysis and multivariate distribution theory . its domain of integrability is often referred to as the gindikin set . in recent years , a variety of useful extensions of the wishart have been proposed in the literature for the purposes of studying markov random fields / graphical models . in particular , generalizations of the wishart , referred to as type i and type ii wishart distributions , have been introduced by letac and massam ( _ annals of statistics _ ) and play important roles in both frequentist and bayesian inference for gaussian graphical models . these distributions have been especially useful in high - dimensional settings due to the flexibility offered by their multiple shape parameters . the domain of integrability of these graphical wisharts are however not fully identified , despite its critical role in determining existence of high dimensional bayes estimators and specifying diffuse proper priors for gaussian graphical models . moreover , these graphical wisharts also serve as statistical models in their own right for matrix - variate distributions defined on sparse convex subsets of the cone . besides its statistical motivation , understanding the domain of integrability is also of independent mathematical interest as these graphical wisharts are extensions of the gamma function on sparse manifolds . in this paper we resolve a long - standing conjecture of letac and massam ( lm ) concerning the domains of the multi - parameters of graphical wishart type distributions . this conjecture , posed in _ annals of statistics _ , also relates fundamentally to the existence of bayes estimators corresponding to these high dimensional priors . to achieve our goal , we first develop novel theory in the context of probabilistic analysis of graphical models . using these tools , and a recently introduced class of wishart distributions for directed acyclic graph ( dag ) models , we proceed to give counterexamples to the lm conjecture , thus completely resolving the problem . our analysis also proceeds to give useful insights on graphical wishart distributions with implications for bayesian inference for such models .

1408.0788

inference for graphical models is a topic of contemporary interest , and in this regard , various tools for inference have been proposed in the statistics literature , including establishing sufficient and/or necessary conditions for existence of high dimensional estimators . one important contribution in the area are the families of type i and type ii graphical wishart distributions introduced by letac and massam ( lm ) @xcite . the families of graphical wishart type distributions of letac - massam have the distinct advantage of being standard conjugate for gaussian graphical models , have attractive hyper markov properties , and have multiple shape parameters . this is in contrast with the classical wishart distribution which has just one shape parameter that is restricted to the one dimensional gindikin set - see . these multi - parameter graphical wishart distributions are therefore useful for flexible high dimensional inference @xcite , and have also been used as flat conjugate priors for objective bayesian inference . since the domain of integrability of these high dimensional priors are not fully identified , it is not clear when these distributions yield proper priors . the lm conjecture aims to address this question formally . the lm conjecture is also critical for understanding when these priors lead to well - defined bayes estimators , since this is not always guaranteed in high dimensional , sample starved settings . in this sense resolving the lm conjecture can be viewed as a bayesian analogue of the frequentist problem of identifying sufficient and necessary conditions for the existence of the maximum likelihood estimator for gaussian graphical models . the primary goal of this paper is therefore to resolve a conjecture of letac and massam ( henceforth the lm conjecture ) which concerns identifying the parameter sets for the families of the so - called type i and type ii wishart distributions . a definitive solution to the lm conjecture has remained elusive to the graphical models community ever since it was formally posed by letac and massam about ten years ago . the conjecture also has deep and profound connections to gindikin s result @xcite on the region of integrability of the p - variate gamma function . this domain is referred to as the gindikin set and is given as follows : @xmath0 though the main goal of this paper is to resolve the lm conjecture , we note that understanding the domain of integrability of these graphical wishart distributions is important for two other reasons beyond bayesian inference and model selection : 1 ) these two classes of distributions also serve as statistical models in their own right for matrix - variate distributions defined on sparse subsets of the cone , and 2 ) the integrals of these graphical wishart densities are extensions of the gamma and multivariate gamma functions on sparse manifolds . thus understanding the domains of integrability of these graphical wishart distributions is of independent mathematical interest that is closely linked to generalizations of the gindikin set . in what follows we shall employ the notation introduced in the work of letac and massam @xcite . the type i wishart and type ii wishart are defined , respectively , on the cones @xmath1 and @xmath2 associated with a decomposable graph @xmath3 , i.e. , an undirected graph that that has no induced cycle of length greater than or equal to four . these cones naturally arise as the set of covariance and inverse - covariance parameters for a gaussian undirected graph model over @xmath3 . i.e. , the family of multivariate gaussian distributions that obey the pairwise or global markov property with respect to @xmath3 @xcite . it is well known that if the vertices of @xmath3 are labeled @xmath4 , then a p - variate gaussian distribution @xmath5 obeys the global markov property with respect to @xmath3 if @xmath6 whenever there exists no edge between @xmath7 and @xmath8 . this property gives a simple characterization of the associated inverse - covariance matrices , i.e. , the elements of the cone @xmath2 . the cone @xmath1 is the dual cone of @xmath2 and its elements are incomplete covariance matrices where only the entries along the edges of @xmath3 are specified , and the rest of the entries are unspecified . however , the specified entries are also the only functionally independent entries of the covariance matrix parameter , and uniquely determine the rest of the entries ( the unspecified entries that is ) . in particular , the space of covariance matrices for the gaussian inverse - covariance graph model over @xmath3 can be identified with the cone @xmath9 . when @xmath3 is complete , i.e. , in the full model , type i and type ii wishart distributions are identical to the classical wishart distribution . moreover by restricting the multi - parameters to a specific one dimensional space , these distributions reduce to the hyper wishart distribution introduced by dawid and lauritzen @xcite and the g - wishart defined by roverato in @xcite respectively ( see @xcite for more details ) . although having multiple shape parameters allow the type i wishart and type ii wishart distributions to be more flexible as prior distributions , there is a trade - off : the sets of multi - parameters are not completely identified . in an attempt to identify the set of multi - parameterss of the type i wisharts , denoted by @xmath10 , and that of type ii wisharts , denoted by @xmath11 , in ( * ? ? ? * section 3.3 ) letac and massam first consider the case when @xmath3 is homogeneous , i.e. , @xmath3 is decomposable and has no induced paths of length greater than or equal to 4 . when @xmath3 is homogeneous letac and massam are able to completely identify @xmath10 and @xmath11 and , furthermore , give algebraic expressions for the elements of both sets . if @xmath3 is non - homogeneous , however , in ( * section 3.4 ) the authors are able to only partially identify the sets @xmath10 and @xmath11 . more specifically , for each perfect order @xmath12 of @xmath3 , they identify a subset @xmath13 of @xmath10 and a subset @xmath14 of @xmath11 . the authors then proceed to conjecture that @xmath10 and @xmath11 are indeed the union of @xmath13 and @xmath14 over all perfect orders of the cliques of @xmath15 , respectively . they demonstrate that the conjecture holds when @xmath3 is the 4-path , the simplest non - homogeneous decomposable graph , @xmath16 . they note that a similar calculation for the 5-path appears insurmountable . on a different route , but motivated by the recent work of letac and massam @xcite and rajaratnam et al . @xcite for concentration graph models , and khare and rajaratnam @xcite for covariance graph models , the authors of this paper undertook a parallel analysis in @xcite for directed acyclic graph models , abbreviated dag models , or bayesian networks . in @xcite , we introduce a new class of multi - parameter wishart type distributions , useful for bayesian inference for gaussian dag models . one of its advantages is that the framework in @xcite applies to all directed acyclic graph models and not just the narrower class of perfect dags . furthermore the normalizing constant for these dag wisharts is available in closed form for all dags . it is also well - known fact that the family of inverse - covariance graph model over a decomposable graph @xmath3 is markov equivalent to the family of dag models over a perfect dag version of @xmath3 . as we shall demonstrate later , this , in particular , implies that both the type ii wisharts of letac - massam in @xcite and the dag wisharts in @xcite are indeed defined on the same cone @xmath2 . therefore , a relevant question is how the functional form and the multi - parameter set of the type ii wishart density compare with those of the dag wishart . a similar comparison arises between the type i wisharts of letac - massam and the riez distributions for decomposable graphs introduced by andersson and klein in @xcite . such comparisons shed light on the lm conjecture since the domains of integrability of the dag wisharts are fully specified in @xcite . in this paper we develop tools which allows a careful comparison of these two types of wisharts on @xmath2 and @xmath1 , leading to counterexamples , which in turn can then be used to conclude that the lm conjecture does not hold in general . the primary key to resolving the part of the lm conjecture that concerns the type ii wishart is theorem [ thm : comparison ] of this paper . in this theorem we show that for any non - homogeneous decomposable graph @xmath15 there exists a perfect order @xmath17 and a perfect dag version of @xmath15 , associated with this order , such that the type ii wishart distribution on @xmath14 is a special case of the dag wishart distribution . using this observation , and depending on the perfect dag version of the underlying graph @xmath3 , we derive a condition in proposition [ prop : b1 ] , which when satisfied , can lead to counterexamples to the lm conjecture . we then proceed to present two graphs ( with their respective perfect dag versions ) where the stated condition in proposition [ prop : b1 ] is satisfied and lead to counterexamples to the lm conjecture . the counterexample to the other part of the lm conjecture concerning the type i wishart distribution is given after proposition [ prop : a1 ] where we prove that the same condition as that in proposition [ prop : b1 ] can lead to resolving the lm conjecture . in addition to disproving the lm conjecture , we also prove that not only for non - homogeneous decomposable graphs , but also for homogeneous graphs , the family of type ii wisharts are a subclass of the family of dag wisharts . the organization of the paper is as follows . in [ sec : pre ] we recall some fundamental notation and concepts in graphical models and , in particular , for gaussian undirected graphical models . in [ sec : wishart_for_decomposable graph ] we provide the reader with definition of type i and type ii wishart distributions and formally state the letac - massam conjecture . in [ subsec : gaussian_over_d ] and [ sub : dag_wishart ] we give a short introduction to guassian dag models and the families of dag wisharts . the main results of the paper are presented in the ensuing four sections . in [ sub : comparison_non_hom ] and [ sub : typeii_dag_wish_hom ] , we develop tools which enable a detailed comparison between type ii wisharts on one hand , and on the other hand , dag wisharts for the corresponding dag versions of the associated undirected graphs . moreover , tools are developed for comparisons of both decomposable and homogeneous type ii wisharts to their dag wishart counterparts . using the tools developed in [ sub : comparison_non_hom ] and [ sub : typeii_dag_wish_hom ] , we formally resolve the letac - massam conjecture in [ sub : lm_partb ] and [ sub : lm_parta ] by providing counterexamples . we now introduce some preliminaries on graph theory and graphical models . this section closely follows the notation and exposition given in @xcite and @xcite . a graph @xmath3 is a pair of objects @xmath18 , where @xmath19 and @xmath20 are two disjoint finite sets representing , respectively , the vertices and the edges of @xmath3 . an edge @xmath21 is said to be undirected if @xmath22 is an unordered pair @xmath23 , or directed if @xmath22 is an ordered pair @xmath24 for some @xmath25 . now a graph is said to be undirected if all its edges are undirected , and directed if all its edges are directed . a directed edge @xmath26 is denoted by @xmath27 . when @xmath27 and @xmath28 we say that @xmath29 is a parent of @xmath30 , and @xmath30 is a child of @xmath31 . the set of parents of @xmath31 is denoted by @xmath32 , and the set of children of @xmath31 is denoted by @xmath33 . the family of @xmath31 is @xmath34 . for an undirected edge @xmath35 the vertex @xmath31 is said to be a neighbor of @xmath30 , or @xmath30 a neighbor of @xmath31 , if @xmath28 . the set of all neighbors of @xmath31 is denoted by @xmath36 . in general two distinct vertices are said to be adjacent , denoted by @xmath37 , if there exists either a directed or an undirected edge between them . a loop in @xmath3 is an ordered pair @xmath38 , or an unordered pair @xmath39 in @xmath20 . for ease of notation , in this paper we shall always assume that the edge set of each graph contains all the loops , however , we draw the graph without the loops . + a graph @xmath40 is a subgraph of @xmath41 , denoted by @xmath42 , if @xmath43 and @xmath44 . in addition , if @xmath42 and @xmath45 , we say that @xmath46 is an induced subgraph of @xmath3 . we shall consider only induced subgraphs in what follows . for a set @xmath47 , the subgraph @xmath48 is said to be the graph induced by @xmath49 . a graph @xmath3 is called complete if every pair of vertices are adjacent . a subset @xmath47 is said to be a clique , if the induced subgraph @xmath50 is a complete subgraph of @xmath3 that is not contained in any other complete subgraphs of @xmath3 . a path in @xmath3 of length @xmath51 from a vertex @xmath31 to a vertex @xmath30 is a finite sequence of distinct vertices @xmath52 in @xmath19 such that @xmath53 or @xmath54 are in @xmath20 for each @xmath55 . a path is said to be directed if at least one of the edges is directed . we say @xmath31 leads to @xmath30 , denoted by @xmath56 , if there is a directed path from @xmath31 to @xmath30 . a graph @xmath41 is said to be connected if for any pair of distinct vertices @xmath57 there exists a path between them . an @xmath58-cycle in @xmath3 is a path of length @xmath58 with the additional requirement that the end points are identical . a directed @xmath58-cycle is defined accordingly . a graph is acyclic if it does not have any cycles . an acyclic directed graph , denoted by dag ( or adg ) , is a directed graph with no cycles of length greater than 1 . henceforth in this paper , we denote an undirected graph by @xmath41 and a dag by @xmath59 . also , otherwise stated , we always assume that the vertex set @xmath60 . the undirected version of a dag @xmath59 , denoted by @xmath61 , is the undirected graph obtained by replacing all the directed edges of @xmath62 by undirected ones . an immorality in @xmath62 is an induced subgraph of the from @xmath63 . moralizing an immorality entails adding an undirected edge between the pair of parents that have the same children . then the moral graph of @xmath62 , denoted by @xmath64 , is the undirected graph obtained by first moralizing each immorality of @xmath62 and then making the undirected version of the resulting graph . naturally there are dags which have no immoralities and this leads to the following definition . a dag @xmath62 is said to be perfect if it has no immoralities ; i.e. , the parents of all vertices are adjacent , or equivalently if the set of parents of each vertex induces a complete subgraph of @xmath62 . given a directed acyclic graph ( dag ) , the set of ancestors of a vertex @xmath31 , denoted by @xmath65 , is the set of those vertices @xmath66 such that @xmath67 . similarly , the set of descendants of a vertex @xmath31 , denoted by @xmath68 , is the set of those vertices @xmath30 such that @xmath56 . the set of non - descendants of @xmath31 is @xmath69 . a set @xmath47 is said to be ancestral when @xmath49 contains the parents of its members . the smallest ancestral set containing a set @xmath70 is denoted by @xmath71 . let @xmath3 be a decomposable graph . the reader is referred to lauritzen @xcite for all the common notions of decomposable graphs that we will use here . one such important notion is that of a perfect order of the cliques . every decomposable graph admits a perfect order of its cliques . let @xmath72 be one such perfect order of the cliques of the graph @xmath3 . the history for the graph is given by @xmath73 and @xmath74 the separators of the graph are given by @xmath75 the residuals are defined as follows : @xmath76 generally , we will denote by @xmath77 the set of cliques of a graph and by @xmath78 its set of separators . let @xmath79 denote the number of distinct separators and @xmath80 denote the multiplicity of @xmath81 , i.e. , the number of @xmath8 such that @xmath82 . decomposable ( undirected ) graphs and ( directed ) perfect graphs have a deep connection . if @xmath3 is decomposable , then there exists a perfect dag version of @xmath3 , i.e. , a perfect dag @xmath62 such that @xmath83 . on the other hand , the undirected version of a perfect dag is necessarily decomposable @xcite . a decomposable graph @xmath3 is said to to be homogeneous if for any two adjacent vertices @xmath84 we have @xmath85 the reader is referred to letac and massam @xcite for all the common notions of homogeneous graphs . let @xmath86 be a undirected graph with @xmath87 and let @xmath88 be a random vector in @xmath89 such that @xmath90 , i.e. , @xmath91 has a @xmath92-variate gaussian distribution with mean zero and covariance @xmath93 . the covariance matrix @xmath93 is assumed to be positive definite ( written as @xmath94 ) with inverse - covariance matrix ( also said to be the precision or concentration matrix ) . now for any two vertices @xmath96 @xmath97 a simple proof of this well - known fact can be found in @xcite . in particular , the distribution @xmath5 is said to be a markov random field over @xmath3 if @xmath98 now let @xmath99 denote the family of all @xmath92-variate gaussian distributions @xmath5 that are markov random fields over @xmath3 . note that equation provides an easy description of the elements of @xmath99 in terms of the pattern of zeros in the associated inverse - covariance matrices . subsequently , @xmath99 is said to be the gaussian concentration graph model over @xmath3 . the set of covariance matrices @xmath100 is the standard parameter set for @xmath99 . in light of equation the distributions in the exponential family @xmath99 can be parametrized by the canonical parameter @xmath101 which lives in the space of inverse - covariance matrices defined as follows : @xmath102 let @xmath103 denote the set of @xmath104 symmetric matrices . then @xmath105 we call each element in @xmath106 a @xmath3-incomplete matrix . one can easily check that @xmath106 is a real linear space , isomorphic to @xmath107 , of dimension @xmath108 . a @xmath3-incomplete matrix @xmath109 is said to be partial positive definite over @xmath3 if for each clique @xmath110 the @xmath111 matrix @xmath112 is positive definite . note that for any decomposable graph @xmath3 , the set of partial positive definite matrices over @xmath3 , denoted by @xmath1 , is the dual cone of the ( open convex ) cone @xmath2 @xcite . when @xmath3 is decomposable grone et al . in @xcite prove that each @xmath113 can be completed to a unique positive definite matrix @xmath114 . this means that @xmath93 is the only element in @xmath115 with the property that @xmath116 , for each @xmath117 . if @xmath118 denotes an element of @xmath1 with the unique positive definite completion @xmath93 in @xmath115 , then grone et al . @xcite explicitly provide a bijective mapping @xmath119 . if we compose this mapping with the inverse mapping @xmath120 , then we obtain the bijective mapping @xmath121 . the corresponding inverse mapping is given as @xmath122 where @xmath123 . we shall frequently invoke these mappings in subsequent sections . henceforth in this paper , we assume that @xmath124 is a decomposable graph and the vertices are labeled @xmath4 . the primary goal of this section is to provide the reader with an overview of the families of wishart - type i and wishart - type ii distributions introduced in @xcite . at the end of this section we shall formally state the lm conjecture concerning the domains of the multi - parameters for these distributions . let @xmath125 be a perfect order of the cliques of @xmath3 and let @xmath126 be the corresponding sequence of separators , with possible repetitions . for each @xmath127 and @xmath128 , the markov ratio @xmath129 is defined as follows : @xmath130 let @xmath131 and @xmath132 where @xmath133 and @xmath134 , respectively . moreover , let @xmath135 denote lebesgue measure on @xmath1 . is the standard lebesgue measure on @xmath106 restricted to the open set @xmath1 . ] then @xmath136 is a measure on @xmath1 . the image of @xmath137 under the mapping @xmath121 is a measure on @xmath2 given by @xmath138 where @xmath139 is lebesgue measure on @xmath2 @xcite . we now introduce the type i and type ii wishart distributions from @xcite . the type i wishart is a distribution defined on the cone @xmath1 . the non - normalized density of this distribution is given by @xmath140 where @xmath141 denotes the multi - shape parameter and @xmath142 is the scale parameter . the normalized version of @xmath143 , denoted by @xmath144 , is defined for pairs of @xmath145 such that for every @xmath142 @xmath146 @xmath147 the type ii wishart is a distribution on the cone @xmath2 with the non - normalized density @xmath148 similarly , the normalized version of @xmath149 , denoted by @xmath150 , is defined for pairs of @xmath145 such that for every @xmath142 @xmath151 @xmath152 the space of multi - shape parameter for the family of type i wisharts , i.e. , the set of pairs @xmath153 that satisfy both conditions @xmath154 and @xmath155 , is denoted by @xmath10 . likewise , the space of multi - shape parameter for the family of type ii wisharts , i.e. , the set of pairs @xmath153 that satisfy conditions @xmath156 and @xmath157 , is denoted by @xmath11 . after defining type i & ii wishart distributions , an important goal of letac @xmath158 massam in @xcite is to identify @xmath10 and @xmath11 , the associated spaces of multi - shape parameters . when the underlying graph @xmath3 is homogeneous both @xmath10 and @xmath11 are completely identified in @xcite , but when @xmath3 is no longer homogeneous , these spaces are only partially identified . more precisely , letac & massam @xcite identify a subset of @xmath10 and a subset of @xmath11 as follows . + let @xmath159 be a given perfect order of the cliques of @xmath3 and @xmath160 the corresponding sequence of separators . for each separator @xmath161 let @xmath162 . a set associated with @xmath17 and @xmath10 , denoted by @xmath163 , is the set of @xmath164 such that : * @xmath165 , for each @xmath166 , where @xmath167 , as before , denotes the multiplicity of the separator @xmath81 ; * @xmath168 , for each @xmath169 ; * @xmath170 , where @xmath171 . similarly , a set associated with @xmath17 and @xmath11 , denoted by @xmath172 , is the set of @xmath145 such that : * @xmath173 , for each @xmath166 ; * @xmath174 , for each @xmath169 and @xmath175 ; * @xmath176 where @xmath177 . theorems 3.3 & 3.4 in @xcite prove that if @xmath3 is a non - complete decomposable graph , then @xmath178 and @xmath179 . therefore , @xmath180 and @xmath181 , where the subscript @xmath17 runs through all perfect orders of the cliques of @xmath3 . when @xmath3 is homogeneous letac and massam in @xcite establish that @xmath182 and @xmath183 , but in the case of an arbitrary non - homogeneous decomposable graph they conjecture that equalities hold . we now proceed to formally state the letac - massam conjecture . + * the letac - massam ( lm ) conjecture*. let @xmath3 be a non - homogeneous decomposable graph and let @xmath184 denote the set of the perfect orders of the cliques of @xmath3 . then @xmath185 @xmath186 note that for each perfect order @xmath187 of the cliques of a decomposable graph @xmath3 the sets @xmath163 and @xmath172 , as manifolds , are of dimension @xmath188 . therefore , the lm conjecture asserts that @xmath10 and @xmath11 are also of dimension @xmath188 . one of the main goals of this paper is to study the lm conjecture and formally demonstrate that it does not hold in general . our goal is slightly broader as we are also interested in understanding when exactly the lm conjecture does not hold . in particular , we aim to identify graph characteristics which lead to a violation of the lm conjecture . our approach is to develop tools which will allow us to compare the type i & ii wishart distributions , respectively , with the generalized versions of riesz distributions , by andersson et al . @xcite for perfect dags , and the dag wishart distributions introduced by the present authors in @xcite . we demonstrate that relating the lm conjecture to the class of dags ( and not just undirected graphical models ) can provide valuable insights . since we are able to completely characterize the domain of integrability of the wisharts associated with dag models . we begin with a compact review of the dag wisharts given in @xcite . inference for gaussian dag models provide the main motivation for developing the dag wisharts in @xcite . we give a brief introduction here . let @xmath59 be a dag with @xmath92 vertices , i.e. , @xmath189 . for each @xmath96 let the relation @xmath190 denote @xmath191 or @xmath192 . the relation @xmath193 clearly defines a partial order on @xmath19 . since every partial order can be extended to a linear order @xcite , without loss of generality , we can assume that the vertices in @xmath19 are labeled @xmath194 , and for each @xmath96 if @xmath195 , then @xmath196 . this order corresponds to the parent order of the vertices of the dag . now let the random vector @xmath197 be a directed markov random field ( or dag ) over @xmath62 . thus @xmath91 obeys the directed local markov property with respect to @xmath62 , i.e. , @xmath198 if , in addition , @xmath199 , then a simple observation in @xcite shows that the directed local markov property in equation is satisfied if and only if @xmath200 and @xmath201}=\sigma_{\nprec j\succ}\left(\sigma_{\prec j\succ}\right)^{-1}\sigma_{\prec j]}\quad\forall j\in v,\ ] ] where @xmath202}:=\left\{\left(i , j\right ) : i : i\in \mathrm{nd}\left(j\right ) , i > j\right\}$ ] , @xmath203 , @xmath204 and @xmath205:=\mathrm{pa}(j)\times \{j\}$ ] . + we define the gaussian dag model , denoted by @xmath206 , to be the family of all centered gaussian distribution @xmath207 which are directed markov random fields over @xmath62 . it is easily seen that the distributions in @xmath206 can be parameterized by the space of covariance matrices @xmath208}=\sigma_{\nprec j\succ}\left(\sigma_{\prec j\succ}\right)^{-1}\sigma_{\prec j ] } , \:\forall j\in v\right\}.\ ] ] these distributions can also be parametrized by the space of inverse - covariance matrices @xmath209 . other important parameterizations of the distributions in @xmath206 are available in terms of the modified cholesky decompositions of the inverse - covariance matrices , i.e. , @xmath210 such that @xmath211 is a lower triangular matrix with all diagonals equal to @xmath212 , and @xmath213 is a diagonal matrix . note that for two distinct vertices @xmath214 , if @xmath7 is not a parent of @xmath8 , then @xmath215 . we refer the reader to @xcite for more details . let @xmath62 be a perfect dag . first note that a random vector @xmath91 in @xmath89 is a dag over @xmath62 if and only if it is an undirected graphical model over @xmath216 , the undirected version of @xmath62 ( which is also necessarily decomposable ) @xcite . this implies that @xmath217 and @xmath218 are , respectively , identical to @xmath219 and @xmath220 ( see [ sub : gaussian_over_g ] for definitions ) . in particular , @xmath217 is an open convex cone . the dag wishart distribution @xmath221 , as we shall define here is a distribution on @xmath217 @xcite . we first define , @xmath222 , the non - normalized version of @xmath221 as follows : @xmath223 where the multi - shape parameter @xmath224 lives in @xmath89 , @xmath225 , @xmath226}^{\top}\left(\sigma_{\prec j \succ}\right)^{-1}\sigma_{\prec j]}$ ] and @xmath227 . from theorem 4.1 in @xcite the domain of integrability of the dag wishart distribution can be fully characterized : @xmath228 moreover , if @xmath229 , then the normalizing constant is given by @xmath230 where @xmath231 . [ rem : cholesky_decomposition ] let @xmath232 and @xmath233 be the modified cholseky decompositions of @xmath234 . then @xmath211 is a lower triangular matrix with all diagonal entries equal to one and @xmath235 and @xmath213 is a diagonal matrix such that @xmath236 @xcite . we now proceed to compare type ii wisharts for decomposable graphs with dag wisharts for perfect dags . we had noted earlier that the class of decomposable graphs are markov equivalent to the class of perfect dags . in particular , every probability distribution that obeys the global markov property with respect to a decomposable graph also obeys the local directed markov property with respect to a perfect dag version and vice versa . in the gaussian setting this means that if @xmath62 is a perfect dag version of @xmath3 , then @xmath237 and @xmath238 define the same family of @xmath92-variate gaussian distributions . consequently , the family of type ii wisharts and the family of dag wisharts are both defined on @xmath239 , the space of inverse - covariance matrices . therefore , a relevant question is how the functional form of the type ii wishart density compares with that of the dag wishart . + first , to facilitate comparison , we re - parameterize the dag wishart @xmath221 as follows . for each @xmath240 let the expressions of the form @xmath241 in equation be replaced by @xmath242 , and let @xmath243 be replaced by @xmath244 . let @xmath245 . under this parametrization , with a slight abuse of notation , we write @xmath246 where the normalizing constant @xmath247 exists if and only if @xmath248 for each @xmath249 and @xmath250 note also that for each perfect directed version @xmath62 of @xmath3 the exponential term @xmath251 is a common term in both the type ii letac - massam wishart @xmath252 and the dag wishart @xmath253 . moreover , the dag wishart @xmath253 has additional terms only of the form @xmath254 . before comparing @xmath252 and @xmath221 more generally , we first illustrate the comparison with an example . [ ex : comparison ] let @xmath3 be the @xmath255-path given in figure [ fig : graph - patha ] . note that @xmath3 is a non - homogeneous decomposable graph . it is clear that the dag given in figure [ fig : graph - pathb ] is a perfect dag version of @xmath3 . the cliques of @xmath3 are @xmath256 and the separators are @xmath257 . let @xmath258 . to compare the corresponding letac - massam type ii wishart and the dag wishart for this graph , we rewrite the markov ratio present in the density of @xmath259 as follows . @xmath260 as shown in section 3.4 of @xcite , one can check that @xmath261 where @xmath262 and @xmath263 are the perfect orders of the cliques of @xmath3 and @xmath264 note that unless @xmath265 and @xmath266 ( i.e. , @xmath145 is restricted to the intersection of @xmath267 and @xmath268 ) the expression in equation contains some terms different from the product of @xmath269 to some powers . since the dag wishart has polynomial terms only of the form @xmath254 , it is clear that for this directed version of @xmath3 , @xmath252 and @xmath221 are not directly comparable . note that we did not need to account for two additional perfect orders @xmath270 and @xmath271 since it has been shown in @xcite that @xmath272 and @xmath273 . now consider the comparison with the perfect dag version given in figure [ fig : graph - pathc ] . in this case the cliques are @xmath274 and the separators are @xmath275 , @xmath276 . by a similar calculation as that in equation we obtain @xmath277 therefore , for this directed version of @xmath3 , the family of type ii wisharts , restricted to @xmath278 , is a subfamily of the family of dag wisharts . in example [ ex : comparison ] we illustrated the fact that although @xmath252 does not , necessarily , compare with @xmath221 for any arbitrary perfect dag version @xmath62 of @xmath3 , it is however comparable with some particular dags . we will show next that this conclusion can be generalized to any decomposable graphs . to this end , we proceed with a few useful lemmas . in particular , we introduce tools that will allow us to relate decomposable graphs with a given perfect ordering of its cliques with perfect dags and vice versa . these tools turn out to be critical ingredients in comparing the undirected letac - massam wisharts to the directed dag wisharts . before we proceed with the next lemma we introduce some convenient notation and a definition . [ not:- > ] let @xmath279 be a dag and let @xmath280 . then @xmath281 denotes the fact that there exist a vertex @xmath282 and a vertex @xmath283 such that @xmath284 . let @xmath285 be a perfect order of the cliques of a decomposable graph @xmath3 . a dag version @xmath62 of @xmath3 is said to be induced by @xmath17 if @xmath286 are all ancestral in @xmath62 . note that in terms of notation [ not:- > ] a dag version @xmath62 of @xmath3 is induced by @xmath285 if and only if @xmath287 implies that @xmath288 . [ lem : combine ] let @xmath3 be a non - complete decomposable graph . 1 . let @xmath289 be a perfect order of the cliques of a decomposable graph @xmath3 . then every dag version of @xmath3 induced by @xmath17 is a perfect dag . moreover , there exists a perfect dag version @xmath62 of @xmath3 induced by @xmath17 such that @xmath290 is ancestral in @xmath62 . 2 . conversely , suppose @xmath62 is a perfect dag version of @xmath3 . then there exists a perfect order @xmath17 , of the cliques of @xmath3 , such that @xmath62 is induced by @xmath17 . a ) suppose , to the contrary , that @xmath62 is not perfect . let @xmath8 be the smallest integer such that @xmath291 , the induced dag on @xmath292 , is not perfect . it is clear that @xmath293 . let @xmath294 be an immorality in @xmath295 . this , in particular , implies that there are two distinct cliques @xmath296 and @xmath297 , with subscript @xmath298 , such that they contain @xmath299 and @xmath300 , respectively . since @xmath301 and @xmath302 are distinct we may assume that @xmath303 . but since @xmath304 is ancestral and @xmath66 is a parent of @xmath305 we must have @xmath306 . this contradicts the fact that the induced dag on @xmath304 is perfect . + now we show that in particular there exists a dag @xmath62 induced by @xmath17 such that @xmath307 is ancestral in @xmath62 . first consider the case where there are only two cliques . we start with relabeling the vertices in @xmath290 , @xmath308 and @xmath309 , respectively , in a decreasing order . if @xmath62 is the dag version of @xmath3 induced by this order , then @xmath290 and @xmath310 are ancestral in @xmath62 . now suppose that such a dag version exists for any decomposable graph with number of cliques less than @xmath311 . by the mathematical induction there exists a dag version @xmath312 of @xmath313 such that @xmath314 are ancestral in @xmath312 . without loss of generality , we can assume that the vertices in @xmath312 are labeled from @xmath315 . let us label the vertices in @xmath316 from @xmath317 and let @xmath62 be the dag version of @xmath3 induced by this order . one can easily check that @xmath62 has the desired properties . + b ) suppose that by mathematical induction the lemma holds for any dag with fewer than @xmath92 vertices . now we assume that @xmath3 is a decomposable graph with @xmath92 vertices . clearly , we can assume that @xmath318 . let @xmath319 be the induced graph on @xmath320 . by our induction hypothesis , there is a perfect order @xmath321 of @xmath319 such that @xmath322 , the induced dag on @xmath323 , is induced by @xmath324 . note that @xmath325 is a clique of @xmath3 . consider two possible cases : 1 . there is an @xmath7 such that @xmath326 is not a clique in @xmath3 . then @xmath327 . this implies that for each @xmath328 , @xmath329 remains a clique in @xmath3 . let us replace @xmath326 with @xmath330 and define @xmath331 one can easily check that @xmath17 is a perfect order of @xmath3 , and @xmath62 is induced by @xmath17 . 2 . for every @xmath332 , @xmath326 is a clique in @xmath3 . let @xmath333 we use the convention that @xmath334 and @xmath335 . first suppose @xmath336 are both finite . thus we have @xmath337 and therefore @xmath338 , because by our induction hypothesis the histories of @xmath339 are ancestral in @xmath312 . one can check that @xmath340 is then a perfect order of @xmath3 and @xmath62 is induced by @xmath17 . if @xmath341 or @xmath342 , then by appending the clique @xmath330 at the end or at the beginning of @xmath343 , respectively , we obtain a perfect order @xmath17 and in either case @xmath62 is induced by such @xmath17 . [ lem : ancestral ] let @xmath62 be a dag and let @xmath344 . suppose @xmath345 is the modified cholesky decomposition of @xmath346 . then we have : * for each @xmath96 if @xmath347 , then @xmath348 , where @xmath349 is the partial regression coefficient of @xmath350 in the linear regression of @xmath351 on @xmath352 and @xmath353 . * if @xmath49 is an ancestral subset of @xmath19 , then @xmath354 . in particular , @xmath355 ( also see @xcite for a related result ) . * this can be proved by using equation ( see @xcite for details ) . * since @xmath49 is ancestral in @xmath62 , by using equation , one can easily show that @xmath356 . now let @xmath357 be the modified cholesky decomposition of @xmath358 . part i ) and the fact that @xmath49 is ancestral imply that @xmath359 whenever @xmath347 , and @xmath360 . this implies that @xmath361 and @xmath362 [ thm : comparison ] for every perfect order @xmath285 of the cliques of a decomposable @xmath3 there exists a perfect dag version @xmath62 of @xmath3 such that the family of distributions @xmath252 , restricted to @xmath14 , is a subfamily of the family of @xmath363 . let @xmath62 be a dag version of @xmath3 induced by @xmath17 . first note that by lemma [ lem : combine ] @xmath62 is perfect . therefore @xmath364 is identical to @xmath2 and @xmath365 is indeed a distribution on @xmath2 . in comparing @xmath252 with @xmath221 it suffices to show that the markov ratio that appears in the density of @xmath252 can be written as products of @xmath269 to some powers . we proceed to rewrite the corresponding markov ratio as follows : @xmath366 let @xmath367 for each @xmath368 . consider the following block - partitioning of @xmath369 . @xmath370 now for each @xmath169 we have @xmath371 , and @xmath372 separates @xmath373 form @xmath374 . by lemma 5.5 @xcite we have @xmath375 by rewriting this and using lemma [ lem : ancestral ] we obtain @xmath376 similarly , lemma [ lem : ancestral ] implies that @xmath377 and @xmath378 . now note that if @xmath379 and @xmath166 , then @xmath380 . thus when the shape parameters are restricted to @xmath172 the markov ratio above is only a product of some powers of @xmath269 . the next proposition is essential for our purposes as it gives us the recipe that we need to construct counterexamples to the lm conjecture . note that if the lm conjecture ( ii ) is true , then the dimension of the set @xmath11 , as a manifold , is @xmath188 . first we introduce a new notation as follows . [ def : ans ] let @xmath62 be a dag version of @xmath3 induced by @xmath17 and let @xmath381 denote the set of all separators @xmath382 that are ancestral in @xmath62 . we denote the number of elements of @xmath381 by @xmath383 . [ prop : b1 ] let @xmath62 be a dag version of @xmath3 induced by @xmath17 . then the dimension of the manifold described by @xmath141 that satisfies equation is greater than or equal to @xmath384 . suppose @xmath385 . by part ii ) of lemma [ lem : ancestral ] the term @xmath386 in equation can be written as products of some powers of @xmath269 . this in turn implies that for each @xmath385 , restricting @xmath145 to the equation @xmath387 is not necessary and , consequently , the dimension of the corresponding set , as a manifold , is at least @xmath384 . [ rem : key ] as we mentioned earlier , the lm conjecture ( ii ) implies that for any decomposable graph @xmath3 the dimension of @xmath11 is @xmath188 . now in light of proposition [ prop : b1 ] the lm conjecture ( ii ) suggests that for any dag version @xmath62 of @xmath3 the number @xmath388 . therefore , the lm conjecture ( ii ) can be shown not to be true if we can construct a decomposable graph @xmath3 and a dag version @xmath62 such that @xmath389 . we shall further exploit this line of reasoning in [ sub : lm_partb ] . henceforth in this section , let @xmath390 denote a homogeneous graph . in this section we show that for any homogeneous graph @xmath3 there is a dag version @xmath62 such that @xmath252 is a special case of @xmath221 on the whole parameter set @xmath11 ( note that , as we discussed in [ subsec : h ] , when a graph is homogeneous the parameter set @xmath11 is completely identified ) . note also that equation defining homogeneous graphs naturally defines a partial order on the vertex set @xmath19 of @xmath391 as follows : @xmath392 see @xcite for more details on rooted hasse trees and on this partial order @xmath393 . we let the linear order @xmath394 ( or simply @xmath395 when there is no danger of confusion ) be a linear extension of the partial order @xmath393 , and let @xmath62 be the dag version of @xmath391 induced by @xmath394 . one can easily check that @xmath62 is perfect , and transitive , i.e. , @xmath396 the above shows that any homogeneous graph has a perfect transitive dag version . now for a homogeneous graph we prove the following generalized form of theorem [ thm : comparison ] . [ prop : comparison_hom ] let @xmath391 be a homogeneous graph and @xmath62 a perfect transitive dag version . then the family of type ii wisharts for @xmath391 is a subfamily of the dag wisharts for @xmath62 . first we claim that all the cliques , and consequently all the separators , of @xmath3 are ancestral in @xmath62 . to see this note the following : let @xmath330 be a clique of @xmath3 and suppose @xmath397 for some @xmath398 . let @xmath399 be any other vertex in @xmath330 . then either @xmath400 or @xmath401 . regardless , since @xmath62 is homogeneous , @xmath402 and @xmath399 must be adjacent . thus @xmath403 . this proves that @xmath330 is ancestral in @xmath62 . now if @xmath81 is a separator of @xmath3 , then the fact that @xmath404 for some @xmath405 implies that @xmath81 is ancestral ( otherwise , it implies that the clique @xmath406 is not ancestral leading to a contradiction ) . therefore by lemma [ lem : ancestral ] we obtain @xmath407 which obviously shows that this markov ratio is a product of powers of @xmath269 , therefore the family of type ii wisharts for @xmath391 is a subfamily of the dag wisharts for @xmath62 . in the following two examples we compare @xmath408 and @xmath221 in more detail . more specifically , we shall explain how the space of shape parameters is identified for each family of distributions . the space of shape parameters for @xmath409 is identified by theorem 3.2 in @xcite . for this purpose , we follow the notion introduced in @xcite . + for any homogeneous graph @xmath391 , let @xmath410 be the hasse tree of @xmath391 ( see @xcite for greater detail ) . note that a vertex in @xmath411 is indeed an equivalent class @xmath412\subseteq v$ ] for some vertex @xmath413 , where @xmath414 $ ] if and only if @xmath415 . if @xmath416 is an internal vertex , i.e. , a vertex that has a child , then @xmath417:\ : [ i]\preceq t \right\}$ ] is a clique of @xmath391 . also if @xmath418 is a leaf vertex , i.e. , has no child , then @xmath419:\ : [ i]\preceq q\right\}$ ] is a separator of @xmath391 . for each @xmath412\in t$ ] define @xmath420}\left(\alpha,\beta\right)&:=\sum_{[i]\preceq t}\alpha\left(c_{t}\right)-\sum_{[i]\preceq q}\nu\left(s_{q}\right)\beta\left(s_{q}\right),\\ m_{[i]}&:=\sum_{t\preceq [ i]}n_{t},\end{aligned}\ ] ] where @xmath421 is the number of the elements in @xmath422 . by theorem 3.2 in @xcite , @xmath423}\left(\alpha,\beta\right)>\left(\sum_{[i]\preceq t}n_{t}-1\right)/2,\ : \ \forall [ i]\in t.\ ] ] we now proceed to compare the space of shape parameters of @xmath252 and @xmath221 in two concrete examples . [ ex ] let @xmath391 be the @xmath424-path given in figure [ fig : path-3a ] and @xmath62 the dag version given in figure [ fig : path-3b ] . it is clear that @xmath391 is a homogeneous graph and @xmath62 is a perfect transitive dag version . first we show that the densities @xmath425 and @xmath221 have the same functional form . using the labeling in @xmath62 , the cliques of @xmath391 are @xmath426 , @xmath427 . the only separator is @xmath428 . thus @xmath429 . replacing these in the corresponding markov ratio that appears in the @xmath252 distribution we obtain @xmath430 therefore , @xmath431 and @xmath432 have exactly the same functional form when @xmath433 , @xmath434 and @xmath435 . to identify the space of shape parameters for @xmath409 we proceed to compute @xmath436}=\alpha_{1 } , & \rho_{[2]}=\alpha_{2 } , & \rho_{[3]}=\alpha_{1}+\alpha_{2}-\beta_{2},\\ n_{[1]}=1 , & n_{[2]}=1 , & n_{[3]}=1 . \end{array}\ ] ] now by using theorem 3.2 in @xcite we obtain @xmath437 . th space of shape parameters @xmath438 for @xmath221 is easily determined by inequalities @xmath248 , which yields the same inequalities @xmath439 and @xmath440 . this shows that up to re - parametrization @xmath431 and @xmath432 are the same distributions . in the next example we shall show that the family of dag wisharts @xmath432 strictly contains the family of type ii wisharts @xmath431 . let @xmath391 be the homogeneous graph given in figure [ fig : subfiga ] . the hasse tree of @xmath391 is given in figure [ fig : subfigb ] . let @xmath62 be the perfect transitive dag version given in figure [ fig : subfigc ] . the cliques of @xmath391 are @xmath441 and @xmath442 . the only separator is @xmath443 . thus @xmath444 and @xmath445 . substituting these in the corresponding markov ratio which appears in the @xmath252 distribution we obtain @xmath446 therefore , the choice of @xmath447 , @xmath448 and @xmath449 the density @xmath432 is equal to @xmath450 . in order to apply theorem 3.2 in @xcite for identifying the space of shape parameters @xmath11 , first we compute @xmath451}=\alpha_{1 } , & \rho_{[2]}=\alpha_{2 } , & \rho_{[3]}=\alpha_{1}+\alpha_{2}-\beta_{2}\\ n_{[1]}=1 , & n_{[2]}=1 , & n_{[3]}=2 . \end{array}\ ] ] from this and after some calculations we obtain @xmath452 if and only if @xmath453 alternatively , with much less computation , from the inequalities @xmath248 we can identify @xmath11 . the examples above demonstrate that for a homogeneous graph @xmath391 , the type ii wishart is a special case of the dag wishart for a perfect transitive dag version of @xmath391 . furthermore , identifying the space of shape parameter @xmath11 under the dag wishart family is computationally less expensive . proposition [ prop : comparison_hom ] for homogeneous graphs is stronger than theorem [ thm : comparison ] for decomposable graphs . the reason is that the latter guarantees the family of @xmath454 distributions is a subfamily of dag wisharts @xmath455 only when @xmath456 are restricted to @xmath457 . proposition [ prop : comparison_hom ] however guarantees that the family of @xmath454 is a subfamily of dag wisharts @xmath455 on the whole parameter set @xmath11 . also note that proposition [ prop : comparison_hom ] is not implied by theorem [ thm : comparison ] . to see this , consider the dag @xmath62 given by figure [ fig : path-3c ] . this dag is a perfect , but non - transitive , dag version of the homogeneous graph @xmath458 given by figure [ fig : path-3a ] and in fact induced by the perfect order @xmath459 of the cliques of @xmath458 . it is easy to check that @xmath460 . therefore using theorem [ thm : comparison ] for this homogeneous graph does not imply that the family of @xmath454 is a subfamily of dag wisharts @xmath455 on the whole parameter set @xmath11 . we now return to the lm conjecture ( ii ) in this section . using the tools we have developed thus far , we proceed to obtain some counterexamples to show that part ( ii ) of the lm conjecture fails . in particular , we show that there exist decomposable graphs where the space of shape parameters @xmath11 for the type ii wisharts , over such a graph @xmath3 , strictly contains @xmath461 . recall that a pair @xmath462 belongs to @xmath11 if and only if it satisfies both equation and equation . as we hinted in remark [ rem : key ] , if there exists a decomposable graph @xmath3 , necessarily non - homogeneous , that has a perfect order @xmath17 such that the dag version @xmath62 induced by @xmath17 yields @xmath463 ( recall definition [ def : ans ] ) , then equation is satisfied on a set that strictly contains @xmath461 . in the following examples we show that in such a situation , simultaneously for the same set , equation can be satisfied as well . [ ex : lm2_a ] let @xmath3 be the decomposable graph given in figure [ fig : c1 ] . clearly , @xmath3 is a non - homogeneous decomposable graph . consider the perfect order @xmath464 of the cliques of @xmath3 . the separators of @xmath3 are @xmath465 and @xmath466 . it is easy to check that the dag @xmath62 given in figure [ fig : cd1 ] is a directed version of @xmath3 induced by @xmath17 . using the labeling in @xmath62 we have @xmath467 note that for the dag version @xmath62 given by figure [ ex : lm2_a ] @xmath468 since @xmath469 and @xmath290 are both ancestral in @xmath62 . thus by proposition [ prop : b1 ] equation is satisfied on a set of @xmath470 that is of dimension greater than or equal to @xmath471 . this is strictly greater than @xmath472 , the dimension of @xmath461 . hence it is clear that the lm conjecture ( ii ) fails for this example . nonetheless , for this specific example , we provide the reader with a self - contained proof . to begin with , we rewrite the markov ratio term that appears in @xmath252 as follows : @xmath473 let @xmath242 be the exponent of @xmath269 in equation . then @xmath474 for every @xmath475 such that @xmath476 now we show that for each @xmath153 that satisfies equation , not only equation is satisfied , but also equation is satisfied , i.e. , @xmath477 by equation and equation we have @xmath478 now similar to our computation in equation we can show @xmath479 therefore , we have @xmath480 this completes our first counterexample . [ ex : lm2_b ] let @xmath3 be the non - homogeneous graph given in [ fig : c2 ] and let @xmath62 be the dag given in [ fig : cd2 ] . one can readily check that @xmath62 is a dag version of @xmath3 induced by @xmath481 since there are @xmath472 cliques and @xmath255 separators @xmath482 and @xmath483 . therefore , the lm conjecture ( ii ) here implies that the dimension of @xmath11 is @xmath484 . now @xmath468 as the separators @xmath485 and @xmath486 are ancestral ( two other separators @xmath487 and @xmath488 are not ancestral ) . therefore , by proposition [ prop : b1 ] , equation is satisfied on a set of dimension greater than or equal to @xmath489 . we leave it to the reader to show that with a similar calculation as in example [ ex : lm2_a ] equation is satisfied on this set . thus the lm conjecture ( ii ) fails . note that in example [ ex : lm2_b ] the separator @xmath490 . therefore @xmath491 . by some calculations similar to those of example [ ex : lm2_a ] , we can show that the markov ratio present in the corresponding type ii wishart is @xmath492 where the exponents @xmath493 and @xmath224 are some linear functions of @xmath494 and @xmath495 . now using equation one can show that both equation and equation are satisfied for every @xmath496 subject to the constraint @xmath497 . this shows that the dimension of @xmath498 . on the other hand , we can see that for any dag version of @xmath3 always two of the separators will be be non - ancestral . therefore , although proposition [ prop : b1 ] identifies a subset of @xmath11 that is strictly larger than the subset identified by the lm conjecture ( ii ) it does not identify the whole set @xmath11 , since for any dag version of @xmath3 proposition [ prop : b1 ] will identify a subset of dimension @xmath499 . in this section we once more use the theory we had developed in previous sections to produce counterexample to part ( i ) of the lm conjecture . first we establish the following lemma . [ lem : mr ] suppose that @xmath62 is a perfect dag version of @xmath3 . then @xmath500 we shall proceed by the method of mathematical induction . suppose this is true for any decomposable graph with number of vertices less than @xmath92 and we prove the lemma for @xmath189 . the equality trivially holds when @xmath501 . let us therefore assume that @xmath502 . as before , let @xmath503 be the number of the cliques and consider the following cases . * suppose that @xmath504 , i.e. , @xmath3 is complete . we write @xmath505 where the last equality uses the induction hypothesis for the induced graph @xmath506 . * suppose that @xmath507 . note that by lemma [ lem : combine ] we can assume that @xmath62 is induced by some perfect order @xmath285 . in particular , there exists a vertex in @xmath316 that has no child . thus , without loss of generality , assume that the vertices in @xmath62 are labeled such that @xmath508 . we now consider two cases : + a ) if the residual @xmath509 , then @xmath510 and @xmath511 are , respectively , a perfect order of the cliques and a perfect dag version of @xmath512 . moreover , @xmath513 . now it follows that @xmath514 b ) if the residual @xmath316 has more than one element , then @xmath515 is a perfect order of the cliques of @xmath512 with associated separators @xmath516 . using the induction hypothesis we obtain @xmath517 the markov ratio in the right - hand - side of the equation is the squared root of the jacobian of the inverse mapping @xmath518 . it can be shown directly that the left - hand - side of the equation is also is the squared root of the jacobian of the inverse mapping @xmath519 ( see @xcite ) . + we illustrate the result of lemma [ lem : mr ] for the decomposable graph @xmath3 and its perfect dag version @xmath62 given in figure [ fig : c1 ] and figure [ fig : cd1 ] , respectively . by using the same perfect order of @xmath17 as in example [ ex : lm2_a ] the corresponding markov ratio becomes @xmath520 next we use lemma [ lem : mr ] to prove an analog of proposition [ prop : b1 ] for the letac - massam type i wisharts . [ prop : a1 ] let @xmath3 be a non - complete ( decomposable ) graph . assume that @xmath62 is a dag version of @xmath3 induced by @xmath17 such that @xmath290 is ancestral . let @xmath381 and @xmath383 be as those defined in definition [ def : ans ] . then the dimension of the set of @xmath141 that satisfies equation is greater than or equal to @xmath384 . first note that by lemma [ lem : combine ] assuming the existence of a such @xmath62 is not vacuous . now using lemma [ lem : mr ] , we begin with rewriting the density of the type i wishart to obtain an expression in terms of relationships in the directed version @xmath62 . we proceed as follows . @xmath521 recall that in the proof of theorem [ thm : comparison ] it was showed that @xmath522 , @xmath523 for @xmath524 , and @xmath525 for each @xmath385 can be written as products of @xmath526 raised to some powers . now define @xmath527 clearly , for each @xmath528 the expression in equation reduces to @xmath529 where @xmath530 is an affine combination of the components of @xmath494 and @xmath495 . the expression in equation is the non - normalized version of the density of the generalized riesz distribution on @xmath531 , defined in @xcite , and is integrable , by equation ( 18 ) in @xcite , if and only if @xmath532 for each @xmath533 . an important observation in proposition [ prop : a1 ] is that if @xmath3 is chosen such that @xmath389 , then part ( i ) of the lm conjecture may fail . in fact we can show that the same decomposable graphs given in figure [ fig : c1 ] and figure [ fig : c2 ] provide two counterexamples to part ( i ) of the lm conjecture . since the calculations are very similar , we provide details only for the second graph . + consider the graph @xmath3 given in figure [ fig : c2 ] and its dag version given in figure [ fig : cd2 ] . note that the lm conjecture ( i ) implies that the dimension of the parameter set @xmath10 is @xmath484 since @xmath482 . now @xmath62 is induced by @xmath534 . we proceed as follows . @xmath535 let @xmath536 be the exponent of @xmath269 in equation for each @xmath537 . if we set @xmath538 , then we obtain @xmath539 the integrand in the right - hand - side of equation corresponds to the non - normalized density of the generalized riesz distribution on @xmath540 and has a finite integral if and only if @xmath541 for each @xmath537 . furthermore , under these conditions , we have @xmath542 see @xcite for details . on the other hand , by very similar calculations as those which lead to equation we can show that @xmath543 therefore , from equation , for any @xmath153 satisfying @xmath544 and @xmath538 we conclude that @xmath545 consequently , the dimension of @xmath10 is greater than or equal to @xmath546 . thus the lm conjecture ( i ) fails . in this paper we develop appropriate tools in order to carefully compare the wishart type ii distributions introduced by letac and massam in @xcite for decomposable graphs and the dag wisharts introduced by the authors in @xcite . the comparison is made when the dag wisharts are restricted to the class of perfect dags , that is where dags are markov equivalent to the class of decomposable graphs . by this comparison , we establish the fact that in general , the family of type ii wisharts is a subfamily of that of dag wisharts when the multi - parameters are restricted to a well identified set @xmath14 . in case of homogeneous graphs we show that the latter restriction is not needed . in light of this result we are led to a condition on the structure of the graphs that yield counterexamples to the second part the lm conjecture . by taking a similar approach we are also able to reject the first part of the lm conjecture and therefore completely resolve the lm conjecture .

we study the thermodynamics of clean , layered superconductor / ferromagnet nanostructures using fully self consistent methods to solve the microscopic bogoliubov - degennes equations . from these self - consistent solutions the condensation free energies are obtained . the trilayer sfs junction is studied in particular detail : first order transitions between and states as a function of the temperature are located by finding where the free energies of the two phases cross . the occurrence of these transitions is mapped as a function of the thickness of the f layer and of the fermi wavevector mismatch parameter . similar first order transitions are found for systems with a larger number of layers : examples are given in the 7 layer ( 3 junction ) case . the latent heats associated with these phase transitions are evaluated and found to be experimentally accessible . the transition temperature to the normal state is calculated from the linearized bogoliubov - degennes equations and found to be in good agreement with experiment . thus , the whole three dimensional phase diagram in space can be found . the first order transitions are associated with dips in the transition temperature to the non - superconducting state , which should facilitate locating them . results are given also for the magnetic moment and the local density of states ( dos ) at the first order transition .

cond-mat0608649

the investigation of systems involving ferromagnet ( f ) and superconductor ( s ) junctions is an active component of superconductor - based spintronics@xcite research . a broad array of interesting effects arises in s / f nanostructures , which opens doors for nanotechnologies and associated devices and applications that may offer benefits beyond current superconducting devices such as standard josephson junctions . advances in fabrication techniques permit growth of ferromagnet and superconductor layers in the form of junctions and heterostructures smooth up to the atomic scale . the arrangement of consecutive f and s layers , as in sfs junctions , results in competition between magnetic and superconducting orderings . superconducting correlations can leak into the ferromagnet while spin polarization can extend into the superconductor : these are the now well established s / f proximity effects.@xcite the phase coherence embodied in the superconducting correlations becomes modified in the f regions . the exchange energy in the ferromagnet shifts the kinetic energies of the quasiparticles constituting the cooper pairs and subsequently a new superconducting state arises whereby the center of mass momentum of the pair is nonzero.@xcite this results in a spatially decaying pair amplitude that oscillates over a characteristic length scale much smaller than the superconducting coherence length . the modulating pair amplitude within the magnet indirectly links adjacent s layers , and thus proximity effects in f cause local oscillations in physically relevant single - particle quantities , including the magnetization@xcite and density of states@xcite ( dos ) . similarly , in the s material the magnet locally polarizes the superconductor , causing a monotonic decline in the pairing correlations near the interface over an extended region . the associated spin - split andreev quasiparticle states also lead to interesting local behavior in the dos and magnetic moment in the superconductor . the nontrivial behavior of the proximity effects in these structures plays a central role in the competition between the magnetic and superconducting order . the modification of the superconducting phase coherence due to proximity effects in clean multilayers consisting of one or more successive sfs junctions is particularly striking . on the atomic level , the pair amplitude is a smoothly varying function of the spatial coordinates . depending on the values of certain parameters ( such as f layer width , @xmath3 ) the damped oscillatory pair amplitude in the f layer may arrange itself in such a manner that is energetically favorable for its sign to change from one of the s layers to the next , yielding a so - called @xmath1-junction , as first proposed long ago.@xcite if the pair amplitude does not change sign between s layers , it is an ordinary or @xmath0-junction . there is a rich and broad parameter space that then enables a certain level of control over the competing magnetic and superconducting orderings , allowing one to increase or diminish the proximity effects that dominate the relative sfs coupling . the actual equilibrium state ( @xmath0 or @xmath1 ) is dependent upon several variables , including predominantly the f region s material characteristics and the temperature , @xmath2 , all of which ultimately determine the pair amplitude modulation in the magnet . a system comprised of a larger number of sfs sequences results in a greater number of possible @xmath0 or @xmath1 junction combinations . the transitions between @xmath0 and @xmath1 states can be explored through the signatures of a variety of physical parameters . experimental study of this question has focused primarily on measurements of the critical current @xmath7@xcite and , thermodynamically , on the critical temperature@xcite of the transition to the normal state , @xmath6 . evidence of @xmath8 transitions can be seen in the sfs josephson coupling , which manifests itself in the vanishing of @xmath7 , although higher order harmonics in the current - phase relationship can modify this.@xcite measurements@xcite as a function of @xmath3 have shown that @xmath6 , which is of course smaller than @xmath9 , the critical temperature for bulk s material , oscillates as a function of @xmath3 , confirming theoretical predictions@xcite based upon the semi - classical usadel equations . intrinsically linked to this phenomenon are damped oscillations in @xmath7 as a function of @xmath3 and exchange energy in the clean@xcite and dirty limits.@xcite these changes in the critical current have been experimentally confirmed@xcite . of particular interest is ref . , which demonstrates the robustness of @xmath8 transitions by providing evidence of switching in samples with interfaces that were not atomically smooth . indeed , despite deviations as large as @xmath10 over @xmath11 of a sample , clear evidence of switching was found . near @xmath6 , and in the diffusive limit , the theory was later extended to include arbitrary interface transparency.@xcite measurements of the superconducting phase@xcite have corroborated the @xmath1 state in sfs junctions , and the predicted oscillations in several thermodynamic quantities have in many cases been found experimentally . direct evidence of dos oscillations was reported in a tunneling spectroscopy experiment,@xcite but not observed@xcite in other cases . such studies give us the valuable insight that the oscillations are correlated with @xmath12 transitions . in this work , we show that there is indeed an intimate relation between the oscillations in @xmath6 as a function of relevant parameters and the transitions from the @xmath1 to the @xmath0 state and we find good quantitative agreement with experimental data . since the possibility of having a particular junction configuration depends fundamentally on the intricate properties of the pair amplitude , the complicated and demanding task of calculating the pair potential , @xmath13 , rigorously and self - consistently becomes absolutely necessary , particularly as the inhomogeneities occur on a microscopic scale . the first step in the self - consistency process often involves an assumed simple piecewise constant form for @xmath13 , which is then iterated through the relevant equations until convergence is achieved . it is not justified to bypass the technical difficulties associated with self - consistency and to use only an assumed form for the pair potential . the final calculated @xmath13 often deviates significantly , even in overall symmetry , from the assumed form : the self - consistent @xmath13 has a complicated spatial behavior that can lead to stable states mixing @xmath0 and @xmath1 junction configurations.@xcite a self - consistently calculated pair potential minimizes , at least locally , the free energy of the system . to determine whether the calculated state is merely a local minimum of the free energy or the global one , the free energies from all possible self - consistent @xmath0- and @xmath1-junction configurations must be compared with high precision . recently developed numerical algorithms@xcite overcome the difficulties that arise in computing the small difference between much larger quantities and enable accurate computation of the differences in the values of the condensation free energy of different minima . for clean sfs junctions , a relevant set of basic parameters to consider includes @xmath3 , the exchange energy @xmath14 and @xmath2 . as these parameters vary , the @xmath0 or @xmath1-state free energies may cross at certain points in parameter space , yielding phase transitions . it has been shown@xcite that at @xmath15 , transitions occur when varying @xmath14 , @xmath3 and also the mismatch parameter @xmath4 , defined as the ratio of fermi energies in the f and s regions . this mismatch can induce a transition because at @xmath16 , when the fermi wavevectors match , the layers couple more strongly , while at small @xmath4 the coupling is effectively weaker . if the temperature varies it is also possible to have a first order transition between @xmath0 and @xmath1 junction states , as recently shown in both the clean,@xcite and dirty@xcite limits , and also predicted for short - period f / s superlattices.@xcite the temperature has been shown to have a pronounced effect on the pair amplitude in the f region of f / s structures,@xcite strongly diminishing its magnitude while maintaining its characteristic period of oscillation as @xmath2 increases . this translates into weaker coupling between adjacent s layers . if the magnet width is such that the junction is near a @xmath8 transition point at @xmath15 , increasing the temperature can result in the critical current of the junction having a non - monotonic temperature dependence.@xcite it has been argued@xcite that the transition is discontinuous in uniform samples but rounded off in samples of variable thickness.@xcite however , a transition can be observed@xcite in just a portion of samples with nonuniform thickness . these results indicate that the temperature can be used to switch between a @xmath0 and a @xmath1 state configuration . it is possible to locate regions of parameter space that give the desired transitions using the @xmath15 results as guides , however the task is still significantly demanding . such temperature transitions were found to occur in one - junction and 3-junction systems for moderate values of @xmath4.@xcite thus , a 1-junction system was found to have a @xmath17 first order transition as @xmath2 was lowered , and a @xmath18 transition was found for a 3-junction system.@xcite in each case , the free energies of a stable and a metastable state crossed at the transition temperature with differing derivatives , and therefore entropies . the existence of metastable states and an entropy discontinuity are hallmarks of first - order phase transitions . moreover , the reported latent heats were reported to be within available experimental resolution . it is therefore desirable to systematically study the coexistence of metastable states and the nature of the transition in sfs and higher - order multilayer structures . the main objective of this paper therefore , is to map out the regions of parameter space in which the different junction states are stable , and to trace the locations of the phase transitions in systems with sfs junctions . an extensive sweep of the geometric and materials parameters including @xmath3 , @xmath4 , as well as @xmath2 , is performed . to start with , it is important to know which @xmath3 and @xmath4 ranges allow more than one self - consistent state at @xmath15 . one can then check if a metastable state at low temperature becomes the equilibrium state at higher @xmath2 . by using this procedure we obtain a complete phase diagram of an sfs junction within the relevant region of @xmath19 space . to accomplish this , we use a method that can accommodate arbitrary values of the above parameters , without recourse to approximations . as discussed above , all calculations involving the pair potential must be performed using fully self - consistent algorithms , starting from the microscopic equations ( bogoliubov - degennes ( bdg ) ) . the need for a fully microscopic theory arises because the characteristic period of the pair potential oscillations approaches the atomic scale . for the nanoscale interlayer widths considered here , geometrical oscillations decisively influence the final results . we present in sec . [ sec : methods ] the microscopic equations and the associated notation relevant for systems containing sfs junctions . we review the numerical procedures involved in calculating the self - consistent pair potential and quasiparticle spectra , and the method used to calculate the primary thermodynamic quantity , the condensation free energy , @xmath20 , from the self - consistent spectrum and pair potential . we also outline a semi - analytic method to calculate @xmath6 through the linearized bdg equations . in sec . [ sec : results ] , we show that first order transitions with measurable latent heat can occur between states containing different numbers of @xmath0 and @xmath1 junctions as the temperature changes . for sfs junctions the transitions we find are from the @xmath1 to the @xmath0 state as @xmath2 increases , as found in experiment,@xcite and occur predominantly in regions where @xmath6 is low . using the @xmath6 calculated from the linearized theory and the @xmath8 phase transitions , we obtain the full phase diagram in an extended region of parameter space spanned by @xmath2 , @xmath3 , and @xmath4 . we compare our calculated oscillations in @xmath6 as a function of @xmath3 with reported nb / co experimental data@xcite and find good agreement . the systems that we study consist of slabs of clean superconductor ( s ) material separated by ferromagnetic ( f ) layers . we will emphasize trilayers consisting of one sfs junction and , as a sample of what can generally occur in multilayers , present also results for seven layer systems consisting of three sfs junctions . the thickness of the s layers in an sfs junction is denoted by @xmath21 , and that of the f layers by @xmath3 . the seven layer system consists of three sfs junctions stacked together , so that the thickness of the two inner s layers is @xmath22 . we assume that the layers are semi - infinite in the directions perpendicular to the interfaces ( the @xmath23 directions ) and that the interfaces are sharp . the spatial inhomogeneity is confined to the @xmath24 direction , allowing us to model the system as quasi one dimensional . we assume parabolic bands , thus in the transverse direction @xmath25 , where @xmath26 is the wavevector in the transverse direction and @xmath27 is the energy corresponding to the @xmath23 variables . we use the microscopic bogoliubov - degennes@xcite equations to study this inhomogeneous system . given a pair potential ( order parameter ) @xmath28 that is to be determined self consistently , the spin - up quasi - particle ( @xmath29 ) and spin - down quasi - hole ( @xmath30 ) amplitudes obey the bdg equations in the following form : @xmath31 here , @xmath32 is a single - particle hamiltonian where @xmath33 is the kinetic energy term . the continuous variable @xmath27 is decoupled from the @xmath24 direction but of course it affects the eigenvalues @xmath34 . we describe the magnetism by an exchange field @xmath35 which takes the value @xmath14 in the f material and vanishes in s. within the superconducting layers , @xmath36 is equal to @xmath37 , the fermi energy of the s layers measured from the bottom of the band , while in the ferromagnet we have @xmath38 so that in the f regions the up and down band widths are @xmath39 and @xmath40 respectively . in the seven layer case we assume parallel orientation of the magnetization in all f layers . one should not assume that @xmath41 and we therefore introduce the dimensionless fermi wavevector mismatch parameter @xmath4 by @xmath42 . usually , one has @xmath43 . it is also convenient to introduce the dimensionless magnetic strength variable @xmath44 by @xmath45 . the @xmath46 limit corresponds to the `` half - metallic '' case . we neglect interfacial scattering . the amplitudes @xmath47 and @xmath48 can be written down from symmetry relations.@xcite the required self - consistency condition for the pair potential @xmath28 is : @xmath49 where here and below the prime indicates a summation over states for which @xmath50 , where @xmath51 is the usual cutoff `` debye '' energy and it is understood that the index @xmath52 includes @xmath26 as well as the longitudinal variables . the bcs coupling @xmath53 is taken to be a constant @xmath54 in the superconductor and zero in the ferromagnet . equations ( [ eq : bdg ] ) and ( [ eq : sc ] ) comprise a non - linear set of equations . exact solutions to this set must be computed in a self - consistent manner . we follow the procedure@xcite used in previous work ; we omit the repetition of the technical details . we begin with an assumed form for @xmath28 , either from a prior calculation with similar parameters or an _ a priori _ guess ( usually a stepwise function ) , and then numerically solve eq . ( [ eq : bdg ] ) for every value of @xmath27 in the appropriate range to compute @xmath29 , @xmath30 , and @xmath34 . an expansion of all quantities in terms of sine waves is used@xcite to carry out the solution . the required matrix elements are given , for our geometries , in ref . . this resulting energy spectrum and quasiparticle amplitudes are used in eq . ( [ eq : sc ] ) to compute a new @xmath28 . we then feed this new @xmath28 back into eqs . ( [ eq : bdg ] ) and repeat this process until the fractional difference between the average of successive solutions for @xmath28 is less than a threshold value that we take to be @xmath55 . solutions obtained in this way are exact up to the chosen numerical precision . the self - consistent solution for a trilayer sfs junction can be of the @xmath1 or the @xmath0 type , with the pair potential either changing or not changing sign across the f layer , respectively . more complicated situations can occur in multilayers : for a three junction system one can encounter four symmetric states ( @xmath56 , @xmath57 , @xmath58 , @xmath59 , with each symbol corresponding to the state of each junction ) . when , for a given temperature and set of geometrical and material parameters such as @xmath44 , @xmath3 , and @xmath4 , several@xcite different self - consistent solutions , that is , local minima in the free energy , exist , the stable state must be determined by comparing the condensation free energies of the competing self - consistent states . as discussed in sec . [ sec : intro ] , when the equilibrium state changes as a function of temperature @xcite a first order phase transition can occur , with a corresponding latent heat . one of the chief goals of this paper is to study an extended region of parameter space , locating where such transitions exist and then mapping out the corresponding phase diagram . to evaluate the free energy , @xmath60 , of the self - consistent states we use the formula from ref . : @xmath61 + \frac{1}{d}\int_0^d\frac{\delta^2(z)}{g(z)}dz , \label{eq : free}\ ] ] where @xmath62 is the total thickness of the system in the @xmath24-direction . in this expression , only the energy eigenvalues appear explicitly , the eigenfunctions appearing only indirectly through the self - consistent @xmath63 it is equivalent to several other expressions found in the literature which contain the quasi - particle amplitudes explicitly , but it is computationally much more convenient . the condensation free energy is defined as @xmath64 , where @xmath65 is the free energy of the superconducting state and @xmath66 that of the non - superconducting system . we compute @xmath66 by setting @xmath67 in equations ( [ eq : bdg ] ) and ( [ eq : free ] ) . calculating @xmath68 is a significant numerical challenge : recall that in a bulk superconductor@xcite @xmath69 , where @xmath70 is the usual density of states and @xmath71 is the order parameter for the bulk superconductor at @xmath15 , which is several orders of magnitude smaller than @xmath72 . hence , to obtain @xmath73 we must subtract two numerically obtained large quantities in order to extract a difference several orders of magnitude smaller than the terms subtracted . furthermore , as we shall see , the difference in condensation free energies of competing self - consistent states ( when they occur ) is only a small fraction of the condensation free energy of each of them . to obtain sufficiently accurate values of @xmath73 requires therefore a very high degree of precision in calculating @xmath65 and @xmath66 so that we can distinguish the relatively small differences between competing states to locate phase transitions . this situation is made more challenging by the need to calculate derivatives of @xmath73 to obtain thermodynamic functions and latent heats . while the transition temperature @xmath6 from the non - superconducting to the superconducting state can be numerically calculated as the temperature at which @xmath73 vanishes , it is much easier to evaluate @xmath6 by treating @xmath28 as a small parameter and linearizing the equations . in this way the calculation is nearly entirely analytic . the amplitudes are written as @xmath74 and @xmath75 ( we have dropped the spin indices for simplicity ) . the @xmath76 and @xmath77 terms are computed from the zeroth order equation , which is obtained by setting @xmath78 in eq . ( [ eq : bdg ] ) . the form of the zeroth order equation implies that @xmath79 and @xmath80 are completely decoupled and have distinct energy spectra , denoted by @xmath81 and @xmath82 respectively . proceeding to calculate the lowest order corrections , we incorporate quasiparticle coupling through the pair potential matrix . one can then obtain @xmath83 and @xmath84 from textbook perturbation formulas . the intermediate sums are in principle over the entire zeroth order spectrum . but as a practical matter it is enough to include in these sums energies @xmath85 and @xmath86 within a few @xmath51 of the fermi level . we then expand the quasiparticle amplitudes and their first order corrections in a sine wave basis @xmath87 , e.g. @xmath88 and @xmath89 , where @xmath90 , with @xmath91 . the range of the sums over @xmath92 is formally infinite , but again it is only necessary to sum up to a wavenumber @xmath93 with an associated energy a few @xmath51 from @xmath94 . inserting these expansions into eq . ( [ eq : sc ] ) gives the lowest order correction to @xmath28 , which we then expand in the @xmath87 basis . upon taking into account the orthogonality of the basis functions , the expanded eq . ( [ eq : sc ] ) is then transformed into the matrix equation @xmath95 where the @xmath96 are the expansion coefficients of @xmath28 in terms of @xmath97 . one finds after straightforward algebra : @xmath98 here we have used @xmath99 . the integral over @xmath27 reflects the dependence of the zeroth order quasiparticle amplitudes and energies on @xmath27 , and the sum over @xmath52 is here only over longitudinal quantum numbers with the prime denoting the limitation indicated below eq . ( [ eq : sc ] ) on the energies @xmath81 and @xmath81 . the transition temperature can then be found , in analogy with standard procedures,@xcite by treating eq . ( [ eigen ] ) as an eigenvalue equation for the matrix @xmath100 . at the transition temperature @xmath6 the largest eigenvalue is unity , while if @xmath101 all eigenvalues are less than unity . unlike the free energy method described in subsection [ sub : self ] , this procedure does not require an iterative process and only the last step ( finding the eigenvalue ) must in practice be performed numerically . therefore this method is much more efficient , and it also provides a check on the numerics of our free energy . from the self - consistent amplitudes and an energy spectrum we can also calculate other quantities of interest such as the density of states ( dos ) and the magnetization . the local density of states is @xmath102 ^ 2f^\prime(\epsilon-\epsilon_n)+ \left[v_n^\sigma(z)\right]^2f^\prime(\epsilon+\epsilon_n)\right ] , \label{eq : dos}\ ] ] where @xmath103 denotes spin and @xmath104 is the first derivative of the fermi function . one can also omit the sum over @xmath103 and obtain the spin dependent dos . similarly , we have the average number density for each spin subband , @xmath105 ^ 2f(\epsilon_n ) + \left[v_n^\sigma(z)\right]^2\left[1-f(\epsilon_n)\right ] \right\}. \label{eq : spol}\ ] ] this leads to the dimensionless magnetization , @xmath106 , @xmath107 which reduces to @xmath108/ \left[(1+i)^{3/2}+(1-i)^{3/2}\right]$ ] for bulk f material , within our assumptions . this expression is in numerical agreement with our results from eq . ( [ eq : mz ] ) in sufficiently thick f layers . in this section we present and discuss our results . as explained above , trilayers consisting of one sfs junction will be emphasized but results for seven layer systems comprised of three stacked sfs junctions will also be given in order to show that the single junction results can indeed be generalized to multilayer samples . we will first discuss the results for the thermodynamics and the phase transitions that ensue . this will include a detailed discussion of the phase diagram for the sfs trilayer in the most interesting region of the three dimensional space spanned by @xmath2 , @xmath4 and the f layer thickness . a discussion of the properties of the transition temperature @xmath6 as a function of @xmath3 and a comparison with experiment follow . we will also discuss other quantities of interest , such as the pair amplitude , the density of states , and the magnetization . the results presented below will be given in terms of convenient dimensionless quantities . we measure all the lengths in units of @xmath109 , the fermi wavevector in s. we fix @xmath110 . we have taken the bcs coherence length @xmath111 equal to @xmath21 . for @xmath21 of order of or larger than @xmath111 , results are only weakly dependent on @xmath21 , hence our results are applicable to a very wide range of values of this variable , provided @xmath21 is not too small . the dimensionless thickness @xmath112 of the ferromagnetic layers will be varied over the range of interest , which corresponds to relatively small values , since at large ones the f / s proximity effects are negligible . similarly , we introduce the notation @xmath113 . the magnetic strength parameter is taken to be @xmath114 unless otherwise noted . the effects of varying @xmath44 are physically similar to those of varying @xmath115 since the pair amplitude oscillations in f are governed@xcite by the difference @xmath116 between fermi wavevectors in the spin bands in f. the fermi wavevector mismatch parameter , @xmath4 , to which results are quite sensitive , is varied over the experimentally relevant range @xmath117 . the temperature is measured in units of @xmath9 , the critical temperature of a bulk sample of the material s. we choose @xmath118 ; the dimensionless quantities calculated are not sensitive to this choice . condensation free energies will be given in units of @xmath119 , twice the absolute value of the condensation free energy of a bulk superconducting sample of the same total thickness at @xmath15 . a dimensionless measure of the latent heats will be given by dividing the corresponding entropy discontinuities by @xmath120 , the specific heat of a sample of the same overall thickness but consisting exclusively of the s material in its normal metal state at @xmath9 . we performed several checks of our numerical methods . we verified that the temperature at which the self - consistent condensation free energy goes to zero is in each case the same as the transition temperature obtained from the linearized solution to the bdg equations . for a sample with @xmath121 and @xmath122 , we quantitatively recovered the well established results@xcite for the thermodynamics , including the second order phase transition at @xmath9 and the associated specific heat discontinuity . furthermore , the spatially averaged dos computed numerically for this system shows a well defined gap at energies within @xmath71 of @xmath94 and the characteristic divergence at @xmath123 . this test is very severe since our numerical method must necessarily be more accurate for smaller systems , where fewer variables are required . thus the ability of our numerical procedures to handle the relatively large systems ( over six superconducting correlation lengths thick ) considered here is verified . the low temperature limit was extensively checked in ref . , and it was also previously verified@xcite that our methods give the correct thickness dependence of @xmath28 for a superconducting slab as found in the literature@xcite . the basic quantity that determines the phase transitions and the entire thermodynamics is the condensation free energy , @xmath68 . we will therefore begin our exposition by describing some of our results for @xmath68 at a few parameter values representative of the regions where the phase transition behavior is richest . we begin with fig . [ fig : ft3l ] which shows , for an sfs trilayer , the self - consistent condensation free energy @xmath20 plotted versus reduced temperature @xmath124 . data points were obtained at @xmath125 intervals . the values of @xmath115 and @xmath4 for which results are shown were chosen to be such that , at @xmath15 , self - consistent solutions of both the @xmath0 and the @xmath1 states are found to exist . in each panel , the free energies of the two competing states are shown . the thermodynamically stable state is of course the one with the lower free energy . the slope of @xmath20 approaches zero as @xmath126 , which indicates that the calculated entropy vanishes at @xmath15 as required by the third law of thermodynamics . the slope also approaches zero as @xmath127 , indicating that the transition to the normal state is of second order , without latent heat . the temperature at which this second order phase transition occurs , @xmath6 , is the temperature at which the lower @xmath20 vanishes . the @xmath6 found this way agrees with the independently calculated @xmath6 using the linearized bdg equations . the inherent finite - size and proximity effects cause @xmath6 to be considerably smaller than @xmath9 in all cases . two outstanding features of the results shown in this figure are the existence of a metastable state at all temperatures from zero up to @xmath6 and a first order phase transition at an intermediate temperature : the free energies of the two competing states cross at @xmath128 . in all cases shown the @xmath1 state is stable below @xmath129 and the @xmath0 state is stable above . the position of @xmath129 is marked by the vertical arrow in each panel . the value of @xmath129 changes smoothly as @xmath115 or @xmath4 are changed . one can see by comparing top and bottom panels how @xmath129 changes with @xmath115 at constant @xmath4 . in fig . [ fig : ft3l ] one can see the difference in slope between the stable state and the metastable state at @xmath129 , particularly apparent in the right panels . the existence of a metastable state and the discontinuity in the slope of the free energy of the stable state ( i.e. , the entropy ) at @xmath129 indicate the existence of a first order phase transition with an associated latent heat . that any such transition should be first order can also be expected from the change of symmetry of the pair amplitude . for a range of parameter values including those shown in this figure , the phase transition behavior is exceptionally rich . in many other regions of @xmath4 and @xmath115 parameter space the behavior is simpler : in some there is only one self consistent solution to the bdg equations at @xmath15 , while for other ranges of @xmath4 and @xmath115 a metastable state is found at low temperature but it never becomes the stable state as @xmath2 increases . it is only in some regions of parameter space that @xmath8 transitions occur as a function of @xmath2 . this question will be discussed in more detail below . vs. reduced temperature ( as in fig . [ fig : ft3l ] ) for a 7 layer system . @xmath129 is indicated by the arrows . the different curves are labeled in the legends . the top panel shows a @xmath130 transition for @xmath131 and @xmath132 . the transition occurs at @xmath133 . the middle panel shows a @xmath134 transition for @xmath131 and @xmath135 , at @xmath136 . the bottom panel shows a more pronounced @xmath137 transition for @xmath131 and @xmath138 , with @xmath139 . , width=288 ] examples of similar results for the 7 layer case are shown in fig . [ fig : ft7l ] . these are all at @xmath131 , for several values of @xmath115 . in all cases shown at least two of the four possible metastable states mentioned above exist over the entire temperature range . the states shown in each panel are the two lowest in free energy . in some cases additional states exist but with higher free energy throughout : any such states are omitted from the plots . the three panels illustrate three different types of phase transitions as @xmath2 increases : @xmath130 ( one junction flipping @xmath140 ) , @xmath134 ( three flips ) , and @xmath141 ( one flip , @xmath142 ) . each of these persists over a range of @xmath115 . the results show all of the same qualitative features as the three layer case : the slope of @xmath68 approaches zero as @xmath126 and as @xmath127 . there is again a change in the slope of the stable state at @xmath129 which shows that the transition is also of first order in multi - junction cases . we will show that the latent heats are of the same magnitude as or larger than in the 3 layer case . there is an important quantitative difference : @xmath129 varies more slowly with @xmath115 or @xmath4 and therefore the range of parameter values for which such transitions are found is wider . one can expect then , that in higher order multilayers these phenomena will be even more general . . the meaning of different curves is indicated in the legend . the location of the first order transition is marked by the bold vertical line . , width=288 ] from the free energy one can obtain the entire thermodynamics . figure [ fig : td3l ] shows some of the thermodynamic functions that can be obtained from the results shown in fig . [ fig : ft3l ] . results are shown for two quantities : the dimensionless condensation entropy @xmath143 , defined as the negative derivative of @xmath68 with respect to the reduced temperature @xmath124 , and the dimensionless condensation energy @xmath144 defined as @xmath145 . results are shown for both the stable and the metastable states as a function of reduced temperature . one sees that @xmath146 smoothly as @xmath126 and @xmath147 in each case , which is an important check on the computation . the condensation entropy , energy , and free energy all vanish at @xmath6 . in each panel a bold vertical line indicates @xmath129 . the free energy crossings correspond neither to crossings in @xmath143 nor to crossings in @xmath144 . the former follows from the phase transitions being of first order with an associated discontinuity in the entropy . both the energy and the entropy , therefore , play important roles in the phase transition . the specific heat is not shown but can be calculated by taking a further derivative . ) for the parameter sets used in fig . [ fig : ft7l ] . the different curves are labeled in the legend , extending the notation introduced in fig . [ fig : td3l ] . the location of @xmath129 is marked by the bold vertical line . , width=288 ] examples of the thermodynamic functions for the 7 layer system are shown in fig . [ fig : td7l ] . the 3 junction case is again qualitatively much like the one junction case . the entropy , energy , and free energy all go to zero in the appropriate limits . there is no crossing in energy or entropy at @xmath129 , indicating the important interplay between energy and entropy . the discontinuity in the entropy at @xmath129 also reflects a latent heat , comparable to or larger than that in the one junction case . and @xmath1 states of the sfs trilayer , at the crossing point @xmath129 , as a function of position @xmath113 . only the middle portion of the sample is shown . the f layer is delimited by the vertical dotted lines . results are presented ( top and bottom panels ) for the two sets of parameter values used , respectively , in the left and right bottom panels of fig . [ fig : ft3l ] . , width=288 ] the behavior of the cooper pair amplitude at the first order transition is illuminating . figure [ fig : pa3l ] shows , for the sfs trilayer at @xmath128 , the pair amplitude ( defined in the usual way as the average of spin up and down creation operators ) normalized to @xmath148 , its value in bulk s material at @xmath15 . results are given versus dimensionless position @xmath149 . the f region is in the middle , set off by vertical dotted lines , and only small portions of the s regions are shown . the two cases shown correspond to the two bottom panels in fig . [ fig : ft3l ] . the absolute value of the pair amplitude is discontinuous at @xmath129 : in both plots it is slightly larger for the @xmath0 state . we recall that for a bulk superconductor at @xmath15 , the free energy is proportional to the average value of the squared pair potential,@xcite and this is also @xcite approximately the case at @xmath15 for sfs layered systems when @xmath150 . even in the bulk system , however , such a relationship is not valid@xcite at finite temperature . it is therefore unreasonable to expect this proportionality in the layered case and indeed it does not occur : the pair amplitudes do not change continuously at @xmath129 . we conclude that the phase transitions are not driven by the pair amplitude . and @xmath56 states of the 7 layer system for @xmath131 and @xmath151 at the crossing point @xmath129 as a function of position @xmath149 . this is the parameter set used in the bottom panel of fig . [ fig : ft7l ] . , width=288 ] the pair amplitude for the three junction system displays properties that are very similar to those of a single junction . a representative example , corresponding to @xmath131 and @xmath151 , is shown in fig . [ fig : pa7l ] . the absolute value of the amplitude is again discontinuous at @xmath129 . it is very important that the 7 layer and the 3 layer systems have qualitatively similar properties , as this shows that the phenomena we discuss are very general . at the same time , in the 7 layer case there is a greater number of possible transitions and the regions of parameter space in which they occur as a function of @xmath2 are wider , indicating that such phenomena can be more readily found in more complicated systems . we can make qualitative predictions for the 7 layer system based on our quantitative ( but computationally less demanding ) calculations for the 3 layer system . thus , the properties of a single junction system can be generalized to systems with many junctions . is the entropy discontinuity in units of @xmath120 ( see text ) . it is plotted against the reduced temperature of the first order phase transition . the symbols joined by lines are for an sfs trilayer : the value of @xmath129 is changed along the horizontal axis by varying @xmath115 , and from curve to curve by varying @xmath4 ( see legend ) . the triangles are for the 7 layer system cases shown in fig . [ fig : ft7l ] . the vertical arrow attached to the topmost triangle indicates that it corresponds to a value @xmath152 ( off the scale ) . , width=312 ] the signature of a first order phase transition is its latent heat . in fig . [ fig : lheat ] we show results for the dimensionless latent heat @xmath153 , defined as the difference between the entropy of the stable states just above and just below @xmath129 divided by @xmath120 , the specific heat of a normal bulk sample of s material at @xmath154 . this is appropriate because @xmath155 is equal to the entropy in the free electron model . results are plotted as a function of @xmath129 . most of the results shown are for a single junction : in that case the crossing temperature is varied by changing @xmath115 for several different values of @xmath4 , as indicated by the symbols connected by straight segments . the three data points indicated by the isolated triangles correspond to the three transitions shown in fig . [ fig : ft7l ] for the 7 layer system . one of them corresponds to a value larger , by over a factor of two , than the upper end of the scale . the latent heats vanish as @xmath129 approaches @xmath0 or @xmath6 , consistent with the smaller condensation entropy of each state in those limits . however , whenever @xmath129 does not approach these limits the latent heat can exceed @xmath156 of @xmath120 for one junction , and even more for the three junction system . since we give @xmath153 in units of @xmath157 , which is an extensive property , it should be easier to observe these latent heats in larger systems . a value of @xmath158 would correspond to picojoules in actual samples of relatively small size . @xcite such latent heats can be readily observed via standard techniques used to measure specific and latent heats in films.@xcite even smaller specific heats can be measured using multiple samples : attojoule level results have been reported@xcite in electronic systems . we see therefore that whenever a first order transition occurs , the associated latent heat is observable . we have seen that in an sfs trilayer there are two kinds of phase transitions . first , there are second order phase transitions from the normal state to a superconducting state of either the @xmath0 or the @xmath1 kind . there are also , at certain ranges of the relevant parameters , first order transitions between the @xmath0 and @xmath1 superconducting states . as a practical matter , observability of the latter transitions through thermodynamic measurements requires an appreciable difference in condensation energies between the two states . this difference is an oscillatory function of @xmath115 at constant @xmath4 and @xmath44 ( see e.g. figure 3 of ref . ) with the oscillations becoming damped at large @xmath115 , since then , at any @xmath159 the proximity effects are reduced and the @xmath0 and @xmath1 states are degenerate . hence the most important regions theoretically and experimentally are at relatively small values of @xmath115 . as to @xmath4 , the entire region @xmath43 is relevant . therefore , we have mapped out the entire phase diagram of an sfs trilayer in this most relevant region of @xmath160 space in fig . [ fig : phase ] . as explained above , varying @xmath44 is equivalent to varying @xmath115 , so we use @xmath115 as the more experimentally relevant parameter . we show two views of the phase diagram to aid in the visualization of this three dimensional figure . there are three regions in this diagram , each representing one of the three possible states : @xmath0 state , @xmath1 state , and normal ( not superconducting ) state . the crossings @xmath129 are calculated from the free energies , and @xmath6 through the linearization method . the top sheet shows the superconductor / normal metal transition . as @xmath161 , @xmath162 approaches unity for all @xmath4 . at small @xmath4 the sheet also flattens , since then the fermi level of the ferromagnet is small compared to @xmath37 so that there is little interaction between the cooper pairs and the ferromagnet . the finite temperature @xmath129 transitions between @xmath0 and @xmath1 regions are located at the sheet or `` wall '' that goes from the @xmath15 plane to the @xmath6 sheet , separating the @xmath0 from the @xmath1 state regions . this wall is of course not completely vertical : its deviation from verticality is what causes first order phase transitions as a function of @xmath2 . on the smaller @xmath115 end , this wall ends because one of the two states becomes unstable : a region of parameter space is entered where only one self consistent solution exists at any temperature . coincident with this , as one can see more clearly in the right panel , @xmath6 is sharply reduced : in other words , the condensation energies of both states rise towards zero , with one actually vanishing . near this region @xmath6 has always a sharp dip . as one proceeds towards the opposite end of the wall , at larger values of @xmath115 , @xmath6 increases and the wall becomes steeper , until it eventually becomes vertical . beyond that , no transition occurs as a function of @xmath2 : the stable state is the same at all temperatures . beyond the portion shown , therefore , the wall would become completely vertical and it is not depicted because it would obscure the diagram . it is sufficient to show its behavior in the @xmath15 plane . the crossings at @xmath15 are not thermodynamic phase transitions , they merely indicate a change in the stable state as various sample parameters are changed . exploration of @xmath6 and @xmath163 for larger values of @xmath115 at several values of @xmath4 indicates the existence of other @xmath0-@xmath1 boundaries at larger values of @xmath115 . thus , one could extend the phase diagram in that direction , but as previously seen@xcite and discussed above , these additional regions are qualitatively similar to the one shown here in detail , and quantitatively less interesting . plane of the phase diagram for the 7 layer system . the regions in which each of the four possible symmetric states is the stable one are indicated by the symbols in the legend . there are also metastable states at most values of @xmath4 and @xmath115 . , width=312 ] computing a complete three dimensional phase diagram such as the one in fig . [ fig : phase ] for a 7 layer system would be very expensive in computational resources and is not necessary . we have already seen in connection with fig . [ fig : ft7l ] that first order transitions not only occur but are more abundant in such systems . further evidence is in fig . [ fig : lowt7l ] where we show the zero temperature plane of the 7 layer phase diagram . a different symbol marks regions where each of the four possible symmetric states is the stable one at very low @xmath2 . the many boundaries between the various states and the presence in many areas of metastable states ( not marked ) reflect that there may be many first order phase transitions in the 7 layer system . we can thus infer that even larger structures will have rather intricate and rich phase diagram we found that the range of parameter space over which each type of transition persists is much broader than in the 3 layer case . for example , for a fixed @xmath131 , we observed phase transitions ( see fig . [ fig : ft7l ] ) between @xmath56 and @xmath57 states for @xmath164 , transitions between @xmath57 and @xmath58 for @xmath165 and transitions between @xmath58 and @xmath59 for @xmath166 . for a fixed @xmath167 , the @xmath168 transition exists@xcite for @xmath169 . we did not search for other transitions at @xmath167 . vs. @xmath115 ( left scale , @xmath170 symbols ) for @xmath171 . also shown ( right scale ) are the energies of the @xmath0 and @xmath1 states at @xmath15 ( @xmath172 and @xmath173 signs respectively ) . note the correlation between the changes in the stable state at @xmath15 and the dips in @xmath6.,width=312 ] by taking a slice of the phase diagram in fig . [ fig : phase ] at fixed @xmath4 , one can discern regular , damped oscillations of @xmath6 with @xmath115 . in fig . [ fig : tc ] we show @xmath6 for @xmath171 over an extended range of @xmath115 . it is clear that as @xmath115 is increased the amplitude of the oscillations decreases . this is in good agreement with experiment , as we shall see in detail below . in addition to @xmath6 , this figure shows @xmath163 for the @xmath0 and @xmath1 states . in a bulk superconductor , the ratio of this dimensionless quantity to the reduced transition temperature is @xmath174 , which is confirmed here by our result for the @xmath0 state at @xmath175 . the @xmath1 state is unstable , for obvious reasons , in the @xmath176 limit . at finite values of @xmath115 this relationship between normalized condensation energy and reduced transition temperature is not strictly obeyed , but there is a qualitative correlation : increases in the absolute value of @xmath163 correspond to increases in @xmath6 . the values of @xmath115 at which the stable state switches between @xmath0 and @xmath1 correspond to the sharp dips in @xmath6 in all cases . this has also been seen in connection with fig . [ fig : phase ] and it indicates that the structure and shape of the oscillations in @xmath6 are strongly correlated with the low temperature state . the free energy data plotted have gaps , notably for the @xmath1 state near @xmath177 and for the @xmath0 state at @xmath178 . these values of @xmath115 delimit regions in which the self - consistent calculation resulted in only one state . the free energy of the vanishing state goes continuously to zero at those boundaries . the pair amplitude is found to go also smoothly to zero . the low temperature crossings at the many different values of @xmath115 suggest the location of more @xmath8 phase transitions . this is in agreement with the direct observations reported in ref . . another corroboration of this claim comes from ref . , in which the related parameter which they denote by @xmath7 ( the overall critical current of their nonuniform thickness junction ) is found to have a significant dip at the @xmath179 transition temperature . remarkably , these transitions were observed in ref . even though their samples did not have layers of uniform thickness . these two experiments and others show that this is an observable and robust phenomenon . a similar analysis for the 7 layer case showed the same correspondence between @xmath163 and @xmath6 . however , the larger number of energy crossings at @xmath15 lead to a closer spacing of energy crossings in @xmath115 , causing several local minima in @xmath163 to appear as a single broad minimum . the result was that multiple dips in @xmath6 often merged . in larger systems , the existence of a broad local minimum in @xmath6 may correspond to multiple crossings at low temperature . , in good agreement with experimental@xcite data for a nb / co system . the experimental points shown are the average of the two series reported in ref . . , width=312 ] many experimental groups @xcite have found oscillations in @xmath6 as a function of f layer thickness . our calculation also finds these oscillations ( see fig . [ fig : tc ] ) . the agreement with experiment is furthermore quantitative . we show in fig . [ fig : exp ] a direct comparison of our results to the experimental data for a nb / co system.@xcite in the experiment , the spontaneous magnetization of the co layer was found to depend on its thickness . this means , in our language , that @xmath44 must be taken as a function of @xmath115 for the purposes of this comparison . to do this , we fitted the spontaneous magnetization reported in ref . and extracted , at each thickness , the value of @xmath44 from the formula below eq . ( [ eq : mz ] ) . we took a constant @xmath180 , which is appropriate to the materials mentioned . the experimental and theoretical values are in excellent quantitative agreement on the vertical scale , and the damped oscillations are are well aligned in the thickness . comparing figures [ fig : exp ] and [ fig : tc ] , we conclude that the dips in fig . [ fig : exp ] must correspond to changes in the stable state at zero temperature . as these changes are , as we have seen , associated with the first order phase transitions , these dips in @xmath6 may also be associated with first order phase transitions , in good agreement with what was reported in ref . . this implies that studies of @xmath6 may be a useful tool for experimental discovery of first order phase transitions and that samples which show dips in @xmath6 are the ones that should be cooled down and studied to locate such phase transitions . for an sfs trilayer . the quantity plotted is the local dos as defined in eq . ( [ eq : dos ] ) , averaged over an s layer , and normalized to the normal state bulk result in s. this is the case shown in the bottom left panel of fig . [ fig : ft3l ] ( @xmath182 ) . , width=312 ] advanced tunneling spectroscopy techniques are a useful experimental tool to measure the local dos , thus probing the single - particle spectrum . it has been found @xcite previously that the local dos results for @xmath0 and @xmath1 states are different , including a modified subgap structure . in such cases , tunneling spectroscopy could be used to distinguish the states . we now investigate whether the density of states is also a suitable technique in locating phase transitions . in fig . [ fig : dos ] we show the dos , defined as the normalized local dos ( from eq . [ eq : dos ] ) averaged over one of the s layers , for a typical 3 layer system at the temperature where the first order transition occurs . the case shown is for the same parameters as in the bottom left panel of fig . [ fig : ft3l ] , with @xmath183 . the energy is normalized to the bulk s gap at zero temperature , @xmath71 , while the dos is normalized to its value in a bulk sample of s material in its normal state . for both @xmath0 and @xmath1 states maxima exist near the bulk gap edge , qualitatively reminiscent of the divergence found in a bulk superconductor . the local dos never quite goes to zero in either state , demonstrating gapless superconductivity induced by the numerous andreev bound states in the gap . the number of states in the gap is clearly larger for the @xmath0 state and the peak is markedly lower . although the dos for both states have some general similarities , the differences that do exist are well within the resolution of current tunneling probes,@xcite making the dos a potentially useful experimental technique in locating the phase transitions or identifying the stable state in the neighborhood of a transition . for an sfs trilayer at @xmath129 only the central part of the sample is plotted . results are given for the two nearly coexisting states . the case plotted is the same as in fig . [ fig : dos ] . the vertical dotted lines delimit the f region . , width=312 ] the last quantity we shall briefly describe is the local dimensionless magnetization , @xmath106 , as defined in eq . ( [ eq : mz ] ) . previous studies @xcite indicate that there is little difference between @xmath106 for the @xmath0 and @xmath1 states at low temperature . in that case it was found that @xmath106 was dominated by the exchange parameter @xmath44 and was rather insensitive to the phase of the superconducting state . there was also little magnetization induced in the s region , as @xmath106 decayed over the fermi length scale.@xcite to illustrate the effect that temperature has on this trend , we show @xmath184 versus the dimensionless length @xmath149 at @xmath128 in fig . [ fig : mz ] . in the figure , the f region is delimited by vertical dotted lines and only a small portion of the s regions is shown . consistent with ref . , there is a quick decay and oscillation of @xmath184 in the s region . there is a rise in the value of @xmath184 to about @xmath185 in the center of the f region , which is consistent with the bulk formula below eq . ( [ eq : mz ] ) for @xmath114 . indeed , as @xmath115 increases @xmath184 flattens to a value that is in good agreement with that estimate . this is not contrary to the experimental results in ref . for which we modeled the change in the saturation magnetization with @xmath115 by allowing for a @xmath115 dependent @xmath44 , since in that case the magnetic properties ( such the saturation magnetization ) of the f layer were experimentally found to depend on @xmath115 . thus , the local magnetization , while interesting for other reasons , is not a good tool for determining the thermodynamically stable state or locating phase transitions . we have rigorously studied the thermodynamics of clean sfs trilayer junctions through self - consistent solutions to the bdg equations , in the clean limit . building on previous work@xcite where @xmath8 transitions in this system were found to be possible , we have computed here the three dimensional phase diagram of a clean sfs junction over an extended and physically relevant region of the space spanned by the parameters @xmath2 , @xmath3 , and @xmath4 . we have found that the transition to the normal state is always of second order , while first order @xmath142 transitions occur , as temperature increases , over a range of @xmath4 and @xmath115 . such transitions have been found experimentally . for systems consisting of three such junctions , we have found here that a variety of first order transitions , involving @xmath8 switching of one or more junctions , occur . the phase transitions were shown to be driven by a delicate balance between the condensation energy and the entropy . the absolute value of the pair amplitude is discontinuous at the first order transition . key elements of our approach are an efficient method to accurately compute free energies and a linearization scheme that calculates @xmath6 . we have shown that dips in @xmath6 overlap with regions in parameter space where phase transitions exist , which suggest that @xmath6 studies should be useful for experimentally locating first order phase transitions . we have also calculated the variation of @xmath6 with @xmath3 and found good quantitative agreement with an experimental study@xcite of a nb / co system . we have demonstrated that the phase transitions will have measurable latent heats , even for relatively small samples , over a broad range of magnet thicknesses . another experimentally relevant quantity , the dos , was calculated and deemed a potentially useful tool in locating phase transitions . the local magnetization , however , shows little difference between two states at the first order transition . the method and results demonstrated here are expected to be applicable to even larger structures . - this work was supported in part by the university of minnesota graduate school and by the arsc at the university of alaska fairbanks ( part of the dod hpcm program ) . i. zuti , j. fabian , and s. das sarma , , 323 ( 2004 ) . andreev , sov . jetp * 19 * , 1228 ( 1964 ) [ z. eksp . . fiz . * 46 * , 1823 ( 1964 ) ] . buzdin , , 935 ( 2005 ) . demler , g.b . arnold , and m.r . beasley , , 15174 ( 1997 ) . v. n. krivoruchko and e. a. koshina , , 014521 ( 2002 ) . k. halterman and o.t . valls , , 014517 ( 2004 ) . k. halterman and o.t . valls , , 014509 ( 2002 ) . a. buzdin , , 11377 ( 2000 ) . bulaevskii , v. kuzii , and a. sobyanin , pisma zh . teor . fiz . * 25 * , 314 ( 1977 ) [ jetp lett . * 25 * , 290 ( 1977 ) ] . ryazanov , v.a . oboznov , a.y . rusanov , a.v . veretennikov , a.a . golubov , and j. aarts , , 2427 ( 2001 ) . h. sellier , c. baraduc , f. lefloch , and r. calemezuk , 257005 ( 2004 ) . frolov , d.j . van harlingen , v.a . oboznov , v. v. bolginov , and v.v . ryazanov , , 144505 ( 2004 ) . w. guichard , m. aprili , o. bourgeois , t. kontos , j. lesueur , and p. gandit , , 167001 ( 2003 ) . b. jin , g. su , and q.r . zheng , j. appl . phys . , * 96 * , 5654 ( 2004 ) . milosevi , g.r . berdiyorov , and f.m . peeters , , 147004 ( 2005 ) . frolov , d.j . van harlingen , v. v. bolginov , v.a . oboznov , and v.v . ryazanov , , 020503(r ) ( 2006 ) . t. kontos , m. aprili , j. lesueur , f. gent , b. stephanidis , and r. boursier , , 137007 ( 2002 ) . oboznov , v. v. bolginov , a. k. feofanov , v. v. ryazanov , and a. i. buzdin , , 197003 ( 2006 ) . v.l . pokrovsky and h. wei , , 104530 ( 2004 ) . y. obi , m. ikabe , t. kubo , and h. fujimori , physica c * 317 - 318 * , 149 ( 1999 ) . jiang , d. davidovi , d.h . reich , and c.l . chien , 314 ( 1995 ) . jiang , d. davidovi , d.h . reich , and c.l . chien , , 6119 ( 1996 ) . mhge , n.n . garifyanov , yu.v . goryunov , g.g . khaliullin , l.r . tagirov , k. westerholt , i.a . garifullin , and h. zabel , , 1857 ( 1996 ) . v. shelukhin , a. tsukernik , m. karpovski , y. blum , k.b . efetov , a.f . volkov , t. champel , m. eschrig , t. lfwander , g. schn , and a. palevski , , 174506 ( 2006 ) . z. radovi , n. lazarides , and n. flytzanis , , 014501 ( 2003 ) . a. radovi , m. lefvij , l. dobrosavljevi - gruji , a. i. buzdin , and j. r. clem , , 759 ( 1991 ) . b. krunavakarn and s. yoksan , physica c * 440 * , 25 ( 2006 ) . buzdin , l.n . bulaevskii , and s.v . panyukov , pisma zh . . fiz . * 35 * , 147 - 148 ( 1982 ) . [ jetp lett * 35 * , 178 - 180 ( 1982 ) ] . a.i . buzdin and m.yu . kuprianov , pisma zh . * 53 * , 308 - 312 ( 1991 ) [ jetp lett . * 53 * , 321 - 326 ( 1991 ) ] a. buzdin and i. baladi , , 184519 ( 2003 ) . t. kontos , m. aprili , j. lesuer , and x. grison , , 304 ( 2001 ) . s. reymond , p. sangiorgio , m.r . beasley , j. kim , t. kim , and k. char , , 054505 ( 2006 ) . l. crtinon , a.k . gupta , h. sellier , f. lefloch , m. faur , a. buzdin , and h.courtois , , 024511 ( 2005 ) . k. halterman and o.t . valls , , 104516 ( 2004 ) . barsic , o.t . valls and k. halterman , , 144514 ( 2006 ) . s. tollis , , 104532 ( 2004 ) . k. halterman and o.t . valls , , 224516 ( 2002 ) . j. cayssol and g. montambaux , , 012507 ( 2005 ) . a. buzdin , , 100501(r ) ( 2005 ) . de gennes , _ superconductivity of metals and alloys _ ( addison - wesley , reading , ma , 1989 ) . i. kosztin , s. kos , m. stone , and a.j . leggett , , 9365 ( 1998 ) . m. tinkham , _ introduction to superconductivity , 2d . ( dover publications , june 2004 ) . allen and r.c . dynes , , 905 ( 1975 ) . k. levin and o.t . valls , , 191 ( 1978 ) . b.p . stojkovi and o.t . valls , 3413 ( 1994 ) ; , 3374 ( 1994 ) , and references therein . see for example , page 451 of a.l . fetter and j.d . walecka , _ quantum theory of many - particle systems _ , mcgraw - hill , new york ( 1971 ) . huang , private communication . o. bourgeois , s.e . skipetrov , f.ong , and j. chaussy , , 057007 ( 2005 ) .

in recent work , li et al . ( comm . math . sci . , 7:81 - 107 , 2009 ) developed a diffuse - domain method ( ddm ) for solving partial differential equations in complex , dynamic geometries with dirichlet , neumann , and robin boundary conditions . the diffuse - domain method uses an implicit representation of the geometry where the sharp boundary is replaced by a diffuse layer with thickness that is typically proportional to the minimum grid size . the original equations are reformulated on a larger regular domain and the boundary conditions are incorporated via singular source terms . the resulting equations can be solved with standard finite difference and finite element software packages . here , we present a matched asymptotic analysis of general diffuse - domain methods for neumann and robin boundary conditions . our analysis shows that for certain choices of the boundary condition approximations , the ddm is second - order accurate in . however , for other choices the ddm is only first - order accurate . this helps to explain why the choice of boundary - condition approximation is important for rapid global convergence and high accuracy . our analysis also suggests correction terms that may be added to yield more accurate diffuse - domain methods . simple modifications of first - order boundary condition approximations are proposed to achieve asymptotically second - order accurate schemes . our analytic results are confirmed numerically in the and norms for selected test problems . karl yngve lervg and john lowengrub numerical solution of partial differential equations , phase - field approximation , implicit geometry representation , matched asymptotic analysis .

1407.7480

there are many problems in computational physics that involve solving partial differential equations ( pdes ) in complex geometries . examples include fluid flows in complicated systems , vein networks in plant leaves , and tumours in human bodies . standard solution methods for pdes in complex domains typically involve triangulation and unstructured grids . this rules out coarse - scale discretizations and thus efficient geometric multi - level solutions . also , mesh generation for three - dimensional complex geometries remains a challenge , in particular if we allow the geometry to evolve with time . in the past several years , there has been much effort put into the development of numerical methods for solving partial differential equations in complex domains . however , most of these methods typically require tools not frequently available in standard finite element and finite difference software packages . examples of such approaches include the extended and composite finite element methods ( e.g. , @xcite ) , immersed interface methods ( e.g. , @xcite ) , virtual node methods with embedded boundary conditions ( e.g. , @xcite ) , matched interface and boundary methods ( e.g. , @xcite ) , modified finite volume / embedded boundary / cut - cell methods / ghost - fluid methods ( e.g. , @xcite ) . in another approach , known as the fictitious domain method ( e.g. , @xcite ) , the original system is either augmented with equations for lagrange multipliers to enforce the boundary conditions , or the penalty method is used to enforce the boundary conditions weakly . see also @xcite for a review of numerical methods for solving the poisson equation , the diffusion equation and the stefan problem on irregular domains . an alternate approach for simulating pdes in complex domains , which does not require any modification of standard finite element or finite difference software , is the diffuse - domain method . in this method , the domain is represented implicitly by a phase - field function , which is an approximation of the characteristic function of the domain . the domain boundary is replaced by a narrow diffuse interface layer such that the phase - field function rapidly transitions from one inside the domain to zero in the exterior of the domain . the boundary of the domain can thus be represented as an isosurface of the phase - field function . the pde is then reformulated on a larger , regular domain with additional source terms that approximate the boundary conditions . although uniform grids can be used , local grid refinement near domain boundaries improves efficiency and enables the use of smaller interface thicknesses than are achievable using uniform grids . a related approach involves the level - set method @xcite to describe the implicitly embedded surface and to obtain the appropriate surface operators ( e.g. , @xcite ) . the diffuse - domain method ( ddm ) was introduced by kockelkoren et al.@xcite to study diffusion inside a cell with zero neumann boundary conditions at the cell boundary ( a similar approach was also used in @xcite using spectral methods ) . the ddm was later used to simulate electrical waves in the heart @xcite and membrane - bound turing patterns @xcite . more recently , diffuse - interface methods have been developed for solving pdes on stationary @xcite and evolving @xcite surfaces . diffuse - domain methods for solving pdes in complex evolving domains with dirichlet , neumann and robin boundary conditions were developed by li et al.@xcite and by teigen et al . @xcite who modelled bulk - surface coupling . the ddm was also used by aland et al . @xcite to simulate incompressible two - phase flows in complex domains in 2d and 3d , and by teigen et al . @xcite to study two - phase flows with soluble surfactants . li et al . @xcite showed that in the ddm there exist several approximations to the physical boundary conditions that converge asymptotically to the correct sharp - interface problem . li et al . presented some numerical convergence results for a few selected problems and observed that the choice of boundary condition can significantly affect the accuracy of the ddm . however , li et al . did not perform a quantitative comparison between the different boundary - condition approximations , nor did they estimate convergence rates . further , li et al . did not address the source of the different levels of accuracy they observed for the different boundary - condition approximations . in the context of dirichlet boundary conditions , franz et al.@xcite recently presented a rigorous error analysis of the ddm for a reaction - diffusion equation and found that the method converges only with first - order accuracy in the interface thickness parameter @xmath0 , which they confirmed numerically . similar results were obtained numerically by reuter et al . @xcite who reformulated the ddm using an integral equation solver . reuter et al . demonstrated that their generalized ddm , with appropriate choices of approximate surface delta functions , converges with first - order accuracy to solutions of the poisson equation with dirichlet boundary conditions . here , we focus on neumann and robin boundary conditions and we present a matched asymptotic analysis of general diffuse - domain methods in a fixed complex geometry , focusing on the poisson equation for robin boundary conditions and a steady reaction - diffusion equation for neumann boundary conditions . however , our approach applies to transient problems and more general equations in the same way as shown in @xcite . our analysis shows that for certain choices of the boundary condition approximations , the ddm is second - order accurate in @xmath0 , which in practice is proportional to the smallest mesh size . however , for other choices the ddm is only first - order accurate . this helps to explain why the choice of boundary condition approximation is important for rapid global convergence and high accuracy . further , inspired by the work of karma and rappel @xcite and almgren @xcite , who incorporated second - order corrections in their phase field models of crystal growth and by the work of folch et al.@xcite who added second - order corrections in phase - field models of advection , we also suggest correction terms that may be added to yield a more accurate version of the diffuse - domain method . simple modifications of first - order boundary condition approximations are proposed to achieve asymptotically second - order accurate schemes . our analytic results are confirmed numerically for selected test problems . the outline of the paper is as follows . in [ sec : dda ] we introduce and present an analysis of general diffuse - domain methods . in [ sec : discretization ] the numerical methods are described , and in [ sec : results ] the test cases are introduced and numerical results are presented and discussed . we finally give some concluding remarks in [ sec : conclusion ] . the main idea of the ddm is to extend pdes that are defined inside complex and possibly time - dependent domains into larger , regular domains . as a model problem , consider the poisson equation in a domain @xmath3 , @xmath4 with neumann or robin boundary conditions . as shown in li et al . @xcite , the results for the poisson equation can be used directly to obtain diffuse - domain methods for more general second - order partial differential equations in evolving domains . the ddm equation is defined in a larger , regular domain @xmath5 as @xmath6 see [ fig : ddadomain ] . here @xmath7 approximates the characteristic function of @xmath3 , @xmath8 and bc is chosen to approximate the physical boundary condition , cf.@xcite . this typically involves diffuse - interface approximations of the surface delta function . a standard approximation of the characteristic function is the phase - field function , @xmath9 here @xmath0 is the interface thickness and @xmath10 is the signed - distance function with respect to @xmath11 , which is taken to be negative inside @xmath3 . ( 0,0 ) rectangle(9,5 ) ; ( 2,1 ) .. controls ( 1,1 ) and ( 1,4 ) .. ( 3,3.5 ) .. controls ( 5,3 ) and ( 4,5 ) .. ( 6,4 ) node[above right ] @xmath11 .. controls ( 8,3 ) and ( 8,1.5 ) .. ( 7,1.5 ) .. controls ( 3,1.5 ) and ( 3,1 ) .. ( 2,1 ) ; at ( 2.3,2.3 ) @xmath3 ; at ( 7.6,0.8 ) @xmath12 ; at ( 4.8,2.4 ) @xmath13 ; at ( 5.4,0.5 ) @xmath14 ; as li et al . @xcite described , there are a number of different choices for bc in [ ddm ] . for example , in the neumann case , where @xmath15 on @xmath11 , one may take : @xmath16 in the robin case , where @xmath17 on @xmath11 , one may use analogous approximations : @xmath18 note that the terms @xmath19 and @xmath20 approximate the surface delta function . following li et al . @xcite we assume that @xmath21 is extended constant in the normal direction off @xmath11 and that @xmath22 is smooth up to @xmath11 and is extended into the exterior of @xmath3 constant in the normal direction . we next perform an asymptotic analysis to estimate the rate of convergence of the corresponding approximations . to show asymptotic convergence , we need to consider the expansions of the diffuse - domain variables in powers of the interface thickness @xmath0 in regions close to and far from the interface . these are called inner and outer expansions , respectively . the two expansions are then matched in a region where both are valid , see [ fig : regions ] , which provides the boundary conditions for the outer variables . we refer the reader to @xcite and @xcite for more details and discussion of the general procedure . [ yscale=0.8 , interface/.style = thick , inner/.style = fill = gray , dotted , fill opacity=0.2 , outer/.style = fill = gray , dashed , fill opacity=0.3 , labels/.style = above right , font= , ] ( 0,0 ) rectangle(9,5 ) ; ( 2,1 ) .. controls ( 1,1 ) and ( 1,4 ) .. ( 3,3.5 ) .. controls ( 5,3 ) and ( 4,5 ) .. ( 6,4 ) .. controls ( 8,3 ) and ( 8,1.5 ) .. ( 7,1.5 ) node ( g1 ) .. controls ( 3,1.5 ) and ( 3,1 ) .. ( 2,1 ) ; at ( 7.4,0.4 ) @xmath12 ; at ( 2.0,2.3 ) @xmath3 ; ( a ) at ( 2.4,1.6 ) ; ( b ) at ( 3.4,1.6 ) ; ( c ) at ( 2.4,0.6 ) ; ( d ) at ( 3.4,0.6 ) ; ( c ) rectangle ( b ) ; \(a ) ( 0 , 3 ) ; ( b ) ( 8 , 3 ) ; ( c ) ( 0,-3 ) ; ( d ) ( 8,-3 ) ; ( 0,-3 ) rectangle ( 8,3 ) ; ( a ) ( 0 , 3 ) ; ( b ) ( 8 , 3 ) ; ( 0,0 ) .. controls ( 2 , 0.5 ) and ( 3 , 0.5 ) .. ( 4,0 ) .. controls ( 5,-0.5 ) and ( 6,-1.0 ) .. ( 8,0 ) ; ( 0,2.0 ) .. controls ( 2,2.5 ) and ( 3,2.5 ) .. ( 4,2.0 ) .. controls ( 5,1.5 ) and ( 6,1.0 ) .. ( 8,2.0 ) ( 8,-2.0 ) .. controls ( 6,-3.0 ) and ( 5,-2.5 ) .. ( 4,-2.0 ) .. controls ( 3,-1.5 ) and ( 2,-1.5 ) .. ( 0,-2.0 ) cycle ; ( 0,3.0 ) ( 8,3.0 ) ( 8,1.0 ) .. controls ( 6,0.0 ) and ( 5,0.5 ) .. ( 4,1.0 ) .. controls ( 3,1.5 ) and ( 2,1.5 ) .. ( 0,1.0 ) cycle ; ( 0,-3.0 ) ( 8,-3.0 ) ( 8,-1.0 ) .. controls ( 6,-2.0 ) and ( 5,-1.5 ) .. ( 4,-1.0 ) .. controls ( 3,-0.5 ) and ( 2,-0.5 ) .. ( 0,-1.0 ) cycle ; at ( 0.1,2.25 ) outer region ; at ( 0.1,1.25 ) overlapping region ; at ( 0.1,0.25 ) inner region ; ( 8.1 , 3.0 ) node[right=0.5em ] @xmath3 ( 8.1 , 0.1 ) ; ( 8.1,-0.1 ) node[right=0.5em ] @xmath12 ( 8.1,-3.0 ) ; at ( 8.1,0 ) @xmath11 ; the outer expansion for the variable @xmath23 is simply @xmath24 the outer expansion of an equation is then found by inserting the expanded variables into the equation . the inner expansion is found by introducing a local coordinate system near the interface @xmath11 , @xmath25 where @xmath26 is a parametrization of the interface , @xmath27 is the interface normal vector that points out of @xmath3 , @xmath28 is the stretched variable @xmath29 and @xmath30 is the signed distance from the point @xmath31 to @xmath11 . in the local coordinate system , the derivatives become @xmath32 where @xmath33 is the curvature of the interface . note that @xmath34 . the inner variable @xmath35 is now given by @xmath36 and the inner expansion is @xmath37 to obtain the matching conditions , we assume that there is a region of overlap where both the expansions are valid . in this region , the solutions have to match . in particular , if we evaluate the outer expansion in the inner coordinates , this must match the limits of the inner solutions away from the interface , that is @xmath38 insert the expansions into [ eq : outer , eq : inner ] to get @xmath39 the terms on the left - hand side can be expanded as a taylor series , @xmath40 where @xmath41 and @xmath42 . now we end up with the matching equation @xmath43 which must hold when the interface width is decreased , that is @xmath44 . in the matching region it is required that @xmath45 . under this condition , if we let @xmath46 , we get the following asymptotic matching conditions : @xmath47 and as @xmath46 , @xmath48 where the quantities on the right - hand side are the limits from the interior ( @xmath49 ) and exterior ( @xmath50 ) of @xmath3 . here @xmath51 means that the expressions approach equality when @xmath52 . that is , @xmath51 is defined such that if some function @xmath53 , then we have @xmath54 . now we are ready to consider the poisson equation with robin boundary conditions , @xmath55 where @xmath56 . consider a general ddm approximation , @xmath57 where @xmath58 represents the bc approximation in the ddm . the scaling factor @xmath59 is taken for later convenience . if we assume that @xmath58 is local to the interface ( e.g. , vanishes to all orders in @xmath0 away from @xmath11 ) and that @xmath22 is independent of @xmath0 ( e.g. , is smooth in a neighbourhood of @xmath11 and is extended constant in the normal direction out of @xmath3 ) , which is the case for the approximations bc1 and bc2 given in [ eq : bc_robin ] , then the outer solution to this equation when @xmath60 satisfies @xmath61 now , if @xmath62 satisfies [ eq : poiss_robin ] and @xmath63 then the outer expansion @xmath64 and the ddm is asymptotically first - order accurate . however , if @xmath65 , then @xmath66 and the ddm is asymptotically second - order accurate . determining which of these is the case requires matching the outer solutions to the solutions of the inner equations . before we analyse the inner expansions , we develop a higher - order matching condition based on [ eq : match2,eq : match3 ] that matches a robin boundary condition for @xmath67 . first we take the derivative of [ eq : match3 ] with respect to @xmath28 and subtract @xmath68 times [ eq : match2 ] , which gives @xmath69 move the terms that make up a robin condition for @xmath67 to the left - hand side , and move the rest to the right - hand side , that is @xmath70 the laplacian can be decomposed into normal and tangential components as @xmath71 which can be shown by writing the gradient vector as @xmath72 . we can therefore write @xmath73 where we have assumed that @xmath62 satisfies the system , as demonstrated below . if we insert this into the matching condition , we get @xmath74 as @xmath75 . now consider the inner expansion of [ eq : general_dda ] , @xmath76 expand @xmath77 , @xmath78 , @xmath79 and @xmath80 in powers of @xmath0 , to get @xmath81 and then collect the leading order terms . note that because @xmath22 is smooth up to @xmath11 and extended constantly outside @xmath3 we have that @xmath82 is independent of @xmath28 . the lowest power of @xmath0 gives @xmath83 suppose that @xmath84 , which is the case as we show below for bc1 and bc2 , then we obtain @xmath85 . by the matching condition , this gives @xmath86 , where @xmath87 is the limiting value of @xmath62 . the next order terms give @xmath88 integrating from @xmath89 to @xmath90 in @xmath28 and using the matching condition , we get @xmath91 to obtain a robin boundary condition for @xmath62 , we need that @xmath92 now consider the zeroth order terms , @xmath93 if we subtract @xmath94 from both sides of [ eq : ddarobin0 ] , we get @xmath95 where we have taken into account the cancellation of terms and used the fact that @xmath82 and @xmath96 are independent of @xmath28 . the latter holds when @xmath21 is extended as a constant in the normal direction off @xmath11 , e.g. , @xmath97 and is independent of @xmath28 and @xmath0 . next , we integrate and use the matching condition on the left - hand side , @xmath98 if the right - hand side of [ order 1 bc ] vanishes , then we obtain @xmath99 from which we can conclude that the outer solution @xmath65 since @xmath67 is harmonic : @xmath100 . next we analyse the boundary condition approximations bc1 and bc2 . the bc1 approximation corresponds to @xmath101 since @xmath102 in the outer region ( interior part of @xmath3 ) , we conclude that @xmath58 vanishes in the outer region . in the inner region , we have @xmath103 where we have used that @xmath97 and is independent of @xmath28 and @xmath0 . since @xmath84 , we conclude from our analysis in [ sec : dda_robin_inner ] that @xmath104 and hence @xmath105 . the next orders of [ expansion of bc1 ] give @xmath106 a direct calculation then shows that @xmath107 as desired . thus , the leading order outer solution @xmath62 satisfies the problem ( [ eq : poiss_robin ] ) . to continue , we must first consider [ eq : first_bc ] , @xmath108 from which we get @xmath109 where @xmath110 is the limiting value of the outer solution ( e.g. , see [ eq : match2 ] ) . combining this with [ order 1 bc ] , we obtain @xmath111 further , it follows from the definition of the phase - field function that @xmath112 is an odd function . therefore the integral on the right - hand side of [ order 1 bc 1 ] is equal to zero . thus @xmath113 and so by our arguments below [ eq : dda_outer ] , the ddm with bc1 is second - order accurate in @xmath0 . when the bc2 approximation is used , we obtain @xmath114 accordingly , in the inner region , we obtain @xmath115 since @xmath116 , we get @xmath117 as desired . from [ eq : first_bc ] we have @xmath118 using that @xmath119 , this gives @xmath120 where @xmath121 combining [ u1 inner solve b2,u1 inner solve b2 a , order 1 bc ] we get @xmath122 direct calculations show that @xmath123 and @xmath124 using these in [ order 1 bc 2 ] , we get @xmath125 this shows that the ddm with bc2 is only first - order accurate because the solution @xmath67 of @xmath126 is in general not equal to @xmath127 , e.g. , @xmath128 . in order to modify bc2 to achieve second - order accuracy , we introduce @xmath129 such that @xmath130 that is , @xmath129 perturbs only the higher order terms in the inner expansion and is chosen to cancel the term on the right - hand side of [ order 1 bc 2 final ] in order to achieve @xmath131 , which in turn implies that @xmath65 and the new formulation is second - order accurate . the correction @xmath129 does not affect the @xmath132 or @xmath133 orders in the system . thus , @xmath134 and @xmath135 are unchanged from the previous subsection . now becomes @xmath136 so we wish to determine @xmath137 such that @xmath138 two simple ways of achieving this are to take @xmath139 putting everything together , we can obtain bc2 m , a second - order version of bc2 , using @xmath140 the resulting ddm is an elliptic system since @xmath141 , as required for the robin boundary condition . in each instance , this is guaranteed if the interface thickness @xmath0 is sufficiently small . thus far , we have taken advantage of integration to achieve second - order accuracy . alternatively , one may try to add correction terms to directly obtain second - order boundary conditions without relying on integration . for example , from [ order 1 bc ] to achieve second - order accuracy we may take @xmath142 where @xmath135 is a functional of @xmath143 . this provides another prescription of how to obtain a second - order accurate boundary condition , which could in principle lead to faster asymptotic convergence since it directly cancels a term in the inner expansion of the asymptotic matching . as an illustration , let us use bc1 as a starting point even though this boundary condition is already second - order accurate . through the prescription in [ local correction ] above , we derive another second - order accurate boundary condition . to see this , write @xmath144 then from [ local correction ] we get @xmath145 this can be achieved by taking @xmath146 where @xmath30 is the signed distance to @xmath11 as defined earlier . note that we can also achieve second - order accuracy by taking instead @xmath147 where we use the fact that the integral involving @xmath148 vanishes in [ order 1 bc ] . we refer to these choices , which are by no means exhaustive , as @xmath149 we remark , however , that this prescription may not always lead to an optimal numerical method . for example , when using bc1m2 , the system is guaranteed to be elliptic when @xmath150 for @xmath151 , which puts an effective restriction on the interface thickness @xmath0 depending on the values of @xmath68 and @xmath152 . when bc1m1 is used , the situation is more delicate since ellipticity can not be guaranteed when @xmath153 due to the @xmath154 term . recall that @xmath153 outside the original domain @xmath3 and so this issue is associated with the extending the modified boundary condition outside @xmath3 . in future work , we plan to consider different extensions that automatically guarantee ellipticity . to summarize , we have shown that the ddm @xmath155 is a second - order accurate approximation of the system when bc1 , bc2 m , and bc1 m are used . when bc2 is used , the ddm is only first - order accurate . since the poisson equation with neumann boundary conditions does not have a unique solution , we instead consider the steady reaction - diffusion equation with neumann boundary conditions , @xmath156 again we consider a general ddm approximation , @xmath157 under the same conditions on @xmath58 as in the previous section , the outer solution now satisfies @xmath158 as in the robin case , if @xmath62 satisfies [ eq : poiss_neumann ] and @xmath63 then the ddm is first - order accurate . however , if @xmath159 , then the ddm is second - order accurate . to construct the boundary condition for @xmath67 , we follow the approach from the robin case and combine [ eq : match3,laplace decomp ] to get @xmath160 assuming that @xmath62 satisfies the system as demonstrated below , and to get @xmath161 as @xmath60 . the inner expansion of [ eq : reacdiff_general_dda ] is analogous to the robin case derived in [ sec : dda_robin_inner ] . as before , if @xmath84 then @xmath162 and @xmath86 , the limiting value of the outer solution . still holds at the next order and so to get the desired boundary condition for @xmath62 , we need @xmath163 analogously to [ eq : ddarobin0 ] the next order equation is @xmath164 subtracting @xmath165 from [ eq : o1 ] we get @xmath166 where we have used @xmath167 as justified below . integrating , we obtain @xmath168 as in the robin case , if the right - hand side of [ order 1 bc neumann ] vanishes then we may conclude that @xmath65 since @xmath67 satisfies @xmath169 with zero neumann boundary conditions . we next analyse the boundary conditions bc1 and bc2 . when the bc1 approximation is used , we obtain @xmath170 and @xmath171 accordingly , we find that @xmath86 and [ integral constraint neumann ] holds . thus @xmath62 satisfies the system as claimed above . at the next order , from [ eq : first_bc , psi expansion for bc1 neumann ] we obtain @xmath172 thus , combining [ hat u1 neumann , order 1 bc neumann ] we get @xmath173 from which we conclude that @xmath65 and the ddm with bc1 is second - order accurate . when the bc2 approximation is used , we obtain @xmath174 and @xmath175 analogously to the case when bc1 is used , @xmath176 , [ integral constraint neumann ] holds , and @xmath62 satisfies the system . at the next order , from [ eq : first_bc , psi expansion for bc2 neumann ] we obtain @xmath177 combining [ hat u1 neumann bc2,order 1 bc neumann ] we get @xmath178 from which we conclude that @xmath65 and the ddm with bc2 is second - order accurate for the neumann problem as well , which is different from the robin case . analogously to the robin case , to achieve second - order accuracy we may also take @xmath179 following the same reasoning , alternative boundary conditions analogous to those in [ bc1 m versions ] may be derived @xmath180 note that as in the robin case , when bc1m1 is used ellipticity can not be guaranteed when @xmath153 due to the @xmath154 term . to summarize , we have shown that the ddm @xmath181 is a second - order accurate approximation of the system when bc1 , bc2 , and bc1 m are used . the equations are discretized on a uniform grid with the second - order central - difference scheme . the discrete system is solved using a multigrid method , where a red - black gauss - seidel type iterative method is used to relax the solutions ( see @xcite ) . the equations are solved in two - dimensions in a domain @xmath182 ^ 2 $ ] for all the test cases . periodic boundary conditions are used on the domain boundaries @xmath183 . the iterations are considered to be converged when the residual of the current solution has reached a tolerance of @xmath184 . since the phase - field function quickly tends to zero outside the physical domain @xmath3 , it must be regularized in order to prevent the equations from becoming ill - posed . we therefore use the modified phase - field function @xmath185 where the regularization parameter is set to @xmath186 unless otherwise specified . in addition , one should note that the chosen boundary condition for the computational domain , @xmath12 , should not interfere with the physical domain . thus one has to make sure that the distance from the computational wall to the diffuse interface of @xmath3 is large enough not to affect the results . as shown in [ sec : asymptotic_analysis ] , the normal vector and the curvature can be calculated from the phase - field function as @xmath187 and @xmath188 the surface laplacian can be found from the identity @xmath189 where @xmath190 in 2d we get @xmath191 below , we verify the accuracy of our numerical implementation on several test problems in which we manufacture a solution to the ddm with different choices of boundary conditions through particular choices of @xmath22 . as suggested by our analysis , we find that when we include the surface laplacian , we are unable to solve the discrete system using the multigrid method even though the correction term and the subsequent loss of ellipticity outside @xmath3 is confined to the interfacial region . as also mentioned in [ sec : dda_robin_other ] , future work involves developing alternative extensions of the boundary conditions outside @xmath3 that maintain ellipticity . nevertheless , as a proof of principle , we still consider the effect of this term by using the surface laplacian of the analytic solution in the ddm equations . we next investigate the performance of the ddm with different choices of boundary conditions and compare the results with the exact solution of the sharp - interface equations for the reaction - diffusion equation with neumann boundary conditions and the poisson equation with robin boundary conditions . we consider four different cases with neumann boundary conditions and three different cases with robin boundary conditions . for each case , we calculate and compare the error between the calculated solution @xmath192 and an analytic solution @xmath193 of the original pde , which is extended from @xmath3 into @xmath12 . the error is defined as @xmath194 where @xmath195 is a norm and @xmath7 is used to restrict the error to the physical domain @xmath3 . the convergence rate in @xmath0 as @xmath196 is calculated as @xmath197 for a decreasing sequence @xmath198 . in the following results we mainly use the @xmath1 norm , @xmath199 where @xmath58 is an array with @xmath200 elements . for a couple of cases , we also present the results with the @xmath2 norm , @xmath201 for a given @xmath0 , the error @xmath202 in both @xmath1 and @xmath2 is calculated by refining the grid spacing until a minimum of two leading digits converge ( e.g. , stop changing under refinement ) in the @xmath1 norm . in some cases for the smallest values of @xmath0 , the error has not yet converged to two digits in the @xmath2 norm . however , due to memory limits on our computers , we were not able to use more than @xmath203 cells in each direction . this limited our ability to obtain grid convergence , particularly when bc2 ( see below ) is used for very small values of @xmath0 . consider the steady reaction - diffusion equation with neumann boundary conditions , @xmath204 in this section we solve the ddm systems @xmath205 where bc refers to selected boundary condition approximations considered in the previous section . in the case of bc1m1 , as remarked above , the surface laplacian term is not solved , rather the surface laplacian of the analytic solution is used and is treated as a known source term . consider the case where @xmath3 is a circle of radius @xmath206 centred at @xmath207 , and where the analytic solution to the reaction - diffusion equation in @xmath3 is @xmath208 this corresponds to @xmath209 , @xmath210 , and @xmath211 . in this case , the curvature is @xmath212 . now consider the case where @xmath3 is the square @xmath213 ^ 2 $ ] . again let the analytic solution in @xmath3 be @xmath214 so that @xmath209 , @xmath210 , and @xmath215 . in this case the curvature is zero almost everywhere . to initialize the square domain @xmath3 , the signed - distance function is defined as @xmath216 the phase - field function is then calculated directly from the signed - distance function in [ eq : characteristic ] . again let @xmath3 be the circle centred at @xmath207 with radius @xmath206 , but now consider the case where the analytic solution is @xmath217 which corresponds to @xmath218 @xmath219 , and @xmath220 note that in the ddm equations , @xmath21 is extrapolated constantly in the normal direction off of the boundary @xmath11 . for the final neumann case we again let @xmath213 ^ 2 $ ] , and we consider the case where the analytic solution is @xmath221 where @xmath222 . this corresponds to @xmath223 the boundary function @xmath21 and the surface laplacian of the analytic function along the boundary are @xmath224 where @xmath225 along the bottom and top boundaries , and @xmath226 along the left and right boundaries . and [ tab : neumann_error_per_eps ] show convergence results in the @xmath1 norm where @xmath0 is reduced for cases 1 to 4 with bc1 , bc1m1 , bc1m2 and bc2 . shows the results for case 4 where the @xmath2 norm is used . although the ddm is most efficient when adaptive meshes are used , here we consider only uniform meshes to more easily control the discretization errors in order to focus on the errors in the ddm . as in all diffuse - interface methods , fine grids are necessary to accurately solve the equations when @xmath0 is small . this is especially apparent for the cases with bc2 where even the finest grid spacing , @xmath227 in each direction , becomes too coarse to obtain results for small @xmath0 that have converged with respect to the grid refinements . the results confirm the second - order accuracy of all the considered boundary condition approximations . note that while the difference between bc1 and bc1m1 tends to be small , bc1m1 consistently performs better than bc1 . in turn , bc1 performs better than bc2 . in case 2 there is a noticeable improvement of bc1m1 over bc1 . case 3 is the first case that has a nonconstant boundary condition , and the surface laplacian of the analytic solution along the boundary is also nonconstant . an unexpected result for case 3 is that bc1m2 performs the best . one possible explanation to this is errors due to grid anisotropy . therefore we also consider a fourth case , which again has a nonconstant boundary condition and nonconstant surface laplacian of the analytic solution . since the domain in this case is a square , the effect of grid anisotropy is lessened . correspondingly bc1m1 performs the best . the cases were also calculated with the @xmath2 norm , which gave similar results , although at the smallest values of @xmath0 , the orders of accuracy of bc1m1 and bc1m2 may deteriorate in @xmath2 , as seen in [ fig : neumann_case4-plot-2,tab : neumann_error_per_eps_infty ] for case 4 . this could be due to the influence of higher order terms in the expansion , or the amplification of error when @xmath0 is small due to the condition number of the system , which should scale like @xmath228 . this is currently under investigation . the difference between bc1 and bc2 is noticeable , especially with regard to the required amount of grid refinement that is needed to obtain a convergent result . this provides practical limits for the use of bc2 . plot coordinates ( 8.000e-01 , 3.4e-01 ) ( 4.000e-01 , 9.9e-02 ) ( 2.000e-01 , 2.6e-02 ) ( 1.000e-01 , 6.4e-03 ) ( 5.000e-02 , 1.6e-03 ) ( 2.500e-02 , 4.2e-04 ) ; plot coordinates ( 8.000e-01 , 4.8e-01 ) ( 4.000e-01 , 1.1e-01 ) ( 2.000e-01 , 2.7e-02 ) ( 1.000e-01 , 6.5e-03 ) ( 5.000e-02 , 1.6e-03 ) ; plot coordinates ( 8.000e-01 , 3.1e-01 ) ( 4.000e-01 , 9.5e-02 ) ( 2.000e-01 , 2.6e-02 ) ; plot coordinates ( 8.000e-01 , 2.5e-01 ) ( 4.000e-01 , 7.3e-02 ) ( 2.000e-01 , 1.9e-02 ) ( 1.000e-01 , 5.2e-03 ) ; plot coordinates ( 8.000e-01 , 2.0e-01 ) ( 4.000e-01 , 2.6e-02 ) ( 2.000e-01 , 5.2e-03 ) ( 1.000e-01 , 1.2e-03 ) ; plot coordinates ( 8.000e-01 , 2.9e-01 ) ( 4.000e-01 , 7.8e-02 ) ( 2.000e-01 , 2.0e-02 ) ( 1.000e-01 , 5.1e-03 ) ; plot coordinates ( 8.000e-01 , 2.2e-01 ) ( 4.000e-01 , 7.0e-02 ) ( 2.000e-01 , 2.0e-02 ) ; plot coordinates ( 8.000e-01 , 1.3e-01 ) ( 4.000e-01 , 3.1e-02 ) ( 2.000e-01 , 7.5e-03 ) ( 1.000e-01 , 1.8e-03 ) ( 5.000e-02 , 4.5e-04 ) ( 2.500e-02 , 1.2e-04 ) ; plot coordinates ( 8.000e-01 , 8.7e-02 ) ( 4.000e-01 , 2.9e-02 ) ( 2.000e-01 , 7.1e-03 ) ( 1.000e-01 , 1.8e-03 ) ( 5.000e-02 , 4.3e-04 ) ( 2.500e-02 , 1.1e-04 ) ; plot coordinates ( 8.000e-01 , 3.2e-02 ) ( 4.000e-01 , 1.3e-02 ) ( 2.000e-01 , 3.4e-03 ) ( 1.000e-01 , 8.6e-04 ) ( 5.000e-02 , 2.1e-04 ) ; plot coordinates ( 8.000e-01 , 1.2e-01 ) ( 4.000e-01 , 2.8e-02 ) ( 2.000e-01 , 8.1e-03 ) ; plot coordinates ( 8.000e-01 , 1.7e-01 ) ( 4.000e-01 , 4.4e-02 ) ( 2.000e-01 , 1.1e-02 ) ( 1.000e-01 , 3.0e-03 ) ; plot coordinates ( 8.000e-01 , 1.4e-01 ) ( 4.000e-01 , 2.0e-02 ) ( 2.000e-01 , 4.6e-03 ) ( 1.000e-01 , 1.2e-03 ) ; plot coordinates ( 8.000e-01 , 3.4e-01 ) ( 4.000e-01 , 5.5e-02 ) ( 2.000e-01 , 1.2e-02 ) ( 1.000e-01 , 3.1e-03 ) ; plot coordinates ( 8.000e-01 , 1.7e-01 ) ( 4.000e-01 , 4.6e-02 ) ( 2.000e-01 , 1.2e-02 ) ; plot coordinates ( 8.000e-01 , 1.8e-01 ) ( 4.000e-01 , 5.2e-02 ) ( 2.000e-01 , 1.5e-02 ) ( 1.000e-01 , 4.3e-03 ) ; plot coordinates ( 8.000e-01 , 1.5e-01 ) ( 4.000e-01 , 2.2e-02 ) ( 2.000e-01 , 1.2e-02 ) ( 1.000e-01 , 5.8e-03 ) ( 5.000e-02 , 2.8e-03 ) ; plot coordinates ( 8.000e-01 , 3.5e-01 ) ( 4.000e-01 , 6.3e-02 ) ( 2.000e-01 , 1.6e-02 ) ( 1.000e-01 , 4.6e-03 ) ( 5.000e-02 , 2.4e-03 ) ; plot coordinates ( 8.000e-01 , 1.8e-01 ) ( 4.000e-01 , 5.1e-02 ) ( 2.000e-01 , 1.4e-02 ) ( 1.000e-01 , 4.7e-03 ) ; rlrlrlrlr & bc1 & & bc1m1 & & bc1m2 & & bc2 + @xmath0 & @xmath229 & @xmath68 & @xmath229 & @xmath68 & @xmath229 & @xmath68 & @xmath229 & @xmath68 + + 0.800 & 3.391 & & 4.771 & & & & 3.091 & + 0.400 & 9.942 & 1.8 & 1.121 & 2.1 & & & 9.522 & 1.7 + 0.200 & 2.572 & 2.0 & 2.682 & 2.1 & & & 2.572 & 1.9 + 0.100 & 6.433 & 2.0 & 6.513 & 2.0 & & & & + 0.050 & 1.613 & 2.0 & 1.593 & 2.0 & & & & + 0.025 & 4.154 & 2.0 & 3.874 & 2.0 & & & & + + 0.800 & 2.461 & & 1.961 & & 2.881 & & 2.221 & + 0.400 & 7.302 & 1.8 & 2.582 & 2.9 & 7.812 & 1.9 & 6.992 & 1.7 + 0.200 & 1.942 & 1.9 & 5.213 & 2.3 & 1.962 & 2.0 & 1.952 & 1.8 + 0.100 & 5.163 & 1.9 & 1.203 & 2.1 & 5.103 & 1.9 & & + + 0.800 & 1.271 & & 8.742 & & 3.162 & & 1.181 & + 0.400 & 3.122 & 2.0 & 2.852 & 1.6 & 1.282 & 1.3 & 2.822 & 2.1 + 0.200 & 7.483 & 2.1 & 7.083 & 2.0 & 3.403 & 1.9 & 8.133 & 1.8 + 0.100 & 1.813 & 2.0 & 1.753 & 2.0 & 8.584 & 2.0 & & + 0.050 & 4.484 & 2.0 & 4.324 & 2.0 & 2.124 & 2.0 & & + 0.025 & 1.154 & 2.0 & 1.064 & 2.0 & & & & + + 0.800 & 1.711 & & 1.381 & & 3.391 & & 1.741 & + 0.400 & 4.422 & 2.0 & 2.042 & 2.8 & 5.512 & 2.6 & 4.612 & 1.9 + 0.200 & 1.142 & 2.0 & 4.583 & 2.2 & 1.242 & 2.2 & 1.202 & 1.9 + 0.100 & 2.953 & 1.9 & 1.183 & 2.0 & 3.093 & 2.0 & & + .the @xmath2 error for the neumann problem as a function of @xmath0 for case 4 . all results are calculated with @xmath227 in each direction on uniform grids . blank results indicate that the solutions require even finer grids to converge . [ cols= " > , < , > , < , > , < , > , < , > " , ] now consider the poisson equation with robin boundary conditions , @xmath230 as in the previous section , we solve the ddm equation @xmath231 using bc1 , bc2 , bc1 m and bc2 m . consider the case where @xmath3 is a circle of radius @xmath206 centred at @xmath207 , and where the analytic solution to the poisson equation in @xmath3 is @xmath208 this corresponds to @xmath232 , @xmath233 and @xmath211 . we will consider the case when @xmath234 , thus @xmath235 . again let @xmath3 be the circle at @xmath207 with radius @xmath206 , but now consider the case where the analytic solution is @xmath236 which corresponds to @xmath237 and @xmath238 again let @xmath234 so that @xmath239 . similar to the neumann case 3 , @xmath21 is extended constantly in the normal direction in the ddm equations . for the final robin case we let @xmath213 ^ 2 $ ] , and we consider a case that corresponds to the neumann case 4 where the analytic solution is @xmath221 where @xmath222 . this corresponds to @xmath240 the boundary function @xmath21 and the surface laplacian of the analytic function along the boundary are @xmath241 where @xmath225 along the bottom and top boundaries , and @xmath226 along the left and right boundaries . the convergence results calculated with the @xmath1 norm are presented in [ fig : robin_cases - plot1,fig : robin_cases - plot2,fig : robin_cases - plot3 ] and [ tab : robin_error_per_eps ] . shows the results for case 2 where the @xmath2 norm is used . again the results indicate that bc1m1 performs better than bc1 , although both methods are second - order accurate , as predicted by our analysis . the results also show that bc1 gives better results than bc2 , which is approximately first - order accurate as also predicted by theory . further , as in the neumann case , bc2 is seen to require very fine grids to converge . for small @xmath0 , the requirement exceeds our finest grid . the modified bc2m1 and bc2m2 schemes are also tested . the results with bc2m2 are almost indistinguishable from the results with bc2m1 , so only the latter results are shown in the following figures . all results are listed in [ tab : robin_error_per_eps ] and [ tab : robin_error_per_eps_infty ] . the bc2 m schemes are shown to perform better than the bc2 scheme , but they also require very fine grids to converge . further , the orders of accuracy of bc2m1 and bc2m2 seem to deteriorate somewhat at the smallest values of @xmath0 . as discussed in [ sec : dda_robin_bc2 m ] , this could be due to the influence of higher order terms in the expansion , or the amplification of error and is under study . plot coordinates ( 8.000e-01 , 2.1e-01 ) ( 4.000e-01 , 4.4e-02 ) ( 2.000e-01 , 9.0e-03 ) ( 1.000e-01 , 2.0e-03 ) ( 5.000e-02 , 4.6e-04 ) ( 2.500e-02 , 1.2e-04 ) ; plot coordinates ( 8.000e-01 , 1.2e-01 ) ( 4.000e-01 , 2.7e-02 ) ( 2.000e-01 , 6.4e-03 ) ( 1.000e-01 , 1.6e-03 ) ( 5.000e-02 , 3.8e-04 ) ; plot coordinates ( 8.000e-01 , 1.7e-01 ) ( 4.000e-01 , 5.2e-02 ) ( 2.000e-01 , 2.2e-02 ) ( 1.000e-01 , 1.0e-02 ) ; plot coordinates ( 8.000e-01 , 1.14e-01 ) ( 4.000e-01 , 2.29e-02 ) ( 2.000e-01 , 6.87e-03 ) ( 1.000e-01 , 2.54e-03 ) ; plot coordinates ( 8.000e-01 , 4.21e-01 ) ( 4.000e-01 , 1.02e-01 ) ( 2.000e-01 , 2.19e-02 ) ( 1.000e-01 , 4.77e-03 ) ( 5.000e-02 , 1.09e-03 ) ( 2.500e-02 , 2.57e-04 ) ; plot coordinates ( 8.000e-01 , 1.27e-1 ) ( 4.000e-01 , 4.87e-2 ) ( 2.000e-01 , 1.30e-2 ) ( 1.000e-01 , 3.27e-3 ) ( 5.000e-02 , 8.15e-4 ) ( 2.500e-02 , 2.04e-4 ) ; plot coordinates ( 8.000e-01 , 3.84e-1 ) ( 4.000e-01 , 9.16e-2 ) ( 2.000e-01 , 2.51e-2 ) ( 1.000e-01 , 9.13e-3 ) ; plot coordinates ( 8.000e-01 , 3.48e-1 ) ( 4.000e-01 , 7.24e-2 ) ( 2.000e-01 , 1.56e-2 ) ( 1.000e-01 , 4.65e-3 ) ; plot coordinates ( 8.000e-01 , 3.80e-1 ) ( 4.000e-01 , 9.39e-2 ) ( 2.000e-01 , 2.14e-2 ) ( 1.000e-01 , 4.90e-3 ) ( 0.500e-01 , 1.15e-3 ) ( 0.250e-01 , 2.78e-4 ) ; plot coordinates ( 8.000e-01 , 1.36e-1 ) ( 4.000e-01 , 4.30e-2 ) ( 2.000e-01 , 1.06e-2 ) ( 1.000e-01 , 2.53e-3 ) ( 5.000e-02 , 6.09e-4 ) ( 2.500e-02 , 1.49e-4 ) ; plot coordinates ( 8.000e-01 , 3.24e-1 ) ( 4.000e-01 , 7.49e-2 ) ( 2.000e-01 , 1.91e-2 ) ( 1.000e-01 , 6.98e-3 ) ; plot coordinates ( 8.000e-01 , 3.04e-1 ) ( 4.000e-01 , 6.94e-2 ) ( 2.000e-01 , 1.72e-2 ) ( 1.000e-01 , 6.40e-3 ) ; plot coordinates ( 8.000e-01 , 7.9e-02 ) ( 4.000e-01 , 1.6e-02 ) ( 2.000e-01 , 3.7e-03 ) ( 1.000e-01 , 9.0e-04 ) ; plot coordinates ( 8.000e-01 , 3.6e-02 ) ( 4.000e-01 , 7.4e-03 ) ( 2.000e-01 , 1.7e-03 ) ( 1.000e-01 , 4.3e-04 ) ; plot coordinates ( 8.000e-01 , 8.2e-01 ) ( 4.000e-01 , 1.9e-02 ) ( 2.000e-01 , 5.9e-03 ) ; plot coordinates ( 8.000e-01 , 6.75e-02 ) ( 4.000e-01 , 1.27e-02 ) ( 2.000e-01 , 2.78e-03 ) ; clrlrlrlrlr & bc1 & & bc1m1 & & bc2 & & bc2m1 & & bc2m2 + @xmath0 & @xmath229 & @xmath68 & @xmath229 & @xmath68 & @xmath229 & @xmath68 & @xmath229 & @xmath68 & @xmath229 & @xmath68 + + 0.800&2.111 & & 1.201 & & 1.701 & & 1.141 & & 1.151 + 0.400&4.402&2.3 & 2.722&2.1 & 5.212&1.7&2.292&2.3&2.312 & 2.3 + 0.200&8.993&2.3 & 6.423&2.1 & 2.182&1.3&6.873&1.7&6.923 & 1.7 + 0.100&1.953&2.2 & 1.573&2.0 & 1.032&1.1&2.543&1.4&2.553 & 1.4 + 0.050&4.574&2.1 & 3.794&2.0 + 0.025&1.234&1.9 + + 0.800&4.211 & & 1.271 & & 3.841 & & 3.481 & & 3.491 & + 0.400&1.021&2.0&4.872&1.4&9.162&2.1 & 7.242&2.3&7.262&2.3 + 0.200&2.192&2.2&1.302&1.9&2.512&1.9 & 1.562&2.2&1.572&2.2 + 0.100&4.773&2.2&3.273&2.0&9.133&1.5 & 4.653&1.7&4.673&1.7 + 0.050&1.093&2.1&8.154&2.0 + 0.025&2.574&2.1&2.044&2.0 + + 0.800&7.892 & & 3.602 & & 8.232 & & 6.752 & & 6.802 & + 0.400&1.642&2.3&7.383&2.3 & 1.892&2.2&1.272&2.3&1.282&2.4 + 0.200&3.703&2.2&1.713&2.1 & 5.903&1.7&2.783&2.2&2.813&2.2 + 0.100&9.044&2.0&4.284&2.0 + clrlrlrlrlr & bc1 & & bc1m1 & & bc2 & & bc2m1 & & bc2m2 + @xmath0 & @xmath229 & @xmath68 & @xmath229 & @xmath68 & @xmath229 & @xmath68 & @xmath229 & @xmath68 & @xmath229 & @xmath68 + + 0.800&3.801 & & 1.361 & & 3.241 & & 3.041 & & 3.051 & + 0.400&9.392&2.0&4.302&1.7&7.492&2.1 & 6.942&2.1&6.952&2.1 + 0.200&2.142&2.1&1.062&2.0&1.912&2.0 & 1.722&2.0&1.732&2.0 + 0.100&4.903&2.1&2.533&2.1&6.983&1.5 & 6.403&1.4&6.423&1.4 + 0.050&1.153&2.1&6.094&2.1 + 0.025&2.784&2.1&1.494&2.0 + shows a plot of the solutions of case 1 with @xmath242 at @xmath243 . the plot shows the solutions with bc1 ( black dashed ) , bc1m1 ( black doted ) , bc2 ( blue dashed ) and bc2m1 ( blue dotted ) . we see that the solutions with the modified schemes bc1m1 and bc2m1 perform better than the corresponding schemes with bc1 and bc2 . . the solutions with bc2m1 and bc2m2 are indistinguishable , so only bc2m1 is shown . ( a ) shows the full slice , where the domain boundary is depicted as thin vertical lines at @xmath244 . ( b ) a zoom - in that shows the solutions near the left boundary.,scaledwidth=80.0% ] . the solutions with bc2m1 and bc2m2 are indistinguishable , so only bc2m1 is shown . ( a ) shows the full slice , where the domain boundary is depicted as thin vertical lines at @xmath244 . ( b ) a zoom - in that shows the solutions near the left boundary.,scaledwidth=80.0% ] we have performed a matched asymptotic analysis of the ddm for the poisson equation with robin boundary conditions and for a steady reaction - diffusion equation with neumann boundary conditions . our analysis shows that for certain choices of the boundary condition approximations , the ddm is second - order accurate in the interface thickness @xmath0 . however , for other choices the ddm is only first - order accurate . this is confirmed numerically and helps to explain why the choice of boundary - condition approximation is important for rapid global convergence and high accuracy . this helps to explain why the choice of boundary - condition approximation is important for rapid global convergence and high accuracy . in particular , the boundary condition bc1 , which arises from representing the surface delta function as @xmath19 , is seen to give rise to a second - order approximation for both the neumann and robin boundary conditions and thus is perhaps the most reliable choice . the boundary condition bc2 , which arises from approximating the surface delta function as @xmath20 yields a second - order accurate approximation for the neumann problem , but only first - order accuracy for the robin problem . in addition , bc2 requires very fine meshes to converge . our analysis also suggests correction terms that may be added to yield a more accurate diffuse - domain method . we have presented several techniques for obtaining second - order boundary conditions and performed numerical simulations that confirm the predicted accuracy , although the order of accuracy may deteriorate at the smallest values of @xmath0 possibly due to amplification errors associated with conditioning of the system or the influence of higher order terms in the asymptotic expansion . this is currently under study . further , the correction terms do not improve the mesh requirements for convergence . a common feature of the correction terms is that the interface thickness must be sufficiently small in order for the ddm to remain an elliptic equation . in addition , one choice of boundary condition involves the use of the surface laplacian of the solution , which could in principle lead to faster asymptotic convergence since it directly cancels terms in the inner expansion of the asymptotic matching . however , the extension of this term outside the domain of interest can cause the loss of ellipticity of the ddm . as such , this is an intriguing but not a practical scheme . nevertheless , as a proof of principle , we still considered the effect of this term , however , by using the surface laplacian of the analytic solution in the ddm . we found that this choice gave the smallest errors in nearly all the cases considered . by incorporating different extensions of the boundary conditions in the exterior of the domain that automatically guarantee ellipticity , we aim to make this method practical . this is the subject of future investigations . we plan to extend our analysis to the dirichlet problem where the boundary condition approximations considered by li et al . @xcite seem only to yield first - order accuracy @xcite . our asymptotic analysis thus has the potential to identify correction terms that can be used to generate second - order accurate diffuse - domain methods for the dirichlet problem . kyl acknowledges support from the fulbright foundation for a visiting researcher grant to fund a stay at the university of california , irvine . kyl also acknowledges support from statoil and gdf suez , and the research council of norway ( 193062/s60 ) for the research project enabling low emission lng systems . jl acknowledges support from the national science foundation , division of mathematical sciences , and the national institute of health through grant p50gm76516 for a center of excellence in systems biology at the university of california , irvine . the authors gratefully thank bernhard mller ( ntnu ) and svend tollak munkejord ( sintef energy research ) for helpful discussions and for feedback on the manuscript . the authors also wish to thank the anonymous reviewers for comments that greatly improved the manuscript . , _ an efficient second - order accurate cut - cell method for solving the variable coefficient poisson equation with jump conditions on irregular domains _ , int . fluids , 52 ( 2006 ) , pp . 723 - 748 . , _ solving partial differential equations on irregular domains with moving interfaces , with applications to superconformal electrodeposition in semiconductor manufacturing _ , j. comput . phys . , 227 ( 2008 ) , pp . 6411 - 6447 . , _ a diffuse - interface approach for modeling transport , diffusion and adsorption / desorption of material quantities on a deformable interface _ , communications in mathematical sciences , 7 ( 2009 ) , pp . 10091037 . , _ a second - order sharp numerical method for solving the linear elasticity equations on irregular domains and adaptive grids application to shape optimization _ , j. comput . phys . , 233 ( 2013 ) , 430 - 448 .

we present the method that the pair production of particles is described as a fock space state vector , where the pair production stems from receiving energy from , as an example , gravitational background field . at the same time we show a sort of new mechanism that in the diffeomorphism - invariant dirac sea the weyl symmetry is broken due to the pair production and this provides us a new physical interpretation of the origin of weyl anomaly . in the present paper at the first place , we consider two kinds of background space - time which are connected each other by the weyl transformation . it is shown that the vacuum , i.e. dirac sea , in the curved " space - time making weyl transformation is nothing but the pair produced state provided we consider the state in terms of the fock space basis in the flat " space - time making the weyl transformation . when summing up with respect to the time the contribution of pair production of each time - slice we obtain the ricci scalar and then it is related to the trace part of the energy - momentum tensor to produce weyl anomaly . oiqp-15 - 03 * a state description of pair production from dirac sea + in gravitational field + physical interpretation of weyl anomaly * yoshinobu habara , holger b. nielsen and masao ninomiya okayama institute for quantum physics , 1 - 9 - 1 kyoyama , okayama 700 - 0015 , japan niels bohr institute , university of copenhagen , 17 blegdamsvej , dk2100 , denmark habara@yukawa.kyoto-u.ac.jp , hbech@nbi.dk , msninomiya@gmail.com pacs numbers : 04.60.kz , 11.10.-z , 11.10.kk , 11.10.gh , 11.30.-j keywords : field theory , pair production , gravitation , dirac sea , dirac fermion , weyl anomaly

1503.05340

it is well known that the virtual pair production and destruction occur always when the external field acts to the vacuum . further these virtual pair produced particles gain the energy from the external field and then these particles appear as real particles . in quantum electrodynamics ( qed ) or similar , in particular the case of electron - positron in the presence of the external electric field this is known as the schwinger effect @xcite and has been well analyzed theoretically . so far the analysis have been done by calculating the vacuum energy density @xmath4 and then the imaginary part @xmath5 is considered as the particle - antiparticle pair production rate defined as @xmath6 it is physically interpreted that the pair production is due to the tunnel effect , because the energy has the imaginary part . a direct physical clarification of the pair production in the background field was made by two of us of the present authors ( h.b.n . and m.n . ) @xcite . when the electric field is applied to the negative energy electrons in the dirac sea , the right - mover particles and left - mover holes gain the energy and are pumped up or down , the pair production occurs . this mechanism is depicted in figure [ pic : pair.prod-0 ] where the blue circle denotes the electron and the white circle is the positron . the purpose of the present article is twofold . the first purpose is to reveal in a quantum field theoretical method the dynamics of the pair production . we show in more detail and precisely that the pair production state is a superposition of all possible particle and antiparticle state . in fact we will show that the pair production state is indeed a superposition of one pair , two pairs , three pairs , ... of particle and antiparticle with various energies . the second theme of the present paper is a kind of new physical derivation of the weyl anomaly . in 1974 , d. m. capper and m. j. duff discovered the weyl anomaly @xcite that the trace of the energy - momentum tensor for a massless weyl fermion in a gravitational background is not as classically expected to be @xmath7 , but rather @xmath8 here @xmath9 is the square root of the determinant of the metric tensor and @xmath0 is the ricci scalar . later many recalculations of this anomaly has been made ( see the review , for example @xcite ) . the present article is an attempt to obtain a physical understanding of how this anomaly comes about , i.e. to reveal the mechanism that the pair production from the dirac sea leads the weyl anomaly . as such it is also inspired by the interpretation of the chiral anomaly as pumping up or down states from or to the dirac sea @xcite of two of us ( m.n and h.b.n ) . an another attempt to understand the @xmath10 dimensional weyl anomaly by our thinking on the dirac sea we presented in @xcite . as setting up we consider @xmath10 dimensional euclidean cylinder type space - time with the conformal flat metric @xmath11 . in this space - time we investigate the vacuum - to - vacuum transition amplitude of the massless dirac fermion which has weyl symmetric action . we then make a basis transformation of ( or , in other words , pull back ) the amplitude into the fock space in the space - time with another metric @xmath12 . then we can show that the dirac sea in the space - time specified by the primed metric @xmath13 ranging from @xmath14-past to @xmath15-future is expressed on the basis in the space - time with @xmath16 as a pair produced state on each time - slice . by summing up these pair production contribution from @xmath14-past to @xmath15-future , the difference between the vacuum ( or casimir ) energy in the space - time of the metric @xmath13 and that of the metric @xmath16 gives rise to the ricci scalar . in this way we finally obtain the weyl anomaly . the organization of this paper is as follows . in the following section [ sec1 ] we summarize the construction of the weyl particles and dirac sea in the gravitational background . in the section [ sec2 ] we then formulate the state description of the pair production and evaluate the weyl anomaly using the dirac sea formalism . in section [ sec3 ] we conclude and give an outlook as to how possibly to extend the use of our formalism . we have in addition three appendices . in order to describe pair production state as a quantum state in a fock space , we consider a complex fermion @xmath17 on the 2-dimensional euclidean cylinder with @xmath18 we adopt the notations that the roman indices @xmath19 and greek indices @xmath20 represent flat and curved space - time respectively . the diffeomorphism - invariant action for @xmath21 reads @xmath22 & = \frac{1}{2\pi}\int d^2\vec{x}\sqrt{g}\ > \psi^{\dagger}(\vec{x } ) \gamma^0\gamma^ie_i^{\ > \mu}(\vec{x})\nabla_{\mu}\psi ( \vec{x } ) \nonumber \\ & = \frac{1}{4\pi}\int d^2\vec{x}\sqrt{g}\ > \left\{\psi^{\dagger}(\vec{x})\gamma^0\gamma^ie_i^{\ > \mu}(\vec{x } ) \nabla_{\mu}\psi ( \vec{x})-\nabla_{\mu}\psi^{\dagger}(\vec{x})\cdot \gamma^0\gamma^ie_i^{\ > \mu}(\vec{x})\psi ( \vec{x } ) \right\ } \nonumber \\ & \equiv \int d^2\vec{x}\ > l[g,\psi ] . \label{eq:1-diffeo.action}\end{aligned}\ ] ] by using the degrees of freedom of diffeomorphism , we may take the metric tensor as the following conformal flat form @xmath23 and thus the zweibeins read @xmath24 in this article we always use the conformal metric ( [ eq:1-conformal.metric ] ) as the background space - time . we assume that the space - time is flat both for infinite past and future , i.e. @xmath25 so that @xmath26 under the background space - time ( [ eq:1-conformal.metric ] ) the action ( [ eq:1-diffeo.action ] ) becomes @xmath22=\frac{1}{4\pi}\int d^2\vec{x}\ > \left\ { \big(e^{\frac{1}{2}\omega ( \vec{x})}\psi^{\dagger}\big)\gamma^0\gamma^i \partial_i\big(e^{\frac{1}{2}\omega ( \vec{x})}\psi \big ) -\partial_i\big(e^{\frac{1}{2}\omega ( \vec{x})}\psi^{\dagger}\big ) \cdot \gamma^0\gamma^i \big(e^{\frac{1}{2}\omega ( \vec{x})}\psi \big)\right\}. \label{eq:1-action}\end{aligned}\ ] ] here @xmath27 and thus @xmath28{g}=\sqrt[4]{e}$ ] . the action ( [ eq:1-action ] ) apparently possesses the following weyl invariance , @xmath29 here we assumed the variation of the metric denoted by @xmath30 also satisfies @xmath31 we investigate the weyl transformation of the hamiltonian density . the hamiltonian density is given from the action ( [ eq:1-diffeo.action ] ) as @xmath32(\vec{x } ) . \label{eq:1-hamiltonian.density}\end{aligned}\ ] ] the form ( [ eq:1-hamiltonian.density ] ) transforms under the weyl transformation ( [ eq:1-weyl.invariance ] ) as @xmath33 this shows that @xmath34 is weyl invariant . we then quantize the system by finding the solutions for @xmath35 and @xmath36 . the equations of motion from ( [ eq:1-action ] ) read @xmath37 and when imposing the periodic boundary condition @xmath38 the general solution is given by @xmath39 } \frac{e^{-\frac{1}{2}\omega ( \vec{x})}}{e^{n(t+ix ) } } , \quad \psi^{\dagger}(\vec{x } ) = \frac{1}{\sqrt{2\pi}}\sum_{n=-\infty}^{+\infty } b_n^{[\psi,\omega]\dagger } e^{-\frac{1}{2}\omega ( \vec{x})}e^{n(t+ix ) } , \nonumber \\ & \bar{\psi}(\vec{x } ) = \frac{1}{\sqrt{2\pi}}\sum_{n=-\infty}^{+\infty } \bar{b}_n^{[\psi,\omega ] } \frac{e^{-\frac{1}{2}\omega ( \vec{x})}}{e^{n(t - ix ) } } , \quad \bar{\psi}^{\dagger}(\vec{x } ) = \frac{1}{\sqrt{2\pi}}\sum_{n=-\infty}^{+\infty } \bar{b}_n^{[\psi,\omega]\dagger } e^{-\frac{1}{2}\omega ( \vec{x})}e^{n(t - ix)}. \label{eq:1-solution}\end{aligned}\ ] ] the suffix @xmath40 $ ] on the expanding coefficients in ( [ eq:1-solution ] ) which represents the original field @xmath21 and the background metric @xmath41 is added for later use . thus the eigenfunctions , i.e. the basis of the hilbert space read @xmath42}(\vec{x})=\frac{1}{\sqrt{2\pi } } \frac{e^{-\frac{1}{2}\omega ( \vec{x})}}{e^{n(t+ix ) } } , \quad f_n^{[\omega]\dagger}(\vec{x } ) = \frac{1}{\sqrt{2\pi}}e^{-\frac{1}{2}\omega ( \vec{x})}e^{n(t+ix ) } , \nonumber \\ & \bar{f}_n^{[\omega]}(\vec{x})=\frac{1}{\sqrt{2\pi } } \frac{e^{-\frac{1}{2}\omega ( \vec{x})}}{e^{n(t - ix ) } } , \quad \bar{f}_n^{[\omega]\dagger}(\vec{x } ) = \frac{1}{\sqrt{2\pi}}e^{-\frac{1}{2}\omega ( \vec{x})}e^{n(t - ix)}. \label{eq:1-basis}\end{aligned}\ ] ] here the hermitian conjugation @xmath43 is understood as the complex conjugation in hilbert space , after making wick rotation @xmath44 , @xmath45 the inner product in the hilbert space is defined on the time - slice @xmath46 as @xmath47 } { } _ { [ \omega]}\langle x|m\rangle = \oint dx\ > \sqrt{g } f_n^{[\omega]\dagger}(\vec{x})e_0^{\ > 0}f_m^{[\omega]}(\vec{x } ) = \delta_{n , m } , \nonumber \\ & \langle \bar{n}|\bar{m}\rangle = \int dx\ > \langle \bar{n}|x\rangle_{[\omega ] } { } _ { [ \omega]}\langle x|\bar{m}\rangle = \oint dx\ > \sqrt{g } \bar{f}_n^{[\omega]\dagger}(\vec{x})e_0^{\ > 0 } \bar{f}_m^{[\omega]}(\vec{x})=\delta_{n , m}. \label{eq:1-inner.product}\end{aligned}\ ] ] here the zweibein component @xmath48 comes from the diagonal component of the matrix @xmath49 which makes the inner product covariant . we make the canonical second quantization @xmath50 and then we have @xmath51},b_m^{[\psi,\omega]\dagger}\right\ } = \left\{\bar{b}_n^{[\psi,\omega]},\bar{b}_m^{[\psi,\omega]\dagger } \right\ } = \delta_{n , m } , \quad \text{others}=0 . \label{eq:1-commutator}\end{aligned}\ ] ] we should notice that the commutation relation ( [ eq:1-commutator ] ) does not depend on the conformal factor function @xmath52 of the metric ( [ eq:1-conformal.metric ] ) . in other word the algebras of the creation and annihilation of the particle are all isomorphic , independent of @xmath52 . in the present article we consider the euclidean space - time , and in particular in the case of the flat space - time @xmath53 , the energy operator is given by @xmath54 , and the momentum one is @xmath55 . thus e.g. the eigenstate @xmath56 has the energy and momentum eigenvalues @xmath57 and @xmath58 respectively . therefore left mover @xmath59 has the dispersion relation @xmath60 while the right mover @xmath61 has @xmath62 . from the above consideration , the vacuum state , i.e. dirac sea is generically defined as @xmath63}= \prod_{n\geq 0}b_n^{[\psi,\omega]\dagger}|0\rangle \otimes \prod_{m<0}\bar{b}_m^{[\psi,\omega]\dagger}|\bar{0}\rangle \label{eq:1-dirac.sea}\end{aligned}\ ] ] such that @xmath64}|0\rangle = \bar{b}_n^{[\psi,\omega]}|\bar{0}\rangle = 0 \quad \text{for all $ n$. } \end{aligned}\ ] ] it should be noticed that the index @xmath65 in @xmath66}$ ] and @xmath67}$ ] is the momentum . needless to say the statement the dirac sea is the state filled by the negative energy particles " is meaningful only when @xmath53 , i.e. the background space - time is flat . in this case the configuration of the particles in the dirac sea is depicted by figure [ pic : dirac ] . the operators @xmath68 } , b_n^{[\psi,0]\dagger } , \bar{b}_n^{[\psi,0]}$ ] and @xmath69\dagger}$ ] are the creation- and annihilation - operators of the particles as is shown in table [ tab : spectrum ] . = -2pt .the interpretation of the creation and annihilation operators [ cols= " < , < " , ] from now on , the pair production states are represented as @xmath70(t)}\end{aligned}\ ] ] in short . strictly speaking , we can use the word pair production " only when @xmath71 , i.e. in the flat space - time . here , we use it symbolically but it expresses the point . by applying the formula @xmath72 of a matrix @xmath73 to ( [ eq:2-weyl.dirac.a ] ) , we obtain @xmath74(t ) } \nonumber \\ & = \exp \left\{\sum_{n\ ! \text { or } \ ! \bar{n}\geq 0}\int dx\ > \frac{1}{2}\omega ( \vec{x } ) \big(\langle n|x\rangle_{[\omega](t ) } { } _ { [ \omega](t)}\langle x|n\rangle + \langle \bar{n}|x\rangle_{[\omega](t ) } { } _ { [ \omega](t)}\langle x|\bar{n}\rangle \big)\right\}|\text{sea}\rangle_{[\tilde{\psi},\omega](t ) } \nonumber \\ & \quad + |\text{pair}\rangle_{[\tilde{\psi},\omega](t ) } \label{eq:2-weyl.dirac.c}\end{aligned}\ ] ] obviously the ket vector ( [ eq:2-weyl.dirac.c ] ) belongs to the fock space @xmath75 and its hermitian conjugation should be taken in terms of the inner product of @xmath75 . the expression ( [ eq:2-weyl.dirac.c ] ) is the dirac sea @xmath76(t)}$ ] at the time - slice @xmath46 arising from the weyl transformation and the factor in front of @xmath77(t)}$ ] will turn out to be a weyl anomaly . such a dirac sea stems from pair production of the particles . strictly speaking the pair production state @xmath78(t)}$ ] does not contribute when taking the expectation value at each time interval of the amplitude ( [ eq:2-identity_g ] ) in the first order of @xmath79 and thus the pair production state does not directly contribute to the weyl anomaly . however by considering ( [ eq:2-weyl.dirac.c ] ) , the pair production state @xmath78(t)}$ ] takes some norm from @xmath76(t)}$ ] and create the state @xmath77(t)}$ ] which contributes directly to the weyl anomaly . in this sense we may say that the pair production state is the origin of the weyl anomaly . furthermore if we take into account of the higher order terms of @xmath79 , the state including states more than two pairs production may directly contribute to the weyl anomaly . the weyl anomaly is produced by summing up through all history from the infinite past to infinite future as ( [ eq:2-identity_d ] ) the exponential factor appearing in front of the dirac sea @xmath77(t)}$ ] at each time - slice @xmath80 of ( [ eq:2-weyl.dirac.c ] ) . in ( [ eq:2-weyl.dirac.c ] ) the summation @xmath81 may diverge and its divergence corresponds to the ultra - violet one of the space coordinate @xmath82 . on the other hand the product @xmath83 in ( [ eq:2-identity_d ] ) may also diverge . as we will see in ( [ eq:2-weyl.dirac.c_temp ] ) , this divergence turns out to be that of the integration over the energy " @xmath84 , i.e. the ultra - violet divergence of the time coordinate @xmath46 . we should notice that this divergence is natural and plausible because we adopt the general covariant theory . in the next subsection we shall indeed perform the derivation of the weyl anomaly in the above method . in this subsection we compute @xmath85 in ( [ eq:2-identity ] ) by the method described in the last part of the previous subsection b. firstly we divide the time direction into an infinitesimal time interval @xmath86 and at each time in the amplitude ( [ eq:2-identity_d ] ) we apply the changing formula of the basis ( [ eq:2-identity_g ] ) of the projections . as is described there , @xmath85 is obtained by summing up over all time the contribution ( [ eq:2-weyl.dirac.c ] ) and its hermitian conjugation coming from the weyl transformation of the intermediate dirac sea at each time - slice . then @xmath85 reads @xmath87(t_n)}{}_{[\omega](t_n)}\langle x|n\rangle + \langle \bar{n}|x\rangle_{[\omega](t_n)}{}_{[\omega](t_n)}\langle x| \bar{n}\rangle \big)\right\}. \label{eq:2-weyl.dirac.c'}\end{aligned}\ ] ] here @xmath86 is defined in ( [ eq:2-discretized_t ] ) . the reader should remember that the product @xmath83 is taken by the step @xmath88 . as we show explicitly in eq . ( [ eq:2-weyl.dirac.c_temp ] ) , this product with respect to @xmath89 is exponentiated and becomes a summation @xmath90 . we furthermore take the continuum limit , @xmath91 , and it becomes the integration over the time @xmath46 , @xmath92 . it should be also noted that the overall factor @xmath93 in the exponent comes from the fact that the bra and ket vectors both contribute . we should be aware the fact that ( [ eq:2-weyl.dirac.c ] ) is non - unitary transformation because of the reality of the exponent which is the consequence of the reality of the matrix @xmath94 in ( [ eq:2-weyl.trans - matrix ] ) . this is rather naturally interpreted similar to the case of schwinger mechanism @xcite and hawking radiation , because the non - unitarity generates the pair production . on the other hand , as will be shown in appendix b , the conformal transformation does not lead to pair production since the contribution by the transformation is unitary ( the exponent is pure imaginary ) . in the course of the calculation of @xmath85 , we face the divergence problem in the sense of @xmath95 . we would like to use the heat kernel regularization to tame this divergence . we now consider the differential operator in the action ( [ eq:2-rescaled.action ] ) @xmath96 and solve the eigenvalue problem @xmath97 with the eigenvalues @xmath98 and a set of complete orthonormal eigenfunctions @xmath99 . then the product @xmath83 appearing in ( [ eq:2-weyl.dirac.c ] ) , we exponentiate to make it the summation over @xmath89 . thus after all it becomes the trace of each time @xmath46 , @xmath100(t_n)}{}_{[\omega](t_n)}\langle x|n\rangle + \langle \bar{n}|x\rangle_{[\omega](t_n)}{}_{[\omega](t_n)}\langle x| \bar{n}\rangle \big)\right\ } \nonumber \\ & = \exp \bigg\{\sum_{n\geq 0}\sum_{n=-\infty}^{+\infty}\int dx\ > \int_{-\frac{\pi}{\delta t}}^{+\frac{\pi}{\delta t}}dk^0 \sqrt{\frac{\delta t}{2\pi}}e^{ik^0t_n}\omega ( t_n , x ) \nonumber \\ & \qquad \qquad \qquad \qquad \qquad \times \big ( { } _ { [ \omega](t_n)}\langle x|n\rangle \langle n|x\rangle_{[\omega](t_n ) } + { } _ { [ \omega](t_n)}\langle x|\bar{n}\rangle \langle \bar{n}| x\rangle_{[\omega](t_n)}\big ) \sqrt{\frac{\delta t}{2\pi}}e^{-ik^0t_n}\bigg\ } \label{eq:2-weyl.dirac.c''_temp}\end{aligned}\ ] ] here @xmath101 denotes the basis of the plane wave with respect to the discretized time @xmath102 and @xmath103 is the continuum limit @xmath91 of it means the trace @xmath104 . however according to the pauli s argument the eigenstate of the time @xmath105 can not exist in our case of the present article . therefore we use the plane wave @xmath106 instead of @xmath107 . if we treat the dirac sea not in @xmath93-dimensional but in @xmath108-dimensional space - time , @xmath105 can exist since the energy spectrum is continuous and unbounded . the normalization " of the basis @xmath109 is written as @xmath110 . when we take the limit @xmath91 , it becomes @xmath111 , i.e. the volume of the time . we take the continuum limit @xmath91 , and the summation @xmath112 and the time step @xmath113 join to the integration over @xmath46 , @xmath92 . then the exponent of ( [ eq:2-weyl.dirac.c_temp ] ) becomes the integration over the whole space - time , @xmath114(t)}\langle x|n\rangle \langle n|x\rangle_{[\omega](t)}+ { } _ { [ \omega](t)}\langle x|\bar{n}\rangle \langle \bar{n}|x\rangle_{[\omega](t)}\big)e^{-ik^0t}\right\ } \nonumber \\ & = \exp \left\{\int \frac{dk^0}{2\pi}\int d^2\vec{x}\ > \omega ( \vec{x } ) e^{ik^0t}\big({}_{[\omega](t)}\langle x|x\rangle_{[\omega](t)}\big ) e^{-ik^0t}\right\ } \nonumber \\ & = \exp \left\{\int \frac{d^2\vec{k}}{(2\pi ) ^2}\int d^2\vec{x}\ > \omega ( \vec{x})e^{i\vec{k}\cdot \vec{x}}e^{-i\vec{k}\cdot \vec{x } } \right\}. \label{eq:2-weyl.dirac.c''}\end{aligned}\ ] ] note that it is here understood that the @xmath115-integration is cut off at @xmath116 as can be seen from ( [ eq:2-weyl.dirac.c_temp ] ) explicitly . this means that @xmath113 is really representing an ultraviolet cut - off . this cut - off interpretation of @xmath113 also explains the mysterious appearance of ( [ eq:2-weyl.dirac.c ] ) in which the various factors the exponentials contain no compensating dependence on the stepping distance in time @xmath88 , so that @xmath85 gets strongly depending on this stepping distance @xmath113 ( interpreted as cut - off this strange dependence may be o.k . ) . the kets @xmath117 and @xmath118 are the complete set in @xmath119 . in fact the quantities of @xmath120 in the exponent of the first line of ( [ eq:2-weyl.dirac.c ] ) is independent of @xmath65 , because @xmath121}\langle x|n\rangle$ ] is just a plane wave essentially . thus we may rewrite @xmath122 and use there the completeness relation @xmath123 . by changing the basis from @xmath124 to @xmath99 , @xmath85 in ( [ eq:2-weyl.dirac.c ] ) reads @xmath125 . \label{eq:2-weyl.dirac.d}\end{aligned}\ ] ] we then next introduce the exponential regulator of @xmath126 that involves the regularization cut - off @xmath127 . the divergent quantity @xmath85 in ( [ eq:2-weyl.dirac.d ] ) is now redefined as @xmath128 . \label{eq:2-weyl.dirac.e}\end{aligned}\ ] ] finally by expanding in powers of the cut - off @xmath127 and inserting @xmath129 and @xmath130 we obtain @xmath131 \nonumber \\ & = \exp \left[\frac{1}{2\pi}\lim_{m\to \infty}\int d^2\vec{x } \ > \omega ( \vec{x } ) \left\{m^2\sqrt{g}+\frac{1}{12}\sqrt{g}r\right\}\right ] . \label{eq:2-weyl.dirac}\end{aligned}\ ] ] the finite part of ( [ eq:2-weyl.dirac ] ) is proportional to the exponentiated ricci scalar and we may expect that it is the weyl anomaly . however , we need the relation with the trace @xmath132 of the energy - momentum tensor . indeed we relate the ricci scalar in the exponent of ( [ eq:2-weyl.dirac ] ) and @xmath132 by using the second and third lines in the identity ( [ eq:2-identity ] ) . we in fact derive the weyl anomaly by using the second and third lines of the identity ( [ eq:2-identity ] ) , @xmath133\right)|\text{sea}\rangle = \langle \text{sea}|j_{\omega}\cdot t\exp \left(-\int d^2\vec{x}\ > h[\tilde{\psi},g]\right ) |\text{sea}\rangle . \label{eq:2-anomaly.identity}\end{aligned}\ ] ] here the limit @xmath134 is taken . thus the dirac sea of the initial and final states in ( [ eq:2-anomaly.identity ] ) are those on the flat space - time @xmath135(t=\pm \infty)}. \end{aligned}\ ] ] the @xmath85 is obtained by renormalizing the divergent term proportional to @xmath136 in ( [ eq:2-weyl.dirac ] ) into the cosmological constant term , @xmath137 . \label{eq:2-jacobian}\end{aligned}\ ] ] remembering the argument in the subsection [ sec2.1 ] the terms in the left and right hands sides are both obtained by integrating the same field @xmath138 . however we rewrite them by using different metrics to produce the different effective actions @xmath139=\int dt\ > e_0^{[\tilde{\psi},\omega]}$ ] , @xmath140 } = j_{\omega}\cdot e^{-s_{\text{eff}}[g = e^{2\omega}]}. \end{aligned}\ ] ] by taking logarithm on both sides , the difference of the effective action @xmath141 becomes to the differentiation with respect to the infinitesimal @xmath79 , @xmath142 this is the equality what we are looking for , i.e. the relation of the weyl anomaly : @xmath143 in the present article we presented that when the dirac sea in the background space - time with the metric @xmath144 is expanded in terms of the fock space basis of the space - time with the metric @xmath41 there appears the state of the sum of the dirac sea and a state called @xmath78}$ ] . in particular , in the case of @xmath71 , the state @xmath145}$ ] describes the one of the particle - antiparticle pairs . in perturbation theory , between the formulation in terms of the empty vacuum and the formulation in terms of the dirac sea in which the negative energy states are all filled , there is no difference in the result . however in the case of the nonperturbative phenomena such as the particle - antiparticle pair production , there appears apparent difference as we have shown . the origin of this comes from the fact that the vacuum is not empty but there is a sea of the infinitely many negative energy particles and these are transformed under the weyl transformation . when the dirac sea , which is defined in a diffeomorphism invariant manner , is acted by the weyl transformation , the pairs of the particle and antiparticle are created and the weyl symmetry is broken . we have shown that this is precisely the origin of the weyl anomaly . in the fujikawa s method based on the path integral formalism @xcite , the path integral measure violates the symmetry and the weyl anomaly stems from the jacobian of the weyl transformation . in our operator formalism , the pair production from the dirac sea on each time - slice , which occurs during the transition from @xmath14-past to @xmath15-future , breaks the weyl symmetry . the weyl anomaly is obtained by summing up the breaking effects over all the time . in this manner , we show the formulation of the physical relation between the pair production state and the appearance of the weyl anomaly . our formulation can be applied to many physical phenomena , irrelevant to the detail of the models . for instance the state expression of the pair production can be used by changing the gravitational field to the electromagnetic field to investigate the dynamics of the electrons and holes in the materials . as for the physics of the high energy particle and the universe we are under investigation and some of the topics will be shown in the following section [ sec3.2 ] . really the main motivation for the present work is the hope that we may extend our formulation to slightly different but related issues : 1 . the pair production of bosons to generate the weyl anomaly : + in our previous works @xcite , we formulated the boson sea which is the vacuum being filled with negative energy bosons . if we set the background metric to the conformally flat one and transform the boson sea by the weyl transformation , we can construct the pair produced state from the boson sea vacuum , which would be the bosonic version of ( [ eq:2-weyl.dirac.a ] ) . all we should do may be parallel to the method shown in the present article . all other anomalies such as the gravitational anomaly being generated via the pair production from the sea : + we can apply our method to the physical interpretations of all anomalies and unify them in terms of the pair production from the sea vacuum . 3 . the hawking radiation and the gravitational wave : + when we consider the sea vacuum in the schwarzschild or de sitter space - time background , we can express the hawking radiation as the pair production state . in particular , in the case of the complex boson and the boson sea , we can represent the gravitational wave by using the pair production state as the hawking radiation , because the number of the degrees of freedom of the gravitational wave is two which is identical to that of the complex boson . then we may be able to evaluate the entropies of the black hole and the inflationary universe using the notion of the typicality @xcite or the formulation of statistical mechanics based on the thermal pure quantum states @xcite . the authors acknowledge yukinori nagatani for useful discussion in the early stage of the present work . one of us ( h.b.n ) thanks for hospitality to stay at the niels bohr institute as emeritus professor . one of the authors ( m.n ) would like to thank the niels bohr institute for hospitality extended to him during his stay at n.b.i . m.n is supported by the jsps grant in aid for scientific research no . the rescaling ( [ eq:2-rescale ] ) in section [ sec2 ] is meant that dirac sea becomes invariant under diffeomorphism . in this appendix we give a proof of the invariance . under the infinitesimal coordinate transformation @xmath146 the rescaled fermion field @xmath147{g}\psi$ ] transforms as @xmath148 as is done for ( [ eq:2-weyl.trans - relation ] ) the transformation rules of the creation and annihilation operators under the diffeomorphism ( [ eq : app - fermion.trans ] ) is written as @xmath149 } & = \oint \frac{dx}{2\pi}\ > e^{-\omega ( \vec{x})}\cdot e^{\frac{1}{2}\omega ( \vec{x})}e^{n(t+ix ) } \left(1+\epsilon^{\mu}\partial_{\mu } + \frac{1}{2}\partial_{\mu}\epsilon^{\mu}\right ) \sum_{m=-\infty}^{+\infty}b_m^{[\tilde{\psi},\omega ] } \frac{e^{\frac{1}{2}\omega ( \vec{x})}}{e^{m(t+ix ) } } \nonumber \\ & \equiv \sum_{m=-\infty}^{+\infty}d_{n , m}b_m^{[\tilde{\psi},\omega ] } , \nonumber \\ \bar{b}_n^{[\tilde{\psi}^{\prime},\omega ] } & = \oint \frac{dx}{2\pi}\ > e^{-\omega ( \vec{x})}\cdot e^{\frac{1}{2}\omega ( \vec{x})}e^{n(t - ix ) } \left(1+\epsilon^{\mu}\partial_{\mu } + \frac{1}{2}\partial_{\mu}\epsilon^{\mu}\right ) \sum_{m=-\infty}^{+\infty}\bar{b}_m^{[\tilde{\psi},\omega ] } \frac{e^{\frac{1}{2}\omega ( \vec{x})}}{e^{m(t - ix ) } } \nonumber \\ & \equiv \sum_{m=-\infty}^{+\infty}\bar{d}_{n , m } \bar{b}_m^{[\tilde{\psi},\omega]}. \label{eq : app - diffeo - relation}\end{aligned}\ ] ] here we defined the representation matrices of the infinitesimal diffeomorphism as @xmath150 } \left(\epsilon^{\mu}\partial_{\mu } + \frac{1}{2}\partial_{\mu}\epsilon^{\mu}\right ) { } _ { [ \omega]}\langle x|m\rangle , \nonumber \\ \bar{d}_{n , m } & \equiv \delta_{n , m}+\oint dx\ > \langle \bar{n}|x\rangle_{[\omega ] } \left(\epsilon^{\mu}\partial_{\mu } + \frac{1}{2}\partial_{\mu}\epsilon^{\mu}\right ) { } _ { [ \omega]}\langle x|\bar{m}\rangle . \label{eq : app - diffeo - matrix}\end{aligned}\ ] ] by using these expressions , we can estimate the transformation law of the dirac sea under the diffeomorphism up to the first order of the infinitesimal parameter @xmath151 as @xmath152(t)}^d & = \prod_{n\geq 0}b_n^{[\tilde{\psi}^{\prime},\omega]\dagger}|0\rangle \otimes \prod_{\bar{n}<0 } \bar{b}_{\bar{n}}^{[\tilde{\psi}^{\prime},\omega]\dagger } |\bar{0}\rangle \nonumber \\ & = \prod_{n\geq 0}\left(\sum_{m=-\infty}^{+\infty}d_{n , m}^{\dagger } b_m^{[\tilde{\psi},\omega]\dagger}\right)|0\rangle \otimes \prod_{\bar{n}<0}\left(\sum_{\bar{m}=-\infty}^{+\infty } \bar{d}_{\bar{n},\bar{m}}^{\dagger } \bar{b}_{\bar{m}}^{[\tilde{\psi},\omega]\dagger}\right)|\bar{0}\rangle \nonumber \\ & = \det \left[\delta_{n , m}+\oint dx\ > { } _ { [ \omega](t)}\langle x|n\rangle \left(\epsilon^{\mu}\partial_{\mu } + \frac{1}{2}\partial_{\mu}\epsilon^{\mu}\right ) \langle m|x\rangle_{[\omega](t)}\right]_{n , m\geq 0 } \nonumber \\ & \qquad \times \det \left[\delta_{\bar{n},\bar{m } } + \oint dx\ > \langle \bar{n}|x\rangle_{[\omega](t ) } \left(\epsilon^{\mu}\partial_{\mu } + \frac{1}{2}\partial_{\mu}\epsilon^{\mu}\right ) { } _ { [ \omega](t)}\langle x|\bar{m}\rangle \right]_{\bar{n},\bar{m}>0}| \text{sea}\rangle_{[\tilde{\psi},\omega](t ) } \nonumber \\ & \quad + |\widetilde{\text{pair}}\rangle_{[\tilde{\psi},\omega](t)}^d \nonumber \\ & = \exp \bigg[\sum_{n\geq 0}\oint dx\ > \bigg\{{}_{[\omega](t)}\langle x|n\rangle \left(\epsilon^{\mu}\partial_{\mu } + \frac{1}{2}\partial_{\mu}\epsilon^{\mu}\right ) \langle n|x\rangle_{[\omega](t ) } \nonumber \\ & \qquad \qquad \qquad \qquad \quad + \langle \bar{n}|x\rangle_{[\omega](t ) } \left(\epsilon^{\mu}\partial_{\mu } + \frac{1}{2}\partial_{\mu}\epsilon^{\mu}\right ) { } _ { [ \omega](t)}\langle x|\bar{n}\rangle \bigg\}\bigg ] |\text{sea}\rangle_{[\tilde{\psi},\omega](t ) } \nonumber \\ & \quad + |\widetilde{\text{pair}}\rangle_{[\tilde{\psi},\omega](t)}^d . \label{eq : app - diffeo.dirac.a}\end{aligned}\ ] ] then , when we consider the vacuum - to - vacuum amplitude , we obtain the following factor @xmath153 for the diffeomorphism instead of @xmath85 in ( [ eq:2-weyl.dirac.c ] ) for the weyl transformation . we also change the basis to @xmath99 and finally @xmath154 \nonumber \\ & = \exp \left[\sum_{r=0}^{+\infty}\int d^2\vec{x}\ > \partial_{\mu } \left\{\epsilon^{\mu}\chi_r^{\dagger}(\vec{x})\chi_r(\vec{x})\right\ } \right ] \nonumber \\ & = 1 . \label{eq : app - diffeo.dirac}\end{aligned}\ ] ] in this way the contribution from the dirac sea to the potential anomaly like the identity ( [ eq:2-identity ] ) vanishes as total derivative . in this appendix b , we consider the transformation of dirac sea under the conformal transformation . as is explained briefly below , in the conformal field theory , it is a field theory on the flat space - time . thus it may be reasonable to suppose that the conformal transformation does not contribute to the pair production unlike to the case of the weyl transformation in curved background space - time . firstly we make the action ( [ eq:1-diffeo.action ] ) by diffeomorphism such that the metric tensor becomes the form of the conformal flat ( [ eq:1-conformal.metric ] ) . we then use the complex coordinate @xmath155 by making the coordinate transformation , @xmath156 the metric ( [ eq:1-conformal.metric ] ) is assumed to be of the form of the product of holomorphic and anti - holomorphic part , @xmath157 it is written in terms of the zweibein @xmath158 such that @xmath159 then the action reads @xmath160 . \label{eq : app - action}\end{aligned}\ ] ] the action ( [ eq : app - action ] ) is further reduced to be of the flat space - time one @xmath161 by applying the weyl invariance @xmath162 we now construct the dirac sea by making the standard canonical second quantization . the field expansion is performed under the periodic boundary condition @xmath163 and given as @xmath164 the commutation relations of the creation and annihilation operators are of the usual one @xmath165 thus the dirac sea is constructed as @xmath166 now the action ( [ eq : app - conformal.action ] ) is invariant under the following infinitesimal conformal transformation @xmath167 this transformation can be rewritten as the expansion coefficients of the fields @xmath35 and @xmath36 . in the same manner as the case of the weyl transformation ( [ eq:2-weyled.field.expansion ] ) and ( [ eq:2-field.expansion ] ) , the fields before and after the conformal transformation are laurent expanded by using the same basis as @xmath168 we then obtain the transformation rule of @xmath169 and @xmath170 , @xmath171 here the matrix representation of the conformal transformation is defined by @xmath172 as a byproduct , it is easily shown that ( [ eq : app - conformal.trans - relation ] ) is indeed the bogolyubov transformation in the following : @xmath173 \nonumber \\ & \qquad \qquad \times \left[\delta_{m , k}-\oint \frac{dz}{2\pi i}\ > z^{-m-\frac{1}{2}}\left\{\frac{1}{2}\partial \epsilon ( z)+\epsilon ( z ) \partial \right\}z^{k-\frac{1}{2}}\right ] \nonumber \\ & = \delta_{n , m}-\oint \frac{dz}{2\pi i}\ > z^{n-\frac{1}{2}}\left\{\frac{1}{2}\partial \epsilon ( z)+\epsilon ( z ) \partial \right\}z^{-m-\frac{1}{2 } } \nonumber \\ & \qquad \qquad -\oint \frac{dz}{2\pi i}\ > z^{-m-\frac{1}{2}}\left\{\frac{1}{2}\partial \epsilon ( z)+\epsilon ( z ) \partial \right\}z^{n-\frac{1}{2 } } \nonumber \\ & = \delta_{n , m}-\oint \frac{dz}{2\pi i}\ > \partial \left\{\epsilon ( z)z^{n - m-1}\right\ } \nonumber \\ & = \delta_{n , m}. \label{eq : app - commutator.conformal}\end{aligned}\ ] ] similarly @xmath170 satisfies the same property as ( [ eq : app - commutator.conformal ] ) . we now consider the transformation of the dirac sea under conformal transformation . the procedure of the calculation is much the same as that of weyl transformation . in the vacuum - to - vacuum amplitude pair production state does not contribute as is shown in weyl transformation so that we abbreviate it in the following : @xmath174_{n , m\geq 0 } \nonumber \\ & \qquad \times \det \left[\delta_{n , m}+\oint \frac{d\bar{z}}{2\pi i}\ > \bar{z}^{-n-\frac{1}{2 } } \left\{\frac{1}{2}\bar{\partial}\bar{\epsilon}(\bar{z})+\bar{\epsilon } ( \bar{z})\bar{\partial}\right\}\bar{z}^{m-\frac{1}{2}}\right]_{n , m>0}| \text{sea}\rangle \nonumber \\ & \quad + |\text{pair}\rangle^c \nonumber \\ & = \exp \left[-\sum_{n\geq 0}\oint \frac{dz}{2\pi i}\ > z^{n-\frac{1}{2}}\left\{\frac{1}{2}\partial \epsilon ( z ) + \epsilon ( z)\partial \right\}z^{-n-\frac{1}{2}}\right ] \nonumber \\ & \qquad \times \exp \left[+\sum_{n>0}\oint \frac{d\bar{z}}{2\pi i}\ > \bar{z}^{-n-\frac{1}{2 } } \left\{\frac{1}{2}\bar{\partial}\bar{\epsilon}(\bar{z})+\bar{\epsilon } ( \bar{z})\bar{\partial}\right\}\bar{z}^{n-\frac{1}{2}}\right ] |\text{sea}\rangle + |\text{pair}\rangle^c \nonumber \\ & = \exp \bigg[-\sum_{n\geq 0}\oint \frac{dz}{2\pi i}\ > \left\{\frac{1}{2}\partial \frac{\epsilon ( z)}{z } -n\frac{\epsilon ( z)}{z^2}\right\ } \nonumber \\ & \qquad \qquad \qquad + \sum_{n>0}\oint \frac{d\bar{z}}{2\pi i}\ > \left\{\frac{1}{2}\bar{\partial } \frac{\bar{\epsilon}(\bar{z})}{\bar{z } } + n\frac{\bar{\epsilon}(\bar{z})}{\bar{z}^2}\right\}\bigg ] |\text{sea}\rangle + |\text{pair}\rangle^c . \tag{\ref{eq : app - conformal.dirac}a } \label{eq : app - conformal.dirac.a}\end{aligned}\ ] ] in ( [ eq : app - conformal.dirac.a ] ) we drop the total derivative term with respect to @xmath175 and @xmath176 and then insert the laurent expansion of @xmath151 and @xmath177 such that @xmath178 the equation ( [ eq : app - conformal.dirac.a ] ) reads @xmath179 |\text{sea}\rangle + |\text{pair}\rangle^c \nonumber \\ & = \exp \left[-\frac{i}{12}\text{im } \epsilon_0\right ] |\text{sea}\rangle + |\text{pair}\rangle^c . \label{eq : app - conformal.dirac}\end{aligned}\ ] ] in the last equality we used the @xmath180-function regularization @xmath181 . in ( [ eq : app - conformal.dirac ] ) the factor in front of the dirac sea @xmath182 is nothing but the phase and we may conclude that the conformal transformation corresponds to the unitary transformation except for the pair production state . furthermore in the factor @xmath183 and @xmath184 are the zero modes of @xmath185 and @xmath186 respectively . among them the physical meaningful terms are only the sum @xmath187 which is the change of the conformal part of the metric . we may set @xmath188 . thus under the conformal transformation we conclude that as far as we concern the vacuum - to - vacuum amplitude ( [ eq:2-identity ] ) , the dirac sea does not produce any contribution like the pair production . this result coincides with the fact that the conformal field theory is just a closed theory in the flat space - time . we start from the second line of ( [ eq:2-weyl.dirac.a ] ) and treat the left and right movers independently . for the left mover , we can assume that the creation oeprators @xmath189\dagger}$ ] of the dirac sea ( see e.g. ( [ eq:1-dirac.sea ] ) ) is arranged in the ascending order as to the index @xmath65 : @xmath190(t ) } = b_0^{[\tilde{\psi},\omega]\dagger}b_1^{[\tilde{\psi},\omega]\dagger } b_2^{[\tilde{\psi},\omega]\dagger}\cdots |0\rangle \otimes \text{[right mover part]}. \end{aligned}\ ] ] then , taking notice of @xmath191 of the summation @xmath192 , cutting off the indices @xmath65 and @xmath193 up to finite @xmath127 and utilizing the fermionic nature of @xmath194\dagger}$ ] , we get @xmath195\dagger}\right)|0\rangle \nonumber \\ & = \sum_{\sigma = \text{\tiny $ \left ( \begin{array}{cccc } 0 & 1 & \cdots & m \\ i_0 & i_1 & \cdots & i_m \end{array}\right)$ } } w_{0,i_0}^{\dagger}b_{i_0}^{[\tilde{\psi},\omega]\dagger}\times w_{1,i_1}^{\dagger}b_{i_1}^{[\tilde{\psi},\omega]\dagger}\times \cdots \times w_{m , i_m}^{\dagger}b_{i_m}^{[\tilde{\psi},\omega]\dagger}|0\rangle \nonumber \\ & = \left[\sum_{\sigma = \text{\tiny $ \left ( \begin{array}{cccc } 0 & 1 & \cdots & m \\ i_0 & i_1 & \cdots & i_m \end{array}\right)$}}\text{sign } ( \sigma ) \ > w_{0,i_0}^{\dagger}w_{1,i_1}^{\dagger}\cdots w_{m , i_m}^{\dagger}\right ] b_0^{[\tilde{\psi},\omega]\dagger}b_1^{[\tilde{\psi},\omega]\dagger } \cdots b_m^{[\tilde{\psi},\omega]\dagger}|0\rangle \nonumber \\ & = \left(\det w_{n , m}^{\dagger}\right)_{n , m\geq 0 } b_0^{[\tilde{\psi},\omega]\dagger}b_1^{[\tilde{\psi},\omega]\dagger } \cdots b_m^{[\tilde{\psi},\omega]\dagger}|0\rangle . \end{aligned}\ ] ] the terms about @xmath196 become the states in the fifth line of ( [ eq:2-weyl.dirac.a ] ) . the same procedure can be followed for the righr mover . after taking the limit @xmath197 , we obtain ( [ eq:2-weyl.dirac.a ] ) . the determinants in ( [ eq:2-identity_e ] ) can be obtain in the same way except for the continuum index @xmath82 . y. habara , h. b. nielsen and m. ninomiya , hep - th/0504173 ; y. habara , y. nagatani , h. b. nielsen and m. ninomiya , int . j. mod a * 23 * ( 2008 ) 2733 - 2769 ; y. habara , y. nagatani , h. b. nielsen and m. ninomiya , int . j. mod . phys . a * 23 * ( 2008 ) 2771 - 2781 . s. sugiura and a. shimizu , phys . * 108 * ( 2012 ) 240401 ; s. sugiura and a. shimizu , phys . * 111 * ( 2013 ) 010401 ; m. hyuga , s. sugiura . k. sakai and a. shimizu , phys . rev . * b 90 * ( 2014 ) 121110 .

we present in this paper the polemica ( polarized expectation - maximization independent component analysis ) algorithm which is an extension to polarization of the smica ( spectral matching independent component analysis ) temperature multi - detectors multi - components ( md - mc ) component separation method . this algorithm allows us to estimate blindly in harmonic space multiple physical components from multi - detectors polarized sky maps . assuming a linear noisy mixture of components we are able to reconstruct jointly the anisotropies electromagnetic spectra of the components for each mode , and , as well as the temperature and polarization spatial power spectra , , , , , and for each of the physical components and for the noise on each of the detectors . polemica is specially developed to estimate the cmb temperature and polarization power spectra from sky observations including both cmb and foreground emissions . this has been tested intensively using as a first approach full sky simulations of the planck satellite polarized channels for a 14-months nominal mission assuming a _ simplified _ linear sky model including cmb , and optionally galactic synchrotron emission and a _ gaussian _ dust emission . finally , we have applied our algorithm to more _ realistic _ planck full sky simulations , including synchrotron , _ realistic _ dust and free - free emissions . cosmic microwave background cosmology : observations methods : data analysis

astro-ph0603044

mapping the cosmic microwave background ( cmb ) polarization is one of the major challenges of future missions of observational cosmology . cmb polarization is linear and therefore can be described by the first three stokes parameters i , q and u which are generally combined to produce three fields ( modes ) , @xmath0 , @xmath1 and @xmath2 @xcite . the polarization of the cmb photons carries extra physical informations that are not accessible by the study of the temperature anisotropies . therefore its measurement helps breaking down the degeneracies on cosmological parameters as encounter with temperature anisotropies measurements only @xcite . furthermore , the study of the cmb polarization is also a fundamental tool to estimate the energy scale of inflation which has been proposed to solve the problems of flatness , of isotropy and of the seed perturbations for the formation of the structures in the universe . inflationary models predict the presence of tensor perturbations of the metric which will lead to an unique signature in the cmb polarization @xmath2 modes . the detection of the latter would be a strong proof of such an epoch and also a way to constrain the energy scale at which inflation occurs by measuring the tensor to scalar ratio , @xmath9 @xcite . since the beginning of the cmb anisotropies observations with the cosmic background explorer ( cobe ) @xcite , a great amount of experiments have been designed to determine the cmb temperature angular power spectrum @xcite . by contrast , the polarization anisotropies , which are between 2 and 5 orders of magnitude weaker than temperature ones are not accurately measured yet . a first detection of the cmb @xmath1 modes has been performed by dasi @xcite , capmap @xcite , cbi @xcite and more recently by boomerang @xcite and wmap @xcite . the @xmath6 temperature - polarization cross correlation has been measured by wmap @xcite and boomerang @xcite . no detection of the cmb @xmath2 modes has been reported yet . nevertheless , constraints on the tensor to scalar ratio , @xmath9 , have been set by the wmap team . they set an upper limit of @xmath10 ( 95% cl ) @xcite for the temperature and polarization analysis and of @xmath11 ( 95@xmath12 cl ) @xcite for a polarization - only analysis . the detection of such low signals is possible by improving the instrumental sensitivity , but this is not the only issue in the determination of the cmb polarization power spectra . other astrophysical emissions as for example the diffuse galactic emission including free - free , dust and synchrotron and the extragalactic - sources emissions also contribute to the sky brightness at the frequencies of interest for cmb studies , and therefore must be efficiently subtracted . these _ foregrounds _ are particularly important for the study of the cmb polarization . excluding the free - free emission which is not polarized , the other contributions are expected to be significantly polarized with similar power on the @xmath1 and @xmath2 modes . recent measurements of the galactic synchrotron polarization emission at 1.41 ghz @xcite and at 23 ghz by wmap @xcite show this emission is significantly polarized at large angular scales . further , archeops measurements at 353 ghz show that the galactic dust diffuse emission is polarized up to a level of 5 to 10 % both in the galactic center @xcite and at high galactic latitudes @xcite . finally , for the polarization of extragalactic point sources the sparsity of the data available makes reliable predictions difficult @xcite . a direct subtraction of these foreground contributions on the cmb data will require an accurate knowledge of their spatial distributions and of the electromagnetic spectra of their anisotropies . for the synchrotron emission a full sky map at 408 mhz in temperature is available @xcite and more recently the wmap team provided a map at 30 ghz from the mem decomposition of the first year observations @xcite . a fake polarized synchrotron emission template was constructed by @xcite based on the parkes 2400 mhz @xcite and haslam 408 mhz @xcite surveys . furthermore , the electromagnetic spectrum of synchrotron anisotropies and its spatial distribution are neither accurately known in temperature nor polarization although a first estimate was produced by @xcite . recently , the 23 ghz polarized wmap data is used as a tracer of the synchrotron polarization @xcite . for the thermal dust emission a full sky map at 100 @xmath13 m as well as templates for cmb use were extracted from the iras and firas data @xcite . no realistic template exists for the dust polarized emission , although a fake one , based on the polarization angles measured at 23 ghz has been constructed by the planck collaborationbernard / psm/ ] . the dust emission in temperature can be approximated by a grey body spectrum of mean temperature 17 k and emissivity between 1.7 and 2.2 @xcite . currently no measurement on the electromagnetic spectrum of the dust polarized emission is available although it is expected to be the same that for temperature @xcite . to try to overcome the above limitations , a great amount of work has been dedicated to design and implement algorithms for component separation which can discriminate between cmb and foregrounds . these methods can also extract , directly from the cmb data , the emission properties of foregrounds . wiener filtering has been successfully tested assuming known gaussian priors for each component and with the electromagnetic spectrum of the anisotropies as an input @xcite . maximum entropy based methods ( mem ) , assuming entropic priors for the spatial distribution of each of the component , have been intensively used for small sky patches @xcite and extended to full sky analysis @xcite . they were adapted to account for spatial anisotropies in the electromagnetic spectra @xcite . more recently , @xcite has developed a new method to perform cmb component separation by parameter estimation and applied it to temperature simulations of the planck satellite experiments . independent component analysis ( ica ) techniques have also been applied to planck simulations in temperature @xcite and extended to polarization @xcite using the fastica algorithm . these methods require no prior on the spectral or spatial distribution of the components but can not make use of the available physical knowledge on the foreground and cmb emissions . in addition , the spectral matching independent component analysis ( smica ) @xcite has been developed to consider both the fully blind analysis for which no prior is assumed and the semi - blind analysis incorporating previous physical knowledge on the astrophysical components . this algorithm , based on the expectation - maximization algorithm ( em ) @xcite , uses the spectral diversity of the components and was developed for temperature only . we present in this paper , polemica ( polarized expectation - maximization independent component analysis ) , an extension of this method to polarization including both the blind and semi - blind analysis . this paper is organized as follows . a simple model of the microwave sky emission in temperature and polarization is described in section [ simus ] . section [ mdmc ] presents the polemica multi - detectors multi - components ( md - mc ) blind component separation algorithm . section [ simus_list ] describe the simulations of the planck satellite experiment used for testing the algorithm . we present in section [ testing ] the application of polemica to the planck simulations with a _ simplified _ model to test the algorithm s performances . finally , in section [ realistic_model ] , we apply our algorithm to more _ realistic _ planck simulations and discuss the separability problem in this case . we summarize and conclude in section [ conclusions ] . to constrain cosmological models , cmb experiments have to reach an accuracy which is well below the expected level of contamination from astrophysical foregrounds , in temperature and even more critically in polarization . therefore , an efficient separation between cmb and foregrounds is crucial for the success of future polarization experiments . to perform such a separation , the diversity of the electromagnetic spectra of the anisotropies and of the spatial spectra of the components is generally used . observations from a multi - band instrument can be modeled as a linear combination of multiple physical components leading to what is called a multi - detectors multi - components ( md - mc ) modeling . assuming an experiment with @xmath14 detector - bands at frequencies @xmath15 and @xmath16 physical components in the data , for each stokes parameter ( @xmath17 , @xmath18 and @xmath19 ) and for each pixel on the sky map we can write y^(*r*)=a^_cs^c(*r*)+n^(*r * ) where @xmath20 is the map of the @xmath21 component , @xmath22 refers to the noise map for each @xmath15 band and @xmath23 which is called the _ mixing matrix _ , gives the electromagnetic spectrum behavior for the component @xmath21 and frequency @xmath15 . beam smoothing and filtering effects are not considered in this work . as in the temperature case , it is more convenient to work in spherical harmonics space , where equation ( [ datareal ] ) can be rewritten independently for each @xmath24 assuming a full sky coverage . thus equation ( [ datareal ] ) reads for @xmath25 and for each frequency band and for each @xmath24 y^,x_m=_c=1^n_ca^,x_cs^c , x _ m+n^,x_m where @xmath26 is a vector of size @xmath27 , @xmath28 is a @xmath29 vector and @xmath30 is a vector of the same size than @xmath26 . @xmath31 is a matrix of @xmath32 elements formed from the mixing matrix of each of the modes , @xmath0 , @xmath1 and @xmath2 . the aim of the component separation algorithm presented in this paper is to extract @xmath31 , @xmath28 and @xmath30 from the @xmath26 sky observations . following the md - mc model discussed above and given an observational setup , we construct , using the healpix pixelization scheme @xcite and in cmb temperature units , fake @xmath17 , @xmath18 and @xmath19 maps of the sky at each of the instrumental frequency bands . for these maps we consider three main physical components in the sky emission : cmb , thermal dust and synchrotron . concerning emissions that are not expected to be significantly polarized , we have optionally considered unpolarized free - free emission and not taken into account the sz emission as we are interested in diffuse emissions . when no free - free is considered and in any cases for the sz emission , we suppose that they have been successfully removed from our sky maps in temperature . we have assumed that the free - free emission is not polarized . instrumental noise is modeled as white noise . the cmb component map is randomly generated from the polarized cmb angular power spectra computed with the camb software @xcite for a set of given cosmological parameters . in the following we have used @xmath33 , @xmath34 , @xmath35 , @xmath36 and @xmath37 that are the values of the cosmological concordance model according to the wmap one year results @xcite . we also consider gravitational lensing effects as described in @xcite . for the diffuse galactic synchrotron emission we use the template maps in temperature and in polarization provided by @xcite . these template maps were derived in temperature directly from the haslam map at 408 mhz @xcite . the polarization maps in @xmath18 and @xmath19 were constructed from the intensity map from a constrained realization of the polarization angles using the parkes 2400 mhz survey @xcite . a template of the spatial variations of the synchrotron spectral index is also provided by @xcite . here we have chosen to use a constant spectral index equal to the mean of the spectral index map , @xmath38 , so that the simple linear model of the data holds . a more realistic treatment of the synchrotron emission will require a specific model of the spectral index spatial variations in order to ensure separability between components . we have used through this article two different dust models . @xmath39 simplified-_dust model _ : we have first considered a simplified dust model which is gaussian and derived from the power - law model from @xcite to describe the dust angular power spectra in temperature and in polarization . this model , although not fully realistic , is not spatially correlated to the synchrotron emission and helps us to extensively test the properties of the separation method . we have renormalized this model to mimic at large angular scales the @xmath6 cross power spectrum measured by archeops at 353 ghz @xcite . the rms of the final dust map is probably overestimated as we do not account for the variation of the dust emission with respect to galactic latitude . the power spectra models are computed at 100 ghz in @xmath40 units . @xmath41 and @xmath19 full - sky maps are generated randomly from these power spectra . we extrapolate them to each of the frequency of interest by assuming a grey body spectrum with an emissivity of 2 . finally , the @xmath40 maps are converted into @xmath42 units . @xmath39 realistic-_dust model _ : secondly we have used the _ _ planck sky model__polarized dust template @xcite . it was modeled using model number 7 of @xcite . this model is normalized to the iras 100 @xmath13 m emission map produced by @xcite . for polarization , a constant polarization degree of 5% is assumed , and the same polarization angles than for the synchrotron model are used . for both temperature and polarization , we assume a grey body emission with an emissivity of 2 . the free - free component is derived from the _ planck sky model_. it is based on the h-@xmath43 template by @xcite . we have assumed a constant spectral index of -2.1 in rayleigh - jeans units . expected to be no significantly polarized except in particular hii regions ( less than 10% @xcite ) , the free - free emission @xmath18 and @xmath19 maps are set to zero . noise maps for each channel are generated from white noise realizations normalized to the nominal level of instrumental noise for that channel . + .electromagnetic spectrum of the anisotropies , in arbitrary units , of the cmb , dust , synchrotron and free - free components at each of the planck channels . relative noise levels in the temperature and in the polarization maps taking as reference the 143 ghz channel are also displayed.[tab : amat_tab ] [ cols="^,^,^,^,^,^,^,^",options="header " , ] the electromagnetic spectra of the anisotropies , in arbitrary units , for the cmb , dust and synchrotron emissions are displayed in table [ tab : amat_tab ] ( these are the values used in the mixing matrix @xmath44 ) for the planck satellite simulations presented in section [ simus_list ] . we also present the relative noise level taking as reference the 143 ghz channel . the noise levels used at 143 ghz are 6.3 @xmath13k@xmath45 ( in temperature ) and 12.3 @xmath13k@xmath45 ( in polarization ) per square pixels of side 7 arcmin and for a 14-months planck mission @xcite . to reduce the number of unknown parameters in the model described by equation ( [ data ] ) , it is interesting to rewrite this equation in terms of the temperature and polarization auto and cross power spectra . this will considerably reduce the computing time with no loss of information . we define the density matrices associated with the data , @xmath46 , the physical components , @xmath47 and the noise , @xmath48 , as follows @xmath49 where @xmath50 represents frequency , @xmath51 , for the data and noise matrices and component , @xmath21 , for the physical - components matrix . averaging over bins on @xmath52 we obtain r_(b)__b^i , x _ m^i,x_m where @xmath53 is the set of @xmath52 values which contributes to bin @xmath54 and @xmath55 is the number of such multipoles . in the following , @xmath56 , represents the total number of bins used in the analysis . for each bin @xmath54 , equation ( [ data ] ) reads r_y(b ) = ar_s(b)a^t+r_n(b ) where @xmath57 and @xmath58 are @xmath59 matrices and @xmath60 is a @xmath61 matrix . to fully understand the component separation algorithm described below it is interesting to have a closer look to the content of the three density matrices defined above ( see appendix for a concrete example ) @xmath57 represents the input density matrix computed from the observed multi - band data . this matrix is composed of @xmath62 symmetric blocks each of them containing in the diagonal the auto - power spectra , @xmath3 , @xmath4 , @xmath5 and in the off - diagonal the cross - power spectra @xmath6 , @xmath8 and @xmath7 . a single block represents either the auto - correlation of a single channel ( for diagonal blocks ) or the cross - correlation between two channels ( for off - diagonal blocks ) . assuming that the physical components in the data are statistically independent and uncorrelated makes the @xmath60 a @xmath63 block diagonal matrix . as above each block , corresponding to the @xmath64 physical component , contains in the diagonal the auto - power spectra , @xmath3 , @xmath4 , @xmath5 and in the off - diagonal the cross - power spectra @xmath6 , @xmath8 and @xmath7 . we also assume that the noise is uncorrelated between channels and therefore , @xmath58 is a diagonal matrix containing the noise auto power spectra @xmath3 , @xmath4 and @xmath5 for each of the channels . from equation ( [ dataspec ] ) we observe that , for each bin @xmath54 , the data density matrix @xmath57 , of size @xmath65 , is fully defined by the set of parameters @xmath66 which corresponds to a total of @xmath67 parameters . this indicates that from the cmb data set and under the hypothesis presented above it is possible to simultaneously estimate the mixing matrix , the physical component s temperature and polarization power spectra and the noise s temperature and polarization power spectra for each of the channels . further , assuming white noise in the maps only three noise parameters per channel ( for tt , ee and bb ) need to be estimated for the entire range in @xmath52 . this reduces the overall set of parameters to @xmath68 where @xmath56 is the number of bins . + to estimate the above parameters from the data we have extended to the case of polarized data the spectral matching algorithm developed by @xcite for temperature only . the key issue of this method is to estimate these parameters , or some of them ( for a semi - blind analysis ) , by finding the best match between the model density matrix , @xmath57 , computed for @xmath69 and the data density matrix @xmath70 obtained from the multi - channel data . assuming that the different physical components and the noise are realizations of gaussian stationary fields ( wittle approximation ) , the log - likelihood function of the form = _ b=1^n_b n_b+c . is a reasonable measure of the mismatch between data and model . the maximization of the likelihood function is achieved via the expectation - maximization algorithm ( em ) @xcite . this algorithm will process iteratively from an initial value of the parameters @xmath71 following a sequence of parameter updates @xmath72 , called ` em steps ' . during the @xmath73 e - step we compute the expectation value for the likelihood from the @xmath74 iteration s parameters . the @xmath73 m - step maximizes the likelihood ( i.e. minimizes the log - likelihood ) to compute the @xmath73 set of parameters . by construction each em step improves the spectral fit by maximizing the likelihood . for a more detailed review of the spectral matching em algorithm used here , see @xcite and @xcite and appendix [ matrix_appendix ] for the formalism used to describe the polarized sky model and data . in the md - mc model presented above there is a scale indetermination on the value of @xmath44 and @xmath75 and only the product @xmath76 is scale invariant . thus , to ease the convergence of the algorithm , we renormalize each column of @xmath44 to unity at each em iteration and correct the @xmath75 density matrix accordingly so that the product @xmath76 is unchanged . to start - up the em algorithm the parameters of the fit , @xmath77 , need to be initialized to reasonable values to avoid exploring local maxima in the likelihood function . in the case of the mixing matrix , @xmath44 , we can consider , in a first approximation , that the electromagnetic spectrum is the same in temperature and polarization for each of the components . therefore , we can concentrate on guessing the electromagnetic spectrum in the temperature data where we expect the signal to noise to be larger . when no physical prior is available this can be obtained by using the @xmath16 dominant eigenvectors of the data density matrix for temperature only , @xmath78 . each of them represent the change of power with frequency for the dominant components in the data . notice that these components and the physical ones are not necessarily the same . on one hand , we can have in the data extra components which have not been identified as for example residual systematics . on the other hand , the electromagnetic spectrum of the physical components may present spatial variations as it is , for example , the case for the galactic synchrotron diffuse emission . in the following , we will consider that the data contain only identified physical components with spatially constant electromagnetic spectrum . if this is not the case , a careful pre - analysis of the initialization parameters is needed and this is not discussed in this paper . assuming the mixing matrix previously initialized , the physical - components density matrix can be obtained from a noiseless fit to the data as follows r_s,0(b)=block diag((a_0^ta_0)^-1a_0^t_y(b)a_0(a_0a_0^t)^-1 ) finally , from @xmath79 and @xmath80 the noise density matrix is given by r_n,0(b)=diag(_b(_y(b)-a_0r_s,0a_0^t ) ) where @xmath81 . here we implicitly assume that the noise is white and not correlated between different channels . we have first performed various sets of simulations of the expected planck satellite data to intensively test the algorithm presented above . for each of those , we performed 300 realizations considering full - sky maps that can contain instrumental noise , cmb , dust , synchrotron and free - free . for each realization the cmb and noise are changed while dust , free - free and synchrotron are kept unchanged . we simulate maps at the lfi and hfi polarized channels , 30 , 40 and 70 ghz for lfi and 100 , 143 , 217 and 353 ghz for hfi . these maps are in healpix pixelization @xcite and correspond to a 14-month survey . 1 . * * : we simulate maps at @xmath82 = 512 ( pixels of area @xmath83 ) . we include in the simulations cmb emission and gravitational lensing . the simulations also contain synchrotron and _ simplified _ thermal dust emissions as described in section [ simus ] . this @xmath82 permits the reconstruction of the angular power spectra up to @xmath84 . the reconstructed spectra will be averaged over bins of size 20 in @xmath52 . with these simulations we test the separation method at small angular scales . the simulations contain also instrumental noise . * : we simulate maps at @xmath82 = 128 ( pixels of area @xmath85 arcmin@xmath86 ) for cmb only . this @xmath82 permits the reconstruction of the power spectra up to @xmath87 . this maximum @xmath52 value is enough to study the effect of gravitational waves in the @xmath5 spectrum which is maximum around @xmath88 for the concordance model . the reconstructed spectra are averaged over bins of size 10 in @xmath52 . these simulations were performed for fast and intense test of the algorithm in the easiest possible case . the simulations contain also instrumental noise . * : we simulate maps at @xmath82 = 128 including cmb , _ simplified _ thermal dust and synchrotron emissions . the reconstructed spectra are averaged over bins of size 10 in @xmath52 . these simulations were performed to check the impact of foregrounds in the reconstruction of the cmb power spectra . the simulations contain also instrumental noise . * : same as * * simulations except that _ simplified _ dust has been replaced by _ realistic _ dust and free - free emission has been included . we have applied the md - mc polemica component separation algorithm to the _ simplified _ simulations presented above . for each set of simulations we have computed the data density matrix @xmath89 and applied the algorithm with different degrees of freedom : 1 . first , we assume that the mixing matrix , @xmath44 , is known . we construct the @xmath44 matrix from the exact electromagnetic spectrum of the anisotropies of each component and fix it in the algorithm . therefore , we take as parameters for the fit @xmath90 for each bin @xmath54 . with this test we want to check the spatial separability of the components . in the following , we refer to this type of component separation as _ a - fixed _ separation . secondly , we have performed what we call a _ cmb semi - blind _ separation . the @xmath44 matrix is fitted as well as @xmath75 and @xmath91 . but assuming a prior on the cmb electromagnetic spectrum . thus , the columns of the @xmath44 matrix corresponding to the cmb are fixed to unity and the set of parameters for the fit is @xmath92 for each bin @xmath54 . the initialization of @xmath44 , except for the cmb , is performed as described in section [ initialization ] . this kind of prior in the cmb electromagnetic spectrum is a reasonable approximation because the former is well - known . finally we have performed a _ blind _ separation fitting all the parameters @xmath66 for each bin @xmath54 including the cmb electromagnetic spectrum . we have performed these three types of analysis in all the _ simplified _ simulated sets . to ensure the reliability of the results we performed 10000 em iterations and checked , for each simulation , the convergence of the em algorithm . in the following we present the main results obtained in reverse order going from ( iii ) to ( i ) and if not stated otherwise we assume white noise @xmath93 we present here a analysis of the * * simulations for which we assume three physical components in the data : cmb and _ simplified_-dust and synchrotron emissions . the noise and physical - components density matrices as well as the mixing matrix are initialized as described in section [ initialization ] . no physical priors are assumed neither for synchrotron nor dust . for cmb , if not stated otherwise , we initialize the electromagnetic spectrum to 1 for temperature and polarization . this is a reasonable approximation as we expect the planck data to be calibrated to better than 1 % @xcite . figure [ fig : planckcmbfgd128 ] shows the _ blind _ reconstructed cmb temperature and polarization power spectra , in @xmath13k@xmath94 , for the 100 ghz channel ( crosses ) . the input model is overplotted in blue . the mean and the error bars are computed from the analysis of 300 simulations . we observe that @xmath95 , @xmath96 and @xmath97 are recovered with no bias , up to @xmath98 , which is the largest accessible @xmath52 value at @xmath99 . in the same way , @xmath100 is accurately reconstructed except for the very high @xmath52 values for which pixelization problems may appear ( see section [ sec : smallscales ] ) . we also reconstruct efficiently @xmath101 although a small bias ( below 10 % ) is introduced at low @xmath52 mainly due to confusion with the synchrotron emission as discussed in the following . the @xmath102 spectrum is not recovered at all and a significant bias is observed . this bias , due mainly to statistical residual noise as discussed in the following section , depends only on the signal to noise ratio and does not affect the reconstruction of the other components . figure [ fig : planckdustfgd128 ] shows the reconstructed _ simplified_-dust emission power spectra , @xmath103 , @xmath104 , @xmath105 , @xmath106 , @xmath107 and @xmath108 , in @xmath109 , at 353 ghz . for comparison we overplot in red the input model . the mean and the error bars displayed were computed using a total of 300 simulations . the reconstruction is fully efficient for @xmath101 , @xmath95 , @xmath96 and @xmath97 up to @xmath98 . the @xmath100 and @xmath102 are accurately reconstructed except at @xmath110 where a small bias ( below 10 % ) appears . the @xmath7 and @xmath8 spectra are compatible with zero as expected from the input model . these results are consistent with the fact that the simulated _ simplified_-dust emission dominate the simulated maps at the hfi channels for which the signal to noise ratio is larger . finally , we present in figure [ fig : plancksyncfgd128 ] the _ blind _ reconstructed synchrotron power spectra , in @xmath109 , at 30 ghz . we overplot in green the power spectrum of the input temperature and polarization synchrotron map from @xcite . here again @xmath100 , @xmath102 , @xmath95 , @xmath96 and @xmath97 are recovered efficiently . a bias at low @xmath52 ( below 20 % ) is observed for @xmath101 . this is due , as discussed in the following , to the slight mixing - up of the synchrotron and cmb emissions in temperature . the slight mixing up between synchrotron and cmb is better observed in the reconstructed mixing matrix . the first row of figure [ fig : amat128 ] shows the _ blind _ recovered @xmath44 matrix for dust , synchrotron and cmb in the case of the * * simulations . the electromagnetic spectrum for @xmath0 , @xmath1 and @xmath2 are respectively traced in green , red and blue . for comparison we overplot the input electromagnetic spectrum for each of the components ( black dashed line ) . for convenience we have renormalized the electromagnetic spectrum so that it is unity at 353 ghz , 30 ghz and 100 ghz for dust , synchrotron and cmb respectively . it is important to remark that the reconstruction of the electromagnetic spectra for temperature and polarization is performed independently . we observe that the error bars are larger for polarization than for temperature as we would expect from the smaller signal to noise ratio in polarization . the _ simplified_-dust electromagnetic spectrum is reconstructed with no bias both in temperature and polarization even at the lowest lfi frequency channels . furthermore , the reconstructed synchrotron electromagnetic spectra in polarization are not biased . in temperature we observe that the spectrum flattens out at intermediate frequencies between 70 and 217 ghz . this is the cause of the slight mixing up between synchrotron and cmb . this mixing up does not happen in polarization for which the synchrotron emission dominates over the cmb emission . finally , the reconstruction of the cmb electromagnetic spectrum from the @xmath0 and @xmath1 modes , although noisy for the latter , is not biased . however for the @xmath2 modes the reconstruction is very poor because of the very low signal to noise ratio ( below @xmath111 for @xmath112 ) . we have repeated the analysis with no prior in the electromagnetic spectrum of the cmb . we use instead the eigenvector corresponding to the third larger eigenvalue of the data density matrix . the results for dust and synchrotron remain unchanged . for cmb the electromagnetic spectrum at 30 and 353 ghz is not reconstructed neither in temperature nor in polarization and the results for @xmath2 are significantly degraded at all frequencies . however , the results on the reconstruction of the spatial power spectra remain unchanged for all the physical components including cmb . this can be easily understood as the reconstruction of the cmb power spectra is mainly dominated by the intermediate frequency maps , from 70 to 217 ghz , where the @xmath44 matrix is accurately reconstructed . in the previous analysis we have computed the electromagnetic spectrum of the physical components independently for each mode @xmath0 , @xmath1 and @xmath2 . in a more realistic approach we should consider a single electromagnetic spectrum for the polarization @xmath1 and @xmath2 modes which may be different from the temperature one . we have repeated the analysis under the above hypothesis and the results remain roughly the same with respect to the reconstruction of the spatial power spectra and of the electromagnetic spectrum . for synchrotron and _ simplified_-dust the @xmath1 and @xmath2 modes have roughly the same power in our simulations and therefore we expect only variations in the error bars . for cmb the @xmath1 mode largely dominates the @xmath2 mode and therefore we expect no significant contribution from the latter to the electromagnetic spectrum reconstruction . the differences between the temperature and polarization electromagnetic spectra are expected to be small for dust and synchrotron @xcite and none for the cmb @xcite . therefore , in the case of a perfectly calibrated experiment , we can consider , in a first approximation , that the polarization and the temperature electromagnetic spectra are the same . the _ blind _ analysis of the * * simulations under this hypothesis shows no evidence of mixing up between synchrotron and cmb . this can be clearly observed in the second row of figure [ fig : amat128 ] where we represent from left to right the reconstructed electromagnetic spectrum of the dust , synchrotron and cmb emissions respectively . no bias is observed for any of the physical components including synchrotron for which the flatten out of the spectrum observed before is not present . figures [ fig : tpqspectplanckcmbfgd128 ] and [ fig : tpqspectplancksyncfgd128 ] show the cmb and synchrotron reconstructed spatial power spectra in temperature and polarization . we do not observe a bias neither on the synchrotron nor on the cmb @xmath3 power spectra . furthermore , we observe that the cmb @xmath5 modes , although biased at large @xmath52 values , are fairly reconstructed up to @xmath113 . for the other modes , the results are similar to those presented in the previous section . the separation method allows us to easily include previous knowledge on the physical components either as priors or as facts . in the previous section we considered a prior on the cmb emission . in the following we move a step forward in the analysis assuming the cmb electromagnetic spectrum known and performing what we call a _ cmb semi - blind _ analysis . for this analysis , the columns of the mixing matrix corresponding to the cmb are initialized to the cmb electromagnetic spectrum and are not updated by the algorithm . for the other components we consider independent electromagnetic spectra for temperature and polarization and they are initialized as for the _ blind _ analysis . the third row of figure [ fig : amat128 ] shows the reconstructed mixing matrix for the _ cmb semi - blind _ analysis considering the * * simulations . the results are similar to those of the _ blind _ analysis discussed before . the dust electromagnetic spectrum is accurately recovered in temperature and polarization . for synchrotron , the polarization spectrum is accurately recovered but the temperature one flattens out at intermediate frequencies with respect to the input model . therefore , the reconstructed _ simplified_-dust and synchrotron spatial power spectra in temperature and polarization are similar to those of the _ blind _ analysis . simplified_-dust power spectra are accurately reconstructed in temperature and polarization . for synchrotron the power spectra are also accurately reconstructed except for the @xmath3 mode which present a slight bias ( below 20% ) at large angular scales ( @xmath114 ) . in general the error bars are smaller for the _ cmb semi - blind _ analysis . the reconstructed cmb power spectra for the _ cmb semi - blind _ analysis are represented in black on figure [ fig : semiblindplanckcmbfgd128 ] . all of them are accurately reconstructed with no bias except for the @xmath5 mode . for the latter the reconstruction is accurate up to @xmath113 and there on is biased . this bias is due to residual noise and is not related to the uncertainties on the reconstruction of the electromagnetic spectrum for the other physical components . to check this we have also performed a _ a - fixed _ analysis assuming the electromagnetic spectrum of all physical components known . the results of this analysis are overplotted in red on the figure . we observe that reconstruction is equivalent to that of the cmb _ semi - blind _ analysis but for the error bars which are smaller this indicates that the bias in the @xmath5 mode is mainly due to residual noise as discussed in section [ sec : colornoisemodel ] . [ sec : discusions ] in the previous section we have fully described the analysis of the * * simulations at @xmath115 for which the reconstruction of the spatial power spectra was limited to @xmath98 . in some cases we have observed small biases in the polarization auto power spectra at large @xmath52 values which may be due to pixelization problems ( we exclude in here the bias observed in the cmb @xmath5 modes which is due to residual noise ) . to check this hypothesis we have also performed the _ blind _ , _ cmb semi - blind _ and _ a - fixed _ analysis on the * * simulations for which we can reconstruct the angular power spectra up to @xmath116 . as the resolution of the planck best channels is @xmath117 arcmin a more realistic analysis will require simulations at @xmath118 which are far too much time demanding for our computational capabilities . the results obtained for the * * simulations are very similar to those for the * * ones . for illustration figures [ fig : cmbfgd512planck ] , [ fig : dustfgd512planck ] and [ fig : syncfgd512planck ] show the reconstructed temperature and polarization power spectra for cmb , _ simplified_-dust and synchrotron in the case of a _ cmb semi - blind _ analysis . the synchrotron and _ simplified_-dust power spectra are accurately reconstructed in temperature and polarization . we observe a slight bias in the dust @xmath4 and @xmath5 modes as for the * * simulations but at much larger @xmath52 values . this indicates that this bias is related to pixelization effects . the same effect is observed for the cmb power spectra . the bias in the @xmath4 modes is present at much larger @xmath52 than for the * * simulations . the reconstruction of the cmb @xmath5 modes is accurate at low @xmath52 ( @xmath114 ) and present a residual noise bias at large @xmath52 values as discussed in the previous section . as seen before , the bias observed in the cmb @xmath5 power spectrum is most probably due to residual noise from the separation . therefore , it is interesting to check both the accuracy of the noise reconstruction and the limitations of the white noise model imposed . with respect to the latter we have repeated all the analysis presented above assuming a color noise model such that the noise power spectra are estimated for each bin @xmath54 in @xmath52 . for the _ blind _ analysis the results are slightly worse in the sense that the mixing up between cmb and synchrotron is more significant . this is not surprising since the noise in the data is white and we are artificially reducing the number of degrees of freedom in the fit . actually , the mismatch between the data and the physical component power spectra can be compensated by changing the noise power spectra . in the case of the _ cmb semi - blind _ and _ a - fixed _ analysis the results for the white and color noise model present no significant differences . figure [ fig : noise128 ] shows , in black , the reconstructed noise angular power spectra , @xmath119 , @xmath120 , @xmath121 , in @xmath122 , at 100 ghz for the * * simulations in the case of a _ cmb semi - blind _ analysis . we overplot , in black , the power spectra of the noise at 100 ghz obtained from 100 realizations of noise - only maps . the noise spectra are reconstructed down to the accuracy of the estimation of the input model both for @xmath3 and @xmath4 , well below 10@xmath123 % . for the @xmath5 noise spectrum there is a small bias which is of the order of @xmath124 % at @xmath125 and around @xmath126 % at @xmath127 . therefore , to improve the reconstruction of the cmb @xmath5 modes we need a better estimation of the @xmath5 noise power spectrum . for this purpose we need to improve the likelihood maximization algorithm . for temperature - only separation , @xcite complemented the em algorithm with a direct maximization of the likelihood function via a newton - raphson algorithm . for polarization similar algorithms can be used but due to the degree of complexity of the problem ( 6 correlated modes per physical component instead of 1 in the temperature only case ) and for the sake of clarity these will be discussed in a forthcoming paper . after testing intensively our algorithm on our _ simplified _ model , deducting the global performances of the spectral matching reconstruction in temperature and polarization we are interested in the performances of the algorithm for significantly spatially correlated components . for this purpose , we have used the * [ planck d ] * simulations . with this set of simulations , we have performed several types of separations . first , we have worked on a joint temperature and polarization analysis , similar to the one presented in section [ testing ] . then we have considered temperature - only separation , and polarization - only separation . for each of the described cases we have applied the algorithm with the different degrees of freedom presented in section [ testing ] . models and recovered data are average over bins of size 10 in @xmath52 beginning at @xmath128 . error bars presented in this section represent the dispersion over 100 simulations . note that @xmath115 allows a theoretical reconstruction up to @xmath129 in temperature and @xmath130 in polarization @xcite . cmb recovered spectra are plotted at 100 ghz while dust at 353 ghz and synchrotron and free - free at 30 ghz . we have first performed a joint temperature and polarization separation on the _ realistic _ model . we present here the results obtained in the _ a - fixed _ separation case , considering the 4 simulated components . in this particular case , the algorithm is very slow to converge and thus 40000 em iterations have been run . the recovered angular power spectra for temperature and polarization are displayed from figure [ 4comp_realistic_cmb ] to figure [ 4comp_realistic_ff ] , respectively for cmb , _ realistic _ dust , synchrotron and free - free emissions and compared to the input model . for the cmb component , on figure [ 4comp_realistic_cmb ] we can see that @xmath103 and @xmath106 are recovered accurately up to @xmath131 . recovered spectra for @xmath107 and @xmath108 are compatible with zero as expected . @xmath104 is recovered with a good accuracy up to @xmath132 and then is slightly biased due to pixelization problems in the healpix scheme . finally , the @xmath105 spectrum is recovered up to @xmath133 and then is biased with residual noise as discussed in section [ sec : colornoisemodel ] . the _ realistic _ dust component recovered spectra are displayed on figure [ 4comp_realistic_dust ] . we can see that @xmath103 , @xmath104 , @xmath105 , @xmath106 , @xmath107 and @xmath103 are recovered with a perfect accuracy up to @xmath129 . the recovered power spectra for the synchrotron component are displayed on figure [ 4comp_realistic_sync ] . polarization only power spectra ( @xmath104 , @xmath105 and @xmath108 ) are recovered in good agreement with the input model . @xmath103 is well recovered up to @xmath134 but then converge sharply to a null signal and therefore a residual noise bias similar to the one of @xmath105 occurs . this is not directly visible in the temperature spectrum but can be seen in the cross temperature and polarization spectra . indeed , @xmath106 and @xmath107 are well recovered up to @xmath134 and further many points are strongly biased . finally , results corresponding to the free - free emission component are displayed in figure [ 4comp_realistic_ff ] . @xmath103 is recovered with a good accuracy up to @xmath129 . as for polarization the input signal is null , we can see in the recovered spectra the overall behavior of our algorithm described in section [ testing ] . the algorithm has also been run for _ cmb - fixed _ and _ blind _ separations . for both of them , excepting the dust component which is well constrained in all cases due to its dominant power at high frequencies , the algorithm fails to converge and then components are mixed and results strongly biased . this may come from the fact that free - free and synchrotron electromagnetic spectra are similar and to the fact that all the galactic emissions have strong spatial correlations . in the following section we will address this problem and show that this mixing that prevents the convergence of the algorithm is mainly due to the separability problem that occurs in temperature . in section [ pol_sep ] , we will see that the separation performed on sets of @xmath18 and @xmath19 maps has not this separability problem . notice that when not considering the free - free emission in the simulations , results are very similar and the same performances of the algorithm with respect to the level of prior we assume are observed . we have performed a temperature - only separation on the _ realistic _ model . for this we consider sets of @xmath17 maps and the algorithm solve the spectral matching equations for @xmath0 modes only , like in the smica algorithm @xcite . we present here the results obtained in the _ a - fixed _ separation case , considering the 4 simulated components . recovered @xmath103 spectra are displayed in figure [ tt_4comp ] for synchrotron , _ realistic _ dust , cmb and free - free . we can see that except the synchrotron spectrum which start to be biased at @xmath134 , for the reason advanced in the last section , spectra are recovered with a good accuracy for the dust , cmb and free - free components up to @xmath129 . we have also performed the separation for the _ cmb - fixed _ and the _ blind _ cases . as for the joint analysis , excepting again the dust component for which the spectrum is recovered efficiently , the algorithm fails to converge . note that in the literature , no method has successfully separated synchrotron from dust on a noisy simulated mixture of cmb and astrophysical foregrounds , working on all sky maps . comparable results and the same performances of the algorithm with respect to the level of prior we assume are observed when no free - free component is included in simulations . finally , we have performed a polarization - only separation on the _ realistic _ model . for this we consider sets of @xmath18 and @xmath19 maps and the algorithm solve the spectral matching equations for @xmath1 and @xmath2 modes ( allowing reconstruction of @xmath104 , @xmath105 and @xmath108 ) . notice that as we suppose that the free - free emission is not polarized , the sets of polarized maps used here only contain cmb , _ realistic _ dust , synchrotron and noise . we have performed a _ cmb semi - blind _ separation on the polarization - only set of maps . in this case , the cmb electromagnetic spectrum is initialized to 1 and kept fixed ( see section [ testing ] ) . the results of this analysis on the cmb power spectra are displayed in figure [ eb_3comp_cmb ] . we can see that @xmath104 are reconstructed without bias up to @xmath135 , as expected with this pixelization scheme and that @xmath108 is compatible with zero . @xmath105 is efficiently recovered up to @xmath136 and then is biased with residual noise from the separation . realistic _ dust and synchrotron emissions recovered power spectra are displayed in figure [ eb_3comp_dust ] and [ eb_3comp_sync ] . in both cases we can see that @xmath104 , @xmath105 and @xmath108 are reconstructed very efficiently up to @xmath129 . we have also performed , for the polarization - only separation , a _ blind _ separation . the results of this analysis for dust and synchrotron are not displayed , but for both , @xmath104 , @xmath105 and @xmath108 are reconstructed very efficiently up to @xmath129 . the reconstruction of the cmb component is displayed in figure [ eb_3comp_cmb_blind ] . for @xmath104 we can see that the reconstruction is similar to the one in the _ cmb semi - blind _ case , only the error bars are larger . however , the @xmath105 spectrum is not recovered . from the previous analyzes we have clearly identified a separability problem when dealing with more than one realistic diffuse galactic emission component . this problem appears both in the joint temperature and polarization and in the temperature - only analyzes , but not in the polarization - only one . this would indicate that it is mainly due to the high level of correlation of the galactic diffuse emission in temperature both in the galactic plane and at high galactic latitudes . the current version of our algorithm assume uncorrelated components and therefore we expect it to behave badly when they are correlated . work is in progress to adapt polemica to account for spatially correlated components . for polarization - only separation the correlation problem seems to be not significant and the _ cmb semi - blind _ and _ blind _ analyzes are possible . therefore , for this case , we can evaluate the loss of accuracy in the reconstruction of the cmb signal due to the foreground contamination . for this , we compare the error bars on the cmb power spectra for the _ a - fixed _ , _ cmb semi - blind _ and _ blind _ separations to the one obtained in the case of a _ a - fixed _ separation on * [ planck b ] * simulations containing cmb and noise only . results are presented in figure [ compa_err_bars_eb ] . @xmath104 , @xmath105 and @xmath108 reconstruction error bars behave similarly with respect to the different algorithm priors . for each of them , _ a - fixed _ and _ cmb semi - blind _ error bars are of the same amplitude and are between 25 and 50 % larger ( respectively a factor 1.26 for @xmath104 , 1.50 for @xmath105 and 1.29 for @xmath108 ) than the reference _ a - fixed _ cmb only case . this means that in the context of our _ realistic _ model , we have no need to put priors on the foregrounds electromagnetic spectra in polarization to perform an efficient separation . in the _ blind _ case , error bars of the reconstruction are increased by a factor 1.59 for @xmath104 , 13.5 for @xmath105 ( but the reconstruction is biased ) and 2.52 for @xmath108 . on the other hand , these results stress the fact that having priors on the cmb electromagnetic spectra inside of our algorithm helps to perform a more accurate separation . we present in this paper the polemica algorithm which is an extension to polarization of the smica temperature md - mc blind component separation method developed by @xcite . both algorithms work in harmonic space and are based on the spectral matching of the data to a noisy linear mixture of uncorrelated physical components using the em algorithm to maximize the likelihood function . by contrast to the temperature data which are described by a single scalar quantity @xmath0 , the combined temperature and polarization data are described in harmonic space by three correlated scalar quantities @xmath0 , @xmath1 and @xmath2 corresponding to the @xmath17 , @xmath18 and @xmath19 stokes parameters in real space . we have developed a new formalism to jointly deal with the 6 resulting auto and cross angular power spectra , @xmath3 , @xmath4 , @xmath5 , @xmath6 , @xmath7 and @xmath8 . using this formalism we have constructed the likelihood function and proved that the em algorithm can be also applied to polarization data . under the assumption of uncorrelated gaussian distributed components and noise , the free parameters in the fit are the mixing matrix describing the electromagnetic spectrum of the physical components for @xmath0 , @xmath1 and @xmath2 , the temperature and polarization angular power spectra of the physical components and the temperature and polarization noise power spectra for each of the detectors . + we have , as a first approach , intensively and successfully tested the polemica method on simulations of the planck satellite experiment considering a 14-months nominal mission and no systematic effects . for these tests , we suppose a _ simplified _ linear model for the sky emission including cmb , synchrotron with constant spectral index and _ simplified_-dust ( gaussian realization ) emissions . we construct full sky maps for all the polarized channels from 30 to 353 ghz including at least one of the above physical components and considering white noise and infinite resolution . the method permits blind separation on these simulations allowing us to reconstruct the noise and physical component s temperature and polarization power spectra as well as the mixing matrix when we consider equal electromagnetic spectrum in @xmath0 , @xmath1 and @xmath2 . when we relax this hypothesis the reconstruction of the electromagnetic spectrum for the cmb @xmath2 modes is significantly degraded as could be expected because of the low signal to noise ratio . these results indicate that the polemica method allows us to both constrain the electromagnetic spectrum of the physical components and also to inter - calibrate the data based on the reconstructed cmb electromagnetic spectrum . + after setting the general performances of the algorithm , we have performed the separation on a more _ realistic _ model that includes _ realistic_-dust , synchrotron and free - free components in section [ realistic_model ] . we have encountered in this case a separability problem , that mixes up components and prevents the algorithm to converge , when performing _ blind _ separations . we have shown that this is due to spatial correlations between galactic components in temperature . thus , when working on sets of @xmath18 and @xmath19 maps and maximizing the likelihood for @xmath4 , @xmath5 and @xmath8 modes only , this separability problem does not occur and _ cmb semi - blind _ and _ blind _ separations are possible . for this _ polarization - only _ case , we have shown that considering our _ realistic _ sky model and our algorithm , in the planck case , we have no need to put priors on the galactic components electromagnetic spectra to reconstruct the cmb polarized power spectra . nevertheless , adding priors on the cmb electromagnetic spectrum helps to perform a more accurate separation . finally , real experiments present finite resolution , partial effective sky coverage , systematic effects and , often , correlated noise . all these issues must be dealt with by the component separation algorithms and will with no doubt significantly limit the precision to which the cmb signal may be reconstructed . polemica , as it was already the case for smica , can account for beam and filtering smoothing . systematic effects and correlated noise can be modeled as extra components in the data for which the spectral dependence can be estimated in a _ blind _ analysis . moreover , the strong spatial correlation in temperature between galactic physical emissions : dust , synchrotron and free - free , is a major problem for blind component separation algorithms which generally assumed uncorrelated components . although not observed yet , we can also imagine spatial correlation of the galactic emissions in polarization . work is in progress to adapt the polemica algorithm to the case of correlated components . in addition , foreground emissions have in general spatially varying electromagnetic spectra far beyond the simple linear model presented here . work is also in progress to adapt polemica to the case of foregrounds with spatially varying electromagnetic spectrum . we would like to thank d. santos and f.x dsert for very useful comments and a careful reading of the paper . special thanks to m. tristram for his comments and power spectrum related procedures and to d. blais for his useful advises on matrix derivation . we acknowledge j.f . cardoso , j. delabrouille and g. patanchon for comments on the technical details of the algorithm . the healpix package @xcite was used extensively in this paper . baccigalupi c. , 2003 , new astron . , 47 , 1127 baccigalupi c. et al . , 2004 , , 354 , 5570 barkats d. , bischoff c. , farese p. et al . , 2005 , , 619,127 bennett c. l. et al . , 2003b , , 148 , 97 benot a. et al . , 2003 , , 399 , 19 benot a. et al . , 2004 , , 424 , 571 bouchet f. r. , prunet s. & sethi s. k. , 1999 , , 302 , 663 challinor a. & chon g. , 2002 , , 66 , 127301 challinor a. & lewis a. , 2005 , , 71 , 103010 debernardis p. et al . , 2000 , , 404 , 955 delabrouille j. , cardoso j .- f . & patanchon g. , 2003 , , 346 , 1089 dempster a. , laird n. & rubin d , 1977 , j. of the roy . b , 39 , 1 dickinson c. , davies r. d. & davis r. j. , 2003 , , 341 , 369 duncan a. r. , haynes r. f. , jones k. l. & stewart r. t. , 1997 , , 291 , 279 eriksen h. k. et al . , 2005 , , submitted , astro - ph/0508268 finkbeiner d. p. , davies m. & schlegel d.j . , 1999 , , 524 , 857 giardino g. , banday a. j. , grski k. m. , bennet k. , jonas j. l. & tauber j. , 2002 , , 387 , 82 grski k. m. , hivon e. & wandelt b. d. , 1999 , astro - ph/9812350 halverson n. w. et al . , 2002 , , 568 , 38 hanany s. et al . , 2000 , , 545 , 5 haslam c. g. t. , stoffel h. , salter c. j. & wilson w. e. , 1982 , , 47 , 1 hildebrand r. h. , 1996 , in polarimetry of the interstellar medium , roberge w.g . , whittet d.c.b . ( eds ) , asp conference series , 97 , 254 hinshaw g. et al . , 2006 , submitted , astro - ph/0603451 hobson m. p. , jones a. w. , lasenby a. n. & bouchet f. , 1998 , , 300 , 1 hu w. , 2000 , , 62 , 043007 jones w. c et al . , 2005 , , submitted , astro - ph/0507494 jones t. j , kelbe d. & dickey j. m. , 1992 , , 389 , 602 keating b. , timbie p. , polnarev a. & steinberger j. , 1998 , , 495 , 580 kovac j. , leitch e. m. , pryke c. et al . , 2002 , , 420 , 772 knox l. & turner s. , 1994 , , 73 , 3347 lagache g. , 2003 , , 405 , 813 lee a. t. et al . , 2001 , , 561 , 1 leitch e. m. , kovac j. , halverson n. et al . , 2005 , , 624 , 10 lewis a. , challinor a. & lasenby a. , 2000 , , 538 , 473 maino d. et al . , 2002 , , 334 , 53 miller a. d. et al . , 1999 , , 524 , 1 montroy t. e. et al . , 2005 , , submitted , astro - ph/0507514 netterfield c. b. et al . , 1997 , , 474 , 47 netterfield c. b. et al . , 2002 , , 571 , 604 okamoto t. & hu w. , 2003 , , 67 , 083002 page l. et al . , 2006 , submitted , astro - ph/0603450 patanchon g. , 2003 , psip03 conference , astro - ph/0302078 piacentini f. et al . , 2005 , , submitted , astro - ph/0507507 planck consortium , 2005 , esa publications division ponthieu n. et al . , 2005 , , 444 , 327 prunet s. , sethi s. k. , bouchet f. r. & miville - deschne m. -a . , 1998 , , 339 , 187 readhead a. , myers s. , pearson t. et al . , 2004 , science , 306 , 836 rubino - martin j. a. et al . , 2003 , , 341 , 1084 seljak u. , hirata c. h. , 2004 , , 69 , 4 schlegel d. j. , finkbeiner d. p. & davies m. , 1998 , , 500 , 525 sievers j. l. et al . , 2003 , , 591 , 599 smoot g. f. et al . 1992 , , 396 , 1 snoussi h. , patanchon g. , macas - prez j .- f.,mohammad - djafari a. & delabrouille j. , 2001 , am . inst . of phys . baltimore , bayesian inference and maximum entropy methods in science enginnering 125 - 140 , maxent 2001 spergel d. n. et al . , 2003 , , 148 , 175 spergel d. n. et al . , 2006 , submitted , astro - ph/0603449 stivoli f. , baccigalupi c. , maino d. & stompor , r. , 2006 , 372 , 561 stolyarov v. , hobson m. p. , ashdown m. a. j. & lasenby a. n. , 2002 , , 336 , 97 stolyarov v. , hobson m. p. , lasenby a. n. & barreiro r. b. , 2005 , , 357 , 145 tegmark m. & efstathiou g. , 1996 , , 281 , 1297 tristram m. et al , 2005 , , 436 , 785 tucci m. , martnez - gonzlez e. , gonzalez - nuevo j. & de zotti g. , 2004 , , 349 , 1267 tucci m. , martnez - gonzlez e. , vielva p. & delabrouille j. , 2005 , , 360 , 926 turner m. s. & white m. , 1996 , , 53 , 6822 wolleben m. , landecker t. l. , reich w. & wielebinski r. , 2005 , , submitted , astro - ph/0510456 zaldarriaga m. & seljak u. , 1997 , , 55 , 1830 zaldarriaga m. , spergel d. n. & seljak u. , 1997 , , 488 , 1 we discuss here the formalism developed to describe the temperature and polarization observations as a noisy mixture of independent components . where @xmath142 and @xmath143 for @xmath144 are the coefficients of the spherical harmonic decomposition of the input sky observations and of the components of the sky model respectively . the coefficients @xmath145 correspond to the electromagnetic spectrum of the component @xmath146 at the frequency of observation @xmath147 . note that the mixing matrix , @xmath44 , has dimensions @xmath148 . + we define the noise , @xmath58 , signal , @xmath60 , and data , @xmath57 , density matrices used in equation ( [ dataspec ] ) for each bin , @xmath54 , as follows @xmath151 & \begin{array}{ccc } & & \\ & 0&\\ & & \\ \end{array } \\ c_2 & \begin{array}{ccc } & & \\ & 0&\\ & & \\ \end{array } & \left[\begin{array}{ccc } s_\ell^{tt}(b)&s_\ell^{te}(b)&s_\ell^{tb}(b)\\ s_\ell^{te}(b)&s_\ell^{ee}(b)&s_\ell^{eb}(b)\\ s_\ell^{tb}(b)&s_\ell^{eb}(b)&s_\ell^{bb}(b)\\ \end{array}\right ] \\ \end{array}\right)\ ] ] @xmath152 & \left[\begin{array}{ccc } y_\ell^{tt}(b)&y_\ell^{te}(b)&y_\ell^{tb}(b)\\ y_\ell^{te}(b)&y_\ell^{ee}(b)&y_\ell^{eb}(b)\\ y_\ell^{tb}(b)&y_\ell^{eb}(b)&y_\ell^{bb}(b)\\ \end{array}\right ] \\ \nu_2 & \left[\begin{array}{ccc } y_\ell^{tt}(b)&y_\ell^{te}(b)&y_\ell^{tb}(b)\\ y_\ell^{te}(b)&y_\ell^{ee}(b)&y_\ell^{eb}(b)\\ y_\ell^{tb}(b)&y_\ell^{eb}(b)&y_\ell^{bb}(b)\\ \end{array}\right ] & \left[\begin{array}{ccc } y_\ell^{tt}(b)&y_\ell^{te}(b)&y_\ell^{tb}(b)\\ y_\ell^{te}(b)&y_\ell^{ee}(b)&y_\ell^{eb}(b)\\ y_\ell^{tb}(b)&y_\ell^{eb}(b)&y_\ell^{bb}(b)\\ \end{array}\right ] \\ \end{array}\right)\ ] ]

i review a number of topics where conventional wisdom relevant to hadron physics at the lhc has been challenged . for example , the initial - state and final - state interactions of the quarks and gluons entering perturbative qcd hard - scattering subprocesses lead to the breakdown of traditional concepts of factorization and universality for transverse - momentum - dependent observables at leading twist . these soft - gluon rescattering effects produce bjorken - scaling single - spin asymmetries , the breakdown of the lam - tung leading - twist relation in drell - yan reactions , as well as diffractive deep inelastic scattering , the antishadowing of nuclear structure functions is predicted to depend on the flavor quantum numbers of each quark and antiquark , thus explaining the anomalous nuclear dependence observed in deep - inelastic neutrino scattering . isolated hadrons can be produced at large transverse momentum directly within a hard higher - twist qcd subprocess , rather than from jet fragmentation , even at the lhc . such direct " processes can explain the observed deviations from pqcd predictions of the power - law fall - off of inclusive hadron cross sections at fixed , as well as the baryon anomaly " , the anomalously large proton - to - pion ratio seen in high - centrality heavy - ion collisions at rhic . the intrinsic charm contribution to the proton structure function at high can explain the large rate for high photon plus charm - jet events observed by d0 at the tevatron . the intrinsic charm and bottom distributions imply a large production rate for charm and bottom jets at high at the lhc , as well as a novel mechanism for higgs and production at high the light - front wavefunctions derived in ads / qcd can be used to calculate jet hadronization at the amplitude level . the elimination of the renormalization scale ambiguity for the qcd coupling using the scheme - independent blm method will improve the precision of qcd predictions and thus increase the sensitivity of searches for new physics at the lhc . the implications of in - hadron condensates " for the qcd contribution to the cosmological constant are also discussed . = 1

1012.0553

the lhc will provide a crucial testing ground for testing qcd , not only at unprecedented energies and momentum transfers , but also at extreme particle densities . in this contribution i will review a number of topics where unexpected new perspectives for qcd physics at the lhc have emerged . 1 . _ high transverse momentum hadron production via direct hard subprocesses _ + it is natural to assume that high transverse momentum hadrons in inclusive high energy hadronic collisions , such as @xmath5 , can only arise from jet fragmentation . in fact , a significant fraction of high @xmath6 isolated hadrons can emerge directly from hard higher - twist subprocess @xcite even at the lhc . the direct production of hadrons can explain @xcite the remarkable baryon anomaly " observed at rhic : the ratio of baryons to mesons at high @xmath6 , as well as the power - law fall - off @xmath7 at fixed @xmath8 both increase with centrality @xcite , opposite to the usual expectation that protons should suffer more energy loss in the nuclear medium than mesons . + a fundamental test of leading - twist qcd predictions in high transverse momentum hadronic reactions is the measurement of the power - law fall - off of the inclusive cross section @xcite @xmath9 at fixed @xmath10 and fixed @xmath11 where @xmath12 . here @xmath13 is the correction to the conformal prediction arising from the qcd running coupling and the dglap evolution of the input parton distribution and fragmentation functions @xcite . the usual expectation is that leading - twist subprocesses will dominate measurements of high @xmath2 hadron production at rhic and tevatron energies . in fact , the data for isolated photon production @xmath14 as well as jet production , agrees well with the leading - twist scaling prediction @xmath15 as shown in fig.[fignew1b ] @xcite . however , as seen in fig.[fignew1b ] , measurements of @xmath16 for hadron production are not consistent with the leading twist predictions . striking deviations from the leading - twist predictions were also observed at lower energy at the isr and fermilab fixed - target experiments @xcite . the high values @xmath17 with @xmath18 seen in the data indicate the presence of an array of higher - twist processes , including subprocesses where the hadron enters directly , rather than through jet fragmentation @xcite . the predicted deviations for the experimental and nlo scaling exponent at rhic and the lhc with phenix preliminary measurements are shown in fig . [ fignew2b ] . + at fixed @xmath18 . the data from r806 , phenix , isr / fnal , e706 are for charged or neutral pion production , whereas the cdf , ua1 data at small @xmath18 are for charged hadrons . the blue curve is the prediction of leading - twist qcd for isolated photon and jet production , including the scale - breaking effects of the running coupling and the evolution of the proton structure functions . the red curve is the qcd prediction for pion production , which also includes the effect from the evolution of the fragmentation function . the dashed line at @xmath19 is the prediction of the scale - invariant parton model . from arleo , et al . @xcite . , width=680 ] + between the experimental and nlo scaling exponent at rhic ( @xmath20 gev ) and the lhc ( @xmath21 tev as compared to @xmath22 tev ) , compared to phenix preliminary measurements . from arleo , et al . @xcite . , width=680 ] + i will discuss further consequences of direct qcd production processes in section ii . 2 . _ breakdown of perturbative qcd factorization and universality _ + the effects of initial and final - state interactions of the quarks and gluons entering hard processes at the lhc are usually assumed to be of higher - twist origin and thus power - law suppressed . however , as emphasized by collins and qiu @xcite , the traditional factorization formalism of perturbative qcd fails in detail for transverse - momentum - dependent observables in hard inclusive reactions because of initial- and final - state gluonic interactions at leading twist . + it is now well - understood that the final - state gluonic interactions of the scattered quark in deep inelastic lepton scattering lead to a @xmath23-odd non - zero spin correlation of the plane of the lepton - quark scattering plane with the polarization of the target proton @xcite . this leading - twist bjorken - scaling sivers effect " is process - dependent since qcd predicts an opposite - sign correlation @xcite in drell - yan reactions due to the initial - state interactions of the annihilating antiquark . the same final - state interactions of the struck quark with the spectators @xcite also lead to diffractive events in deep inelastic scattering ( ddis ) at leading twist , such as @xmath24 where the proton remains intact and isolated in rapidity ; in fact , approximately 10% of the deep inelastic lepton - proton scattering events observed at hera are diffractive @xcite . the presence of a rapidity gap between the target and the diffracted proton requires that the target remnant emerges in a color - singlet state ; this is made possible in any gauge by the soft rescattering incorporated in the wilson line or by augmented light - front wavefunctions @xcite . + in the case of hadron - hadron collisions , the quark and antiquark in the drell - yan subprocess @xmath25 will interact with the spectators of the other hadron ; this leads to an anomalous @xmath26 planar correlation in unpolarized drell - yan reactions @xcite . this double boer - mulders effect " can account for the large @xmath27 correlation and the corresponding violation @xcite of the lam tung relation for drell - yan processes observed by the na10 collaboration . another important signal for factorization breakdown at the lhc will be the observation of a @xmath27 planar correlation in dijet production . + it is usually assumed following the intuition of the parton model that the structure functions measured in deep inelastic scattering can be computed in the bjorken - scaling leading - twist limit from the absolute square of the light - front wavefunctions , summed over all fock states . in fact , the dynamical effects , such as the sivers spin correlation and diffractive deep inelastic lepton scattering due to final - state gluon interactions , contribute to the experimentally observed dis cross sections . diffractive events also lead to the interference of two - step and one - step processes in nuclei which in turn , via the gribov - glauber theory , lead to the shadowing and the antishadowing of the deep inelastic nuclear structure functions @xcite ; such phenomena are not included in the light - front wavefunctions of the nuclear eigenstate . this leads to an important distinction between dynamical " vs. static " ( wavefunction - specific ) structure functions @xcite . + gev@xmath28 . from i. schienbein et al . @xcite , width=680 ] 3 . _ non - universal nuclear distributions _ + it is usually assumed that the nuclear modifications to the structure functions measured in deep inelastic lepton - nucleus and neutrino - nucleus interactions are identical ; in fact , the gribov - glauber theory predicts that the antishadowing of nuclear structure functions is not universal , but depends on the quantum numbers of each struck quark and antiquark @xcite . this observation can explain the recent analysis of schienbein et al . @xcite which shows that the nutev measurements of nuclear structure functions obtained from neutrino charged current reactions differ significantly from the distributions measured in deep inelastic electron and muon scattering in the @xmath29 domain . [ fignew9 ] . i will discuss this in further detail in section iv . intrinsic heavy - sea quark distributions _ + most parametrizations of the charm and bottom quark distributions in the proton structure functions only have support at low @xmath1 since it is conventionally assumed that they only arise from gluon splitting @xmath30 this erroneous assumption has led to many incorrect predictions ; it is especially misleading for heavy hadron production at the lhc . + in fact , one can show from first principles that the proton light - front wavefunction contains _ ab initio _ intrinsic heavy quark fock state components such as @xmath31 @xcite . in contrast to the usual extrinsic " contribution from gluon - splitting ( i.e , dglap evolution ) , the intrinsic contributions are connected by gluons to at least two of the valence quarks of the proton . the intrinsic heavy quarks carry most of the proton s light - cone momentum since this minimizes the off - shellness of the fock state . one can also associate the @xmath31 fock state with meson - baryon fluctuations such as @xmath32 . thus , as is the case for intrinsic strangeness , the charm and anti - charm quarks can have different momentum and spin distributions @xcite . + the probability of the intrinsic component falls as @xmath33 due to the non - abelian qcd couplings of the gluons @xcite . the heavy - quark pair @xmath34 in the intrinsic fock state is thus primarily a color - octet , and the ratio of intrinsic charm to intrinsic bottom scales scales as @xmath35 as can be verified from the operator product expansion in non - abelian qcd @xcite . + intrinsic charm and bottom explain the origin of high @xmath36 open - charm and open - bottom hadron production , as well as the single and double @xmath37 hadroproduction cross sections observed at high @xmath36 . the factorization - breaking nuclear @xmath38 dependence of hadronic @xmath37 production cross sections is also explained . kopeliovich , schmidt , soffer , goldhaber , and i @xcite have proposed a novel mechanism utilizing intrinsic heavy quarks for both diffractive @xmath39 and inclusive higgs production @xmath40 in which the higgs boson carries a significant fraction of the projectile proton momentum . the production mechanism is based on the subprocess @xmath41 where the @xmath34 in the @xmath42 intrinsic heavy quark fock state of the colliding proton has approximately @xmath43 of the projectile protons momentum . i discuss this in further detail in section vi . _ eliminating the renormalization scale ambiguity _ + it is often stated that the renormalization scale of the qcd running coupling @xmath44 can not be fixed , and thus it has to be chosen in an _ ad hoc _ fashion . this statement is clearly false . for example , in qed the renormalization scale is simply the photon virtuality @xmath45 in the conventional gell - mann low scheme since this sums all vacuum polarization corrections to all orders . in fact , as in qed , the renormalization scale for perturbative qcd can be fixed unambiguously in any scheme by shifting @xmath46 so that all terms associated with the qcd @xmath47 function vanish . in general , each set of skeleton diagrams has its respective scale . the result is independent of the choice of the initial renormalization scale @xmath48 as well as the scheme , thus satisfying the callan - symanzik equation and renormalization group invariance . + the conventional procedure where one guesses the renormalization scale , such as @xmath49 and a range such as @xmath50 is clearly problematic since it depends on the choice of renormalization scheme . this heuristic choice and range is wrong for the simplest example in qed , @xmath51 scattering , since the exact answer which sums all vacuum contributions to all orders has the scale @xmath52 in the gell - mann low scheme [ or @xmath53 in @xmath54 scheme @xcite ] ; in fact , the exact resummed answer at forward cm angles always lies outside the guessed heuristic range . in other cases , the choice of a heuristic scale leads to nonsensical physical results @xcite or even negative cross sections at nlo @xcite . + clearly the elimination of the renormalization scale ambiguity would greatly improve the precision of qcd predictions and increase the sensitivity of searches for new physics at the lhc . further discussion of the renormalization scale setting problem is given in section iii . 6 . _ qcd condensates _ + it is conventionally assumed that the vacuum of qcd contains quark @xmath55 and gluon @xmath56 vacuum condensates , although the resulting vacuum energy density leads to a @xmath57 order - of - magnitude discrepancy with the measured cosmological constant . @xcite however , a new perspective has emerged from bethe - salpeter and light - front analyses where the qcd condensates are identified as in - hadron " condensates , rather than vacuum entities ; the gell - mann - oakes - renner relation is still satisfied @xcite . the in - hadron " condensates become realized as higher fock states of the hadron when the theory is quantized at fixed light - front time @xmath58 i discuss this in further detail in section vii . 7 . _ hidden color _ + in nuclear physics nuclei are composites of nucleons . however , qcd provides a new perspective : @xcite six quarks in the fundamental @xmath59 representation of @xmath60 color can combine into five different color - singlet combinations , only one of which corresponds to a proton and neutron . the deuteron wavefunction is a proton - neutron bound state at large distances , but as the quark separation becomes smaller , qcd evolution due to gluon exchange introduces four other hidden color " states into the deuteron wavefunction @xcite . the normalization of the deuteron form factor observed at large @xmath61 @xcite , as well as the presence of two mass scales in the scaling behavior of the reduced deuteron form factor @xcite , suggest sizable hidden - color fock state contributions in the deuteron wavefunction @xcite . hidden color can also play an important role in nuclear collisions involving quark distributions at high @xmath4 8 . _ light - front holography and hadronization at the amplitude level _ + a long - sought goal in hadron physics is to find a simple analytic first approximation to qcd analogous to the schrdinger - coulomb equation of atomic physics . this problem is particularly challenging since the formalism must be relativistic , color - confining , and consistent with chiral symmetry . de teramond and i have shown that the correspondence between theories in a positive dilaton - modified five - dimensional anti - de sitter space and confining field theories in physical space - time , leads to a simple schrdinger - like light - front wave equation and a remarkable one - parameter description of nonperturbative hadron dynamics @xcite . the model predicts a zero - mass pion for zero - mass quarks and a regge spectrum of linear trajectories with the same slope in the ( leading ) orbital angular momentum @xmath62 of the hadrons and their radial quantum number @xmath63 . + light - front holography " @xcite allows one to map the amplitudes @xmath64 in ads space directly to the light - front wavefunctions defined at fixed light - front time in 3 + 1 space . the resulting lorentz - invariant relativistic light - front wave equations are functions of an invariant impact variable @xmath65 which measures the separation of the quark and gluonic constituents within the hadron at equal light - front time . this correspondence was derived by showing that the polchinski - strassler formula @xcite for form factors in ads space is equivalent to the drell - yan west light - front matrix element both for external electromagnetic and gravitational currents . one then finds an exact mapping between @xmath66 in ads space and the invariant impact separation @xmath65 in 3 + 1 space - time . in the case of two - parton wavefunctions , one has @xmath67 , where @xmath68 is the usual impact separation conjugate to @xmath69 and @xmath70 is the light - front fraction . this correspondence agrees with the intuition that @xmath66 is related inversely to the internal relative momentum , but the relation @xmath71 is precise and exact . this relation also provides a direct connection between light - front hamiltonian equations for bound state systems and the ads wave equations . one thus obtains a semi - classical frame - independent first approximation to the spectra and light - front wavefunctions of meson and baryon light - quark bound states , which in turn predicts the behavior of the pion and nucleon form factors . the theory implements chiral symmetry in a novel way : the effects of chiral symmetry breaking increase as one goes toward large interquark separation . the the hadron eigenstates generally have components with different orbital angular momentum ; e.g. , the proton eigenstate in ads / qcd with massless quarks has @xmath72 and @xmath73 light - front fock components with equal probability . the ads / qcd soft - wall model also predicts the form of the non - perturbative effective coupling @xmath74 and its @xmath47-function @xcite . + -matrix is computed in light - front hamiltonian perturbation theory , evolving in light - front time @xmath75 . the hadronic light - front wavefunctions convert the off - shell quarks and gluons to on - shell hadrons . , width=680 ] + the ads / qcd light - front wavefunctions obtained from ads / qcd and light - front holography provide a method for computing the hadronization of quark and gluon jets at the amplitude level @xcite . this is illustrated for @xmath76 hadrons in fig . [ hadronize ] . an analogous method was used for qed to predict the production of relativistic antihydrogen @xcite . it should be emphasized that the existence of dynamical higher - twist processes in which a hadron interacts directly within a hard subprocess is a prediction of qcd . for example , the subprocess @xmath77 where the pion is produced directly through the pion s @xmath78 distribution amplitude @xmath79 underlies deeply virtual meson scattering @xmath80 the corresponding timelike subprocess @xmath81 dominates the drell - yan reaction @xmath82 at high @xmath36 @xcite , thus accounting for the change in angular distribution from the canonical @xmath83 distribution , for transversely polarized virtual photons , to @xmath84 , corresponding to longitudinal photons ; the virtual photon thus becomes longitudinally polarized at high @xmath36 , reflecting the spin of the pion entering the direct qcd hard subprocess . crossing predicts reactions where the final - state hadron appears directly in the subprocess such as @xmath85 at @xmath86 . the nominal power - law fall - off at fixed @xmath18 is set by the number of elementary fields entering the hard subprocess @xmath87 the power - law fall - off @xmath88 at high @xmath18 is set by the total number of spectators @xmath89 @xcite , up to spin corrections . the direct higher - twist subprocesses , where the trigger hadron is produced within the hard subprocess avoid the waste of same - side energy , thus allowing the target and projectile structure functions to be evaluated at the minimum values of @xmath90 and @xmath91 where they are at their maximum . examples of direct baryon and meson higher - twist subprocesses are : @xmath92 these direct subprocesses involve the distribution amplitude of the hadron which has dimension @xmath93 for mesons and @xmath94 for baryons ; thus these higher - twist contributions to the inclusive cross section @xmath95 at fixed @xmath18 nominally scale as @xmath96 for mesons and @xmath97 for baryons . . , width=680 ] at fixed @xmath98 as a function of centrality versus peripheral collisions , using spectra at @xmath99 gev and @xmath100 gev @xcite . the positive - charged hadron trigger is dominated by protons at high @xmath2 for central collisions , consistent with the color transparency of direct higher - twist baryon production processes . , width=680 ] the behavior of the single - particle inclusive cross section will be a key test of qcd at the lhc , since the leading - twist prediction for @xmath101 is independent of the detailed form of the structure and fragmentation functions . the fixed-@xmath18 scaling of the proton production cross section @xmath102 is particularly anomalous , far from the @xmath103 to @xmath104 scaling predicted by pqcd @xcite . [ fignew1b ] . sickles and i @xcite have argued that the anomalous features of inclusive high @xmath2 proton production is due to hard subprocesses @xcite where the proton is created directly within the hard reaction , such as @xmath105 , such as the mechanism illustrated in fig . [ figdirectproton ] . the fragmentation of a gluon or quark jet to a proton requires that the underlying 2 to 2 subprocess occurs at a higher transverse momentum than the @xmath2 of the observed proton because of the fast - falling quark - to - proton fragmentation function @xmath106 at high momentum fraction @xmath66 ; in contrast , the direct subprocess is maximally energy efficient . such direct " reactions thus can explain the fast - falling power - law falloff observed at fixed @xmath18 and fixed-@xmath107 at the isr , fermilab and rhic @xcite . since the proton is initially produced as a small - size @xmath108 color - singlet state , it is color transparent " @xcite , and it can thus propagate through dense nuclear matter with minimal energy loss . in contrast , the pions which are produced from jet fragmentation have a normal inelastic cross section . this provides an explanation @xcite of the rhic data @xcite , which shows a dramatic rise of the @xmath109 to @xmath110 ratio with increasing @xmath2 when one compares peripheral with central heavy ion collisions , as illustrated in fig . [ fignew2 ] . the color transparency of the proton produced in the direct process also can explain why the index @xmath17 rises with centrality , as seen in fig . [ fignew6 ] , the higher - twist color - transparent subprocess dominates in the nuclear medium @xcite . in addition , the fact that the proton tends to be produced alone in a direct subprocess explains why the yield of same - side hadrons along the proton trigger is diminished with increasing centrality . thus the qcd color transparency of directly produced baryons can explain the baryon anomaly " seen in heavy - ion collisions at rhic : the color - transparent proton state is not absorbed , but a pion produced from fragmentation is diminished in the nuclear medium @xcite . the increase of @xmath17 with centrality is consistent with the nuclear survival of direct higher - twist subprocesses for both protons and antiprotons , and to a lesser extent , for mesons . predicted by light - front holographic mapping for @xmath111 is compared with effective qcd couplings extracted from different observables and lattice results . details on the comparison with other effective charges are given in ref . in the blm method @xcite , the qcd scale @xmath46 is chosen , just as in qed , such that in principle all of the terms associated with the @xmath47 function are summed into the qcd coupling @xcite quark loops can be used to identify the @xmath47 terms , at least to two loops the remaining terms are identical to that of a conformal theory . unlike heuristic scale - setting procedures , the blm method gives results which are independent of the choice of renormalization scheme , as required by the transitivity property of the renormalization group . the divergent renormalon terms of order @xmath112 are transferred to the physics of the running coupling . furthermore , one retains sensitivity to conformal " effects which arise in higher orders , physical effects which are not associated with qcd renormalization . in contrast , the factorization scale @xmath113 which sets the separation between nonperturbative dynamics of hadrons and the perturbative evolution of their parton distribution functions is arbitrary ; it is unrelated to the factorization scale @xmath46 since it is present even in a conformal theory . the blm method also provides scale - fixed , scheme - independent connections between observables , such as the generalized crewther relation " @xcite , in which crewther s conformal prediction @xcite between the bjorken sum rule and the annihilation cross section is effectively restored , as well as other commensurate scale relations " @xcite . the ratio of scales in the two observables is scheme - independent and ensures that the same number of quark flavors are active . such relations between observables provide high precision test of qcd . the blm method was used in ref . @xcite to unambiguously fix the pomeron intercept predicted by the bfkl analysis . the blm method is also correct in the abelian limit @xmath114 at fixed @xmath115 @xcite , where @xmath116 is the casimir constant for @xmath117 the consistent application of the blm method to pqcd leads in most cases to multiple renormalization scales associated with different skeleton diagrams ; for example in electron - electron scattering in qed there are separate scales @xmath118 and @xmath119 for photon exchange in the @xmath120 and @xmath121-channels respectively . in the case of @xmath122 the renormalization scales are @xmath123 and @xmath124 where @xmath125 is complex . the scale controlling the three gluon coupling in qcd is particularly interesting . in the case of lhc processes such as @xmath126 where the 3-gluon coupling enters at very different gluon virtualities , the blm scale can be shown to be proportional to @xmath127 where @xmath128 is the largest of the gluon virtualities @xcite . this scale controls the onset of the new quark flavors that enter the triangle graph . thus , in many cases , the actual renormalization scale is much smaller than the largest hard scale in the process . two distinctly different scales arise in heavy quark production at threshold : the relative momentum of the quarks governing the soft gluon exchange responsible for the coulomb potential , and a large momentum scale approximately equal to twice the quark mass for the corrections induced by transverse gluons @xcite . since one probes the qcd coupling at small virtuality even at the lhc , it is imperative to have good control on the qcd coupling in the soft domain . it is usually assumed that the qcd coupling @xmath129 diverges at @xmath130 ; i.e. , infrared slavery " . in fact , determinations from lattice gauge theory , bethe - salpeter methods , effective charge measurements , gluon mass phenomena , and ads / qcd all lead ( in their respective schemes ) to a finite value of the qcd coupling in the infrared @xcite . because of color confinement , the quark and gluon propagators vanish at long wavelength : @xmath131 , and consequently , the quantum - loop corrections underlying the qcd @xmath47-function decouple in the infrared , and the coupling freezes to a finite value at @xmath132 @xcite . [ alphas ] . this observation underlies the use of conformal methods in ads / qcd . clearly the application of the blm method to eliminate of the renormalization scale ambiguity would greatly improve the precision of qcd predictions and increase the sensitivity of searches for new physics at the lhc . the shadowing of the nuclear structure functions : @xmath133 at small @xmath134 can be readily understood in terms of the gribov - glauber theory . consider a two - step process in the nuclear target rest frame . the incoming @xmath135 dipole first interacts diffractively @xmath136 on nucleon @xmath137 leaving it intact . this is the leading - twist diffractive deep inelastic scattering ( ddis ) process which has been measured at hera to constitute approximately 10% of the dis cross section at high energies . the @xmath135 state then interacts inelastically on a downstream nucleon @xmath138 @xmath139 . the phase of the pomeron - dominated ddis amplitude is close to imaginary , and the glauber cut provides another phase @xmath140 , so that the two - step process has opposite phase and destructively interferes with the one - step dis process @xmath141 where @xmath137 acts as an unscattered spectator . the one - step and - two - step amplitudes can coherently interfere as long as the momentum transfer to the nucleon @xmath137 is sufficiently small that it remains in the nuclear target ; _ i.e. _ , the ioffe length @xcite @xmath142 is large compared to the inter - nucleon separation . in effect , the flux reaching the interior nucleons is diminished , thus reducing the number of effective nucleons and @xmath143 the bjorken - scaling diffractive contribution to dis arises from the rescattering of the struck quark after it is struck ( in the parton model frame @xmath144 ) , an effect induced by the wilson line connecting the currents . thus one can not attribute ddis to the physics of the target nucleon computed in isolation @xcite . one of the novel features of qcd involving nuclei is the _ antishadowing _ of the nuclear structure functions as observed in deep inelastic lepton - nucleus scattering . empirically , one finds @xmath145 in the domain @xmath29 ; _ i.e. _ , the measured nuclear structure function ( referenced to the deuteron ) is larger than the scattering on a set of @xmath146 independent nucleons . ivan schmidt , jian - jun yang , and i @xcite have extended the analysis of nuclear shadowing to the shadowing and antishadowing of the electroweak structure functions . we note that there are leading - twist diffractive contributions @xmath147 arising from reggeon exchanges in the @xmath120-channel @xcite . for example , isospin non - singlet @xmath148 reggeons contribute to the difference of proton and neutron structure functions , giving the characteristic kuti - weisskopf @xmath149 behavior at small @xmath1 . the @xmath1 dependence of the structure functions reflects the regge behavior @xmath150 of the virtual compton amplitude at fixed @xmath61 and @xmath151 the phase of the diffractive amplitude is determined by analyticity and crossing to be proportional to @xmath152 for @xmath153 which together with the phase from the glauber cut , leads to _ constructive _ interference of the diffractive and nondiffractive multi - step nuclear amplitudes . the nuclear structure function is predicted to be enhanced precisely in the domain @xmath154 where antishadowing is empirically observed . the strength of the reggeon amplitudes is fixed by the regge fit to the nucleon structure functions , so there is little model dependence . since quarks of different flavors will couple to different reggeons ; this leads to the remarkable prediction that nuclear antishadowing is not universal ; it depends on the quantum numbers of the struck quark . this picture implies substantially different antishadowing for charged and neutral current reactions , thus affecting the extraction of the weak - mixing angle @xmath155 . the ratio of nuclear to nucleon structure functions is thus process dependent . we have also identified contributions to the nuclear multi - step reactions which arise from odderon exchange and hidden color degrees of freedom in the nuclear wavefunction . schienbein et al . @xcite have recently given a comprehensive analysis of charged current deep inelastic neutrino - iron scattering , finding significant differences with the nuclear corrections for electron - iron scattering . [ fignew9 ] . the nuclear effect measured in the nutev deep inelastic scattering charged current experiment is distinctly different from the nuclear modification measured at slac and nmc in deep inelastic scattering electron and muon scattering . this implies that part of of the anomalous nutev result @xcite for @xmath155 could be due to the non - universality of nuclear antishadowing for charged and neutral currents . this effect could also explain the absence of antishadowing observed in drell - yan reactions @xcite . a new understanding of nuclear shadowing and antishadowing has emerged based on multi - step coherent reactions involving leading twist diffractive reactions @xcite . the nuclear shadowing of structure functions is a consequence of the lepton - nucleus collision ; it is not an intrinsic property of the nuclear wavefunction . the same analysis shows that antishadowing is _ not universal _ , but it depends in detail on the flavor of the quark or antiquark constituent @xcite . detailed measurements of the nuclear dependence of individual quark structure functions are thus needed to establish the distinctive phenomenology of shadowing and antishadowing and to make the nutev results definitive . a comparison of the nuclear modification in neutrino versus anti - neutrino interactions is clearly important . there are other ways in which this new view of antishadowing can be tested ; for example , antishadowing can also depend on the target and beam polarization . the nontrivial effects from rescattering and diffraction highlight the need for a fundamental understanding the dynamics of hadrons in qcd at the amplitude level . this is essential for understanding phenomena such as the quantum mechanics of hadron formation , the remarkable effects of initial and final interactions , the origins of diffractive phenomena and single - spin asymmetries , and manifestations of higher - twist semi - exclusive hadron subprocesses . a central tool in these analyses is the light - front wavefunctions of hadrons , the frame - independent eigensolutions of the heisenberg equation for qcd @xmath156 quantized at fixed light - front . given the light - front wavefunctions @xmath157 , one can compute a large range of exclusive and inclusive hadron observables . for example , the valence , sea - quark and gluon distributions are defined from the squares of the lfwfs summed over all fock states @xmath158 . form factors , exclusive weak transition amplitudes @xcite such as @xmath159 and the generalized parton distributions @xcite measured in deeply virtual compton scattering are ( assuming the handbag " approximation ) overlaps of the initial and final lfwfs with @xmath160 and @xmath161 . it is thus important to distinguish static " structure functions which are computed directly from the light - front wavefunctions of a target hadron from the nonuniversal dynamic " empirical structure functions which take into account rescattering of the struck quark in deep inelastic lepton scattering . [ fignew17 ] . the real wavefunctions underlying static structure functions can not describe diffractive deep inelastic scattering nor single - spin asymmetries , since such phenomena involve the complex phase structure of the @xmath162 amplitude . one can augment the light - front wavefunctions with a gauge link corresponding to an external field created by the virtual photon @xmath135 pair current @xcite , but such a gauge link is process dependent @xcite , so the resulting augmented wavefunctions are not universal . @xcite . the physics of rescattering and nuclear shadowing is not included in the nuclear light - front wavefunctions , and a probabilistic interpretation of the nuclear dis cross section is precluded . intrinsic heavy quark distributions are a rigorous feature of qcd , arising from diagrams in which two or more gluons couple the valence quarks to the heavy quarks . the probability for fock states of a light hadron to have an extra heavy quark pair decreases as @xmath163 in non - abelian gauge theory @xcite . the relevant matrix element is the cube of the qcd field strength @xmath164 in contrast to qed where the relevant operator is @xmath165 and the probability of intrinsic heavy leptons in an atomic state is suppressed as @xmath166 the maximum probability occurs at @xmath167 where @xmath168 ; _ i.e. _ , when the constituents have minimal invariant mass and equal rapidity . thus the heaviest constituents have the highest momentum fractions and the highest @xmath169 . intrinsic charm thus predicts that the charm structure function has support at large @xmath170 in excess of dglap extrapolations @xcite ; this is in agreement with the emc measurements @xcite . intrinsic charm can also explain the @xmath171 puzzle @xcite . it also affects the extraction of suppressed ckm matrix elements in @xmath172 decays @xcite . the dissociation of the intrinsic charm @xmath31 fock state of the proton can produce a leading heavy quarkonium state at high @xmath173 in @xmath174 since the @xmath175 and @xmath176 can readily coalesce into the charmonium state . since the constituents of a given intrinsic heavy - quark fock state tend to have the same rapidity , coalescence of multiple partons from the projectile fock state into charmed hadrons and mesons is also favored . for example , one can produce a leading @xmath177 at high @xmath36 and low @xmath2 from the coalescence of the @xmath178 constituents of the projectile @xmath31 fock state . production , width=680 ] the operator product analysis of the ic matrix element shows that the ic fock state has a dominant color - octet structure : @xmath179 . the color octet @xmath180 converts to a color singlet by gluon exchange on the front surface of a nuclear target and then coalesces to a @xmath37 which interacts weakly through the nuclear volume @xcite . thus the rate for the ic component has @xmath181 dependence corresponding to the area of the front surface . this is illustrated in fig [ fignew10 ] . this forward contribution is in addition to the @xmath182 contribution derived from the usual perturbative qcd fusion contribution at small @xmath4 because of these two components , the cross section violates perturbative qcd factorization for hard inclusive reactions @xcite . this is consistent with the two - component cross section for charmonium production observed by the na3 collaboration at cern @xcite and more recent experiments @xcite . the diffractive dissociation of the intrinsic charm fock state leads to leading charm hadron production and fast charmonium production in agreement with measurements @xcite . the hadroproduction cross sections for double - charm @xmath183 baryons at selex @xcite and the production of @xmath37 pairs at na3 are consistent with the diffractive dissociation and coalescence of double ic fock states @xcite . these observations provide compelling evidence for the diffractive dissociation of complex off - shell fock states of the projectile and contradict the traditional view that sea quarks and gluons are always produced perturbatively via dglap evolution or gluon splitting . it is also conceivable that the observations @xcite of @xmath184 at high @xmath36 at the isr in high energy @xmath185 collisions could be due to the dissociation and coalescence of the intrinsic bottom " @xmath186 fock states of the proton . and @xmath187 in the rapidity regions @xmath188 and @xmath189 . the uncertainties for the data include both statistical ( inner line ) and full uncertainties ( entire error bar ) . also shown are the uncertainties on the theoretical pqcd scales and the cteq6.6 m parameterization of the parton distribution functions ( pdfs ) . the scale uncertainties are shown as dotted lines and the pdf uncertainties by the shaded regions . the ratio of the standard cteq6.6 m prediction to two models of intrinsic charm is also shown . from the d0 collaboration @xcite . , width=680 ] as emphasized by lai , tung , and pumplin @xcite , there are strong indications that the structure functions used to model charm and bottom quarks in the proton at large @xmath170 have been underestimated , since they ignore intrinsic heavy quark fluctuations of hadron wavefunctions . the anomalous growth of the @xmath190 inclusive cross section observed by d0 collaboration @xcite at the tevatron and shown in fig . [ figratio ] also suggests that the charm distribution has been underestimated at @xmath191 beyond what is included in the cteq6.6 m pdfs . the neglect of the intrinsic - heavy quark component in the proton structure function will lead to an incorrect assessment of the gluon distribution at large @xmath1 if it is assumed that sea quarks always arise from gluon splitting . the intrinsic charm and bottom distributions at high @xmath1 in the proton structure functions also lead to novel heavy hadron production processes such as anomalous production of charm and bottom jets at high @xmath2 at the lhc , as well as novel mechanisms for higgs and @xmath3 production at high @xmath36 @xcite it is astonishing that the original emc experiment which first observed a large signal for charm at large @xmath1 in 1983 has never been repeated . it is thus critical for experiments such as compass to definitively establish the phenomenology of the charm structure function at large @xmath192 the vacuum in quantum field theories is remarkably simple in light - front quantization because of the restriction @xmath193 for example in qed , vacuum graphs such as @xmath194 which are normally associated with a zero - point energy do not arise in the light - front vacuum . in the higgs theory , the usual higgs vacuum expectation value is replaced with a @xmath195 zero mode ; @xcite however , the resulting phenomenology is identical to the standard analysis . the usual assumption that non - zero vacuum condensates exist and possess a measurable reality has long been recognized as posing a conundrum for the light - front formulation of qcd . in the light - front formulation of qcd , the ground - state is a structureless fock space vacuum , in which case it would seem to follow that dynamical chiral symmetry breaking ( csb ) is impossible . in fact , as first argued by casher and susskind @xcite , dynamical csb must be a property of hadron wavefunctions , not of the vacuum . ( they used an infinite momentum framework which is equivalent to the front - form . ) this thesis has also been explored in a series of recent articles @xcite . it is widely held that quark and gluon vacuum condensates have a physical existence , independent of hadrons , as measurable spacetime - independent configurations of qcd s elementary degrees - of - freedom in a hadron - less ground state . however , a non - zero spacetime - independent qcd vacuum condensate poses a critical dilemma for gravitational interactions , because it would lead to a cosmological constant some 45 orders of magnitude larger than observation . as noted in ref . @xcite , this conflict is avoided if the strong interaction condensates are properties of the light - front wavefunctions of the hadrons , rather than the hadron - less ground state of qcd . conventionally , the quark and gluon condensates are considered to be properties of the qcd vacuum and hence to be constant throughout spacetime . a new perspective on the nature of qcd condensates @xmath196 and @xmath197 , particularly where they have spatial and temporal support , has recently been presented . @xcite the spatial support of condensates is restricted to the interior of hadrons , since condensates in qcd arise due to the interactions of quarks and gluons which are confined within hadrons . for example , consider a meson consisting of a light quark @xmath198 bound to a heavy antiquark , such as a @xmath172 meson . one can analyze the propagation of the light @xmath198 in the background field of the heavy @xmath199 quark . solving the dyson - schwinger equation for the light quark one obtains a nonzero dynamical mass and evidently a nonzero value of the condensate @xmath196 . but this is not a true vacuum expectation value ; in fact , it is the matrix element of the operator @xmath200 in the background field of the @xmath199 quark . the change in the ( dynamical ) mass of the light quark in a bound state is somewhat reminiscent of the energy shift of an electron in the lamb shift , in that both are consequences of the fermion being in a bound state rather than propagating freely . it is clearly important to use the equations of motion for confined quarks and gluon fields when analyzing current correlators in qcd , not free propagators , as has often been done in traditional analyses . since the distance between the quark and antiquark can not become arbitrarily large , one can not create a quark condensate which has uniform extent throughout the universe . thus in a fully self - consistent treatment of a qcd bound state , the condensate phenomenon occurs in the background field of the @xmath199-quark , whose influence on light - quark propagation is primarily concentrated in the far infrared and whose presence ensures the manifestations of light - quark dressing are gauge invariant . in the case of the pion , one can show from the bethe - salpeter equation that the vacuum quark condensate that appears in the gell - mann - oakes - renner formula , is , in fact , a chiral - limit value of an ` in - pion ' condensate @xcite . this condensate is no more a property of the `` vacuum '' than the pion s chiral - limit leptonic decay constant . one can connect the bethe - salpeter formalism to the light - front formalism , by fixing the light - front time @xmath201 . this then leads to the fock state expansion . in fact , dynamical csb in the light - front formulation , expressed via ` in - hadron ' condensates , can be shown to be connected with sea - quarks derived from higher fock states . this solution is similar to that discussed in ref . moreover , ref . @xcite establishes the equivalence of all three definitions of the vacuum quark condensate : a constant in the operator product expansion , @xcite via the banks - casher formula , @xcite and the trace of the chiral - limit dressed - quark propagator . i have reviewed a number of qcd topics where conventional wisdom relevant to hadronic physics at the lhc has been challenged . for example , the initial - state and final - state interactions of the quarks and gluons entering perturbative qcd hard - scattering subprocesses lead to the breakdown of traditional concepts of factorization and universality at leading twist . these soft - gluon rescattering , which are associated with the wilson line of the propagating partons , lead to bjorken - scaling single - spin asymmetries , diffractive deep inelastic scattering , the breakdown of the lam - tung leading twist relation in drell - yan reactions , as well as nuclear shadowing . furthermore , the gribov - glauber theory applied to the antishadowing domain predicts that nuclear structure functions are not universal , but instead depend on the flavor quantum numbers of each quark and antiquark , thus explaining the anomalous nuclear dependence recently observed in deep - inelastic neutrino scattering . surprisingly , isolated hadrons can be produced at large transverse momentum at a significant rate at the lhc directly within a hard higher - twist qcd subprocess , rather than from jet fragmentation . direct " processes can explain the observed deviations from perturbative qcd predictions in measurements of inclusive hadron cross sections at fixed @xmath0 , as well as the baryon anomaly " , the anomalously large proton - to - pion ratio seen in high centrality heavy - ion collisions . the intrinsic charm contribution to the proton structure function at high @xmath1 can explain the anomalously large rate for high @xmath2 photon plus charm jet events observed by d0 at the tevatron . intrinsic charm and bottom distributions also imply anomalously large production of charm and bottom jets at high @xmath2 at the lhc , as well as a novel mechanism for higgs and @xmath3 production at high @xmath4 the correspondence between theories in a warped anti - de sitter space and light - front quantization in physical space - time is very powerful , and leads to much insight into qcd dynamics , including a nonperturbative qcd running coupling and a remarkably accurate relativistic lf schrodinger equation which reproduces much of light - quark spectroscopy and dynamics , using a soft - wall model of with a positive sign dilaton . other novel features of qcd have also been discussed , including the consequences of confinement for quark and gluon condensates and the implications for the qcd contribution to the cosmological constant . i have also emphasized that setting the renormalization scale of the qcd coupling using the scheme - independent blm method will greatly improve the precision of qcd predictions , and thus greatly increase the sensitivity of searches for new physics at the lhc . invited talk , presented at the 5th workshop on high @xmath203 physics at the lhc held at the instituto de ciencias nucleares of the universidad national automata de mexico in mexico city , september 27-october 1 , 2010 . i am grateful to the organizers for their invitation to this meeting , and i thank all of my collaborators whose work has been cited in this report . this research was supported by the department of energy , contract de ac0276sf00515 . slac - pub-14302 . f. arleo , s. j. brodsky , d. s. hwang and a. m. sickles , phys . lett . * 105 * , 062002 ( 2010 ) [ arxiv:0911.4604 [ hep - ph ] ] . f. arleo , s. j. brodsky , d. s. hwang and a. m. sickles , arxiv:1006.4045 [ hep - ph ] . s. j. brodsky and a. sickles , phys . lett . b * 668 * , 111 ( 2008 ) [ arxiv:0804.4608 [ hep - ph ] ] . s. s. adler _ et al . _ [ phenix collaboration ] , phys . lett . * 91 * , 172301 ( 2003 ) [ arxiv : nucl - ex/0305036 ] . d. w. sivers , s. j. brodsky and r. blankenbecler , phys . * 23 * , 1 ( 1976 ) . s. j. brodsky and m. rijssenbeek , arxiv : hep - ph/0511178 . j. w. cronin , h. j. frisch , m. j. shochet , j. p. boymond , p. a. piroue and r. l. sumner , phys . lett . * 31 * , 1426 ( 1973 ) . d. antreasyan , j. w. cronin , h. j. frisch , m. j. shochet , l. kluberg , p. a. piroue and r. l. sumner , phys . d * 19 * , 764 ( 1979 ) . r. blankenbecler , s. j. brodsky and j. f. gunion , phys . d * 12 * , 3469 ( 1975 ) . j. collins and j. w. qiu , phys . d * 75 * , 114014 ( 2007 ) [ arxiv:0705.2141 [ hep - ph ] ] . s. j. brodsky , d. s. hwang and i. schmidt , phys . b * 530 * , 99 ( 2002 ) [ arxiv : hep - ph/0201296 ] . j. c. collins , phys . b * 536 * , 43 ( 2002 ) [ arxiv : hep - ph/0204004 ] . s. j. brodsky , d. s. hwang and i. schmidt , nucl . b * 642 * , 344 ( 2002 ) [ arxiv : hep - ph/0206259 ] . s. j. brodsky , p. hoyer , n. marchal , s. peigne and f. sannino , phys . d * 65 * , 114025 ( 2002 ) [ arxiv : hep - ph/0104291 ] . c. adloff _ et al . _ [ h1 collaboration ] , z. phys . c * 76 * , 613 ( 1997 ) [ arxiv : hep - ex/9708016 ] . j. breitweg _ et al . _ [ zeus collaboration ] , eur . j. c * 6 * , 43 ( 1999 ) [ arxiv : hep - ex/9807010 ] . s. j. brodsky , b. pasquini , b. w. xiao and f. yuan , phys . b * 687 * , 327 ( 2010 ) [ arxiv:1001.1163 [ hep - ph ] ] . d. boer , s. j. brodsky and d. s. hwang , phys . d * 67 * , 054003 ( 2003 ) [ arxiv : hep - ph/0211110 ] . d. boer , phys . d * 60 * , 014012 ( 1999 ) [ arxiv : hep - ph/9902255 ] . s. j. brodsky , i. schmidt and j. j. yang , phys . d * 70 * , 116003 ( 2004 ) [ arxiv : hep - ph/0409279 ] . s. j. brodsky , nucl . a * 827 * , 327c ( 2009 ) [ arxiv:0901.0781 [ hep - ph ] ] . i. schienbein , j. y. yu , c. keppel , j. g. morfin , f. i. olness and j. f. owens , arxiv:0806.0723 [ hep - ph ] . s. j. brodsky , p. hoyer , c. peterson and n. sakai , phys . lett . b * 93 * , 451 ( 1980 ) . s. j. brodsky , j. c. collins , s. d. ellis , j. f. gunion and a. h. mueller , published in _ design and utilization of the ssc , snowmass , 1984 _ : proceedings . edited by rene donaldson and jorge g. morfin . batavia , ill . , fermilab , 1985 . . b. w. harris , j. smith and r. vogt , nucl . b * 461 * , 181 ( 1996 ) [ arxiv : hep - ph/9508403 ] . m. franz , m. v. polyakov and k. goeke , phys . d * 62 * , 074024 ( 2000 ) [ arxiv : hep - ph/0002240 ] . s. j. brodsky and b. q. ma , phys . b * 381 * , 317 ( 1996 ) [ arxiv : hep - ph/9604393 ] . s. j. brodsky , b. kopeliovich , i. schmidt and j. soffer , phys . d * 73 * , 113005 ( 2006 ) [ arxiv : hep - ph/0603238 ] . s. j. brodsky , m. s. gill , m. melles and j. rathsman , phys . d * 58 * , 116006 ( 1998 ) [ arxiv : hep - ph/9801330 ] . g. kramer and b. lampe , z. phys . a * 339 * , 189 ( 1991 ) . c. f. berger _ _ , arxiv:1009.2338 [ hep - ph ] . s. j. brodsky and r. shrock , arxiv:0905.1151 [ hep - th ] . s. j. brodsky , c. d. roberts , r. shrock and p. c. tandy , phys . rev . c * 82 * , 022201 ( 2010 ) [ arxiv:1005.4610 [ nucl - th ] ] . s. j. brodsky and b. t. chertok , phys . d * 14 * , 3003 ( 1976 ) . v. a. matveev and p. sorba , lett . nuovo cim . * 20 * , 435 ( 1977 ) . s. j. brodsky , c. r. ji and g. p. lepage , phys . lett . * 51 * , 83 ( 1983 ) . r. g. arnold _ et al . _ , phys . lett . * 35 * , 776 ( 1975 ) . g. r. farrar , k. huleihel and h. y. zhang , phys . * 74 * , 650 ( 1995 ) . g. f. de teramond and s. j. brodsky , phys . lett . * 94 * , 201601 ( 2005 ) [ arxiv : hep - th/0501022 ] . g. f. de teramond and s. j. brodsky , phys . lett . * 102 * , 081601 ( 2009 ) [ arxiv:0809.4899 [ hep - ph ] ] . s. j. brodsky and g. f. de teramond , arxiv:1009.4232 [ hep - ph ] . j. polchinski and m. j. strassler , phys . lett . * 88 * , 031601 ( 2002 ) [ arxiv : hep - th/0109174 ] . s. j. brodsky , g. f. de teramond and a. deur , phys . d * 81 * , 096010 ( 2010 ) [ arxiv:1002.3948 [ hep - ph ] ] . s. j. brodsky , g. de teramond and r. shrock , aip conf . proc . * 1056 * , 3 ( 2008 ) [ arxiv:0807.2484 [ hep - ph ] ] . c. t. munger , s. j. brodsky and i. schmidt , phys . d * 49 * , 3228 ( 1994 ) . e. l. berger and s. j. brodsky , phys . * 42 * , 940 ( 1979 ) . a. m. sickles , nucl . a * 830 * , 131c ( 2009 ) [ arxiv:0907.4921 [ nucl - ex ] ] . s. j. brodsky and a. h. mueller , phys . b * 206 * , 685 ( 1988 ) . s. j. brodsky , g. p. lepage and p. b. mackenzie , phys . d * 28 * , 228 ( 1983 ) . g. grunberg and a. l. kataev , phys . b * 279 * , 352 ( 1992 ) . s. j. brodsky , g. t. gabadadze , a. l. kataev and h. j. lu , phys . b * 372 * , 133 ( 1996 ) [ arxiv : hep - ph/9512367 ] . a. l. kataev and s. v. mikhailov , arxiv:1011.5248 [ hep - ph ] . r. j. crewther , phys . rev . lett . * 28 * , 1421 ( 1972 ) . d. j. broadhurst and a. l. kataev , phys . b * 315 * , 179 ( 1993 ) [ arxiv : hep - ph/9308274 ] . s. j. brodsky and h. j. lu , phys . d * 51 * , 3652 ( 1995 ) [ arxiv : hep - ph/9405218 ] . s. j. brodsky , e. gardi , g. grunberg and j. rathsman , phys . d * 63 * , 094017 ( 2001 ) [ arxiv : hep - ph/0002065 ] . s. j. brodsky , v. s. fadin , v. t. kim , l. n. lipatov and g. b. pivovarov , jetp lett . * 70 * , 155 ( 1999 ) [ arxiv : hep - ph/9901229 ] . s. j. brodsky and p. huet , phys . b * 417 * , 145 ( 1998 ) [ arxiv : hep - ph/9707543 ] . m. binger and s. j. brodsky , phys . d * 74 * , 054016 ( 2006 ) [ arxiv : hep - ph/0602199 ] . s. j. brodsky , a. h. hoang , j. h. kuhn and t. teubner , phys . b * 359 * , 355 ( 1995 ) [ arxiv : hep - ph/9508274 ] . s. j. brodsky and g. f. de teramond , phys . d * 77 * , 056007 ( 2008 ) [ arxiv:0707.3859 [ hep - ph ] ] . s. j. brodsky and r. shrock , phys . b * 666 * , 95 ( 2008 ) [ arxiv:0806.1535 [ hep - th ] ] . b. l. ioffe , phys . b * 30 * , 123 ( 1969 ) . s. j. brodsky and h. j. lu , phys . * 64 * , 1342 ( 1990 ) . g. p. zeller _ et al . _ [ nutev collaboration ] , phys . lett . * 88 * , 091802 ( 2002 ) [ erratum - ibid . * 90 * , 239902 ( 2003 ) ] [ arxiv : hep - ex/0110059 ] . d. m. alde _ et al . _ , phys . lett . * 64 * , 2479 ( 1990 ) . s. j. brodsky and d. s. hwang , nucl . b * 543 * , 239 ( 1999 ) [ arxiv : hep - ph/9806358 ] . s. j. brodsky , m. diehl and d. s. hwang , nucl . b * 596 * , 99 ( 2001 ) [ arxiv : hep - ph/0009254 ] . a. v. belitsky , x. ji and f. yuan , nucl . b * 656 * , 165 ( 2003 ) [ arxiv : hep - ph/0208038 ] . j. c. collins and a. metz , phys . lett . * 93 * , 252001 ( 2004 ) [ arxiv : hep - ph/0408249 ] . j. c. collins , acta phys . b * 34 * , 3103 ( 2003 ) [ arxiv : hep - ph/0304122 ] . s. j. brodsky and m. karliner , phys . lett . * 78 * , 4682 ( 1997 ) [ arxiv : hep - ph/9704379 ] . s. j. brodsky and s. gardner , phys . d * 65 * , 054016 ( 2002 ) [ arxiv : hep - ph/0108121 ] . p. hoyer , m. vanttinen and u. sukhatme , phys . b * 246 * , 217 ( 1990 ) . j. badier _ et al . _ [ na3 collaboration ] , phys . b * 104 * , 335 ( 1981 ) . m. j. leitch _ et al . _ [ fnal e866/nusea collaboration ] , phys . lett . * 84 * , 3256 ( 2000 ) [ arxiv : nucl - ex/9909007 ] . j. c. anjos , j. magnin and g. herrera , phys . b * 523 * , 29 ( 2001 ) [ arxiv : hep - ph/0109185 ] . a. ocherashvili _ et al . _ [ selex collaboration ] , phys . b * 628 * , 18 ( 2005 ) [ arxiv : hep - ex/0406033 ] . r. vogt and s. j. brodsky , phys . b * 349 * , 569 ( 1995 ) [ arxiv : hep - ph/9503206 ] . g. bari _ et al . _ , nuovo cim . a * 104 * , 1787 ( 1991 ) . v. m. abazov _ et al . _ [ d0 collaboration ] , phys . lett . * 102 * , 192002 ( 2009 ) [ arxiv:0901.0739 [ hep - ex ] ] . j. pumplin , h. l. lai and w. k. tung , phys . d * 75 * , 054029 ( 2007 ) [ arxiv : hep - ph/0701220 ] . s. j. brodsky , a. s. goldhaber , b. z. kopeliovich and i. schmidt , nucl . b * 807 * , 334 ( 2009 ) [ arxiv:0707.4658 [ hep - ph ] ] . srivastava and s. j. brodsky , phys . d * 66 * , 045019 ( 2002 ) [ arxiv : hep - ph/0202141 ] . a. casher and l. susskind , phys . d * 9 * , 436 ( 1974 ) . s. j. brodsky and r. shrock , arxiv:0803.2541 [ hep - th ] . s. j. brodsky and r. shrock , arxiv:0803.2554 [ hep - th ] . maris and c. d. roberts , phys . c * 56 * , 3369 ( 1997 ) [ arxiv : nucl - th/9708029 ] . k. langfeld , h. markum , r. pullirsch , c. d. roberts and s. m. schmidt , phys . c * 67 * , 065206 ( 2003 ) [ arxiv : nucl - th/0301024 ] . k. d. lane , phys . d * 10 * , 2605 ( 1974 ) . h. d. politzer , nucl . b * 117 * , 397 ( 1976 ) . t. banks and a. casher , nucl . b * 169 * , 103 ( 1980 ) .

applications of perturbative qcd to deeply virtual compton scattering and hard exclusive electroproduction processes require a generalization of usual parton distributions for the case when long - distance information is accumulated in nonforward matrix elements of quark and gluon light - cone operators . we describe two types of nonperturbative functions parametrizing such matrix elements : double distributions and skewed distribution functions , discuss their properties , and basic uses in the qcd description of hard exclusive processes .

hep-ph0011383

the standard feature of applications of perturbative qcd to hard processes is the introduction of phenomenological functions accumulating information about nonperturbative long - distance dynamics . the well - known examples are the parton distribution functions @xmath2 @xcite used in perturbative qcd approaches to hard inclusive processes and distribution amplitudes @xmath3 , which naturally emerge in the asymptotic qcd analyses of hard exclusive processes @xcite . the cases of deeply virtual compton scattering ( dvcs ) and hard exclusive electroproduction processes @xcite involve nonforward matrix elements @xmath4 . the important feature ( noticed long ago @xcite ) is that kinematics of hard elastic electroproduction processes ( dvcs can be also treated as one of them ) requires the presence of the longitudinal component in the momentum transfer @xmath5 from the initial hadron to the final : @xmath6 . for dvcs and @xmath7-electroproduction in the region @xmath8 , the longitudinal momentum asymmetry ( or `` skewedness '' ) parameter @xmath9 coincides with the bjorken variable @xmath10 associated with the virtual photon momentum @xmath11 @xcite . this means that the nonperturbative matrix element @xmath4 is nonsymmetric ( skewed ) , and the distributions which appear in the hard elastic electroproduction amplitudes differ from those studied in inclusive processes . in the latter case , one has a symmetric situation when the same momentum @xmath12 appears in both brackets of the hadronic matrix element @xmath13 . to parametrize nonforward matrix elements @xmath14 of quark and gluon light - cone operators one can use two basic types of nonperturbative functions . the double distributions ( dds ) @xmath0 @xcite specify the sudakov light - cone `` plus '' fractions @xmath15 and @xmath16 of the initial hadron momentum @xmath12 and the momentum transfer @xmath17 carried by the initial parton . the other possibility is to treat the proportionality coefficient @xmath9 as an independent parameter and introduce an alternative description in terms of the nonforward parton distributions ( nfpds ) @xmath1 with @xmath18 being the total fraction of the initial hadron momentum taken by the initial parton . the shape of nfpds explicitly depends on the parameter @xmath9 characterizing the _ skewedness _ of the relevant nonforward matrix element . this parametrization of nonforward matrix elements by @xmath1 is similar to that proposed by x. ji @xcite who introduced off - forward parton distributions ( ofpds ) @xmath19 in which the parton momenta and the skewedness parameter @xmath20 are measured in units of the average hadron momentum @xmath21 . ofpds and nfpds @xcite can be treated as particular forms of _ skewed _ parton distributions ( spds ) . one can also introduce the version of dds ( `` @xmath22-dds '' @xcite ) in which the active parton momentum is written in terms of symmetric variables @xmath23 . the basics of the pqcd approaches incorporating skewed parton distributions were formulated in refs.@xcite . a detailed analysis of pqcd factorization for hard meson electroproduction processes was given in ref . our goal in the present lectures is to give a description of the approach outlined in our papers @xcite . in the pqcd factorization treatment of hard electroproduction processes , the nonperturbative information is accumulated in the nonforward matrix elements @xmath24 of light cone operators @xmath25 . for @xmath26 the matrix elements depend on the relative coordinate @xmath27 through two lorentz invariant variables @xmath28 and @xmath29 . in the forward case , when @xmath30 , one obtains the usual quark helicity - averaged densities by fourier transforming the relevant matrix element with respect to @xmath28 @xmath31 where @xmath32 is the gauge link , @xmath33 are the dirac spinors and we use the notation @xmath34 . the parameter @xmath35 in this representation has an evident interpretation : it characterizes which the fraction of the initial hadron momentum is carried by the active parton . in the nonforward case , we can use the double fourier representation with respect to both @xmath28 and @xmath29 : @xmath36 where @xmath37 is the nucleon mass and @xmath38 specify the nucleon polarization . we use the `` hat '' ( rather than `` slash '' ) convention @xmath39 . the parametrization of nonforward matrix elements must include both the nonflip term described here by the functions @xmath40 and the spin - flip term characterized by the functions @xmath41 . the parameters @xmath42 tell us that the active parton carries the fractions @xmath35 of the initial momentum @xmath12 and the fraction @xmath43 of the momentum transfer @xmath17 . using the approach @xcite based on the @xmath22-representation analysis it is possible to prove @xcite that double distributions @xmath44 have a natural property that both @xmath35 and @xmath43 satisfy the `` parton '' constraints @xmath45 , @xmath46 for any feynman diagram contributing to @xmath44 . a less obvious restriction @xmath47 guarantees that the argument @xmath18 of the skewed distribution @xmath48 also changes between the limits the support area for the _ double distribution _ @xmath49 is shown on fig.[fg : support]a . in principle , we can not exclude also the possibility that the functions @xmath50 have singular terms at @xmath51 proportional to @xmath52 or its derivative(s ) . such terms have no projection onto the usual parton densities . we will denote them by @xmath53 @xmath54 they may be interpreted as coming from the @xmath55-channel meson - exchange type contributions ( see fig.[fg : exchange]b ) . in this case , the partons just share the plus component of the momentum transfer @xmath17 : information about the magnitude of the initial hadron momentum is lost if the exchanged particle can be described by a pole propagator @xmath56 . hence , the meson - exchange contributions to a double distribution may look like @xmath57 where @xmath58 are the functions related to the distribution amplitudes of the relevant mesons @xmath59 . the two examples above correspond to @xmath35-even and @xmath35-odd parts of the double distribution @xmath60 . another type of terms in which the dependence on @xmath28 is lost can be produced by diagrams containing a quartic pion vertex ( fig.[fg : exchange]c ) . as shown by polyakov and weiss @xcite , such terms correspond to an independent @xmath61 type contribution which can be parametrized by a single integral over @xmath43 involving an effective distribution amplitude @xmath62 . the meson - exchange terms in @xmath63 and @xmath64 as well as polyakov - weiss terms are invisible in the forward limit , hence the existing knowledge of the usual parton densities can not be used to constrain these terms . later , describing the models for skewed distributions , we discuss only the `` forward visible parts '' of spds which are obtained by scanning the @xmath65 parts of the relevant dds . comparing eq . ( [ 33 ] ) with the @xmath66 limit of the dd definition ( [ 31 ] ) gives the `` reduction formulas '' relating the double distribution @xmath67 to the quark and antiquark parton densities @xmath68 hence , the positive-@xmath35 and negative-@xmath35 components of the double distribution @xmath69 can be treated as nonforward generalizations of quark and antiquark densities , respectively . if we define the `` untilded '' dds by @xmath70 then @xmath35 is always positive and the reduction formulas have the same form @xmath71 in both cases . the new antiquark distributions also `` live '' on the triangle @xmath72 . taking @xmath27 in the lightcone `` minus '' direction , we arrive at the parton interpretation of functions @xmath73 as probability amplitudes for an outgoing parton to carry the fractions @xmath15 and @xmath16 of the external momenta @xmath17 and @xmath12 . the double distributions @xmath0 are universal functions describing the flux of @xmath74 and @xmath75 independently of the ratio @xmath76 . note , that extraction of two separate components @xmath77 and @xmath78 from the quark dd @xmath79 as its positive-@xmath35 and negative-@xmath35 parts is unambiguous . taking the @xmath80 term of the taylor expansion gives the sum rules @xmath81 \ , dy = f_1^a ( t ) \ , , \label{2}\ ] ] @xmath82 \ , dy = f_2^a ( t ) \ , , \label{222}\ ] ] relating the double distributions @xmath40 , @xmath41 to the @xmath83-flavor components of the dirac and pauli form factors : @xmath84 respectively . a common element of the reduction formulas given above is an integration over @xmath43 . hence , it is convenient to introduce intermediate functions @xmath85 they satisfy the reduction formulas @xmath86 \ , dx = f_1(t ) \ , \label{3b}\ ] ] @xmath87 \ , dx = f_2(t ) \ , , \label{3c}\ ] ] which show that these functions are the simplest hybrids of the usual parton densities and form factors . for this reason , one can call them _ nonforward parton densities _ ( nds ) @xcite . the spin - flip terms disappear only if @xmath30 . in the weaker @xmath88 limit , they survive , @xmath89 @xmath90 is the @xmath83-flavor contribution to the nucleon anomalous magnetic moment . in other words , the @xmath91 limit of the `` magnetic '' nds exists : @xmath92 , and the integral @xmath93 \ , dx = \kappa_p \ , \label{4}\ ] ] gives the anomalous magnetic moment of the proton . the knowledge of the @xmath35-moment of @xmath94 s is needed to determine the contribution of the quark orbital angular momentum to the proton spin @xcite . since the @xmath95-type dds are always accompanied by the @xmath96 factor , they are invisible in deep inelastic scattering and other inclusive processes related to strictly forward @xmath30 matrix elements . there are also parton - helicity sensitive double distributions @xmath97 and @xmath98 . the first one reduces to the usual spin - dependent densities @xmath99 in the @xmath30 limit and gives the axial form factor @xmath100 after the @xmath42-integration . the second one involves hadron helicity flip and is related to the pseudoscalar form factor @xmath101 . it is worth mentioning here that for a massive target ( nucleons in our case ) there is a kinematic restriction @xcite @xmath102 hence , for fixed @xmath9 , the formal limit @xmath103 is not physically reachable . however , many results ( evolution equations being the most important example ) obtained in the formal @xmath104 , @xmath105 limit are still applicable . to make the description more symmetric with respect to the initial and final hadron momenta , we can treat nonforward matrix elements as functions of @xmath106 and @xmath29 , where @xmath21 is the average hadron momentum . the relevant double distributions @xmath107 [ which we will call @xmath22-dds to distinguish them from @xmath43-dds @xmath0 ] are defined by @xmath109 the support area for @xmath110 is shown in fig.[fg : support]b . again , the usual forward densities @xmath111 and @xmath112 are given by integrating @xmath113 over vertical lines @xmath114 for @xmath115 and @xmath116 , respectively . hence , we can split @xmath117 into three components @xmath118 where @xmath119 is a singular term with support at @xmath51 only . due to hermiticity and time - reversal invariance properties of nonforward matrix elements , the @xmath22-dds are even functions of @xmath22 : @xmath120 for our original @xmath43-dds @xmath121 , this corresponds to symmetry with respect to the interchange @xmath122 ( `` munich '' symmetry , established in ref . @xcite ) . in particular , the functions @xmath123 for singular contributions @xmath124 are symmetric @xmath125 both for @xmath35-even and @xmath35-odd parts . the @xmath83-quark contribution into the flavor - singlet operator @xmath126\ ] ] can be parametrized either by @xmath43-dds @xmath127 or by @xmath22-dds @xmath128 @xmath129 in the second and third lines here we have used the fact that positive-@xmath35 and negative-@xmath35 parts in this case are described by the same untilded function @xmath130 the @xmath22-dds @xmath128 are even functions of @xmath22 and , according to eq . ( [ 311q ] ) , odd functions of @xmath35 : @xmath131 finally , the valence quark functions @xmath132 related to the operators @xmath133\ ] ] are even functions of both @xmath22 and @xmath35 : @xmath134 the reduction formulas and interpretation of the @xmath35-variable as the fraction of @xmath12 ( or @xmath135 ) momentum suggest that the profile of @xmath44 ( or @xmath136 ) in @xmath35-direction is basically determined by the shape of @xmath137 . on the other hand , the profile in @xmath43 ( or @xmath22 ) direction characterizes the spread of the parton momentum induced by the momentum transfer @xmath17 . in particular , since the @xmath22-dds @xmath138 are even functions of @xmath22 , it make sense to write @xmath139 where @xmath140 is an even function of @xmath22 normalized by @xmath141 we may expect that the @xmath22-profile of @xmath140 is similar to that of a symmetric distribution amplitude ( da ) @xmath142 . since @xmath143 , to get a more complete analogy with da s , it makes sense to rescale @xmath22 as @xmath144 introducing the variable @xmath145 with @xmath35-independent limits : @xmath146 . the simplest model is to assume that the profile in the @xmath145-direction is a universal function @xmath147 for all @xmath35 . possible simple choices for @xmath147 may be @xmath148 ( no spread in @xmath145-direction ) , @xmath149 ( characteristic shape for asymptotic limit of nonsinglet quark distribution amplitudes ) , @xmath150 ( asymptotic shape of gluon distribution amplitudes ) , etc . in the variables @xmath151 , this gives @xmath152 ^ 2}{(1-|x|)^5 } \ , \ . \label{mod123 } \end{aligned}\ ] ] these models can be treated as specific cases of the general profile function @xmath153^b}{(1-|x|)^{2b+1 } } \ , , \label{modn}\ ] ] whose width is governed by the parameter @xmath154 . the coefficient of proportionality @xmath155 ( or @xmath156 ) between the plus components of the momentum transfer and initial ( or average ) momentum specifies the _ skewedness _ of the matrix elements . the characteristic feature implied by representations for double distributions [ see , e.g. , eq.([31 ] ) ] is the absence of the @xmath9-dependence in the dds @xmath44 and @xmath157-dependence in @xmath136 . an alternative way to parametrize nonforward matrix elements of light - cone operators is to use @xmath9 ( or @xmath157 ) and the _ total _ momentum fractions @xmath158 ( or @xmath159 ) as independent variables . integrating each particular double distribution over @xmath43 gives the nonforward parton distributions @xmath160 where @xmath161 . the two components of nfpds correspond to positive ( @xmath162 ) and negative ( @xmath163 ) values of the fraction @xmath164 associated with the `` returning '' parton . as explained in refs . @xcite , the second component can be interpreted as the probability amplitude for the initial hadron with momentum @xmath12 to split into the final hadron with momentum @xmath165 and the two - parton state with total momentum @xmath166 shared by the partons in fractions @xmath167 and @xmath168 , where @xmath169 . the relation between `` untilded '' nfpds and dds can be illustrated on the `` dd - life triangle '' defined by @xmath170 ( see fig.[fg : spds]a ) . specifically , to get @xmath48 , one should integrate @xmath44 over @xmath43 along a straight line @xmath171 . fixing some value of @xmath9 , one deals with a set of parallel lines intersecting the @xmath35-axis at @xmath172 . the upper limit of the @xmath43-integration is determined by intersection of this line either with the line @xmath173 ( this happens if @xmath174 ) or with the @xmath43-axis ( if @xmath175 ) . the line corresponding to @xmath176 separates the triangle into two parts generating the two components of the nonforward parton distribution . in a similar way , we can write the relation between ofpds @xmath19 and the @xmath22-dds @xmath177 @xmath178 the delta - function in eq.([710 ] ) specifies the line of integration in the @xmath179 plane . for definiteness , we will assume below that @xmath157 is positive . information contained in spds originates from two physically different sources : meson - exchange type contributions @xmath180 coming from the singular @xmath51 parts of dds and the functions @xmath181 , @xmath182 obtained by scanning the @xmath65 parts of dds @xmath183 , @xmath184 . the support of exchange contributions is restricted to @xmath185 . up to rescaling , the function @xmath180 has the same shape for all @xmath9 . for any nonvanishing @xmath186 , these exchange terms become invisible in the forward limit @xmath187 . on the other hand , the support of functions @xmath181 , @xmath182 in general covers the whole @xmath188 region . furthermore , the forward limit of such spds as @xmath189 is rather well known from inclusive measurements . hence , information contained in the usual ( forward ) densities @xmath190 , @xmath191 can be used to restrict the models for @xmath181 , @xmath182 . let us consider spds constructed using simple models of dds specified above . in particular , the model @xmath192 ( equivalent to @xmath193 ) , gives the simplest model @xmath194 in which ofpds at @xmath91 have no @xmath157-dependence . for nfpds this gives @xmath195 i.e. , nfpds for non - zero @xmath9 are obtained from the forward distribution @xmath196 by shift and rescaling . in case of the @xmath197 and @xmath198 models , simple analytic results can be obtained only for some explicit forms of @xmath137 . for the `` valence quark''-oriented ansatz @xmath199 , the following choice of a normalized distribution @xmath200 is both close to phenomenological quark distributions and produces a simple expression for the double distribution since the denominator @xmath201 factor in eq . ( [ mod123 ] ) is canceled . as a result , the integral in eq . ( [ 710 ] ) is easily performed and we get @xmath202 \ , \theta ( \tilde x ) + ( \tilde x \to -\tilde x ) \biggr \ } \label{outs } \end{aligned}\ ] ] for @xmath203 and @xmath204 \nonumber \\ + ( \tilde x \to -\tilde x ) \biggr \ } \label{middles}\end{aligned}\ ] ] in the middle @xmath205 region . we use here the notation @xmath206 and @xmath207 @xcite . to extend these expressions onto negative values of @xmath157 , one should substitute @xmath157 by @xmath208 . one can check , however , that no odd powers of @xmath208 would appear in the @xmath209 moments of @xmath210 . furthermore , these expressions are explicitly non - analytic for @xmath211 . this is true even if @xmath83 is integer . discontinuity at @xmath211 , however , appears only in the second derivative of @xmath210 , i.e. , the model curves for @xmath210 look very smooth ( see fig.[q - at - diff - z ] ) . the explicit expressions for nfpds in this model were given in ref.@xcite . the relevant curves are also shown in fig.[q - at - diff - z ] . for @xmath212 , the @xmath213 part of ofpd has the same @xmath35-dependence as its forward limit , differing from it by an overall @xmath157-dependent factor only : @xmath214 the @xmath215 behaviour can be trivially continued into the @xmath216 region . however , the actual behaviour of @xmath217 in this region is given by a different function . in other words , @xmath218 can be represented as a sum of a function analytic at border points and a contribution whose support is restricted by @xmath219 . it should be emphasized that despite its da - like appearance , this contribution should not be treated as an exchange - type term . it is generated by regular @xmath65 part of dd , and , unlike @xmath220 functions changes its shape with @xmath157 and becomes very small for small @xmath157 . for the singlet quark distribution , the @xmath22-dds @xmath221 should be odd functions of @xmath35 . still , we can use the model like ( [ 74 ] ) for the @xmath115 part , but take @xmath222 . note , that the integral ( [ 710 ] ) producing @xmath223 in the @xmath219 region would diverge for @xmath224 if @xmath225 , which is the usual case for standard parametrizations of singlet quark distributions for sufficiently large @xmath226 . however , due to the antisymmetry of @xmath221 wrt @xmath227 and its symmetry wrt @xmath228 , the singularity at @xmath229 can be integrated using the principal value prescription which in this case produces the @xmath230 antisymmetric version of eqs.([outs ] ) and ( [ middles ] ) . for @xmath212 , its middle part reduces to @xmath231 the shape of singlet spds in this model is shown in fig . [ qs - at - diff - z ] in the lowest order , the dvcs amplitude @xmath232 is given by two handbag diagrams . in particular , the invariant amplitude containing the @xmath233 functions is given by @xmath234 { \cal f}^{a+ \bar a}_{\zeta}(x;t ) \ , dx \ , . \label{123 } \end{aligned}\ ] ] an important feature of the dvcs amplitude is that for large @xmath226 and fixed @xmath55 it depends only on the ratio @xmath235 : dvcs is an _ exclusive _ process exhibiting the bjorken scaling . note that the imaginary part of the dvcs amplitude is proportional to @xmath236 . in this function , the parameter @xmath9 appears twice : first as the skewedness of the process and then as the fraction @xmath237 at which the imaginary part is generated . one may ask which @xmath226 are large enough to ensure the dominance of the lowest - twist handbag contribution . in dis , approximate bjorken scaling starts at @xmath238gev@xmath239 . another example is given by the exclusive process @xmath240 studied on @xmath241 colliders . if one of the photons is highly virtual @xmath242 while another is ( almost ) real @xmath243 , the process is kinematically similar to dvcs . in the leading order , the @xmath244 transition form factor is given by a handbag diagram again . the recent measurements by cleo @xcite show that the pqcd prediction @xmath245 again works starting from @xmath238gev@xmath239 . the @xmath246 vertex ( for a virtual pion ) can be also measured on a fixed - target machine like cebaf in which case it is just a part of the dvcs amplitude corresponding to the 4th skewed distribution @xmath247 ( which is related to the pseudoscalar form factor @xmath248 of the nucleon ) . hence , cleo data give an evidence that dvcs may be handbag - dominated for @xmath226 as low as 2gev@xmath239 . the main problem for studying dvcs is the contamination by the bethe - heitler process in which the final photon is emitted from the initial or final electron . the bethe - heitler amplitude is enhanced at small @xmath55 . on the other hand , the virtual photon flux for fixed @xmath226 and @xmath249 increases when the electron beam energy increases . hence , the energy upgrade would make the dvcs studies at jefferson lab more feasible . experimental aspects of virtual compton scattering studies at jefferson lab are discussed in refs . @xcite . the skewed parton distributions can be also measured in hard meson electroproduction processes @xcite . the leading - twist pqcd contribution in this case involves a one - gluon exchange , which means that the hard subprocess is suppressed by @xmath250 factor . the competing soft mechanism corresponds to a triple overlap of hadronic wave functions and has a relative suppression @xmath251 by a power of @xmath226 , with @xmath252 gev@xmath239 being a characteristic hadronic scale . hence , to clearly see the one - gluon - exchange signal one needs @xmath226 above 10gev@xmath239 . numerical pqcd - based estimates and comparison of dvcs and hard meson electroprodution cross sections can be found in ref . the imaginary part of hard exclusive meson electroproduction amplitude is determined by the skewed distributions at the border point . for this reason , the magnitude of @xmath253 [ or @xmath254 and its relation to the forward densities @xmath137 has a practical interest . this example also gives a possibility to study the sensititivity of the results to the choice of the profile function . assuming the infinitely narrow weight @xmath255 , we have @xmath256 and @xmath257 . hence , both @xmath253 and @xmath258 are given by @xmath259 since @xmath260 and @xmath261 . since the argument of @xmath137 is twice smaller than in deep inelastic scattering , this results in an enhancement factor . in particular , if @xmath262 for small @xmath35 , the ratio @xmath263 is @xmath264 . the use of a wider profile function @xmath265 produces further enhancement . for example , taking the normalized profile function @xmath266 and @xmath262 we get @xmath267 which is larger than @xmath264 for any finite @xmath154 and @xmath268 . the @xmath264 enhancement appears as the @xmath269 limit of eq.([rhon ] ) . for small integer @xmath154 , eq.([rhon ] ) reduces to simple formulas obtained in refs . @xcite . for @xmath197 , we have @xmath270 which gives the factor of 3 for the enhancement if @xmath271 . for @xmath198 , the ratio ( [ rhon ] ) becomes @xmath272 producing a smaller enhancement factor @xmath273 for @xmath271 . calculating the enhancement factors , one should remember that the gluon spd @xmath274 reduces to @xmath275 in the @xmath276 limit . hence , to get the enhancement factor corresponding to the @xmath277 small-@xmath35 behavior of the forward gluon density , one should take @xmath278 in eq.([rhon ] ) , i.e. , despite the fact that the @xmath279 behavior of the singlet quark distribution gives the factor of 3 for the @xmath280 ratio , the same shape of the gluon distribution results in no enhancement . due to evolution , the effective parameter @xmath83 characterizing the small-@xmath35 behavior of the forward distribution is an increasing function of @xmath226 . as a result , for fixed @xmath154 , the @xmath281 ratio increases with @xmath226 . in general , the profile of @xmath282 in the @xmath22-direction is also affected by the pqcd evolution . in particular , in ref . @xcite it was shown that if one takes an ansatz corresponding to an extremely asymmetric profile function @xmath283 , the shift of the profile function to a more symmetric shape is clearly visible in the evolution of the relevant spd . recently , it was demonstrated @xcite that evolution to sufficiently large @xmath226 enforces a direct relation @xmath284 between the parameter @xmath83 characterizing the small-@xmath35 behavior of dds and the parameter @xmath154 governing the shape of their @xmath22 profile . this gives @xmath285 for the @xmath286 ratio . for @xmath271 , e.g. , the spd enhancement factor in this case equals 3 . the compton scattering in its various versions provides a unique tool for studying hadronic structure . the compton amplitude probes the hadrons through a coupling of two electromagnetic currents and in this aspect it can be considered as a generalization of hadronic form factors . in qcd , the photons interact with the quarks of a hadron through a vertex which , in the lowest approximation , has a pointlike structure . however , in the soft regime , strong interactions produce large corrections uncalculable within the perturbative qcd framework . to take advantage of the basic pointlike structure of the photon - quark coupling and the asymptotic freedom feature of qcd , one should choose a specific kinematics in which the behavior of the relevant amplitude is dominated by short ( or , being more precise , lightlike ) distances . the general feature of all such types of kinematics is the presence of a large momentum transfer . for compton amplitudes , there are several situations when large momentum transfer induces dominance of configurations involving lightlike distances : + @xmath287 both photons are far off - shell and have equal spacelike virtuality : virtual forward compton amplitude , its imaginary part determines structure functions of deep inelastic scattering ( dis ) ; + @xmath288 initial photon is highly virtual , the final one is real and the momentum transfer to the hadron is small : deeply virtual compton scattering ( dvcs ) amplitude ; + @xmath289 both photons are real but the momentum transfer is large : wide - angle compton scattering ( wacs ) amplitude . our main statement made in ref . @xcite is that , at accessible momentum transfers @xmath290 gev@xmath239 , the wacs amplitude is dominated by handbag diagrams , just like in dis and dvcs . in the most general case , the nonperturbative part of the handbag contribution is described by double distributions ( dds ) @xmath291 , etc . , which can be related to the usual parton densities @xmath137 , @xmath292 and nucleon form factors like @xmath293 . among the arguments of dds , @xmath35 is the fraction of the initial hadron momentum carried by the active parton and @xmath43 is the fraction of the momentum transfer @xmath17 . the description of wacs amplitude simplifies when one can neglect the @xmath43-dependence of the hard part and integrate out the @xmath43-dependence of the double distributions . in that case , the long - distance dynamics is described by nonforward parton densities ( nds ) @xmath294 etc . the latter can be interpreted as the usual parton densities @xmath137 supplemented by a form factor type @xmath55-dependence . we proposed in @xcite a simple model for the relevant nds which both satisfies the relation between @xmath295 and usual parton densities @xmath137 and produces a good description of the @xmath296 form factor up to @xmath297 gev@xmath239 . we have used this model to calculate the wacs amplitude and obtained the results which are rather close to existing data . the forward virtual compton amplitude whose imaginary part gives structure functions of deep inelastic scattering ( see , e.g. , @xcite ) is the classic example of a light cone dominated compton amplitude . in this case , the `` final '' photon has momentum @xmath298 coinciding with that of the initial one . the momenta @xmath299 of the initial and final hadrons also coincide . the total cm energy of the photon - hadron system @xmath300 should be above resonance region , and the bjorken ratio @xmath10 is finite . the light cone dominance is secured by high virtuality of the photons : @xmath301 gev@xmath239 . in the large-@xmath226 limit , the leading contribution in the lowest @xmath302 order is given by handbag diagrams in which the perturbatively calculable hard quark propagator is convoluted with parton distribution functions @xmath111 ( @xmath303 ) which describe / parametrize nonperturbative information about hadronic structure . the condition that both photons are highly virtual may be relaxed by taking a real photon in the final state . keeping the momentum transfer @xmath304 to the hadron as small as possible , one arrives at kinematics of the deeply virtual compton scattering ( dvcs ) the importance of which was recently emphasized by x. ji @xcite ( see also @xcite ) . having large virtuality @xmath226 of the initial photon is sufficient to guarantee that in the bjorken limit the leading power contributions in @xmath305 are generated by the strongest light cone singularities @xcite , with the handbag diagrams being the starting point of the @xmath302 expansion . the most important contribution to the dvcs amplitude is given by a convolution of a hard quark propagator and a nonperturbative function describing long - distance dynamics , which in the most general case is given by double distributions ( dds ) @xmath306 @xcite . in the dvcs kinematics , @xmath307 is assumed to be small compared to @xmath226 , and for this reason the @xmath55- and @xmath308-dependence of the short - distance amplitude in refs . @xcite was neglected and @xmath308 analogous to the nachtmann - georgi - politzer @xmath309 target mass corrections in dis @xcite . ] . this is equivalent to approximating the active parton momentum @xmath310 by its plus component alone : @xmath311 . our final goal in the present paper is to get an estimate of the handbag contributions for the large-@xmath55 real compton scattering . since the initial photon in that case is also real : @xmath312 ( and hence @xmath313 ) , it is natural to expect that the nonperturbative functions which appear in wacs correspond to the @xmath314 limit of the skewed parton distributions - dependence of the hard part . ] it is easy to see from eq.([3a ] ) that in this limit the spds reduce to the nonforward parton densities @xmath316 introduced above : @xmath317 note that nds depend on `` only two '' variables @xmath35 and @xmath55 , with this dependence constrained by reduction formulas ( [ 3b]),([3c ] ) . furthermore , it is possible to give an interpretation of nonforward densities in terms of the light - cone wave functions . consider for simplicity a two - body bound state whose lowest fock component is described by a light cone wave function @xmath318 . choosing a frame where the momentum transfer @xmath17 is purely transverse @xmath319 , we can write the two - body contribution into the form factor as @xcite @xmath320 where @xmath321 . comparing this expression with the reduction formula ( [ 3b ] ) , we conclude that @xmath322 is the two - body contribution into the nonforward parton density @xmath323 . assuming a gaussian dependence on the transverse momentum @xmath324 ( cf . @xcite ) @xmath325 we get @xmath326 where @xmath327 is the two - body part of the relevant parton density . within the light - cone approach , to get the total result for either usual @xmath137 or nonforward parton densities @xmath328 , one should add the contributions due to higher fock components . by no means these contributions are small , e.g. , the valence @xmath329 contribution into the normalization of the @xmath330 form factor at @xmath91 is less than 25% @xcite . in the absence of a formalism providing explicit expressions for an infinite tower of light - cone wave functions we choose to treat eq.([8 ] ) as a guide for fixing interplay between the @xmath55 and @xmath35 dependences of nds and propose to model them by @xmath331 the functions @xmath111 here are the usual parton densities assumed to be taken from existing parametrizations like grv , mrs , cteq , etc . in the @xmath91 limit ( recall that @xmath55 is negative ) this model , by construction , satisfies the first of reduction formulas ( [ 3b ] ) . within the gaussian ansatz ( [ 13 ] ) , the basic scale @xmath332 specifies the average transverse momentum carried by the quarks . in particular , for valence quarks @xmath333 where @xmath334 are the numbers of the valence @xmath83-quarks in the proton . to fix the magnitude of @xmath332 , we use the second reduction formula in ( [ 3b ] ) relating @xmath335 s to the @xmath296 form factor . to this end , we take the following simple expressions for the valence distributions @xmath336 @xmath337 they closely reproduce the relevant curves given by the grv parametrization @xcite at a low normalization point @xmath338 gev@xmath239 . the best agreement between our model @xmath339 e^{\bar x t / 4x \lambda^2 } dx \label{14}\ ] ] and experimental data @xcite in the moderately large @xmath55 region 1 gev@xmath239 @xmath340 gev@xmath239 is reached for @xmath341 gev@xmath239 ( see fig.[fig : ff ] ) . this value gives a reasonable magnitude @xmath342 for the average transverse momentum of the valence @xmath343 and @xmath344 quarks in the proton . similarly , building a model for the parton helicity sensitive nds @xmath345 one can take their @xmath91 shape from existing parametrizations for spin - dependent parton distributions @xmath99 and then fix the relevant @xmath332 parameter by fitting the @xmath346 form factor . the case of hadron spin - flip distributions @xmath347 and @xmath348 is more complicated since the distributions @xmath94 , @xmath349 are unknown . at @xmath91 , our model by construction gives a correct normalization @xmath350 for the form factor . however , if one would try to find the derivative @xmath351 at @xmath91 by expanding the exponential @xmath352 $ ] into the taylor series under the integral ( [ 14 ] ) , one would get a divergent expression . an analogous problem is well known in applications of qcd sum rules to form factors at small @xmath55 @xcite . the divergence is related to the long - distance propagation of massless quarks in the @xmath55-channel . formally , this is revealed by singularities starting at @xmath91 . however , @xmath353 should not have singularities for timelike @xmath55 up to @xmath354 , with the @xmath7-meson peak at @xmath355 being the most prominent feature of the @xmath55-channel spectrum . technically , the singularities of the original expression are singled out into the bilocal correlators @xcite which are substituted by their realistic version with correct spectral properties ( usually the simplest model with @xmath7 and @xmath356 terms is used ) . an important point is that such a modification is needed only when one calculates form factors in the small-@xmath55 region : for @xmath357 , the correction terms should vanish faster than any power of @xmath358 @xcite . in our case , the maximum deviation of the curve for @xmath353 given by eq.([14 ] ) from the experimental data in the small-@xmath55 region @xmath359 is 15@xmath360 . hence , if one is willing to tolerate such an inaccuracy , one can use our model starting with @xmath91 . our curve is within 5% from the data points @xcite for @xmath361 gev@xmath239 and does not deviate from them by more than 10% up to 9 gev@xmath239 . modeling the @xmath55-dependence by a more complicated formula ( e.g. , assuming a slower decrease at large @xmath55 , and/or choosing different @xmath332 s for @xmath343 and @xmath344 quarks and/or splitting nds into several components with different @xmath332 s , etc . , see ref.@xcite for an example of such an attempt ) or changing the shape of parton densities @xmath111 one can improve the quality of the fit and extend agreement with the data to higher @xmath55 . such a fine - tuning is not our goal here . we just want to emphasize that a reasonable description of the @xmath296 data in a wide region was obtained by fixing just a single parameter @xmath332 reflecting the proton size . moreover , we could fix @xmath332 from the requirement that @xmath362 and present our curve for @xmath296 as a successful prediction of the model . we interpret this success as an evidence that the model correctly catches the gross features of the underlying physics . since our model implies a gaussian dependence on the transverse momentum , it includes only what is usually referred to as an overlap of soft wave functions . it completely neglects effects due to hard pqcd gluon exchanges generating the power - law @xmath363 tail of the nonforward densities at large @xmath55 . it is worth pointing out here that though we take nonforward densities @xmath335 with an exponential dependence on @xmath55 , the @xmath296 form factor in our model has a power - law asymptotics @xmath364 dictated by the @xmath365 behavior of the parton densities for @xmath35 close to 1 . this connection arises because the integral ( [ 14 ] ) over @xmath35 is dominated at large @xmath55 by the region @xmath366 . in other words , the large-@xmath55 behavior of @xmath367 in our model is governed by the feynman mechanism @xcite . one should realize , however , that the relevant scale @xmath368 gev@xmath239 is rather large . for this reason , when @xmath290 gev@xmath239 , it is premature to rely on asymptotic estimates for the soft contribution . indeed , with @xmath369 , the asymptotic estimate is @xmath370 , in an apparent contradiction with the ability of our curve to follow the dipole behavior . the resolution of this paradox is very simple : the maxima of nonforward densities @xmath335 for @xmath290 gev@xmath239 are at rather low @xmath35-values @xmath371 . hence , the @xmath35-integrals producing @xmath372 are not dominated by the @xmath373 region yet and the asymptotic estimates are not applicable : the functional dependence of @xmath374 in our model is much more complicated than a simple power of @xmath358 . the fact that our model closely reproduces the experimentally observed dipole - like behavior of the proton form factor is a clear demonstration that such a behavior may have nothing to do with the quark counting rules @xmath375 @xcite valid for the asymptotic behavior of the hard gluon exchange contributions . our explanation of the observed magnitude and the @xmath55-dependence of @xmath376 by a purely soft contribution is in strong contrast with that of the hard pqcd approach to this problem . with both photons real , it is not sufficient to have large photon energy to ensure short - distance dominance : large-@xmath377 , small-@xmath55 region is strongly affected by regge contributions . hence , having large @xmath378gev@xmath239 is a necessary condition for revealing short - distance dynamics . the simplest contributions for the wacs amplitude are given by the @xmath377- and @xmath343-channel handbag diagrams fig.[fig : cohan]b , c . the nonperturbative part in this case is given by the proton dds which determine the @xmath55-dependence of the total contribution . just like in the form factor case , the contribution dominating in the formal asymptotic limit @xmath379 , is given by diagrams corresponding to the pure sd regime , see fig.[fig : cosubno]a . the hard subraph then involves two hard gluon exchanges which results in a suppression factor @xmath380 absent in the handbag term . the total contribution of all two - gluon exchnange contributions was calculated by farrar and zhang @xcite and then recalculated by kronfeld and nii @xcite . a sufficiently large contribution is only obtained if one uses humpy das and @xmath381 propagators with no finite - size effects included . even with such propagators , the wacs amplitude calculated with the asymptotic da is negligibly small @xcite compared to existing data . as argued in ref.@xcite , the enhancements generated by the humpy das should not be taken at face value both for form factors and wide - angle compton scattering amplitudes . for these reasons , we ignore hard contributions to the wacs amplitude as negligibly small . another type of configurations containing hard gluon exchange is shown in fig.[fig : cosubno]b . there are also the diagrams with photons coupled to different quarks ( `` cat s ears '' , fig.[fig : cosubno]c ) . such contributions have both higher order and higher twist . this brings in the @xmath382 factor and an extra @xmath383 suppression . the latter is partially compensated by a slower fall - off of the four - quark dds with @xmath55 since only one valence quark should change its momentum . for simplicity , we neglect all the suppressed terms and deal only with the handbag contributions fig.[fig : cohan]b , c in which the highly virtual quark propagator connecting the photon vertices is convoluted with proton dds parametrizing the overlap of soft wave functions . since the basic scale @xmath384 characterizing the @xmath55-dependence of dds in our model is 2.8 gev@xmath239 , while existing data are all at momentum transfers @xmath55 below 5 gev@xmath239 , we deal with the region where the asymptotic estimate ( feynman mechanism ) for the overlap contribution is not working yet . in the coordinate representation , the sum of two handbag contributions to the compton amplitude can be written as @xmath385 where @xmath386 and @xmath387 is the hard quark propagator ( throughout , we use the `` hat '' notation @xmath388 ) . the summation over the twist-0 longitudinal gluons adds the usual gauge link between the @xmath389,@xmath390 fields which we do not write down explicitly ( gauge link disappears , e.g. , in the fock - schwinger gauge @xmath391 ) . because of the symmetry of the problem , it is convenient to use @xmath392 ( cf . @xcite ) and @xmath393 as the basic momenta . applying the fiertz transformation and introducing the double distributions by @xmath394 \ , d\tilde y + \frac1{4m_p } \ , \bar u(p ' ) ( \hat z \hat r - \hat r \hat z ) u(p ) \label{17a } \\ & & \times \int_0 ^ 1 \ , dx \int_{-\bar x}^{\bar x } \ , \left [ e^{-i(kz ) } { k}^a(x,\alpha ; t ) - e^{i(kz ) } { k}^{\bar a } ( x,\alpha ; t ) \right ] \ , d \alpha + o(z^2 ) \ { \rm terms}\nonumber \end{aligned}\ ] ] ( we use here the shorthand notation @xmath395 ) and similarly for the parton helicity sensitive operators @xmath396 \ , d\alpha + \ , \frac{(rz)}{m_p } \ , \bar u(p ' ) \gamma_5 u(p ) \label{18}\\ & & \times \int_0 ^ 1 \ , dx \int_{-\bar x}^{\bar x } \left [ e^{-i(kz ) } { p}^a(x,\tilde y;t ) + e^{i(kz ) } { p}^{\bar a}(x,\alpha ; t ) \right ] \ , d\alpha + o(z^2 ) \ { \rm terms } \ , , \nonumber \end{aligned}\ ] ] we arrive at a leading - twist qcd parton picture with @xmath22-dds serving as functions describing long - distance dynamics . the @xmath22-dds @xmath397 , etc . , are related to the original @xmath43-dds @xmath398 by the shift @xmath399 . integrating @xmath400 over @xmath22 one obtains the same nonforward densities @xmath401 . the hard quark propagators for the @xmath377 and @xmath343 channel handbag diagrams in this picture look like @xmath402 and @xmath403 respectively . we denote @xmath404 and @xmath405 . since @xmath22-dds are even functions of @xmath22 @xcite , the @xmath406 terms in the numerators can be dropped . it is legitimate to keep @xmath407 and @xmath408 terms in the denominators : the dependence of hard propagators on target parameters @xmath308 and @xmath55 can be calculated exactly because of the effect analogous to the @xmath157-scaling in dis @xcite ( see also @xcite ) . note that the @xmath55-correction to hard propagators disappears in the large-@xmath55 limit dominated by the @xmath373 integration . the @xmath55-corrections are the largest for @xmath409 . at this value and for @xmath410 and @xmath411 ( cm angle of 90@xmath412 ) , the @xmath55-term in the denominator of the most important second propagator is only 1/8 of the @xmath343 term . this ratio increases to 1/3 for @xmath413 . however , at nonzero @xmath22-values , the @xmath55-corrections are smaller . hence , the @xmath55-corrections in the denominators of hard propagators can produce @xmath414 effects and should be included in a complete analysis . here , we consider an approximation in which these terms are neglected and hard propagators are given by @xmath415-independent expressions @xmath416 and @xmath417 . as a result , the @xmath22-integration acts only on the dds @xmath418 and converts them into nonforward densities @xmath328 . the latter would appear then through two types of integrals @xmath419 and similarly for @xmath420 . the functions @xmath421 are the flavor components of the usual @xmath296 form factor while @xmath422 are the flavor components of a new form factor specific to the wide - angle compton scattering . in the formal asymptotic limit @xmath423 , the @xmath35 integrals for @xmath421 and @xmath422 are both dominated in our model by the @xmath373 region : the large-@xmath55 behavior of these functions is governed by the feynman mechanism and their ratio tends to 1 as @xmath307 increases ( see fig.[fig : rcuux]a ) . however , due to large value of the effective scale @xmath368 gev@xmath239 , the accessible momentum transfers @xmath424 gev@xmath239 are very far from being asymptotic . in fig.[fig : rcuux]b we plot @xmath425 and @xmath426 at @xmath427 gev@xmath239 . it is clear that the relevant integrals are dominated by rather small @xmath35 values @xmath428 which results in a strong enhancement of @xmath429 compared to @xmath430 for @xmath431 gev@xmath239 . note also that the @xmath432 matrix elements can produce only @xmath55 as a large variable while @xmath433 gives @xmath377 . as a result , the enhanced form factors @xmath422 are accompanied by extra @xmath434 enhancement factors compared to the @xmath421 terms . in the cross section , these enhancements are squared , i.e. , the contributions due to the non - enhanced form factors @xmath421 are always accompanied by @xmath435 factors which are smaller than 1/4 for cm angles below 90@xmath412 . because of double suppression , we neglect @xmath421 terms in the present simplified approach . the contribution due to the @xmath436 functions appears through the flavor components @xmath437 of the @xmath438 form factor and their enhanced analogues @xmath439 . the major part of contributions due to the @xmath436-type nds appears in the combination experimentally , @xmath440 . since @xmath441 , @xmath442 is similarly suppressed compared to @xmath443 , and we neglect contributions due to the @xmath439 form factors . we also neglect here the terms with another spin - flip distribution @xmath444 related to the pseudoscalar form factor @xmath248 which is dominated by the @xmath55-channel pion exchange . our calculations show that the contribution due to the parton helicity sensitive densities @xmath445 is suppressed by the factor @xmath446 compared to that due to the @xmath447 densities . this factor only reaches 1/8 for the cm angle of 90@xmath412 , and hence the @xmath445 contributions are not very significant numerically . for simplicity , we approximate @xmath345 by @xmath335 . after these approximations , the wacs cross section is given by the product @xmath448 \ , r_1 ^ 2(t ) \ , , \ ] ] of the klein - nishina cross section ( in which we dropped @xmath449 and @xmath450 terms ) and the square of the @xmath443 form factor @xmath451 \ , .\ ] ] in our model , @xmath443 is given by @xmath452 e^{\bar x t / 4x \lambda^2 } \frac{dx}{x } \ , .\ ] ] we included here the sea distributions assuming that they are all equal @xmath453 and using a simplified parametrization @xmath454 which accurately reproduces the grv formula for @xmath338 gev@xmath239 . due to suppression of the small-@xmath35 region by the exponential @xmath455 $ ] , the sea quark contribution is rather small ( @xmath456 ) even for @xmath457 gev@xmath239 and is invisible for @xmath458 gev@xmath239 . comparison with existing data @xcite is shown in fig.[fig : rct ] . our curves follow the data pattern but are systematically lower by a factor of 2 , with disagreement becoming more pronounced as the scattering angle increases . since we neglected several terms each capable of producing up to a @xmath459 correction in the amplitude , we consider the agreement between our curves and the data as encouraging . the most important corrections which should be included in a more detailed investigation are the @xmath55-corrections in the denominators of hard propagators and contributions due to the `` non - leading '' @xmath460 nonforward densities . the latter , as noted above , are usually accompanied by @xmath461 and @xmath462 factors , i.e. , their contribution becomes more significant at larger angles . the @xmath55-correction in the most important hard propagator term @xmath463 $ ] also enhances the amplitude at large angles . the angular dependence of our results for the combination @xmath464 is shown on fig.[fig : rctheta ] . all the curves for initial photon ehergies 2,3,4,5 and 6 gev intersect each other at @xmath465 . this is in good agreement with experimental data of ref.@xcite where the differential cross section at fixed cm angles was fitted by powers of @xmath377 : @xmath466 with @xmath467 . our curves correspond to @xmath468 and @xmath469 which also agrees with the experimental result @xmath470 . this can be compared with the scaling behavior of the asymptotic hard contribution : modulo logarithms contained in the @xmath302 factors , they have a universal angle - independent power @xmath471 . for @xmath472 , the experimental result based on just two data points is @xmath473 , while our model gives @xmath474 . clearly , better data are needed to draw any conclusions here . the hard exclusive electroproduction processes provide new information about hadronic structure accumulated in skewed parton distributions . the spds are universal hybrid functions having the properties of parton densities , hadronic form factors and distribution amplitudes . they give a unified description of various hard exclusive and inclusive reactions . the basic supplier of information about skewed parton distributions is deeply virtual compton scattering which offers a remarkable example of bjorken scaling phenomena in exclusive processes . furthermore , wide - angle real compton scattering is an ideal tool to test angle - dependent scaling laws characteristic for soft overlap mechanism . i am grateful to a. bialas and m. praszalowicz for hospitality in zakopane and support . i thank k. golec - biernat for discussions . this work was supported by the u.s . department of energy under contract no . de - ac05 - 84er40150 . r. p. feynman , _ the photon - hadron interaction _ ( benjamin , reading 1972 ) . v. l. chernyak and a. r. zhitnitsky , jetp letters , * 25 * , 510 ( 1977 ) ; + v. l. chernyak , a. r. zhitnitsky and v. g. serbo , jetp letters * 26 * , 594 ( 1977 ) . d. r. jackson , thesis , caltech , 1977 ( unpublished ) ; + g. r. farrar and d. r. jackson , phys . * 43 * , 246 ( 1979 ) . a. v. radyushkin , jinr report p2 - 10717 , dubna , 1977 ( unpublished ) . a. v. efremov and a. v. radyushkin , jinr preprint e2 - 11983 , dubna ( october 1978 ) , + published in theor . phys . * 42 * , 97 ( 1980 ) . s. j. brodsky and g. p. lepage , phys . * 87b * , 359 ( 1979 ) . s. j. brodsky and g. p. lepage , phys . rev . * d22 * , 2157 ( 1980 ) . x. ji , phys . * 78 * , 610 ( 1997 ) ; phys.rev . d * 55 * ( 1997 ) 7114 . radyushkin , phys . b * 380 * , 417 ( 1996 ) ; a.v . radyushkin , phys . b * 385 * , 333 ( 1996 ) . radyushkin , phys . d * 56 * , 5524 ( 1997 ) . collins , l. frankfurt and m. strikman , phys . d * 56 * ( 1997 ) 2982 . d. mller , d. robaschik , b. geyer , f. -m . dittes and j. hoeji , fortschr . phys . * 42 * , 101 ( 1994 ) . j. bartels and m. loewe , z. phys . * c12 * , 263 ( 1982 ) . l. v. gribov , e. m. levin and m. g. ryskin , phys . reports , * 100 * , 1 ( 1983 ) . h. abramowicz , l. frankfurt and m. strikman , slac summer institute , stanford , p. 539 ( 1994 ) ; hep - ph/9503437 . radyushkin , phys . d * 59 * ( 1999 ) 014030 . radyushkin , phys . b * 449 * ( 1999 ) 81 . radyushkin , phys . b * 131 * ( 1983 ) 179 . m.v . polyakov and c. weiss , rub - tpii-1 - 99 , hep - ph/9902451 . radyushkin , phys . d * 58 * ( 1998 ) 114008 . l. mankiewicz , g. piller and t. weigl , eur . j. c * 5 * ( 1998 ) 119 . x.ji , j. phys . g * 24 * ( 1998 ) 1181 . brodsky , t.huang and g.p.lepage , in _ particles and fields 2 _ , proceedings of the banff summer institute , banff , alberta , 1981 , edited by a.z . capri and a.n . kamal ( plenum , new york,1983 ) p.143 . m. gluck , e. reya and a. vogt , z.phys . c * 67 * , 433 ( 1995 ) . l. andivahis et al . phys . rev.d * 50 * , 5491 ( 1994 ) . b.l . ioffe and a.v . smilga , nucl . b * 232 * , 109 ( 1984 ) . i.i . balitsky and a.v . yung , phys . b * 129 * , 328 ( 1983 ) . v.a . nesterenko and a.v . radyushkin , jetp lett . * 39 * , 707 ( 1984 ) . v.m . belyaev and ya . i. kogan , int . a , * 8 * , 153 ( 1993 ) . balitsky , phys . b * 114 * , 53 ( 1982 ) . farrar and h. zhang , phys.rev . d * 41 * , 3348 ( 1990 ) ; erratum : phys.rev . d * 42 * , 2413 ( 1990 ) a. kronfeld and b. nii , phys.rev . d * 44 * , 3445 ( 1991 ) ; erratum : phys.rev . d * 46 * , 2272 ( 1992 ) . m. vanderhaeghen , dapnia - sphn-97 - 42 , saclay ( 1997 ) . a.v . radyushkin and r. ruskov , nucl . b * 481 * , 625 ( 1996 ) . shupe et al . d * 19 * , 1921 ( 1979 ) .

we have studied the properties of the stellar populations in the field of the ngc346 cluster in the small magellanic cloud , using a novel self - consistent method that allows us to reliably identify pre - main sequence ( pms ) objects actively undergoing mass accretion , regardless of their age . the method does not require spectroscopy and combines broad - band and photometry with narrow - band imaging to identify all stars with excess h emission and derive the accretion luminosity and mass accretion rate for all of them . the application of this method to existing hst / acs photometry of the ngc346 field has allowed us to identify and study 680 bona - fide pms stars with masses frommtomand ages in the range frommyr tomyr . previous investigations of this region , based on the same data , had identified young (myr old ) candidate pms stars on the basis of their broad - band colours . in this study we show that there are at least two , almost equally numerous , young populations with distinct ages of respectively and myr . we provide for all of them accurate physical parameters . we take advantage of the unprecedented size of our pms sample and of its spread in mass and age to study the evolution of the mass accretion rate as a function of stellar parameters . we find that , regardless of stellar mass , the mass accretion rate decreases with roughly the square root of the age , or about three times slower than predicted by current models of viscous disc evolution , and that more massive stars have systematically higher mass accretion rate in proportion to their mass . a multivariate linear regression fit reveals that , where is the age of the star , its mass and a quantity that is higher at lower metallicity . this result is consistent with measurements of the mass accretion rate in the 30 dor region and in the milky way and suggests that longer duration for mass accretion could be related to lower metallicity . the high mass accretion rates that we find suggest that a considerable amount of mass is accreted during the pms phase , of ordermor possibly of the final mass for stars with massmif their discs are eroded by 20myr , i.e. before they reach the main sequence . therefore , pms evolutionary models that do not account for this effect will systematically underestimate the true age when compared with the observations .

1104.4494

ngc346 is presently the region of most intense star formation in the entire small magellanic cloud ( smc ) . it contains over 30 o type stars ( massey , parker & garmany 1989 ; evans et al . 2006 ) that ionise the n66 nebula , the largest hii region in the smc ( henize 1956 ) . the presence of these very young ( @xmath20myr ) massive stars in ngc346 has long been established with ground - based spectroscopy ( e.g. walborn 1978 ; walborn & blades 1986 ; niemela et al . 1986 ; massey et al . 1989 ) , but more recently also very young stars of low mass have been detected . observations with the hubble space telescope and spitzer space telescope have revealed respectively a multitude of pre - main sequence ( pms ) star candidates down to the subsolar mass ( nota et al . 2006 ; sabbi et al . 2007 ; gouliermis et al . 2007 ; hennekemper et al . 2008 ) and a large number of candidate young stellar objects ( yso ) with masses as low as @xmath21m@xmath7(bolatto et al . 2007 ; simon et al . 2007 ) . what makes ngc346 and the surrounding regions particularly interesting is that they allow us to study the properties of star formation in a nearby galaxy with a metallicity similar to those in place in the high redshift universe at @xmath22 . the distance modulus for the smc is @xmath23 corresponding to @xmath24kpc ( hilditch et al . 2005 ; keller & wood 2006 ) , whereas the currently accepted values for its metallicity range from @xmath25 to @xmath26z@xmath27 ( see russell & dopita 1992 ; rolleston et al . 1999 ; lee et al . 2005 ; perez montero & diaz 2005 ) . moreover , besides this recent burst , previous star formation episodes have been detected in this area . heap et al . ( 2006 ) and mokiem et al . ( 2006 ) reported the presence of massive stars with an age of @xmath28myr , while massey et al . ( 1989 ) discovered five red supergiants and two b type supergiants that form a spatially distinct subgroup located about @xmath29 sw of the centre of ngc346 and that have an estimated age of @xmath30myr . more recently , sabbi et al . ( 2007 ) detected a small star cluster , located about @xmath31 or @xmath32pc ne of the centre whose colour magnitude diagram ( cmd ) is compatible with an age of @xmath33myr . finally , older populations are also present : sabbi et al . ( 2007 ) find evidence for star formation in this region dating as far back as @xmath34gyr with a moderate enhancement @xmath35myr ago . in addition , an intermediate - age cluster , bs90 , is also present in the field , with an age of @xmath36gyr . it is not easy to establish whether there is any relationship between the very old generations in this field and the more recent star formation episodes . however , it is interesting to investigate whether the @xmath30myr old massive stars of massey et al . ( 1989 ) and the @xmath37myr old small cluster of sabbi et al . ( 2007 ) , although spatially unrelated to one another , could be the precursors of the 30 young massive o type stars and of the recently detected pms stars and yso in this region . it is conceivable that a large number of these @xmath38myr old objects are actually present in ngc346 , many more than those known so far , but that they could belong to a more diffuse population spread throughout the region because of appreciable velocity dispersion motions and , therefore , might not be easy to identify . if the members of such a @xmath39myr old population could be detected , including those of low mass , their spatial distribution and physical properties would allow us to understand whether and how they may have triggered the most recent bursts , as theories of sequential star formation suggest ( e.g. elmegreen & lada 1977 ) . the task of detecting a @xmath39myr old population in a region heavily contaminated by older field stars and younger objects might appear daunting . hennekemper et al . ( 2008 ) attempted to study the presence of an age spread in the pms population of ngc346 through the analysis of the cmd . they showed that the observations are in principle compatible with an upper limit of @xmath34myr to the age , but they concluded that differential reddening and unresolved binaries could be at the origin of the observed colour spread . actually , if these objects have not yet completed their pre - main sequence ( pms ) phase , it should be possible to detect them through the distinctive excess emission features in their spectra that originate from the accretion process , particularly in the uv continuum and in recombination lines such as h@xmath3 , pa@xmath40 and br@xmath41 ( e.g. calvet et al . 2000 ) . although in nearby star forming regions the fraction of pms stars with discs appears to decline quite rapidly during the first 10myr ( e.g. haisch , lada & lada 2001 ; fedele et al . 2010 ) , there is evidence that sustained mass accretion is present also in objects older than this age , particularly above 1m@xmath7 . this is the case for instance of tr37 where sicilia aguilar et al . ( 2006 ) find g - type dwarfs still accreting at ages in excess of 10myr . in the more distant ( @xmath42kpc ) and massive galactic star forming region ngc3603 beccari et al . ( 2010 ) recently found a conspicuous number of pms stars older than 10myr still undergoing mass accretion . in even more distant and dense stellar fields , where spectroscopy of individual stars is limited to the brightest members , romaniello ( 1998 ) , panagia et al . ( 2000 ) and romaniello , robberto & panagia ( 2004 ) have shown that objects undergoing active mass accretion can be efficiently detected with accurate multi - colour photometry and have identified in the regions around sn1987a about 500 pms stars with h@xmath3 equivalent width in excess of 8 and with an age of @xmath43myr . more recently , de marchi , panagia & romaniello ( 2010 , hereafter paperi ) showed that through a suitable combination of broad- and narrow - band photometry it is also possible to derive the mass accretion rate of these objects , with an accuracy comparable to that allowed by spectroscopy . in this work , we build on the method developed in paperi to securely identify bona - fide pms objects in dense stellar fields and apply it to the high - quality hst photometry of the ngc346 region ( sabbi et al . our goal is to measure the physical properties of these objects and study how they depend on age , in order to understand how star formation has proceeded in this area over the past @xmath44myr . the paper is organised as follows : in section[obser ] we briefly describe the observations and address how we correct for differential reddening , whereas section[pmsst ] is devoted to the identification of pms stars via their h@xmath3 excess emission and on the determination of their h@xmath3 luminosity . in section[physi ] we derive the other important physical parameters for these objects , including age , mass and mass accretion rate , while in section[evolu ] we study how the latter evolves in time , how it could affect stellar evolution , and we compare the results to those for similar objects in the milky way and large magellanic cloud . a summary of the most important conclusions of the paper is offered in section[summa ] . the observations used in this work were collected with the wide field channel of the advanced camera for surveys ( ford et al . 2003 ) on board the hubble space telescope in 2004 july as part of proposal go10248 ( with antonella nota as principal investigator ) . the data and their reduction were extensively described in sabbi et al . ( 2007 ) , to which we refer the reader for details . in summary , the observations of interest here cover an area of @xmath45 on a side centered on the cluster ngc346 in the small magellanic cloud ( smc ) , at ra@xmath46 , dec=@xmath47 ( j2000 ) , itself immersed in n66 , the largest and brightest hii region in the smc . the data comprise both deep and short exposures in the broad bands f555w ( hereafter @xmath0 ) and f814w ( hereafter @xmath1 ) , as well as observations in the 80 wide band f658n ( hereafter h@xmath3 ) , which were not described in sabbi et al . the total combined exposure times in each band are , respectively , 4148s , 4124s and 1542s . the data reduction in the @xmath0 and @xmath1 bands was carried out by sabbi et al . ( 2007 ) using a combination of psf fitting and aperture photometry and their final catalogue , in the central @xmath48 of interest here , includes about 53550 stars reliably detected ( at the @xmath49 level ) in @xmath0 and @xmath1 . as regards the h@xmath3 band , we have adopted the same combination of psf fitting and aperture photometry to search for and measure all objects with a signal in the central pixel exceeding the local background level by at least 3 times the standard deviation of the latter ( i.e. @xmath49 detection ) . of all the stars detected in the @xmath0 and @xmath1 bands , about 19800 satisfy this condition also in the h@xmath3 band . we show in figure[fig1 ] the resulting colour magnitude diagram ( cmd ) of all the stars in the selected field as light grey dots and with darker symbols stars whose mean error @xmath50 in the three bands does not exceed @xmath51mag , where @xmath52 and @xmath53 , @xmath54 and @xmath55 are the photometric uncertainties in each individual band . a total of @xmath56 stars satisfy this condition , with the most stringent constraint being set by the uncertainty in the h@xmath3 band . the cmd , as already discussed by nota et al . ( 2006 ) , gouliermis et al . ( 2006 ; 2007 ) and sabbi et al . ( 2007 ) , reveals the presence in this field of both young and old populations . the former includes the massive stars on the pronounced upper main sequence ( ms ) and for the candidate pms objects to the right of and brighter than the lower ms , while the latter accounts for the lower ms and for the sub - giant and red giant branches ( sgb , rgb ) . in general , however , it is not possible to classify an object as being young or old just on the basis of its position on the cmd , not only because there are regions such as the lower ms that are common to both populations , but also because there is evidence for variable reddening in the field , as sabbi et al . ( 2007 ) already pointed out and as one can see from the conspicuous broadening of the upper ms in figure[fig1 ] . differential reddening makes it particularly difficult to understand the true nature of the large group of objects below the sgb and redder than the ms . while their colours and magnitudes , as originally pointed out by nota et al . ( 2006 ) , are compatible with those of young pms stars , depending on the amount of reddening and on the specific extinction law for this field , some of them may in reality be reddened ms or sgb stars . furthermore , even if all these objects were pms stars , without accounting for differential reddening one could not properly assign to them a mass and age through the comparison with pms evolutionary tracks ( see e.g. hennekemper et al . it is therefore important to address the issue of differential reddening and to correct the photometry accordingly , to the extent possible . we do so in the following section . sabbi et al . ( 2007 ) concluded that the young population in this field , corresponding to the upper ms , is compatible with an age of @xmath57myr . therefore , one can determine the reddening towards each of those stars if one assumes that they are all ms objects and one looks at their colour and magnitude displacement with respect to a theoretical isochrone for an age of 4myr , a metallicity of @xmath58 and a distance modulus of @xmath23 ( see introduction ) . hennekemper et al . ( 2008 ) followed a similar approach to derive an extinction map for this region , albeit with a set of isochrones for a higher metallicity ( @xmath59 ) . we note here that although the metallicity that we adopt , @xmath58 , is at the lower end of the currently accepted values for the smc ( see introduction ) , it appears better suited to describe the properties of young pms stars , as we will show in section[physi ] . as for the extinction law in this field , we decided to use the one derived by scuderi et al . ( 1996 ) for the field of sn1987a in the large magellanic cloud , rather than the galactic law as adopted by hennekemper et al . ( 2008 ) , in light of the more similar metallicity . the total extinction values for the specific hst filters used here were derived by averaging over the entire bandpass the dust attenuation ( @xmath60 ) , as explained in romaniello et al . the solid lines cutting through the upper ms in figure[fig2]a correspond to the reddening vectors drawn from a 4myr old isochrone for @xmath58 ( marigo et al . 2008 ) , in steps of @xmath61 mag for stars brighter than @xmath62 . the length of the lines corresponds to @xmath63 . a total of about 600 stars selected in this way and bluer than @xmath64 are marked with thicker symbols in figure[fig2]a . the corresponding reddening distribution ranges from @xmath65 to @xmath66 ( respectively 17% and 83% limits corresponding to @xmath67 values of @xmath68 and @xmath69 ) , with a median value of @xmath70 or @xmath71 , indicating that a considerable spread in reddening is present in this field . we have used the reddening values towards each of the selected upper ms stars to derive a reddening correction for all other objects in their vicinities . the analysis of the h@xmath3 image indicates that the typical scale of the brightness fluctuations is of order @xmath72 radius . we have therefore assumed that stars projected within this region have similar extinction , even though we have no information on the depth of their distribution nor on the uniformity of the absorbing material . in order to validate the procedure , we first applied the reddening correction to the upper ms stars themselves , using for each object the average of the extinction values of at most 20 other upper ms stars within @xmath72 of their location . we took the straight average without weighting them for the distance from the object , since we have no information on the distribution of the stars along the line of sight . we found that the average and standard deviation of the reddening distribution derived with the nearest neighbours are in excellent agreement with those measured directly , namely @xmath73 . this has convinced us that the average reddening value measured in a small area surrounding each star can be used to correct for extinction . objects with no reference upper ms stars within @xmath72 were assigned the average reddening value of all reference stars ( @xmath74 ) . the upper ms corrected in this way appears considerably tighter in the cmd than the raw data ( see figure[fig2]b ) and this convinced us that the method , albeit statistical in nature , can be applied to the entire data set . we also experimented with a different number of neighbours , namely 5 , 10 , 20 and 50 . we found that , by using the 20 nearest upper ms stars , the red clump ( rc ) feature in the red giant branch of old ( @xmath9gyr ) in the cmd at @xmath75 appeared the tightest . since our goal is to correct for reddening all candidate pms stars , which are typically red , this choice seemed the most appropriate . the de - reddened cmd obtained in this way is shown in figure[fig2]b . one of the characteristic signatures of the accretion process on to pms stars is the presence of excess emission in h@xmath3 ( see e.g. calvet et al . recently , we showed in paperi that pms stars undergoing active accretion can be reliably identified , and their h@xmath3 luminosity measured , by using a combination of broad - band ( @xmath76 ) and narrow - band ( h@xmath3 ) photometry , with no need for spectroscopy . this way of identifying pms stars is more accurate and reliable than the simple classification based on the position of the objects in the hertzsprung russell diagram , i.e. stars placed well above the main sequence ( e.g. gilmozzi et al . 1994 ; hunter et al . 1995 ; see in particular nota et al . 2006 and gouliermis et al . 2007 for the specific case of ngc346 ) . the true advantages of the method presented in paperi are that : _ ( i ) _ it allows us to discriminate between bona - fide pms stars and interlopers that occupy the same region of the cmd ; _ ( ii ) _ it provides a secure detection of relatively old pms stars , already close to their ms , whose colours and magnitudes would otherwise be indistinguishable from those of ms stars ; _ ( iii ) _ by measuring the @xmath2 luminosity @xmath77 of these pms stars we can obtain their mass accretion rates @xmath78 from @xmath77 and the stellar parameters ( mass and radius ) ; _ ( iv ) _ finally , this approach is more practical , efficient and economical than slitless spectroscopy for studying pms stars in dense star forming regions , although it can not provide constraints for magnetospheric accretion models that require information on the profile of the h@xmath3 emission line ( e.g. muzerolle et al . 2001 ) . as explained in detail in paperi , the method hinges on the simple consideration that , at any time , the largest majority of stars of a given effective temperature @xmath79 in a stellar population will have no excess h@xmath3 emission . therefore , for stars of that effective temperature , the _ median _ value of any colour index involving h@xmath3 ( e.g. @xmath80 ) effectively defines a spectral reference template with respect to which the h@xmath3 colour excess should be sought . in figure[fig3 ] we show with small dots the @xmath80 colour as a function of @xmath81 for all 18764 stars that have a combined photometric uncertainty @xmath82mag ( see equation[eq1 ] ) . the value of @xmath50 is dominated by the uncertainty on the h@xmath3 magnitude , while the median value of the uncertainty in the other two bands is @xmath83 and @xmath84 , respectively . the thick dashed line in figure[fig3 ] represents the median @xmath80 colour obtained as the running median with a box - car size of 100 points . the thin solid line shows the colours in these filters for the model atmospheres of bessell et al . ( 1998 ) . as discussed in paperi , the @xmath85mag discrepancy between models and observations around @xmath86 is most likely due to the coarse spectral sampling of the h@xmath3 line in the models . where the density of observed points decreases considerably , at @xmath87 , we reduced the box - car size to 10 points and the result is shown by the thick dot - dashed line ( which is extrapolated for @xmath88 ) . a total of 891 objects have a @xmath80 index exceeding that of the reference template ( i.e. the photospheric colour ) at the same @xmath81 colour by more than four times the uncertainty on their @xmath80 values . the arrow in the figure corresponds to the reddening vector for @xmath89 or @xmath90 for the specific bands used in this study , derived from the extinction law of scuderi et al . the fact that the arrow is almost parallel to the reference template implies that even relatively large uncertainties in the reddening correction would not significantly affect the selection of stars with h@xmath3 excess . on the other hand , it is possible that for some objects surrounded by non homogeneous nebular emission the h@xmath3 excess that we measure is spurious . we explain in section[nebul ] how we have removed these objects from our list . the effects of nebular contamination on the detection of stars with h@xmath3 excess emission are addressed in detail in paperi ( and in beccari et al . 2010 ) , to which we refer the reader . we show there that gas emission acts like a `` pedestal , '' or an additional source of background . as such , it does not affect our photometry because it increases the measured flux in the same way both inside the central aperture and inside the background annulus used for the determination of the sky level . therefore the gas emission is automatically removed in the sky subtraction procedure . if the diffuse emission is strong compared to the intensity of the objects , the uncertainty on the flux of the star will be larger , but this is fully taken into account in our photometry . a possible complication , not discussed in paperi , arises when the nebular emission is not homogeneous across the field . for example , when a filamentary structure in the nebular gas overlaps with a star , the magnitude of the star can be affected in different ways . if the filament is wider than the diameter of the annulus that we use to measure the local background , this only results in a higher photometric uncertainty , because we fall in the case explained above for diffuse emission . the situation is more complex when the filament only partly covers the background annulus ( which in our case extends from 3 to 5 pixel radius around the object ) and/or the central star ( for which an aperture of 2 pixel radius is used ) , since in those cases the signal from the star and/or the background could be incorrectly determined . we , therefore , inspected the h@xmath3 image around the location of all stars with an apparent h@xmath3 excess . in order to enhance the contrast between neighbouring pixels in the image , we applied an unsharp - masking filter ( e.g. malin 1977 ) to the h@xmath3 frame , with a smoothing radius of 10 pixel meant to subtract the low - frequency signal , i.e. structures in the image with a characteristic scale larger than the size of our background annulus . unsharp masking makes structures and filaments in the nebular gas easier to see than in the direct image . in this way we compared the structures and filaments with the positions of the stars with halpha excess emission ( see beccari et al . 2010 for an application of this method to the case of ngc3603 ) . we adopted a very conservative approach and excluded from our pms candidate list all stars whose photometric background annulus is partially covered by a gas filament . in total , about 70 stars were excluded in this way , or less than 10% of the original population ( an additional 30 stars were excluded because they are closer than 10 pixels to the image edges in the halpha frame , although their magnitudes are well defined in the other bands ) . a total of 791 objects are left after this conservative selection and they are indicated with large dots in figure[fig3 ] ( shown in red in the on - line version ) and , in light of their @xmath80 excess at the @xmath91 level , they are excellent pms candidates . as we shall see later , most of them are indeed bona - fide pms stars . following the detailed procedure outlined in paperi , we can derive the h@xmath3 emission line luminosity @xmath77 for these stars from the @xmath92 colour corresponding to the excess emission , namely : @xmath93 where the superscript _ obs _ refers to the observations and _ ref _ to the reference template . @xmath77 is then obtained from @xmath94 , from the photometric zero point and absolute sensitivity of the instrumental set - up and from the distance to the sources . the assumed distance modulus is @xmath95 ( see introduction ) , whereas the photometric properties of the instrument are as listed in the acs instrument handbook ( maybhate et al . the derived median @xmath2 luminosity of the 791 objects with h@xmath3 excess at the @xmath91 level is @xmath96erg / s or @xmath97l@xmath27 . this value is similar to that of a group of 189 stars with h@xmath3 excess emission in the field of sn1987a that were studied in paperi with the same method applied to hst wide field planetary camera2 data . the uncertainty on @xmath77 is typically @xmath30% and is completely dominated by the statistical uncertainty on the h@xmath3 photometry ( 10% ) , while the distance accounts for 5% ( hilditch et al . 2005 ; keller & wood 2006 ) and the absolute sensitivity of the instrumental setup for another 3% . owing to the rather wide passband of the f658n filter , @xmath92 includes a small contribution from the [ nii ] emission features at @xmath98 and @xmath99 . even though such a contribution is small ( see paperi , where the effect is explained in detail ) , it is a good practice to correct for it as it is a systematic effect . the adopted average correction factor to the intensity for the acs f658n filter , based on the spectra of classical t tauri stars is @xmath100 . besides the luminosity of the h@xmath3 emission line , also its equivalent width @xmath101 can be derived from our photometry . to this aim , as shown in paperi , we used the model atmospheres of bessell et al . ( 1998 ) to derive the magnitude @xmath102 corresponding to the sole continuum in the h@xmath3 band from the @xmath0 and @xmath1 magnitudes . the corresponding relationships , properly validated by comparison with spectro - photometric measurements , are listed in the appendix of paperi and allow us to derive @xmath103 from the h@xmath3 magnitude via the equation : @xmath104 \label{eq3}\ ] ] where @xmath105 is the rectangular width of the filter ( similar in definition to the equivalent width of a line ) , which depends on the characteristics of the filter and corresponds to @xmath106for the f658n filter used here . the equivalent widths obtained in this way are shown in figure[fig4 ] as a function of the @xmath81 colour . the 791 stars with a @xmath91 h@xmath3 excess are indicated with thicker symbols ( circles and squares ) . we take as bona - fide pms stars all those with @xmath107 , indicated as thick dots ( in red in the online version ) , corresponding to a total of 694 stars . note that , as customary , negative values of the equivalent width indicate emission lines . since at temperatures @xmath108k or colours @xmath109 the sample could be contaminated by be stars that are evolving off the ms , there we set a more stringent condition on the equivalent width , namely @xmath110 . this limit is suggested by a survey of the h@xmath3 equivalent width of about 100 be stars in the galaxy ( cot & waters 1987 ) in which only one star is found with @xmath110 and the largest majority have values in the range from @xmath111 to @xmath112 . in fact , as we discuss in section[physi ] , all the objects with @xmath113k and an h@xmath3 excess above @xmath91 are most likely pms objects with masses around 3m@xmath7 . although we do not have spectroscopic observations of our stars to directly confirm our photometrically inferred equivalent widths , similar studies in the literature have demonstrated the success of this approach . white & ghez ( 2001 ) carried out a detailed study of 44 t tauri stars in binary systems in taurus auriga , using both photometry and spectroscopy with the hst . their analysis shows that photometric and spectroscopic measurements of the equivalent width of the h@xmath3 emission line in these objects are in good agreement with one another . more recently , barentsen et al . ( 2011 ) carried out a h@xmath3 survey of t tauri stars in ic1396 , a region with considerable nebular emission . their photometric determination of @xmath101 for 109 objects in common with the spectroscopic survey of sicilia aguilar et al . ( 2005 ) also shows an excellent agreement , particularly for stars with @xmath114 , which are the objects of interest in our study . when the bona - fide pms objects identified in the previous section are placed on the cmd ( see figure[fig5 ] ) , they clearly reveal two distinct groups made up of a similar number of objects . one is located above ( i.e. brighter and redder than ) the ms and thus is indicative of a population of young pms objects . this population was originally noticed by nota et al . ( 2006 ) and later studied by gouliermis et al . ( 2007 ) and sabbi et al . ( 2007 ) , although these earlier works could only identify candidate young ( @xmath115myr ) pms stars since no information was available to verify and characterise the pms nature of each individual object . this is the reason why the works above could not identify the second group of bona - fide pms objects , which overlap in the cmd with older ms stars in the field of the smc . hennekemper et al . ( 2008 ) detected some of these objects using a preliminary version of the colour transformations between @xmath0 , @xmath1 and @xmath116 that we developed for paperi to search for stars with @xmath117 excess in this field , but they did not realise what was the nature of these objects . as we show here below , these objects are older pms stars with an age in excess of @xmath118myr , of the same type as those that we discovered in the field around sn1987a in the lmc ( see paperi ) . the physical parameters of the bona - fide pms stars that we have identified are obtained as explained in the following . the radius @xmath119 comes from the luminosity and effective temperature of the stars , which in turn follow from the observed colour @xmath120 and magnitude ( @xmath0 ) , properly corrected for interstellar extinction as explained in section[redde ] . the stellar mass @xmath121 and age were derived by comparing the location of each star in the hertzsprung russell ( h r ) diagram of figure[fig6 ] with the pms evolutionary tracks and corresponding isochrones , via interpolation over a finer grid than the one shown in the figure . we followed the interpolation procedure developed by romaniello ( 1998 ) , which does not make assumptions on the properties of the population , such as the functional form of the imf . on the basis of the measurement errors , and taking into account its position in the h r diagram , this procedure provides the probability distribution for each individual star to have a given value of the mass and age ( the method is conceptually identical to the one presented recently by da rio et al . 2010 ) . as for the theoretical models , we adopted those of the pisa group ( deglinnocenti et al . 2008 ; tognelli et al . 2011 ) , for metallicity @xmath58 or about @xmath122 . these new pms tracks were specifically computed for the mass range from @xmath123m@xmath7to 5m@xmath7with an updated version of the franec evolutionary code ( see chieffi & straniero 1989 and deglinnocenti et al . 2008 for details and cignoni et al . 2009 for an application ) . as mentioned in the introduction , we opted for a lower value of the metallicity , namely @xmath58 in place of the more customarily adopted @xmath59 , because the corresponding tracks provide a better match to the observations , particularly where they reach the ms . as previously mentioned , @xmath58 is still compatible with the metallicity range found in the literature for the smc and is in any case better suited to ngc346 than the value adopted in previous studies of its pms population that made use of the evolutionary tracks of siess et al . ( 2000 ) for @xmath124 . figure[fig6 ] reveals that the masses of the bona - fine pms stars span over a decade , from @xmath125m@xmath7for the coolest objects to @xmath8m@xmath7for the hottest ones . as mentioned above , some of the objects with @xmath126k and @xmath127 could be be stars and we have marked all objects of this type with squares to distinguish them from the rest . figure[fig6 ] also provides a more quantitative characterisation of the age difference between the two populations of pms stars that we identified earlier in the cmd ( figure[fig5 ] ) . the isochrones in the figure , corresponding from right to left to ages of @xmath128 , @xmath129 , @xmath61 , 1 , 2 , 4 , 8 , 16 and 32myr , show that the young population has a median age of @xmath9myr , whereas the older bona - fide pms stars span a wider age range from @xmath34myr to 30myr , with a broad peak around 20myr . we derived a reliable measure of the mass and age for all stars indicated as thick dots in figure[fig6 ] , except for the coolest objects that would have required a more uncertain extrapolation beyond the limits of the theoretical models . histograms with the mass and age distribution for the 680 stars measured in this way are shown in figures[fig7]a and [ fig7]b , respectively . as figure[fig6 ] already implied , the histogram in figure[fig7]b shows that there is a remarkable paucity of stars with ages around 48myr . in particular , the solid line in figure[fig7]b provides a histogram of the age distribution with constant logarithmic step ( a factor of 2 ) and shows two peaks , at @xmath130 and @xmath131myr the two peaks remain visible also when the number of stars in each bin is rescaled by the size of the bin , as indicated by the dot - dashed line ( note that at the extremes of the distribution it becomes more difficult to assign an age to the stars , so the first and last bin are drawn with a dotted line to indicate a larger uncertainty ) . in a companion paper ( de marchi , paresce & sabbi 2011 ) , we discuss the age and spatial distributions of these stars in more detail and show that between 30 and 10myr ago a large number of pms stars formed , similar to that produced by the current burst , which however appears to be much more intense . besides the h@xmath3 luminosity @xmath77 already discussed in the previous section , the other physical parameters of interest in this work are the accretion luminosity @xmath4 and the mass accretion rate @xmath78 , that we derived following the procedure described in paperi . from an analysis of the literature values of @xmath4 and @xmath77 measurements as recently summarised by dahm ( 2008 ) for pms stars in the taurus auriga association , that paper concluded that the ratio @xmath132 can be assumed to be constant . the proportionality constant , obtained from an elementary fit to the data in the compilation of dahm ( 2008 ) , is : @xmath133 where the large uncertainty on the proportionality factor arises because the two quantities were determined from non - simultaneous observations . considering that the intensity of the h@xmath3 line from pms sequence stars is known to vary by about 20% on a time scale of a few hours and by as much as a factor of 2 3 in a few days ( e.g. fernandez et al . 1995 ; smith et al . 1999 ; alencar et al . 2001 ) , the large spread can be understood . however , as we will show in section[evolu ] , the small dispersion of the mass accretion rate with stellar age suggests an uncertainty smaller than @xmath129 . the mass accretion rate is related to @xmath4 via the free - fall equation , that links the luminosity released in the impact of the accretion flow with the rate of mass accretion @xmath78 , according to the relationship : @xmath134 where @xmath135 is the gravitational constant , @xmath121 and @xmath119 the mass and photospheric radius of the star as derived above and @xmath136 the inner radius of the accretion disc . the value of @xmath136 is rather uncertain and depends on how exactly the accretion disc is coupled with the magnetic field of the star . following gullbring et al . ( 1998 ) , we adopt @xmath137 for all pms objects . by substituting equation[eq4 ] in equation[eq5 ] , the mass accretion rate @xmath78 , in units of m@xmath7 yr@xmath138 , can be directly expressed as a function of @xmath77 : @xmath139 the histograms showing the distribution of @xmath77 and @xmath5 as derived with equation[eq6 ] for the 680 bona - fide pms stars with well defined masses are shown in figures[fig7]c and [ fig7]d , respectively . the median values of @xmath77 and @xmath140 are , respectively , @xmath141erg / s ( or @xmath142l@xmath27 ) and @xmath143m@xmath7 yr@xmath138 . the latter value is about 50% higher than the median mass accretion rate of a population of 133 pms stars that we detected in the field of sn1987a ( see paperi ) and this is fully consistent with the much younger average age of the pms stars in ngc346 . paperi provides an extensive discussion of the statistical and systematic uncertainties involved in determining @xmath78 with this method and we refer the reader to that work for more details . the typical combined statistical uncertainty on @xmath78 for the ngc346 field is 13% . it is largely dominated by the uncertainty on @xmath77 of 10% which in turn is mostly due to random errors in the photometry ( we note that this uncertainty is lower than for the study of the field around sn1987a in paperi because of the higher quality of the acs photometry ) . as for systematic effects , those on @xmath119 , @xmath121 and reddening add up to 15% , but it is the rather uncertain relationship linking @xmath77 to the accretion luminosity @xmath4 that plays a dominant role . we stress here that our method adopts the calibration provided by dahm ( 2008 ) , which is based on a spectroscopic study of galactic pms stars . if the uncertainty on the relationship between @xmath77 and @xmath4 is as large as dahm s ( 2008 ) work suggests ( but see section[evolu ] ) , it could possibly cause a systematic effect of up to a factor of 3 on the value of @xmath140 that we obtain . this is the unfortunate consequence of our incomplete understanding of the physics of the accretion process . improving this situation would require direct measurements of the bolometric accretion luminosity @xmath4 and simultaneous measurements of the hot continuum excess emission and of the emission line fluxes in h@xmath3 or other lines ( e.g. pa@xmath40 and br@xmath41 ) . it is also true , however , that these systematic effects plague in the same way all determinations of @xmath140 that make use of emission lines diagnostics . this includes detailed studies of the profile and intensity of the h@xmath3 line ( e.g. muzerolle et al . 1998 ) , since they rely on the same measurements for their calibration . therefore , it is possible and meaningful to compare the values of @xmath78 that we obtain here to one another and to those derived with other methods , as we do in the following section . in this section we investigate how @xmath78 varies with stellar parameters such as stellar mass and age . it is important to underline that this analysis is based only on stars that , at the time of the observations , were undergoing active mass accretion as revealed by h@xmath3 excess emission above our detection threshold . actually , pms objects are known to display large variations in their h@xmath3 emission over hours or days ( e.g. fernandez et al . 1995 ; smith et al . 1999 ; alencar et al . 2001 ) , with only about one third of them at any given time being active h@xmath3 emitters with @xmath144in the field around sn1987a ( panagia et al . 2000 ) . in fact , a similar requirement of a high accretion activity is common to _ all _ studies in which pms stars are identified based on single - epoch observations , including those conducted spectroscopically on t tauri objects in the milky way ( e.g. sicilia aguilar et al . this implies that the accretion rates measured for the detected objects will , in fact , represent the highest values possible for that class of objects , while many more stars may well be accreting at more modest rates . as a result , the true _ average _ accretion rate appropriate for a given class of objects will be just a fraction of the values measured for the members of that class that can be detected . on the other hand , since all studies made with different techniques suffer from the same kind of selection on the accretion efficiency , a direct comparison between their results is still possible and meaningful . let s start with a qualitative analysis of the h r diagram of the bona - fide pms stars shown in figure[fig8]a , where the size of the symbols is set to be proportional to @xmath78 according to the legend . besides the 680 bona - fide pms stars in ngc346 ( darker dots , red in the online version ) we have included in the figure for comparison also the 133 bona - fide pms stars in the field of sn1987a discussed in paperi ( lighter dots , gray in the online version ) . the theoretical pms isochrones are the same as shown in figure[fig6 ] and correspond to a metallicity @xmath58 and to ages from @xmath128myr to 32myr increasing from the right with a constant logarithmic step of a factor of 2 . while not directly applicable to the pms stars in the sn1987a field , due to the different metallicity , these isochrones serve nonetheless the purpose to illustrate the effects of increasing pms age . it is noteworthy that the mass accretion rate varies by over three orders of magnitude across the h r diagram in a very regular way : younger stars , which are farther away from the ms , have typically higher mass accretion rates , as expected from the theory of stellar evolution . also the stellar mass appears to affect the value of @xmath78 , in the sense that less massive stars tend to have smaller accretion rates . it is also quite remarkable that even pms stars very close to the location of the zero - age ms ( zams ) , which for the masses of interest here can be considered equivalent to the 32myr pms isochrone , are still undergoing substantial mass accretion . to verify that this is not a spurious effect due to errors in our derivation of @xmath78 , we show in figure[fig8]b the same h r diagram but with the size of the symbols proportional to the measured h@xmath3 luminosity . there one sees that @xmath77 can be rather large for objects at or near the zams , suggesting that the accretion process remains relatively active when the star approaches the ms . more precisely , this demonstrates that , even if @xmath78 decreases as a pms star approaches the ms , its shrinking radius allows the star to extract more potential energy per unit rate of mass accretion ( i.e. for a given rate in gcm@xmath138 ) , thereby making the object easier to detect in h@xmath3 . finally , figure[fig8]c shows that the @xmath146 values of pms objects near or at the ms are typically much larger than those caused by chromospheric activity for stars in this mass range ( see e.g. ulmschneider 1979 ; pasquini et al . 2000 ) , confirming that they are bona - fide young objects . for a more quantitative analysis of the evolution of the mass accretion rate , we turn to figure[fig9 ] where the variation of @xmath5 is shown as a function of stellar age . all 680 bona - fide pms stars in ngc346 are indicated as diamonds , with a symbol size proportional to their mass . besides being very homogeneous , this sample is about five times larger than all @xmath78 measurements today available for galactic pms objects taken together ( e.g. sicilia aguilar et al . furthermore , it spans over a decade in mass and two orders of magnitudes in age while including objects from the same star forming region with very accurate and uniform measurements . all this makes the sample particularly well suited for a quantitative analysis of the dependence of the mass accretion rate on stellar parameters . the dashed line in figure[fig9 ] represents the best fit to the evolution of the mass accretion rate in ngc346 , with a slope @xmath147 and intercept @xmath148 ( at 1 myr ) . the fit appears considerably less steep than the prediction of hartmann et al . ( 1998 ; see also muzerolle et al . 2000 ) , based on models of viscous disc evolution and represented here by the solid line , which however seems to adequately reproduce the trend of decreasing @xmath78 with stellar age for low - mass galactic t - tauri stars as compiled by sicilia aguilar et al . ( 2006 ) and also shown in the figure . the different slopes of the observed and model relationships translate into predicted @xmath5 values that are about two order of magnitudes smaller than what we actually observe for an age of 10myr . this discrepancy had already been noticed in the field of sn1987a studied in paperi , although it is pointed out there that the @xmath140 value measured at 10myr is in excellent agreement with that of the most massive pms stars ( with spectral type g ) in the trumpler37 sample of sicilia aguilar et al . ( 2006 ) shown as large circles in the figure . even though the age of the g - type stars in trumpler37 is very uncertain ( sicilia aguilar et al . 2006 ) and could be much younger ( j. muzerolle , priv . see also hartmann 2003 ) , we can not exclude that the slower decline of @xmath78 with time that we observe in ngc346 and around sn1987a could be due to the typically higher mass of our objects compared with that of the galactic t tauri stars that were the subject of recent studies . a dependence of @xmath78 on the stellar mass @xmath149 is indeed expected and could be as steep as @xmath150 according to muzerolle et al . ( 2003 ; 2005 ) , natta et al . ( 2004 ; 2006 ) and calvet et al . ( 2004 ) , although the uncertainties are large ( see also clarke & pringle 2006 for a different interpretation ) . owing to the large size and wide mass range of the objects in our sample , we can investigate the mass dependence of @xmath78 in a very robust way . as a first step , we divide out sample in four mass groups , namely @xmath151m@xmath7 , @xmath152m@xmath7 , @xmath153m@xmath7and @xmath154m@xmath7 , and plot for each one separately the temporal evolution of @xmath78 in the four panels of figure[fig10 ] . the slope @xmath3 and intercept @xmath155 ( at 1myr ) of the best linear fit to the data , according to the relationship @xmath156 , are given in each panel together with their uncertainty . it appears that the slopes of the four mass groups are in very good agreement with one another , within the uncertainty of the fits , but remain significantly different from the trend predicted by hartmann et al . ( 1998 ) and muzerolle et al . ( 2000 ) for viscous disc evolution ( solid lines ) . but for stars in four mass groups , as indicated in each panel . we also provide the slope @xmath3 and intercept @xmath155 ( at 1myr ) of the best fits ( thick long - dashed lines ) . the thin short - dashed lines correspond to our @xmath77 detection limits , which appear to not significantly affect the determination of the slopes , except possibly for the least massive stars at ages @xmath157myr . ] in principle , the shallower slopes of our fits could be affected by our very strict criteria for selecting bona - fide pms stars . requiring that the h@xmath3 excess be at the @xmath91 level or higher and that @xmath158 will set a lower limit to the h@xmath3 luminosity that we can accept as significant and , therefore , to the smallest @xmath78 value that we can detect at a given mass or age . if this lower limit is too high , we could be ignoring too many objects with low mass accretion rate , particularly at older ages , and therefore skew the measurement of the slope towards low values . on the other hand , it appears that our detection limits ( shown by the thin short - dashed lines for the average mass of each group ) do not significantly affect the determination of the slope , except possibly for the least massive stars at ages @xmath159myr . for this group , however , the slope is determined by the much more numerous young stars . more importantly , the upper envelopes of the distribution of all four mass groups appear to be fully consistent with the slope of the best fit . this is shown graphically by the thin long - dashed lines that represent the best - fitting line shifted vertically by @xmath61dex . while our conservative limits on the h@xmath3 excess emission may prevent us from detecting objects with low @xmath77 , no selection effects are to be expected for stars with high h@xmath3 luminosity . since the distribution of these objects is consistent with the slope of the best fit , the latter must be correct . therefore , within the uncertainties quoted in figure[fig10 ] , we can conclude that all four mass groups are compatible with the same slope @xmath160 ( formally @xmath161 ) . an interesting feature of figure[fig10 ] is that the spread of the @xmath162 data values around the best fitting lines is small , typically @xmath163dex ( rms ) . this proves quite convincingly that the true uncertainty in the relationship between @xmath164 and @xmath165 given by equation[eq4 ] can not be as large as the @xmath166dex value suggested by our elementary fit to the data in the compilation of dahm ( 2008 ) , but it has to be intrinsically smaller than the @xmath163dex factor that we observe in our own ngc346 data . we conservatively assume @xmath129dex ( or a factor of @xmath167 in linear units ) . having established that the data in all four panels of figure[fig10 ] are consistent with the same slope @xmath168 , we can use the intercept @xmath155 at 1myr to study how @xmath5 varies with the stellar mass @xmath15 . it appears that the value of @xmath155 in the four panels of figure[fig10 ] decreases with increasing mass approximately as @xmath169 . the formal uncertainty on @xmath155 may seem too large to reach this conclusion , but the scatter is due to the grouping of stars of different mass in the same bin . it is , therefore , more appropriate to perform a multivariate fit to the three variables simultaneously ( @xmath78 , age and mass ) . we opt for a linear multivariate least - square fit of the type : @xmath170 where @xmath14 is in years , @xmath15 the mass in solar units and @xmath16 is a constant . the formal best fit gives @xmath171 , @xmath172 and @xmath173 . however , since the set of parameters @xmath174 , @xmath175 and @xmath176 would give almost equally small residuals , we will take them hereafter as a good as well as convenient representation of the dependence of @xmath78 on stellar mass and age in ngc346 . as a function of age for the pms objects in ngc346 ( circles ) and for galactic stars in taurus ( crosses ) and in trumpler 37 ( squares ) . the hatched band show the best fit ( @xmath177 ) to the ngc346 data , the dark - shaded band is the fit to the pms stars in the 30dor region as derived in paperiii and the light - shaded band is the fit to the galactic stars ( including also the orion nebula cluster ; da rio et al . in preparation ) . ] from this analysis it appears that the ngc346 data are not consistent with a steep dependence of @xmath78 on the stellar mass of the type reported by muzerolle et al . ( 2003 ) and calvet et al . ( 2004 ) in nearby star forming regions . however , the gentler decline of @xmath140 with mass that we find here is fully compatible with that obtained in a study of about 900 pms stars recently identified in the regions around 30dor ( spezzi et al . , in preparation , paper iii ) as well as with the preliminary results of a ground - based study of about 500 pms objects in the orion nebula cluster ( da rio et al . , in preparation ) . in all these cases , values of @xmath174 and @xmath175 in equation[eq7 ] give a good fit to the data . this suggests that we could use the same functional form and parameter values to fit the observed evolution of @xmath78 for galactic pms stars and compare the result to the data in the magellanic clouds . we do so in figure[fig11 ] , where we show the run of the _ specific _ mass accretion rate , defined as @xmath178 , as a function of time for the pms objects in ngc346 ( circles ) and for galactic stars in taurus ( crosses ; from calvet et al . 2000 ) and trumpler 37 ( squares ; from sicilia - aguilar et al . the hatched band represents the best fit to the ngc346 data , the dark - shaded band is the fit to the pms stars in the 30dor region as derived in paperiii , whereas the light - shaded band refers to galactic stars ( including also the orion nebula cluster ; panagia et al . in preparation ) . the hatched and solid bands correspond to the @xmath179 scatter ( i.e. the range from 17 to 83 percentile ) and indicate that the smc and lmc data are compatible with one another , but that the difference with galactic pms stars is significant . the logarithmic value of the specific mass accretion rate at 1myr , @xmath180 calculated using the best fitting line as given in equation[eq7 ] , can serve as a measure of this difference . @xmath181 takes on the values of @xmath182 , @xmath183 and @xmath184 respectively for ngc346 , the 30dor region and the galaxy . as already pointed out in paperi , the higher @xmath78 values of pms stars in the lmc compared to galactic objects of the same age could result from the lower metallicity of the clouds . the radiation pressure of the forming star is expected to be less strong on lower - metallicity disc material and this will delay the dissipation of the disc , thereby keeping the accretion process active for a longer time . this scenario is compatible with the pms stars in the galaxy having a smaller value of the specific mass accretion rate @xmath181 than those in the magellanic clouds . a possibility would be that the shallower photometric depth of the lmc data may cause the @xmath77 detection threshold for pms stars to be higher ( see paperiii for more details ) and thus could skew the median value of @xmath78 . we will address this possibility in detail through the analysis of recently obtained hst observations of the same region ( de marchi et al . , in preparation ) . alternatively , it is also possible that below a certain metallicity threshold the radiation pressure on the disc material no longer changes . investigating this possibility requires the analysis of larger and homogeneous data set of @xmath78 measurements in low - metallicity environments . finally , we would like to highlight an interesting corollary result stemming from the large mass accretion rate values that we find for the objects in ngc346 . using the rates as expressed by equation[eq7 ] we can estimate the total mass accreted by a star over the time span from birth to the zams . already in 1993 palla & stahler ( 1993 ) showed that the timescale to reach the zams is a strong function of the stellar mass . fitting the more modern evolutionary calculations ( deglinnocenti et al 2008 ; tognelli et al 2011 ) for metallicities @xmath185z@xmath27 and stellar masses in the range @xmath186 up to 4m@xmath7 , to within 8% accuracy we can express that time in years as : @xmath187 equation[eq7 ] provides an analytical expression for the average mass accretion rate of the detected stars . on the other hand , there must be an appreciable fraction of pms objects that , at any given time , are not detectable as strong h@xmath3 emitters . therefore , in order to properly estimate the amount of accreted mass over the pms phase we have to make allowance for the appropriate duty cycle . while doing this would require a detailed knowledge of the time evolution of the h@xmath3 emission for a large sample of pms stars in this region , for the sake of simplicity we will assume that the evolution consists in a series of two - level transitions that recur in time . in this simplified model the high state value corresponds to the average value measured for the detected pms stars and the low state of the h@xmath3 undetected stars is zero . thus , the true mass accretion rate of any star of mass @xmath15 at the time @xmath14 will simply be the average value measured for the detected pms stars multiplied by an efficiency factor @xmath188 that in principle is a function of the stellar mass , the time and the metallicity of the region . an approximate value of this function is given by the ratio of the number of pms stars with detected h@xmath3 emission and the total number of candidate pms stars as defined by their location in the h r diagram . in a companion paper , de marchi , panagia & sabbi ( 2011 ) show that for ngc346 such a ratio is essentially constant over time up to pms ages of 8myr at a level of @xmath189 . note that similar values have been found in the field of sn1987a for pms stars with median age of 14myr ( @xmath190 ; panagia et al 2000 ) . therefore , adopting a constant value of @xmath191 for pms stars in ngc346 allows us to integrate the accretion rate over time from 0 up to @xmath192 in order to estimate the total amount of mass ( in m@xmath7units ) accreted by these objects during their pms phase . from equations[eq7 ] and 8 we can express it in units of m@xmath7as follows : @xmath193 showing that the amount of accreted mass is virtually insensitive to the stellar mass . in the range @xmath194m@xmath7@xmath195m@xmath7as studied here , equation[eq9 ] implies @xmath196m@xmath7 . taken at face value , such a high @xmath197 figure would imply that a considerable amount of mass is accreted during the pms phase , at least in low - metallicity environments such as the magellanic clouds . if this is the case , the pms evolution of moderate - mass stars , say @xmath198m@xmath7 , should be reconsidered and recalculated taking into account the high @xmath78 values . one of the effects of equation9 is that the appropriate @xmath192 for a star that reaches the ms with a given mass will be longer than the value estimated by evolutionary models that assume @xmath199 after the first few @xmath200yr ( e.g. baraffe et al . since the star had an appreciably lower mass for a very long time ( several myr ) at the early epochs , it was necessarily evolving more slowly . extrapolating these findings to lower masses than those that we have reached here ( i.e. @xmath201m@xmath7 ) would carry the implication that no star could form with a mass below a suitable minimum value . for ngc346 this would correspond to a lower cut - off of the initial mass function at @xmath202m@xmath7 . in the milky way , where @xmath78 is about five times lower but the metallicity is about 8 times higher so that the pms evolutionary times are a factor of about @xmath203 longer , the cut - off would occur at @xmath204m@xmath7 . on the other hand , it is also possible that for low - mass stars circumstellar discs are eroded in a finite time that is appreciably shorter than their @xmath192 . for example , if for stars below 1m@xmath7we were to limit the integrals in equation9 between 0 and the smaller of @xmath192 and 20myr , the accreted mass would decrease proportionally to the mass , i.e. @xmath205 in practice this would still require some refinements to models of stellar evolution in the pms phase , but the increase in mass would never be higher than 20% of the final zams mass . note that the choice of 20myr as the upper limit for the integration stems from the fact that near the ms we still find 1m@xmath7stars that have strong h@xmath3 excess , thus indicating the presence of considerable accretion . deeper observations of ngc346 reaching well below 1m@xmath7on the ms can easily show whether the discs of lower - mass objects are indeed eroded before they reach the zams . it would be sufficient to verify whether strong h@xmath3 excess emission is still shown by stars smaller than 1m@xmath7near the ms , or , alternatively , whether for objects of these lower masses strong h@xmath3 emission is only seen at younger ages , when they are still well above the ms . if at these masses the discs are indeed eroded before reaching the zams , h@xmath3 emission should systematically taper off for objects older than a cut - off age . in any case , these new results can have important implications for some of the assumptions and boundary conditions used in theories of star formation and early evolution . we have studied the properties of the stellar populations in the field of the ngc346 cluster in the smc as observed with the acs camera on board the hubble space telescope ( sabbi et al . to this aim , we have employed a novel , self - consistent method that allows us to reliably identify pms stars undergoing active mass accretion , regardless of their age . the method , fully described in paperi , does not require spectroscopy and combines broad - band @xmath0 and @xmath1 photometry with narrow - band @xmath2 imaging to identify all stars with excess h@xmath3 emission and allows us to derive the accretion luminosity @xmath4 and mass accretion rate @xmath78 for all of them . the main results of this work can be summarised as follows . 1 . from the photometry of sabbi et al . ( 2007 ) , we have selected all objects in a region of @xmath48 around the centre of ngc346 that have a combined mean error @xmath50 in the three bands ( @xmath0 , @xmath1 and h@xmath3 ) of less than @xmath51mag , as defined in equation[eq1 ] . a total of 18764 stars satisfy this condition . 2 . the cmd reveals the presence of both young and old populations , but the conspicuous broadening of the upper ms proves that there is also variable extinction . we have determined the reddening towards each of these stars from their colour and magnitude displacements with respect to a 4myr isochrone for metallicity @xmath58 and a distance modulus of @xmath206 . we found that extinction is in the range @xmath207 or @xmath208 ( respectively 17% and 83% limits ) , with a median value of @xmath70 or @xmath71 . we have then used the reddening values towards each of the selected stars to derive a reddening correction for all objects in their vicinity . 3 . following the method developed in paperi , we have identified 791 pms candidates as being objects with h@xmath3 excess above the @xmath91 level with respect to the reference provided by normal stars . their average h@xmath3 luminosity is @xmath96ergs@xmath138 or @xmath97l@xmath27 . we have also determined the equivalent width of the h@xmath3 emission line of these objects and have classified as bona - fide pms stars all those with @xmath158 ( or @xmath209 for stars with @xmath126k ) . these conditions guarantee that our sample is free from contamination due to the chromospheric activity of older objects or to the rotational winds of be stars . a total of 694 objects satisfy these conditions . 4 . by comparing the locations of these objects in the h r diagram with the pms evolutionary models of the pisa group for metallicity @xmath58 , we have been able to derive an accurate value of the mass and age of 680 bona - fide pms stars ( note that previous determinations of these parameters for candidate pms stars in ngc346 that made use of evolutionary models for @xmath124 are necessarily less accurate ) . the masses of these objects range from @xmath194m@xmath7to @xmath210m@xmath7 , with an average value of @xmath9m@xmath7 . their ages show a clear bimodal distribution with two peaks at @xmath9myr and @xmath12myr and very few objects around @xmath118myr , revealing the presence of two distinct but equally populous generations of stars . we address in a companion paper ( de marchi , panagia & sabbi 2011 ) the properties of and the relationships between the two generations . 5 . from the h@xmath3 luminosity and the other physical parameters that we have measured , we have derived the mass accretion rate @xmath5 of all bona - fide pms stars . the median value of @xmath78 is @xmath143m@xmath7yr@xmath138 . this value is about 50% higher than that measured in paperi for a population of 133 pms stars in the field of sn1987a , owing to the much younger median age of pms objects in ngc346 . thanks to the unprecedented size of our pms sample and of its spread in mass and age , we have been able to study the evolution of the mass accretion rate as a function of stellar parameters . regardless of the mass of the star , our analysis shows that the mass accretion rate decreases with roughly the square root of the age , or about three times slower than predicted by current models of viscous disc evolution , and that more massive stars have systematically higher mass accretion rate in proportion to their mass . a multivariate linear regression fit reveals that @xmath211 , where @xmath14 is the age of the star , @xmath15 its mass and @xmath16 a quantity that is higher at lower metallicity . the large mass accretion rates that we find imply that a considerable amount of mass is accreted during the pms phase , at least in low - metallicity environments such as the magellanic clouds . at face value , the observed rates integrated over the entire pms phase would give a total accreted mass of @xmath17m@xmath7 , with negligible dependence of the final mass on the zams , thereby possibly suggesting a lower cut - off mass in the stellar mass function of ngc346 . if , however , the circumstellar discs are eroded on a time scale shorter than the pms lifetime , the total accreted mass will be of course smaller but not negligible . in any case , the pms evolution of moderate - mass stars ( @xmath212m@xmath7 ) should be reconsidered and recalculated taking into account the high @xmath78 values , since for a given ms mass the evolutionary time needed to reach the zams will be longer than what is currently estimated by models that assume @xmath199 after the first few @xmath200yr . we wish to thank an anonymous referee for useful comments that have greatly helped us to improve the presentation of this work . np acknowledges partial support by hst - nasa grants go-11547.06a and go-11653.12a , and stsci - ddrf grant d0001.82435 .

[ txt : abstract ] we develop a grid method for multi - channel scattering of atoms in a waveguide with harmonic confinement . this approach is employed to extensively analyze the transverse excitations and deexcitations as well as resonant scattering processes . collisions of identical bosonic and fermionic as well as distinguishable atoms in harmonic traps with a single frequency permitting the center - of - mass ( c.m . ) separation are explored in depth . in the zero - energy limit and single mode regime we reproduce the well - known confinement - induced resonances ( cirs ) for bosonic , fermionic and heteronuclear collisions . in case of the multi - mode regime up to four open transverse channels are considered . previously obtained analytical results are extended significantly here . series of feshbach resonances in the transmission behaviour are identified and analyzed . the behaviour of the transmission with varying energy and scattering lengths is discussed in detail . the dual cir leading to a complete quantum suppression of atomic scattering is revealed in multi - channel scattering processes . possible applications include , e.g. , cold and ultracold atom - atom collisions in atomic waveguides and electron - impurity scattering in quantum wires .

0802.3682

during the last years , the field of ultracold few - body confined systems has progressed remarkably . by employing optical dipole traps @xcite and atom chips @xcite it is possible to fabricate mesoscopic structures in which the atoms are freezed to occupy a single or a few lowest quantum states of a confining potential such that in one or more dimensions the characteristic length possesses the order of the atomic debroglie wavelength . these configurations can be well described by effective one - dimensional systems . well - known examples are quantum wires and atom waveguides or quasi two - dimensional systems such as 2d electronic gases . the quantum dynamics of such systems is strongly influenced by the geometry of the confinement . to control the coherent propagation of particle beams through 1d or 2d traps it is crucial to very well understand the impact of the confinement on the collisional properties ( see refs . @xcite and refs.therein ) . free - space scattering theory is no longer valid in such systems and a new theory is needed . this stimulated the development of quantum scattering theory in low dimensions . for a sufficiently dilute gas under a strong transverse confinement , one may expect both the chemical potential and the thermal energy @xmath1 being less than the transverse level spacing . with this assumption , the dynamics of ultracold atoms in low dimensional structures e.g. tight wave guides has been studied using the simplification that the atoms occupy only the ground state of the transverse confining potential . nevertheless , the virtual transverse excitations in the course of the collision process can play a crucial role in the scattering leading to the so - called confinement - induced resonances ( cirs ) , predicted by olshanii @xcite . s - wave zero - energy scattering of bosons is mapped to an effective longitudinal zero - range pseudopotential @xmath2 approximating the 1d atom - atom interaction in a transverse harmonic confinement . it was shown@xcite that cirs appear if the binding energy of the pseudopotential , approximating the two - atom molecular state in the presence of the confinement , coincides with the energy spacing between the levels of the transverse harmonic potential . in the vicinity of the cir the coupling constant @xmath3 can be tuned from @xmath4 to @xmath5 by varying the strength of the confining potential over a small range . this can result in a total atom - atom reflection , thereby creating a gas of impenetrable bosons . cirs have also been studied for the three - body @xcite and the four - body @xcite scattering under confining potential , as well as for a pure p - wave scattering of fermions @xcite . experimental evidence for the cirs for bosons @xcite and fermions @xcite has recently been reported . a general analytical treatment of low - energy scattering under action of a general cylindrical confinement involving all partial waves and their coupling , was first provided in ref.@xcite for a spherically symmetric short - range potential . the effect of the c.m . motion on the s - wave collision of two distinguishable atoms ( i.e. a two - species mixture ) in a harmonic confinement as well as for two identical atoms in a non - parabolic confining potential has been investigated in ref.@xcite in the zero - energy limit neglecting the s and p wave mixing . a detailed study including the effect of the c.m . nonseparability and taking into account the s and p wave mixing for harmonic confinement was performed in the single - mode regime in ref . recently a so - called dual - cir was discovered @xcite , which is characterized by a complete transmission ( suppression of quantum scattering ) in the waveguide due to destructive interference of s and p waves although the corresponding collisions in free space involve strong interactions . the problem of atomic pair collisions under the action of a harmonic trap in the multi - mode regime when the energy of the atoms exceeds the level spacing of the transverse trapping potential is much more intricate than the single - mode regime due to several open transverse channels . it demands the development of a multi - channel scattering theory accounting for the possible transitions between the levels of the confining potential . using as a starting point the formalism for scattering in restricted geometries suggested in refs . @xcite , the multi - channel scattering problem for bosons in a harmonic confining potential has been analyzed analytically by olshanii el al . @xcite in the s - wave pseudopotential approximation and the zero - energy limit . without detailization of the interatomic interaction but using only two input parameters - the s - wave two - body scattering length in free space and the trap frequency - they have derived an approximate formula for the scattering amplitude describing two boson collisions confined by transverse harmonic trap in the multi - mode regime . in the present work we develop a general grid method for multi - channel scattering of identical as well as distinguishable atoms confined by a transverse harmonic trap . the method applies to arbitrary atomic interactions , permitting a rich spectral structure where several different partial waves are participating in the scattering process or even the case of an anisotropy of the interaction . the only limitation is that we consider harmonic traps with a single frequency for every atom causing a separation of the c.m . and relative motion . with our approach we analyze transverse excitations / deexcitations in the course of the collisional process ( distinguishable or identical atoms ) including all important partial waves and their couplings due to the broken spherical symmetry . special attention is paid to the analysis of the cirs in the multi - mode regimes for non - zero collision energies , i.e. , to suggest a non - trivial extension of the cirs theory developed so far only for the single - mode regime and zero - energy limit . in detail we proceed as follows . section ii contains the derivation of the hamiltonian , definitions of the interatomic interaction and the scattering asymptotics in the confined geometry , and a discussion of the scattering parameters and the transition probabilities characterizing the two - body collisions in the trap . our computational method is outlined in sec . iii , important technical details are given in the appendix . in section iv our results are presented and analyzed . a summary and conclusions are given in sec . let us consider collisions of two atoms under the action of the transverse harmonic confinement . we address both cases , distinguishable and indistinguishable atoms under the action of the same confining potential , i.e. the trap is characterized by a single frequency @xmath0 for every atom . the corresponding hamiltonian is given by @xmath6 where @xmath7 is the mass of the @xmath8 atom , @xmath9 is the two - body potential describing the interaction between two colliding atoms in free space and @xmath10 are the coordinates of the @xmath8 atom . the hamiltonian is separable with respect to the relative @xmath11 and c.m . @xmath12 variables @xmath13 where @xmath14 and @xmath15 here @xmath16 and @xmath17 are the total and reduced masses respectively . the problem , thus , reduces to scattering of a single effective particle with the reduced mass @xmath18 and collision energy @xmath19 , off a scatterer @xmath20 at the origin , under transverse harmonic confinement with frequency @xmath0 @xmath21 \psi ( \mathbf{r } ) = \epsilon\psi(\mathbf{r})\,\,.\ ] ] here the energy of the relative two - body motion @xmath22 is a sum of the transverse @xmath23 and longitudinal collision @xmath24 energies . due to our definition of the confining potential , the transverse excitation energies @xmath23 can take the possible values @xmath25 of the discrete spectrum of the 2d oscillator @xmath26 . for the two - body interaction we choose a screened coulomb potential @xmath27 already employed in ref.@xcite for analyzing the ultracold scattering in cylindrical waveguides in the single - mode regime as @xmath28 . the chosen potential ( [ potential ] ) depends on two parameters - the depth @xmath29 and the screening length @xmath30 . following the computational scheme already developed in ref.@xcite , we implement different spectral structures of the atomic interaction by varying the single parameter @xmath31 for a fixed length scale @xmath32 . obvious advantages of the screened coulomb potential ( [ potential ] ) compared to the s - wave pseudopotential used in ref . @xcite that is devoted to the multi - channel scattering of bosons are the following . first , by varying @xmath31 one can vary the number of bound states of s - wave character in the interaction potential , and second , one can create new bound and resonant states of higher partial wave character . as a consequence we can consider not only bosonic but also fermionic as well as mixed collisions in a trap and including the case of higher energies . this will , as we shall see below , permit us to investigate new regimes and effects of multi - channel confined scattering . performing the scale transformation @xmath33 with the units @xmath34 , @xmath35 , and @xmath36 , it is convenient to rewrite the equation ( [ poth ] ) in the rescaled form @xmath37\psi ( \mathbf{r } ) = \epsilon\psi(\mathbf{r})\,\,,\ ] ] where @xmath38 is now the correspondingly scaled potential and we fix @xmath39 and @xmath40 in the subsequent consideration . in the asymptotic region , @xmath41 , where the transverse trapping potential dominates the interaction potential , the axial and transverse motions decouple and the asymptotic wavefunction can be written as a product of the longitudinal and transverse @xmath42 components with @xmath43^{-1/2}(\frac{\rho}{a_{\bot}})^{|m|}e^{-\rho^2/(2a^2_{\bot } ) } l^{|m|}_{n}(\rho^2/a^2_{\bot})e^{im\varphi}\ ] ] being the eigenfunctions of the transverse trapping hamiltonian @xmath44 with the corresponding eigenvalues @xmath45 and the angular momentum projection @xmath46 onto the @xmath47-axis . here @xmath48 are the generalized laguerre polynomials and @xmath49 , is the transverse oscillator length . the radial and azimuthal quantum numbers @xmath50 and @xmath46 of the 2d harmonic oscillator independently take the values @xmath51 . due to the axial symmetry of the transverse confinement and the spherical symmetry of the interatomic interaction ( [ potential ] ) the angular momentum component along the @xmath47-axis is conserved . therefore , the problem ( [ rescaled scheroedinger equation ] ) is reducible to a 2d one by separating the @xmath52-variable and can be solved for every @xmath46 independently . the quantum number @xmath50 is a good one only in the asymptotic region @xmath41 and used for the definition of the initial ( incident ) asymptotic state i.e. channel @xmath53 of two infinitely separated atoms confined in a transverse state @xmath54 . in addition to the quantum numbers @xmath50 and @xmath46 the asymptotic scattering state is defined also by the momentum @xmath55 of the channel @xmath56 which we express through the dimensionless energy @xmath57 . it is clear that the integer part of the dimensionless energy @xmath58 coincides with the number @xmath59 of open excited transverse channels . for @xmath60 we have @xmath61 . the spherically symmetric interatomic interaction ( [ potential ] ) mixes different transversal channels and leads to transitions @xmath62 between the open channels @xmath63 , i.e. to transverse excitation / deexcitation processes during the collisions . assuming the system to be initially in the channel @xmath50 , the asymptotic wavefunction takes at @xmath64 the form @xcite @xmath65e^{ik_{n'}|z|}\phi_{n',m}(\rho , \varphi),\end{aligned}\ ] ] where @xmath66 and @xmath67 are the matrix elements of the inelastic scattering amplitudes for the even and odd partial waves , respectively , which describe transitions between the channels @xmath50 and @xmath68 . for a bosonic ( fermionic ) collision just the symmetric ( antisymmetric ) part of ( [ full asymptotic wave ] ) should be considered . it is clear that the scattering amplitude depends also on the index @xmath46 which , however , remains unchanged during the collision due to the axial symmetry of the problem . hereafter we consider only the case @xmath69 and the index is omitted in the following . using the asymptotic wavefunction ( [ full asymptotic wave ] ) and the total current conservation one obtains @xmath70 for the inelastic transmission ( reflection ) coefficients @xmath71 ( @xmath72 ) . @xmath55 ( @xmath73 ) is the initial ( final ) relative wave vector and @xmath50 ( @xmath68 ) the transverse excitation numbers according to eq.([wave vector k ] ) . we have @xcite @xmath74\frac{k_{n'}}{k_{n}}|\delta_{n , n'}+f^e_{nn'}+f^o_{nn'}|^2\ ] ] @xmath75\frac{k_{n'}}{k_{n}}|f^e_{nn'}+f^o_{nn'}|^2\,\,,\ ] ] where @xmath76 is the heavyside step - function . the transition probability @xmath77 , characterizing the transverse excitation / deexcitation , into a particular channel @xmath68 from the initial state @xmath50 is given by the sum of the corresponding transmission and reflection coefficients @xmath78 due to the time reversal symmetry of the hamiltonian we have @xmath79 , @xmath80 and @xmath81 . the total transmission ( reflection ) coefficient @xmath82 ( @xmath83 ) is given by the sum of the transmission ( reflection ) coefficients of all the open channels . eq.([current conservation ] ) leads to @xmath84 . to obtain the observable quantities @xmath85 and @xmath77 of the scattering process , we have to calculate the matrix elements @xmath86 of the scattering amplitude @xmath87 by matching the numerical solution of the schrdinger equation ( [ rescaled scheroedinger equation ] ) with the scattering asymptotics ( [ full asymptotic wave ] ) . to integrate this multi - channel scattering problem in two dimensions @xmath88 and @xmath89 ( @xmath47 and @xmath90 ) we adopt the discrete - variable method suggested in ref.@xcite for solving nonseparable 2d scattering problems . this approach was applied in ref.@xcite to the case of a 3d anisotropic scattering problem of ultracold atoms in external laser fields . first , we discretize the 2d schrdinger equation ( [ rescaled scheroedinger equation ] ) on a 2d grid of angular @xmath91 and radial @xmath92 variables . the angular grid points @xmath93 are defined as the zeroes of the legendre polynomial @xmath94 of the order @xmath95 . using the completeness property of the normalized legendre polynomials which remains valid also on the chosen angular grid @xmath96 where @xmath97 are the weights of the gauss quadrature , we expand the solution of equation ( [ rescaled scheroedinger equation ] ) in the basis @xmath98 according to @xmath99 here @xmath100 is the inverse of the @xmath101 matrix @xmath102 with the matrix elements defined as @xmath103 . due to this definition one can use the completeness relation ( [ completeness ] ) in order to determine the matrix elements @xmath104 as @xmath105 . it is clear from ( [ grid expansion of wave function ] ) that the unknown coefficients @xmath106 in the expansion are the values @xmath107 of the two - dimensional wave function @xmath108 at the grid points @xmath93 multiplied by @xmath109 . near the origin @xmath110 we have @xmath111 due to the definition ( [ grid expansion of wave function ] ) and the demand for the probability distribution @xmath112 to be bounded . substituting ( [ grid expansion of wave function ] ) into ( [ rescaled scheroedinger equation ] ) results a system of @xmath95 schrdinger - like coupled equations with respect to the @xmath95-dimensional unknown vector @xmath113 @xmath114\mathbf{u}(r)=0,\ ] ] where @xmath115 @xmath116 and @xmath117 is the unit matrix . we solve the system of equations ( [ sheroedinger - like equation ] ) on the quasi - uniform radial grid @xcite @xmath118 of @xmath119 grid points @xmath120 defined by mapping @xmath121 $ ] onto the uniform grid @xmath122 $ ] with the equidistant distribution @xmath123 . by varying @xmath119 and the parameter @xmath124 one can choose more adequate distributions of the grid points for specific interatomic and confining potentials . by mapping the initial variable @xmath88 in eq.([sheroedinger - like equation ] ) onto @xmath125 we obtain @xmath126\mathbf{u}(r(x))=0\,\,,\ ] ] with @xmath127 where @xmath128 the uniform grid with respect to @xmath125 gives 6-order accuracy for applying a 7-point finite - difference approximation of the derivatives in the equation ( [ sheroedinger - like equation2 ] ) . thus , after the finite - difference approximation the initial 2d schrdinger equation ( [ rescaled scheroedinger equation ] ) is reduced to the system of n algebraic matrix equations @xmath129\mathbf{u}_j+\sum_{p=1}^3\mathbb{a}^j_{j+p}\mathbf{u}_{j+p}=0 \,\,,\,\ , ,n-3\nonumber\\ \mathbf{u}_j+\alpha_j^{(1)}\mathbf{u}_{j-1}+\alpha_j^{(2)}\mathbf{u}_{j-2}+ \alpha_j^{(3)}\mathbf{u}_{j-3}+\alpha_j^{(4)}\mathbf{u}_{j-4}=\mathbf{g}_j\quad\quad j = n-2 , n-1 , n\end{aligned}\ ] ] where each coefficient @xmath130 is a @xmath131 matrix , each @xmath132 is a diagonal @xmath131 matrix and each @xmath133 is a @xmath134-dimensional vector . here the functions @xmath135 , @xmath136 , @xmath137 and @xmath138 in the first three equations of the system ( for @xmath139 and 3 ) are eliminated by using the `` left - side '' boundary conditions : @xmath140 and @xmath141 ( @xmath142 ) . the last three equations in this system for @xmath143 and @xmath144 are the `` right - side '' boundary conditions approximating at the edge points @xmath145 and @xmath146 of the radial grid , the scattering asymptotics ( [ full asymptotic wave ] ) for the desired wave function @xmath147 . in order to construct the `` right - side '' boundary conditions ( [ lu ] ) at @xmath148 and @xmath119 we used an idea of ref.@xcite i.e. the asymptotic behaviour ( [ full asymptotic wave ] ) at the edge points @xmath145 and @xmath146 are considered as a system of vector equations with respect to the unknown vector @xmath86 of the scattering amplitude for a fixed @xmath50 . by eliminating the unknowns @xmath86 from this system we implement the `` right - side '' boundary conditions defined by eqs.([lu ] ) at @xmath148 and @xmath119 ( see appendix a ) . the reduction of the 2d multi - channel scattering problem to the finite - difference boundary value problem ( [ lu ] ) permits one to apply efficient computational methods . here we use , in the spirit of the @xmath149-decomposition @xcite , and the sweep method @xcite ( or the thomas algorithm @xcite ) , a fast implicit matrix algorithm which is briefly described in appendix b. the block - diagonal structure of the matrix of the coefficients in the system of equations([lu ] ) with the width of the diagonal band equal to @xmath150 makes this computational scheme an efficient one . solving the problem ( [ lu ] ) for the defined initial vector @xmath55 and a fixed @xmath50 from the possible set @xmath151 we first calculate the vector function @xmath152 . then , by matching the calculated vector @xmath153 with the asymptotic behaviour ( [ full asymptotic wave ] ) at @xmath154 , we calculate the @xmath50-th row of the scattering amplitude matrix @xmath86 describing all possible transitions @xmath155 . this procedure is repeated for the next @xmath50 from @xmath151 . after calculating all the elements @xmath86 of the scattering amplitude we obtain any desired scattering parameter @xmath156,@xmath157 or @xmath158 . with the above - described method being implemented we have analyzed the two - body scattering under the transverse harmonic confinement for both cases of identical and distinguishable colliding atoms . for confined scattering of identical atoms one has to distinguish the bosonic and fermionic cases . in the case of two colliding bosons the two - body wave function must be symmetric and only even scattering amplitude provides us with a nonzero contribution . first , we show that our result for the special case @xmath159 of a single - channel scattering is in agreement at @xmath160 with the s - wave pseudopotential approach @xcite , and , particularly , reproduces s - wave cir predicted and analyzed in refs . @xcite for bosons . then we extend our consideration to the multi - channel scattering @xmath161 . we demonstrate that our results are in a good agreement in the limit of a long - wavelength trap @xmath162 with the analytical expression given in @xcite which has been obtained in the s - wave pseudopotential approach for the zero - energy limit . the range of validity of the analytical investigation in ref . @xcite is explored . next we present results for multi - channel scattering of two fermions under transverse harmonic confinement . for a fermionic collision the two - body wave function is antisymmetric , i.e. only the odd scattering amplitude is nonzero . in the special case of single - channel scattering we reproduce the p - wave cir for fermions @xcite . these results are also in agreement with our previous investigations in refs.@xcite performed within a wave - packet propagation method @xcite . finally we consider the confined multi - channel scattering of two distinguishable atoms . in this case both even and odd amplitudes contribute to the scattering process . @xmath163{1.jpg}\\ \end{array}$ ] for modeling different interatomic interactions in the subsequent sections we vary the depth @xmath31 of the potential ( [ potential ] ) in the wide range @xmath164 for a fixed width @xmath40 . in free space , this potential being superimposed with the centrifugal term @xmath165 makes an effective potential which may support some _ even _ or _ odd _ bound states depending on the value of the quantum number @xmath166 and the parameter @xmath31 . for @xmath167 there might be also some _ shape resonances _ for certain relative energies . these _ shape resonances _ may enhance strongly the contribution of @xmath167 partial waves in the energy domain where one would have expected a pure @xmath168 scattering . it is known that in the zero - energy limit , when the scattering process does not depend on the details of the potential , the collision can be described by a single parameter : the s - wave scattering length @xmath169 for @xmath168 ( bosonic collision ) and the p - wave scattering volume @xmath170 for @xmath171 ( fermionic collision ) . for sufficiently low collision energy the contribution of the partial waves with larger @xmath166 can be neglected . in fig.[fig1 ] we have plotted @xmath172 and @xmath173 in the region @xmath164 . fig.[fig1 ] demonstrates the rich spectral structure of the chosen form of the interatomic interaction : the scattering parameters @xmath172 and @xmath173 can be positive or negative and they diverge for the values of @xmath31 corresponding to the appearance of new bound states . in the case of p - wave scattering the increase of the depth @xmath31 of the potential first leads to a shape resonance which approaches zero energy and finally transforms to a p - wave bound state . for two bosons colliding in a transverse harmonic confinement the scattering wave function is symmetric with respect to the exchange @xmath174 , i.e. @xmath175 in eq.([full asymptotic wave ] ) . we consider multi - channel scattering with the dimensionless energy @xmath176 ( [ wave vector k ] ) permitting the collisional transverse excitations / deexcitations @xmath177 up to four open channels ( four - mode regime ) . for comparison with analytical results @xcite obtained in the s - wave pseudopotential approach , we have extracted the effective quasi-1d coupling constant @xmath178 as well as the transmission coefficient @xmath179([transmission ] ) and the scattering amplitude @xmath180([full asymptotic wave ] ) , as a function of the scattering length @xmath172 in the single - mode regime ( @xmath181 ) . the calculated parameters are presented along with the analytical results in fig.[fig2 ] for @xmath182 and the longitudinal relative energy @xmath183 . fig.[fig2](a ) shows the coupling constant @xmath3 as a function of the scattering length @xmath172 . our numerical result clearly exhibit a singularity at @xmath184 with @xmath185 , which corresponds to the well - known s - wave cir @xcite . in fig.[fig2](b ) we present the transmission coefficient @xmath156 versus @xmath172 . the transmission coefficient @xmath156 goes to unity ( total transmission ) when @xmath172 tends to zero ( i.e. no interaction between the atoms ) , while at the cir position , it exhibits the well - known minimum ( blocking of the atomic current by the cir ) . fig.[fig2](c ) shows the scattering amplitude @xmath180 as a function of @xmath172 . the amplitude approaches zero at @xmath186 and @xmath187 at the cir position , which results in total transmission and total reflection respectively . in general the presented values of @xmath180 are in very good agreement with the analytical results . @xmath188{2a.jpg } & \includegraphics[height=6cm , width=8cm]{2b.jpg}\\ \includegraphics[height=6cm , width=8cm]{2c.jpg } \\ \end{array}$ ] in the multi - mode regime our results are in good agreement with the analytical ones obtained within the s - wave pseudopotential approach in the zero - energy limit@xcite for long - wavelength traps ( tight confinement : @xmath162 ) . in fig.[fig3 ] we present our scattering amplitude @xmath180 along with the analytical results as a function of the dimensionless energy @xmath58 for @xmath189 and @xmath190 . apart from energies close to the channel thresholds , the real and imaginary part of the scattering amplitude @xmath180 show a monotonous behaviour : @xmath191 is monotonically increasing and approaching zero asymptotically whereas @xmath192 decays monotonically and also approaches zero for large values of @xmath58 . the peak structure located at integer values of @xmath58 is due to the resonant scattering once a new previously closed channel opens with increasing energy . there is a good agreement between our results and the analytical ones given by eq.(6.9 ) in ref . @xcite for the complete range @xmath193 . however , we encounter major deviations with increasing @xmath0 except for narrow regions close to the channel thresholds ( as @xmath58 approaches integer values ) for the real parts of the scattering amplitudes ( see fig.[fig4 ] ) . these deviations are most presumably due to the energy dependence of the s - wave scattering length which is neglected in ref.@xcite . @xmath188{3a.jpg}&\includegraphics[height=6cm , width=8cm]{3b.jpg}\\ \end{array}$ ] @xmath188{4a.jpg } \includegraphics[height=6cm , width=8cm]{4b.jpg}\\ \end{array}$ ] for the multi - channel regime ( @xmath161 ) we find a strong dependence of the total transmission coefficient @xmath156 on the population of the initial state @xmath50 , except for the case @xmath194 of noninteracting bosons in free space . fig.[fig5 ] shows the calculated transmission coefficients([transmission ] ) as a function of @xmath195 for @xmath196 for the ( a ) two - mode regime with @xmath197 and ( b ) the three - mode regime with @xmath198 . note , that all these cases correspond to near - threshold collision energies if the maximal integer is subracted from @xmath58 . similar to the single - mode regime , @xmath156 goes to unity ( total transmission ) when @xmath172 tends to zero . we also encounter a minimum . however , the value of the transmission at the minimum is not zero anymore in the multi - mode regime , the larger the number of open channels , the position of the minima will be more shifted to the left . for a fixed value of the ratio @xmath199 , a lower initially populated transverse level @xmath50 , leads to a larger total transmission . @xmath188{5a.jpg } \includegraphics[height=6cm , width=8cm]{5b.jpg}\\ \end{array}$ ] in fig.[fig6 ] we present the transmission as a function of the dimensionless energy @xmath58 for the two cases of the system being initially in the ground @xmath200 ( a ) or first excited @xmath201 ( b ) transverse states , for different values of the ratio @xmath202 . with increasing collision energy the transmission coefficient exhibits a nonmonotonic behaviour : we observe a sequence of minima and peaks . for @xmath203 the value of @xmath156 at the minimum is zero . as the number of open channels increases with increasing collision energy @xmath204 , the transmission values at the corresponding minima increase strongly . the corresponding transmission peaks @xmath205 are located at the channel thresholds . the shape of the transmission valleys in between two integer values of @xmath58 as well as the positions of the minima strongly depend on the ratio @xmath202 which can be changed by varying the strength of the interatomic interaction @xmath31 or the trap frequency @xmath0 . both , changing @xmath31 and/or @xmath0 leads to energetical shifts of the bound states ( in particular for the excited transversal channels ) of the atoms in the presence of the confinement . if the collision energy coincides with a bound state of the corresponding closed channel we encounter an occupation of the closed channel in the course of the scattering process , i.e. a feshbach - resonance occurs . the interpretation of the minimum of the transmission @xmath156 in terms of a feshbach - resonance at the point @xmath206 for the zero - energy limit was provided in refs . @xcite , where it was shown that the origin of the cir is an intermediate occupation of a bound state belonging to an excited transverse ( closed ) channel . to demonstrate that the above - discussed behaviour ( minima ) of the transmission coefficient @xmath207 in certain regions of @xmath58 is due to feshbach resonances we have analyzed the probability density of the scattering wave function of the atoms in the trap . in the single - mode regime and in the zero - energy limit we encounter the well - known cir : fig.[fig6e ] shows the corresponding probability density @xmath208 for an initial transverse ground state @xmath200 and @xmath209 as well as @xmath210 . for small @xmath211 one observes additional two pronounced peaks along the transverse ( @xmath125- ) direction corresponding to the occupation of the bound state ( with the binding energy @xmath212 ) in the first excited closed channel of the transverse potential . the probability density tends to zero as @xmath213 . this leads to a zero of the transmission @xmath207 for @xmath160 ( see fig.[fig6](a ) ) corresponding to the zero - energy cir . in fig.[fige6 ] we show the probability densities @xmath208 at @xmath214 for several values of the dimensionless energy @xmath58 for ( a ) the single- and ( b ) two - mode regimes for collisions with initial transverse state @xmath200 . the corresponding probability density exhibits for small values of @xmath211 additional two ( for the single - mode regime ) and four ( for the two - mode regime ) pronounced peaks with respect to the transverse ( @xmath125-)direction as @xmath58 approaches the cir - position . this demonstrates the occupation of bound states of higher , namely first excited ( with binding energy @xmath215 ) and second excited ( with binding energy @xmath216 ) channels in the course of the scattering process . the corresponding transmission values are also indicated in fig.[fige6 ] . zero transmission is observed also for @xmath214 and @xmath217 with no probability density being present for large positive values of @xmath47 , see fig.[fige6](a ) . @xmath188{6a.jpg}\includegraphics[height=6cm , width=8cm]{6b.jpg}\\ \end{array}$ ] @xmath188{6ea.jpg}\includegraphics[height=6cm , width=8cm]{6ed.jpg}\\ \includegraphics[height=6cm , width=8cm]{6eb.jpg}\includegraphics[height=6cm , width=8cm]{6ee.jpg}\\ \includegraphics[height=6cm , width=8cm]{6ec.jpg}\includegraphics[height=6cm , width=8cm]{6ef.jpg}\\ \end{array}$ ] the energy of the bound state @xmath215 , that leads to a minimal transmission due to a resonant scattering process , changes with varying @xmath202 as follows . it is below @xmath218 for @xmath219 and consequently no minimum is encountered for @xmath207 in the range @xmath220 ( see fig . [ fig6 ] ) . at the position of the zero - energy cir it is located just above the threshold @xmath221 leading to a zero transmission for @xmath222 . for @xmath223 the bound state energy is somewhere in between the channel thresholds @xmath221 and @xmath224 whereas for @xmath225 it is below but close to @xmath224 leading again to a corresponding minimum of @xmath156 . for both cases , @xmath225 and @xmath223 , an increase of @xmath195 leads to a narrow transmission well and the corresponding transmission minimum is shifted towards the next higher channel threshold . the dependence of @xmath215 on the parameter @xmath202 is in agreement with the pseudopotential analysis given in ref.@xcite ( see fig . 2 of this ref . ) . in fig.[fig7](a ) we show the transition probabilities @xmath77 as a function of @xmath195 for @xmath196 and @xmath226 corresponding to four open channels . we observe that the probability of remaining at the same initial state , @xmath227 ( i.e. elastic scattering ) is in the complete range of the ratio @xmath202 much larger than the probability of a transition into a different state , @xmath77 ( i.e. inelastic scattering ) . with increasing @xmath50 or @xmath68 the inelastic transition probabilities @xmath77 increase but the elastic probabilities @xmath227 decrease . in an inelastic ( elastic ) collision @xmath77 ( @xmath227 ) goes to zero ( unity ) as @xmath172 tends to zero . @xmath77 ( @xmath227 ) possess a maximum ( minimum ) at the resonance position @xmath228 consistent with the minimum of @xmath207 at @xmath226 . it is instructive to see how the distribution of the initial flux among the open channels changes due to pair collisions as a function of the collision energy . fig.[fig7](b ) shows the transition probabilities @xmath77 as a function of the dimensionless energy up to four open channels . the probability of elastic scattering remains larger than that of inelastic scattering in the complete range of the energy . for two open channels the elastic collision probability @xmath227 is independent of the initial state ( @xmath229 ) . for a higher number of open channels @xmath227 is decreasing with increasing initial value of @xmath50 . near the thresholds , the probabilities of the inelastic ( elastic ) transitions @xmath77 ( @xmath227 ) go to zero ( unity ) . @xmath188{7a.jpg}\includegraphics[height=6cm , width=8cm]{7b.jpg}\\ \end{array}$ ] in this section we focus on fermionic collisions in harmonic traps . in this case the interatomic wave function is anti - symmetric with respect to the interchange of the two fermions and the even amplitude @xmath66 in eqs.([full asymptotic wave ] ) is zero . we have analyzed the multi - channel scattering of fermions up to four open transverse channels for different interatomic interactions , by varying the potential strength @xmath31 in the vicinity of the value @xmath230 ( see fig.1 ) generating a resonant p - wave state in free space . fig . [ fig8 ] shows corresponding results for the single - mode regime . [ fig8](a ) shows the _ mapped coupling constant _ @xmath231 @xcite as a function of the p - wave scattering length @xmath232{v_p}$ ] . the _ mapped coupling constant _ @xmath233 goes to zero at the position of the mapped cir @xcite , which for @xmath182 and @xmath234 is equal to @xmath235 . the position of the mapped cir obviously depends on the values of @xmath0 and @xmath58 . in fig.[fig8](b ) we have plotted the transmission coefficient @xmath156 as a function of @xmath236 . the transmission coefficient exhibits a minimum and an accompanying well ( i.e. the blocking due to the resonance ) at the position of the cir , and tends to unity ( i.e. total transmission ) for @xmath236 far from the cir - position . for larger @xmath0 the well becomes wider , and its minimum is shifted to the left , see also ref.@xcite . [ fig8](c ) shows the scattering amplitude @xmath237 as a function of @xmath236 for @xmath238 and @xmath239 . the amplitude approaches zero far from the cir position and @xmath187 at the cir position , which results in total transmission , @xmath240 ( i.e. the fermions do not scatter each other ) and total reflection , @xmath241 ( i.e. strongly interacting and impenetrable fermions ) , respectively . @xmath163{8a.jpg}\\ \includegraphics[height=6cm , width=8cm]{8b.jpg}\\ \includegraphics[height=6cm , width=8cm]{8c.jpg}\\ \end{array}$ ] in fig.[fig9 ] we present the total transmission coefficient as a function of @xmath236 in the two - mode regime for @xmath242 ( a ) and three - mode regime for @xmath243 ( b ) for @xmath196 . similar to the single - mode regime , @xmath156 exhibits a minimum and accompanying well , however the position of the minimum is shifted to the right and the value at the minimum is nonzero . for the lower degree of transversal excitation , we observe a deeper and narrower transmission well . with increasing energy the transmission well becomes wider and more shallow and its position is shifted to larger values of @xmath244 . this is also demonstrated in fig.[fig10 ] , where the transmission coefficient is plotted as a function of @xmath58 for several values of @xmath244 for the two cases of being initially in the ground @xmath200 ( a ) and the first excited @xmath201 ( b ) transverse states for @xmath182 . we observe that for any number of open channels , @xmath156 exhibits a minimum for some value of @xmath244 . with increasing @xmath244 the transmission well becomes more shallow ( except for the single - mode regime ) and wider . in contrast to the bosonic case , there is no specific threshold behaviour . this is a consequence of the fact that the relative motion does not feel the interatomic interaction in the closed channels which are strongly screened by the centrifugal repulsion playing a dominant role for near - threshold collision energies or in other words : we encounter a weak coupling of the different scattering channels . @xmath188{9a.jpg}\includegraphics[height=6cm , width=8cm]{9b.jpg}\\ \end{array}$ ] @xmath188{10a.jpg}\includegraphics[height=6cm , width=8cm]{10b.jpg}\\ \end{array}$ ] in fig.[fig11 ] we show the transition probabilities @xmath77 as a function of @xmath244 for four open channels . we see that the probability of an elastic scattering process ( i.e. to remain in the same transversal state ) is much larger than that of an inelastic collision ( i.e. the transition to a different transversal state ) . for an elastic collision the probability @xmath227 shows a minimum and corresponding well which becomes wider and more shallow with increasing initial quantum number @xmath50 ( i.e. , with the population of a higher excited initial transversal state ) . for an inelastic collision @xmath77 ( @xmath245 ) exhibits a peak which becomes less pronounced as the quantum numbers @xmath50 or @xmath68 increase . @xmath163{11.jpg}\\ \end{array}$ ] in this section we analyze the multi - channel scattering of two distinguishable particles in a harmonic trap with the same trap frequency @xmath246 allowing the separation of the c.m . such a case corresponds to the different atomic species confined by the same potential as it is the case e.g. for two different isotopes in the same optical dipole trap . the two - body scattering wave - function of the distinguishable atoms does not possess a well - defined symmetry with respect to the reflection @xmath174 , i.e. both s- and p - state contributions ( @xmath247 and @xmath248 ) must be taken into account for the scattering amplitude . fig.[fig12 ] shows our results for the transmission coefficient @xmath156 in the single - mode regime ( @xmath249 ) , which is plotted as a function of the tuning parameter @xmath250 of the interparticle interaction . in general the scattering process can not be described by a single scattering length @xmath172 or @xmath236 for this case . in regions with a negligible p - wave contribution ( see the regions in fig.[fig1 ] with @xmath251 ) , @xmath156 exhibits a behavior similar to bosonic scattering . we observe the well - known s - wave cirs which lead to zeros of the transmission @xmath156 at the positions @xmath252 and tend to unity when @xmath172 goes to zero ( together with @xmath236 ) . in regions , where @xmath172 and @xmath236 are comparable we observe the effect reported in ref . @xcite : remarkable peaks of the transmission @xmath253 i.e. almost complete transmission in spite of the strong interatomic interaction in free space . this is the so - called dual cir : quantum suppression of scattering in the presence of confinement due to destructive interference of odd and even scattering amplitudes . equally minima of @xmath254 due to the interference of even and odd scattering amplitudes under the action of the transverse confinement can occur ( see fig.[fig12 ] ) . complete transmission corresponds to @xmath255 while total reflection corresponds to @xmath256 . fig.[fig13 ] shows the corresponding amplitudes @xmath257 as a function of @xmath250 . @xmath163{12.jpg}\\ \end{array}$ ] @xmath163{13.jpg}\\ \end{array}$ ] fig.[fig14 ] shows the total transmission coefficient versus @xmath250 in the three - mode regime for @xmath182 and @xmath243 . similar to the single - mode regime , when @xmath236 is negligible compared to @xmath172 , @xmath156 behaves analogously to the case of a bosonic collision , and tends to unity ( complete transmission ) when @xmath172 tends to zero , while at the s - wave cir position , it exhibits a minimum with a nonzero value . for a lower degree of transversal excitation of the initial state , we encounter a larger transmission coefficient . for the same reasons as in the single - mode regime , we observe in the regions of @xmath31 where the p - wave scattering length @xmath236 is comparable to @xmath172 , sharp peaks of @xmath156 . however in contrast to the single - mode regime we do not observe complete transmission i.e. @xmath258 ( see e.g. @xmath156 at @xmath259 in fig.[fig14 ] ) . @xmath163{14.jpg}\\ \end{array}$ ] fig.[fig15 ] presents the transmission coefficients as a function of the dimensionless energy @xmath58 for @xmath182 . fig.[fig15](a ) shows the results for several ratios of @xmath202 if @xmath260 is small compare to @xmath261 . here the system is initially in the transversal ground state . the behaviour of the transmission coefficient is similiar to the case of two - boson scattering ( see [ fig6]-a ) . fig.[fig15](b ) shows the transmission for the scattering of two distinguishable atoms together with the results for bosonic- ( blue ) and fermionic- ( black ) collisions , for @xmath262 when both scattering length @xmath172 and @xmath236 are large . in the limit @xmath222 the behaviour of the transmission for the distinguishable and the bosonic case are very similar while in the vicinity of the energy of the p - wave shape - resonance @xmath263 , apart from a small shift to larger energies , we can find a complete coincidence of the transmission behaviour for the case of distinguishable and fermionic scattering . at the threshold energy @xmath224 the latter two transmission curves cross . fig.[fig15](c ) shows the transmission coefficient versus @xmath58 for @xmath262 initially occupying the ground state @xmath200 ( blue ) , the first excited state @xmath201 ( black ) , the second excited state @xmath264 ( red ) and the third excited state @xmath265 ( green ) . in the limit @xmath266 , @xmath156 drops rapidly to zero . apart from the single mode regime , it is for @xmath267 approximately constant with @xmath268 . in fig.[fig16 ] we present the transition probabilities @xmath77 as a function of @xmath250 for four open channels . for the regions of @xmath31 where @xmath236 is negligible compared to @xmath172 , a very good agreement of the transmission behaviour for distinguishable and bosonic atoms is observed . in regions where @xmath172 and @xmath236 are comparable , similar to the fermionic case , we observe narrow and deep wells ( for elastic collision ) and narrow as well as strongly pronounced peaks ( for inelastic collision ) . @xmath163{15a.jpg}\\\includegraphics[height=6cm , width=8cm]{15b.jpg}\\\includegraphics[height=6cm , width=8cm]{15c.jpg}\\ \end{array}$ ] @xmath163{16.jpg}\\ \end{array}$ ] we have analyzed atomic multi - channel scattering in a 2d harmonic confinement . identical bosonic and fermionic scattering as well as scattering of distinguishable atoms in traps with the same frequency for the different species have been explored . equal frequencies allows us to separate the c.m . and relative motion . firstly we reproduced the well - known s - wave cir for bosonic collision in the single - mode regime@xcite . next bosonic collisions in the multi - channel regime including elastic and inelastic processes i.e. transverse excitations and deexcitations have been investigated . transmission coefficients as well as transition probabilities for energies covering up to four open channels are reported on . it is shown that the transmission coefficient as a function of the scattering length exhibits a minimum at the cir position . except for @xmath269 in the first channel , the transmission curves show , with varying energy , a minimum and accompanying well . for a single open channel the value of the minimum of the transmission is zero . increasing the degree of transverse excitation leads to an increase of the transmission minima . a distinct threshold behaviour is observed at every @xmath270 . for @xmath271 the accompanying transmission well is more pronounced compared to the case @xmath272 and it is also deeper in the case of several open channels . the position of the minimum is closer to the upper threshold again when compared to the case @xmath272 . with increasing ratio @xmath202 for either @xmath272 or @xmath271 , the corresponding transmission well becomes also more pronounced i.e. we encounter a deeper and narrower well for more open channels and the energetical position of the minimum moves to the neighboring upper threshold . our results for the transition probabilities @xmath77 show that the probability of remaining in the same state ( elastic collision ) is larger than transitions into a different transversal state ( inelastic collision ) . in the case of elastic collisions ( inelastic collisions ) it has to go to unity ( zero ) as @xmath172 tends to zero . the inelastic transition probability @xmath77 increases as the numbers @xmath50 or @xmath68 increase while the elastic transition probability @xmath77 decreases . our next focus has been the multi - channel scattering of fermions addressing up to four open channels in the waveguide . in the regime of a single open channel we reproduced the well - known p - wave mapped cir in the zero - energy limit @xcite . we have analyzed the p - wave cir dependence on the collision energy @xmath204 and the trapping frequency @xmath0 . consequently the multi - mode regime has been explored by determining the behaviour of the transmission as well as the transition probabilities . the dependence of the transmission on the p - wave scattering length exhibits a minimum with zero transmission and an accompanying well for the single - mode regime . increasing the energy , this well becomes more shallow ( except for the single - mode regime),wider and its position is shifted to the region of higher energies . in the multi - mode regime the transmission well depends also on the degree of the initial transverse excitation . increasing the degree of initial transverse excitation the transmission well becomes shallower and wider and the transmission is overall increased . for a fixed number of open channels the transmission exhibits as a function of the energy a minimum for some value of the p - wave scattering length . increasing @xmath273 leads to a shift of the transmission minimum to higher energies and an increasingly shallower well . this holds except for the single mode regime in which the minimal value is zero . in contrast to the bosonic case , we do not observe a distinct threshold behavior . the transition probabilities @xmath77 for fermionic collisions show that elastic collisions are more probable than inelastic ones . the probability @xmath227 for an elastic process shows a well which becomes wider and shallower as the initial population of the excited states @xmath50 increases , while for an inelastic collision it exhibits a peak which becomes smaller as the channel numbers @xmath50 or @xmath68 increase . by varying the potential parameter @xmath31 in the viccinity of the value @xmath230 , we found a set of shape - resonances . finally we analyzed the multi - channel-1d scattering of distinguishable atoms in the trap . here both s- and p - wave contributions have been taken into account . we have studied the transmission coefficient in the single - mode regime and reproduced the cir @xcite as well as the dual - cir @xcite corresponding to total reflection and transmission , respectively . in the multi - mode regime the transmission versus @xmath250 shows except for the regions where @xmath236 is comparable with @xmath172 a behaviour analogous to the situation of bosonic scattering . we encounter @xmath274 for @xmath275 . at the position of the cir , @xmath156 exhibits a minimum with a nonzero value . the lower is the transverse excitation of the initial state , the larger is the transmission coefficient . in regions of @xmath31 where the p - wave scattering length @xmath236 is comparable to @xmath172 , @xmath156 exhibits sharp peaks or dips near the dual - cirs due to comparable contributions of both s- and p - waves . in contrast to the single - mode regime there is no a complete transmission at the dual cir points . our results for the transmission coefficient versus energy show that in the limit of zero longitudinal relative energy the distinguishable atoms behave like bosons , while in the vicinity of the position of a shape - resonance , they behave like fermions . for larger energies @xmath276 @xmath156 is close to unity . finally we have analyzed the transition probability as a function of @xmath250 . for values of @xmath31 where @xmath236 is negligible compared to @xmath172 we find an excellent agreement with the results for bosonic collisions . in regions where these two scattering parameters are comparable , we observe , similar to the fermionic case , sharp downward- ( for elastic collision ) and upward- ( for inelastic collision ) peaks but with different values . we conclude with the general statement that our multi - channel scattering results in waveguides are of immediate relevance to cold or ultracold atomic collisions in atomic waveguides or impurity scattering in quantum wires . s.s . acknowledges a scholarship by the ministry of science , research and technology of iran . v.s.m . acknowledges financial support by the landesstiftung baden - wrttemberg in the framework of a guest program . financial support by the heisenberg - landau program is also acknowledged . thanks the deutsche forschungsgemeinschaft for financial support . in the limit @xmath88 tends to zero , @xmath277 goes to zero . this boundary condition can be satisfied easily by putting @xmath278 and @xmath279 in the vicinity of @xmath280 . the latter is needed for approximating the derivatives @xmath281 and @xmath282 near the point @xmath280 . the boundary condition for @xmath283 approximating the asymptotic form ( [ full asymptotic wave ] ) at large @xmath88 can be written in the form @xmath284 where @xmath132s are diagonal @xmath131 matrices and @xmath133 is a @xmath134-dimensional vector . the above boundary - conditions are constructed by eliminating the unknown amplitudes @xmath86 from the asymptotic equations ( [ full asymptotic wave ] ) written for a few @xmath285s neighbouring to the point @xmath146 . this gives the values for the coefficients @xmath132s and @xmath133 . in general for up to four open channels we have @xmath286_{mm'}=(-1)^i\left ( t_j^{l-1,m}+t_j^{l , m}\right ) \frac{r_j}{r_{j - l}}\frac{\phi_{n_{e}}(\rho^m_{j})}{\phi_{n_{e}}(\rho^m_{j - l})}e^{ik_{n_{e}}(|z^m_j|-|z^m_{j - l}|)}\delta_{mm ' } , \quad\quad l=1,2,3,4\ ] ] and @xmath287_{m}=2\sqrt{\lambda_m}r_j\phi_{n_{e}}(\rho^m_{j})e^{ik_{n_{e}}|z^m_j|}\sum_{n=0}^{n_e}\sum_{l=0}^4(-1)^ie^{-ik_{n_{e}}|z^m_{j - l}|}e^{ik_nz^m_{j - l } } ( t_j^{l-1,m}+t_j^{l , m})\frac{\phi_n(\rho^m_{j - l})}{\phi_{n_{e}}(\rho^m_{j - l})}\ ] ] for the case of bosonic ( fermionic ) collisions we must consider just the even ( odd ) part of the equation ( [ g ] ) , i.e. we just need to replace the term @xmath288 by @xmath289 ( @xmath290 ) . here @xmath291 is the eigenfunction of the transverse trapping hamiltonian ( see eq.([transvers hamil . eigenfunction ] ) ) , @xmath292 , @xmath293 , @xmath50 is the channel number of the initial state , and @xmath294 is the number of transversely excited open channels . the complex numbers @xmath295s are given through @xmath296 if @xmath297 ( a single - mode regime ) @xmath298 , otherwise @xmath299 if @xmath300 , @xmath301 , otherwise @xmath302 if @xmath303 , @xmath304 , otherwise @xmath305 there @xmath306 . @xmath55 and @xmath307 are given by eq.([wave vector k ] ) . following the idea of the lu - decomposition @xcite and the sweep method @xcite ( or the thomas algorithm @xcite ) , we search the solution of the system of @xmath119 vector equations with any coefficient being a @xmath131 matrix ( [ lu ] ) in the form @xmath308 here we define @xmath309 as @xmath310 for simplicity . the @xmath311s are unknown @xmath131 matrices . to find the solution , first we should calculate the unknown @xmath311 matrices . the plan is the following : due to eq.([vector solution ] ) we have @xmath312 then one obtains @xmath313 @xmath314\mathbf{u}_{j}+\left [ c^{(1)}_{j-2}c^{(2)}_{j-1}+c^{(3)}_{j-2}\right ] \mathbf{u}_{j+1 } + c^{(1)}_{j-2}c^{(3)}_{j-1}\mathbf{u}_{j+2}\ ] ] and @xmath315\mathbf{u}_{j}\nonumber\\ { } + \left [ c^{(1)}_{j-3}c^{(1)}_{j-2}c^{(2)}_{j-1}+c^{(1)}_{j-3}c^{(3)}_{j-2}+c^{(2)}_{j-2}c^{(2)}_{j-1}\right ] \mathbf{u}_{j+1 } \nonumber\\ { } + \left [ c^{(1)}_{j-3}c^{(1)}_{j-2}c^{(3)}_{j-1}+c^{(2)}_{j-2}c^{(3)}_{j-1}\right ] \mathbf{u}_{j+2}\end{aligned}\ ] ] by substituting @xmath316 defined by eqs.([uj-1]-[uj-3 ] ) into eq.([lu ] ) one can calculate @xmath310 in terms of @xmath317 , @xmath318 and @xmath319 . then , by comparing with eq.([vector solution ] ) we find a recurrence formula for calculating the unknown matrices @xmath311 : @xmath320\nonumber\\ { } + \mathbb{a}^j_{j-2}\left [ c^{(1)}_{j-2}c^{(2)}_{j-1}+c^{(3)}_{j-2}\right ] + \mathbb{a}^j_{j-1}c^{(2)}_{j-1}+\mathbb{a}^j_{j+1}\end{aligned}\ ] ] @xmath321\nonumber\\ { } + \mathbb{a}^j_{j-2}c^{(1)}_{j-2}c^{(3)}_{j-1 } + \mathbb{a}^j_{j-1}c^{(3)}_{j-1}+\mathbb{a}^j_{j+2}\end{aligned}\ ] ] and @xmath322 here @xmath323\nonumber\\ { } + \mathbb{a}^j_{j-2}\left [ c^{(1)}_{j-2}c^{(1)}_{j-1}+c^{(2)}_{j-2}\right ] + \mathbb{a}^j_{j-1}c^{(1)}_{j-1}+\mathbb{a}^j_{j}+2(\epsilon i -v_j)\end{aligned}\ ] ] by using the left - side boundary conditions and eq.([vector solution ] ) one can calculate the @xmath311 matrices for @xmath324 and @xmath325 . then by using eqs.([c4]-[c1 ] ) we calculate all the matrices @xmath311 . subsequently by using the right - side boundary conditions ( [ right - side boundary condition ] ) and recurrence formula ( [ vector solution ] ) we first calculate @xmath310 for @xmath326 and @xmath119 and then @xmath310 for @xmath327 . 40 r. grimm , m. weidemller , and y.b . ovchinnikov , adv . at . * 42 * , 95 ( 2000 ) . r. folman _ et al._,adv . at . . phys . * 42 * , 95 ( 2000 ) . j. reichel , appl . b : laser opt . * 74 * , 469 ( 2002 ) . j. fortagh _ et al . * 79 * , 235 ( 2007 ) . bolda , e. tiesinga and p.s . julienne , phys.rev.a * 66 * , 013403 ( 2002 ) . r. stock , i.h . deutsch and e.l . bolda , phys.rev.lett.*91 * , 183201 ( 2003 ) . yurovsky , phys.rev.a*71 * , 012709 ( 2005 ) . r. stock and i.h . deutsch , phys.rev.a * 73 * , 32701 ( 2006 ) . v.a . yurovsky and y.h . band , phys.rev.a * 75 * , 012717 ( 2007 ) . p. naidon _ et al _ , new j. phys . * 9 * , 19 ( 2007 ) . bhongale , s.j.j.m.f . kokkelmans and i.h . deutsch , arxiv:0712.2070v1 , physics : atom - ph . yurovsky , m. olshanii and d.s . weiss , adv.at.mol.opt.phys . * 55 * , 61 ( 2007 ) . m. olshanii , phys . * 81 * , 938 ( 1998 ) . t. bergeman , m.g . moore , and m. olshanii , phys . 91 * , 163201 ( 2003 ) . c. mora , r. egger , a.o . gogolin , and a. komnik , phys . . lett . * 93 * , 170403 ( 2004 ) . c. mora , r. egger , and a.o . gogolin , phys . rev . * a71 * 052705 ( 2005 ) . c. mora , a. komnik , r. egger , and a.o . gogolin , 2005 preprint cond - mat/0501641 . b.e . granger and d. blume , . lett . * 92 * , 133202 ( 2004 ) . t. kinoshita , t. wenger and d.s . weiss , science * 305 * , 1125 ( 2004 ) . b. paredes et al , nature * 429 * , 277 ( 2004 ) . k. gnter et al , phys . 95 * , 230401 ( 2005 ) . j.i . kim , j. schmiedmayer , and p. schmelcher , phys . rev . * a72 * , 042711 ( 2005 ) . v. peano , m. thorwart , c. mora , and r. egger , new j. phys . * 7 * , 1 ( 2005 ) . kim , v.s . melezhik , and p. schmelcher , phys . * 97 * , 193203 ( 2006 ) . kim , v.s . melezhik , and p. schmelcher , rep . . supp . * 166 * , 159 ( 2007 ) . melezhik , j.i . kim , and p. schmelcher , phys . rev . * a76 * , 053611 ( 2007 ) . a. lupu - sax , quantum scattering theory and applications , phd thesis , harvard university ( 1998 ) . m.g.e . da luz , a.s . lupu - sax and e.j . heller , phys.rev.e * 56 * , 2496 ( 1997 ) m.g . moore , t. bergeman and m. olshanii , j. phys . iv * 116 * , 69 ( 2004 ) , lecture courses les houches school on quantum gases in low dimensions ( 2003 ) . melezhik , j. comput . phys . * 92 * , 67 ( 1991 ) melezhik and chi - yu hu , phys . lett . * 90 * , 083202 ( 2003 ) . melezhik , phys . a230 * , 203 ( 1997 ) . w.h . press , s.a . teukolsky , w.t . vetterling , and b.p . flannery , _ numerical recipes _ , cambridge university press , 1992 . gelfand and s.v . fomin,_calculus of variations _ , dover publications , 2000 . bruse et al . , petrol . aime * 198 * , 79 ( 1953 ) .

a quantum compiling algorithm is an algorithm for decomposing ( compiling " ) an arbitrary unitary matrix into a sequence of elementary operations ( seo ) . suppose is an-bit unstructured unitary matrix ( a unitary matrix with no special symmetries ) that we wish to compile . for , expressing as a seo requires more than a million cnots . this calls for a method for finding a unitary matrix that : ( 1)approximates well , and ( 2 is expressible with fewer cnots than . the purpose of this paper is to propose one such approximation method . various quantum compiling algorithms have been proposed in the literature that decompose into a sequence of-multiplexors , each of which is then decomposed into a seo . our strategy for approximating is to approximate these intermediate-multiplexors . in this paper , we will show how one can approximate a-multiplexor by another-multiplexor that is expressible with fewer cnots .

quant-ph0412072

in quantum computing , elementary operations are operations that act on only a few ( usually one or two ) qubits . for example , cnots and one - qubit rotations are elementary operations . a quantum compiling algorithm is an algorithm for decomposing ( compiling " ) an arbitrary unitary matrix into a sequence of elementary operations ( seo ) . a quantum compiler is a software program that implements a quantum compiling algorithm . one measure of the inefficiency of a quantum compiler is the number of cnots it uses to express an unstructured unitary matrix ( a unitary matrix with no special symmetries ) . we will henceforth refer to this number as @xmath4 . although good quantum compilers will also require optimizations that deal with structured matrices , unstructured matrices are certainly an important case worthy of attention . minimizing the number of cnots is a reasonable goal , since a cnot operation ( or any 2-qubit interaction used as a cnot surrogate ) is expected to take more time to perform and to introduce more environmental noise into the quantum computer than a one - qubit rotation . ref.@xcite proved that for matrices of dimension @xmath5 ( @xmath6 number of bits ) , @xmath7 . this lower bound is achieved for @xmath8 by the 3 cnot circuits first proposed in ref.@xcite . it is not known whether this bound can be achieved for @xmath9 . [ cols=">,>,>",options="header " , ] suppose @xmath0 is an @xmath1-bit unstructured unitary matrix that we wish to compile . as the above table illustrates , compiling @xmath0 is hopeless for @xmath2 unless we approximate @xmath0 . we need a method for finding a unitary matrix that : ( 1 ) approximates @xmath0 well , and ( 2 ) is expressible with fewer cnots than @xmath0 . the purpose of this paper is to propose one such approximation method . the use of approximations in quantum compiling dates back to the earliest papers in the field . for example , refs.@xcite and @xcite contain discussions on this issue . as ref.@xcite points out , even when compiling a highly structured matrix like the discrete fourier transform matrix , some gates that contribute negligibly to its exact seo representation can be omitted with impunity . refs.@xcite@xcite and @xcite discuss a quantum compiling algorithm that decomposes an arbitrary unitary matrix into a sequence of @xmath3-multiplexors , each of which is then decomposed into a seo . other workers have proposed@xcite@xcite alternative quantum compiling algorithms that also generate @xmath3-multiplexors as an intermediate step . the strategy proposed in this paper for approximating @xmath0 is to approximate the intermediate @xmath3-multiplexors whose product equals @xmath0 . in this paper , we will show how one can approximate a @xmath3-multiplexor by another @xmath3-multiplexor ( the approximant " ) that has fewer controls , and , therefore , is expressible with fewer cnots . we will call the reduction in the number of control bits the * bit deficit * @xmath10 . fig.[fig - approx ] is emblematic of our approach . it shows a @xmath3-multiplexor with 3 controls being approximated by either a @xmath3-multiplexor with 2 controls or one with 1 control . in this section , we will define some notation that is used throughout this paper . for additional information about our notation , we recommend that the reader consult ref.@xcite . ref.@xcite is a review article , written by the author of this paper , which uses the same notation as this paper . let @xmath11 . as usual , let @xmath12 represent the set of integers ( negative and non - negative ) , real numbers , and complex numbers , respectively . for integers @xmath13 , @xmath14 such that @xmath15 , let @xmath16 . for @xmath17 equal to @xmath18 or @xmath19 , let @xmath20 and @xmath21 represent the set of positive and non - negative @xmath17 numbers , respectively . for any positive integer @xmath22 and any set @xmath23 , let @xmath24 denote the cartesian product of @xmath22 copies of @xmath23 ; i.e. , the set of all @xmath22-tuples of elements of @xmath23 . for any ( not necessarily distinct ) objects @xmath25 , let @xmath26 denote an ordered set . for some object @xmath14 , let @xmath27 . let @xmath28 be the empty set . for an ordered set @xmath23 , let @xmath29 be @xmath23 in reverse order . we will use @xmath30 to represent the truth function " ; @xmath30 equals 1 if statement @xmath23 is true and 0 if @xmath23 is false . for example , the kronecker delta function is defined by @xmath31 . for @xmath32 , _ k=0 ^ 1(-1)^kx=2 ( x,0 ) . [ eq - bool - delta - fun ] for any positive integer @xmath33 , we will use @xmath34 where @xmath35 to denote the standard basis vectors in @xmath33 dimensions ; i.e. , @xmath36_j = \delta(i , j)$ ] for @xmath37 . @xmath38 and @xmath39 will represent the @xmath22-dimensional unit and zero matrices . for any matrix @xmath40 and positive integer @xmath22 , let a^n = _ n a , a^n = _ n a . for any matrix @xmath41 and @xmath42 , @xmath43 will represent the @xmath44-norm of @xmath40 , and @xmath45 its frobenius norm . see @xcite for a discussion of matrix norms . let @xmath46 . as is customary in the physics literature , @xmath47 will also be denoted by @xmath48 and called the magnitude of @xmath49 . for any complex matrix @xmath40 , we will use @xmath50 to denote the matrix that is obtained from @xmath40 by replacing each of its entries by its absolute value . in other words , @xmath51_{ij}= abs(a_{ij } ) = |a_{ij}|$ ] . ( careful : ref.@xcite and many other mathematical books call @xmath52 what we call @xmath50 ) . any @xmath53 can be expressed as a doubly infinite power series in powers of a base @xmath54 : @xmath55 . this expansion can be represented by : @xmath56 , which is called the base @xmath57 representation of @xmath58 . the plus or minus in these expressions is chosen to agree with the sign of @xmath58 . it is customary to omit the subscript @xmath59 when @xmath60 . for example @xmath61 suppose @xmath62 . note that division by 2 shifts the binary representation of @xmath58 one space to the right : @xmath63 . likewise , multiplication by 2 shifts the binary representation of @xmath58 one space to the left : @xmath64 . in general , for any @xmath65 , @xmath66 . define the action of an overline placed over an @xmath67 by @xmath68 @xmath69 . call this bit negation . define the action of an oplus placed between @xmath70 by @xmath71 . call this bit addition . one can extend the bit negation and bit addition operations so that they can act on non - negative reals . suppose @xmath72 , and @xmath73 are non - negative real numbers . then define the action of an overline over @xmath58 so that it acts on each bit individually ; i.e. , so that @xmath74_\alpha = \overline{x_\alpha}$ ] . this overline operation is sometimes called bitwise negation . likewise , define the action of an oplus placed between @xmath58 and @xmath75 by @xmath76 . this oplus operation is sometimes called bitwise addition ( without carry ) . for any @xmath53 , the floor function is defined by @xmath77 , and the ceiling function by @xmath78 . for example , if @xmath72 , then @xmath79 . we will often use @xmath1 to denote a number of bits , and @xmath80 to denote the corresponding number of states . we will use the sets @xmath81 and @xmath82 interchangeably , since any @xmath83 can be identified with its binary representation @xmath84 . for any @xmath85 , define @xmath86 ; i.e. , @xmath87 is the result of reversing the binary representation of @xmath58 . suppose @xmath88 is a 1 - 1 onto map . ( we use the letter @xmath89 to remind us it is a permutation ; i.e. , a 1 - 1 onto map from a finite set onto itself ) . one can define a permutation matrix @xmath90 with entries given by @xmath91 for all @xmath92 . ( recall that all permutation matrices @xmath90 arise from permuting the columns of the unit matrix , and they satisfy @xmath93 . ) in this paper , we will often represent the map @xmath89 and its corresponding matrix @xmath90 by the same symbol @xmath89 . whether the function or the matrix is being alluded to will be clear from the context . for example , suppose @xmath40 is an @xmath94 dimension matrix , and @xmath89 is a permutation on the set @xmath82 . then , it is easy to check that for all @xmath95 , @xmath96 and @xmath97 . suppose @xmath98 is a 1 - 1 onto map ( i.e. , a bit permutation ) . @xmath99 can be extended to a map @xmath100 as follows . if @xmath101 , then let @xmath102_\alpha = x_{\pi_b(\alpha)}$ ] for all @xmath103 . the function @xmath100 is 1 - 1 onto , so it can be used to define a permutation matrix of the same name . thus , the symbol @xmath99 will be used to refer to 3 different objects : a permutation on the set @xmath104 , a permutation on the set @xmath82 , and an @xmath94 dimensional permutation matrix . all permutations on @xmath104 generate a permutation on @xmath82 , but not all permutations on @xmath82 have an underlying permutation on @xmath104 . an example of a bit permutation that will arise later is @xmath105 ; it maps @xmath106 for all @xmath107 and @xmath108 for all @xmath109 . in this section , we will review some well known facts about gray code@xcite . ( gray code was named after a person named gray , not after the color . ) for any positive integer @xmath1 , we define a * grayish code * to be a list of the elements of @xmath81 such that adjacent @xmath1-tuples of the list differ in only one component . in other words , a grayish code is a 1 - 1 onto map @xmath110 such that , for all @xmath111 , the binary representations of @xmath112 and @xmath113 differ in only one component . for any @xmath114 , there are many functions @xmath115 that satisfy this definition . next we will define a particular grayish code that we shall refer to as the " gray code and denote by @xmath116 . the gray code for @xmath117 is : 0&0&0 + 1&1&1 + 0&00&00 + 1&01&01 + 2&10&11 + 3&11&10 + 0&000&000 + 1&001&001 + 2&010&011 + 3&011&010 + 4&100&110 + 5&101&111 + 6&110&101 + 7&111&100 + . the * gray code * can be defined recursively as follows . let @xmath118 . for @xmath119 , let @xmath120 equal the set @xmath81 ordered in the gray code order . in other words , @xmath121 . then , _ + 1 = \{0 _ , 1^r_}_ord for @xmath122.(see section [ sec - notation ] for ordered set notation . ) from the recursive definition of the gray code , it is possible to prove that if @xmath123 and @xmath124 are nonnegative integers such that @xmath125 , then [ eq - gk - comps ] g_= k_k_+1 , [ eq - g - in - k - comps ] for all @xmath126 . ( for all @xmath127 , @xmath128 ) . eq.([eq - g - in - k - comps ] ) specifies @xmath1 linear equations for the @xmath1 components of @xmath129 expressed in terms of the @xmath1 components of @xmath130 . these equations can be easily inverted using gauss elimination to get : k_= g_g_+1 g_+2 g_+3 . [ eq - k - in - g - comps ] eqs.([eq - gk - comps ] ) can also be written in terms of the floor function : [ eq - gk - floor ] g = k , k = g . as in section [ sec - notation ] , suppose @xmath99 represents a permutation on @xmath104 which generates a permutation on @xmath82 of the same name . clearly , @xmath131 is a grayish code . indeed , @xmath131 is a 1 - 1 onto map , and permuting bits the same way for all elements of a list preserves the property that adjacent @xmath1-tuples differ in only one component . ( note , however , that it is easy to find @xmath99 s such that @xmath132 is not a grayish code . hence , to preserve grayishness , one must apply the bit permutation after @xmath116 , not before ) . in this section , we will review some well known facts about the so called hadamard , paley and walsh matrices ( a.k.a . transforms ) @xcite . for any positive integer @xmath1 , we define the @xmath1-bit hadamard matrix by ( h_)_k , r= ( -1)^ _ = 0 ^ -1 k_r _ , the @xmath1-bit paley matrix by ( p_)_k , r= ( -1)^ _ = 0 ^ -1 _ = 0 ^ -1 k_r__+ ^-1 , and the @xmath1-bit walsh matrix by ( w_)_k , r= ( -1)^ _ = 0 ^ -1 _ = 0 ^ -1 k_r _ , where @xmath133 , @xmath123 and @xmath134 . we will often omit the subscript @xmath1 from @xmath135 in contexts where doing this does not lead to confusion . note that @xmath136 are real symmetric matrices . for @xmath137 , define the reversal " function @xmath138 , and the negation " function @xmath139 . the function @xmath140 for @xmath1-bit gray code has been defined previously . the functions @xmath141 , @xmath142 and @xmath140 are 1 - 1 onto so they can be used to define permutation matrices of the same name ( see section [ sec - notation ] . ) we will often write @xmath105 , @xmath143 and @xmath116 instead of @xmath141 , @xmath142 and @xmath140 in contexts where this does not lead to confusion . note that @xmath105 and @xmath143 are symmetric matrices but @xmath116 is nt . next we will show that the @xmath1-bit hadamard , paley and walsh matrices all have the same columns , except in different orders . more specifically , @xmath144 are related to each other by the following equations : h_r = p , p_g = w . [ eq - h - p - w - relations ] a more pictorial way of expressing eqs.([eq - h - p - w - relations ] ) is : h p w . taking the transpose of both sides of eqs.([eq - h - p - w - relations ] ) leads to _ r h = p , _ g^t p = w . [ eq - h - p - w - transp - relations ] in the last equation , we have used the fact that matrices @xmath145 , @xmath146 are symmetric but @xmath116 is nt . comparing eqs.([eq - h - p - w - relations ] ) and ( [ eq - h - p - w - transp - relations ] ) , we see that h_r = _ r h , [ eq - hs - r - sym ] and p_g = _ g^t p . [ eq - ps - g - sym ] in fact , eq.([eq - hs - r - sym ] ) can be generalized as follows . suppose @xmath99 is a bit permutation on @xmath104 . then ( _ b^t h _ b)_kr & = & h__b(k),_b(r ) + & = & ( -1)^ _ = 0 ^ -1 [ _ b(k ) ] _ _ + & = & ( -1)^ _ = 0 ^ -1 k _ ( ) r _ ( ) + & = & h_kr , so h_b = _ b h . [ eq - hs - b - sym ] eq.([eq - hs - b - sym ] ) becomes eq.([eq - hs - r - sym ] ) when @xmath147 . to prove eqs.([eq - h - p - w - relations ] ) , note that ( h_r)_ik&= & _ j=0 ^ -1 h_ij(_r)_jk + & = & _ j(-1)^_i_j _ ( j = k^r ) + & = & p_ik . similarly , ( p_g)_ik&= & _ j=0 ^ -1 p_ij(_g)_jk + & = & _ j(-1)^ _ , i_j__+^-1 ( j=_g(k ) ) + & = & ( -1)^ _ , i_(k_k_+1 ) _ + ^-1 + & = & ( -1)^ _ i_(k_-1- + k_- ) + & = & w_ik . the square of @xmath148 and @xmath149 is one . indeed , using eq.([eq - bool - delta - fun ] ) we get ( h^2)_ik&= & _ j=0 ^ -1 h_ijh_jk + & = & _ j_-1=0 ^ 1 _j_1=0 ^ 1 _ j_0=0 ^ 1 ( -1)^ _ = 0 ^ -1 ( i_+k_)j _ + & = & _ _ i_^k _ = _ i^k , p^2=(_r h)(h_r)=1 , and w^2=(_g^t p)(p_g)=1 . since their square equals one , and they are real symmetric matrices , @xmath148 and @xmath149 are also orthogonal matrices . from the definitions given above for @xmath150 , one can prove by induction on @xmath1 that these matrices obey the following recursive equations : _ r(0)=1 , _ r(+1)= , _ n ( ) , and _ g(0)=1 , _ g(+1)= . similarly , from the definitions given above for @xmath136 , one can prove by induction on @xmath1 that these matrices obey the following recursive equations : h_0 = 1 , h_+1 = h _ , p_0 = 1 , p_+1= , and w_0 = 1 , w_+1= . by virtue of eqs.([eq - h - p - w - relations ] ) , w = h _ r _ g , [ eq - w - fun - h ] eq.([eq - w - fun - h ] ) means that the permutation @xmath151 will permute the columns of @xmath152 to give @xmath149 . expressing eq.([eq - w - fun - h ] ) in component form , we find w_ij & = & _ r , k h_ir ( _ r)_rk ( _ g)_kj + & = & _ r , k h_ir ( r=_r(k ) ) ( k=_g(j ) ) + & = & h_i,_r_g(j ) . thus , if we denote the columns of @xmath153 and @xmath154 by @xmath155 and @xmath156 , respectively , then _ j = _ _ r_g(j ) , [ eq - wi - def ] for @xmath157 . in this section , we will define a property of vectors called constancy . the columns of @xmath153 can be conveniently classified according to their constancy . consider the 3-bit hadamard matrix : h_3 = h_1 ^ 3= & & & & & & & + & & & & & & & + & & & & & & & + & & & & & & & + & & & & & & & + & & & & & & & + & & & & & & & + & & & & & & & + , [ eq - had3 ] where we have labelled the columns of @xmath158 by @xmath155 , where the index @xmath159 is given in its binary representation . according to eq.([eq - wi - def ] ) , to get @xmath160 from eq.([eq - had3 ] ) , one can simply reorder the columns of @xmath158 in bit - reversed gray code . the columns of @xmath158 ( and of @xmath160 ) can be classified according to their constancy . we define the * constancy * @xmath161 of a vector @xmath162 to be the smallest number of identical adjacent entries of @xmath162 . for example , @xmath163^t)=1 $ ] and @xmath164^t)=2 $ ] . the next table gives the constancy of the columns of @xmath158 , with the columns listed in the order in which they appear in @xmath160 . 000&000&000&8 + 001&001&100&4 + 010&011&110&2 + 011&010&010&2 + 100&110&011&1 + 101&111&111&1 + 110&101&101&1 + 111&100&001&1 + . [ tab - const ] it is clear from eq.([tab - const ] ) that the columns of @xmath160 are listed in order of non - increasing constancy , and that the constancies of the columns of @xmath160 are all powers of 2 . the literature on walsh matrices often refers to the index that labels the columns of @xmath149 as the * sequency * of that column . thus , as sequency increases , constancy decreases or stays the same . sequency and constancy are analogous to frequency and period , respectively , in fourier analysis . note that given any matrix @xmath40 , more than one of the columns of @xmath40 may have the same constancy . we will refer to : the number of columns of @xmath40 with the same constancy @xmath165 , as : the * multiplicity * * of the constancy * @xmath165 * in the matrix * @xmath40 , and denote it by @xmath166 . in this paper , we are only concerned with the case where @xmath40 equals the @xmath1-bit hadamard matrix so we will henceforth omit the subscript @xmath40 from @xmath166 . sometimes we will also abbreviate @xmath167 by @xmath168 , if doing this does not lead to confusion . the next table gives the multiplicity of the constancy @xmath165 in the @xmath1-bit hadamard matrix : l|lllll ( = k)&k=1 & k=2 & k=4 & k=8 & + = 1 & 1 & 1 & 0 & 0 & + = 2 & 2 & 1 & 1 & 0 & + =3 & 4 & 2 & 1 & 1 & + & & & & & + . [ tab - multi ] it is also convenient to define ( )= _ k ( k ) . [ eq - cum - multi ] we shall call this the * cumulative multiplicity of the constancy*. the next table can be easily obtained from eq.([tab - multi ] ) and eq.([eq - cum - multi ] ) . it gives @xmath169 for the @xmath1-bit hadamard matrix . l|lllll ( )&=2 ^ 0 & =2 ^ 1 & =2 ^ 2 & =2 ^ 3 & + = 1 & 2 & 1 & 0 & 0 & + = 2 & 4 & 2 & 1 & 0 & + =3 & 8 & 4 & 2 & 1 & + & & & & & + . [ tab - cum - multi ] it is clear from eq.([tab - cum - multi ] ) that @xmath170 , where @xmath171 . in this section , we discuss some symmetries of exact decompositions of @xmath3-multiplexors . for simplicity , we will first consider the @xmath172-multiplexors used in ref.@xcite . ref.@xcite uses @xmath3-multiplexors that are more general than the @xmath172-multiplexors used in ref.@xcite . at the end of the paper , we will discuss how to generalize our results for @xmath172-multiplexors so that they apply to the more general multiplexors used in ref.@xcite . below , we will present some quantum circuit diagrams . besides the circuit notational conventions defined in refs.@xcite and @xcite , the circuits below will use the following additional notation . a square gate with an angle @xmath173 below the square will represent @xmath174 applied at that wire " . typically , we will consider a seo consisting of alternating one - qubit rotations and cnots . the seo will always have a one - qubit rotation at one end and a cnot at the other . the angle for the one - qubit rotation that either begins or ends the seo will be denoted by @xmath175 . given two adjacent angles @xmath176 and @xmath177 in the seo , @xmath178 and @xmath179 will differ only in one component , component @xmath180 , where @xmath180 is the position of the control bit of the cnot that lies between the @xmath176 and @xmath177 gates . if we take the hermitian conjugate of the multiplexor @xmath181 , and then we replace the angles @xmath182 by their negatives ( and also the angles @xmath176 , hadamard transforms of the @xmath183 , by their negatives ) , we get the same multiplexor back . henceforth , we will refer to this symmetry transformation as * time reversal*. thus , an @xmath172-multiplexor is invariant under time reversal . suppose @xmath99 is a bit permutation on @xmath184 bits . if we replace @xmath182 by @xmath185 ( and also @xmath176 by @xmath186 ) and @xmath187 by @xmath188 in the multiplexor @xmath181 , we get the same multiplexor back . henceforth , we will refer to this symmetry transformation as * bit permutation*. thus , an @xmath172-multiplexor is also invariant under bit permutation . fig.[fig - sym - ops ] shows how time reversal and bit permutation act on a sequence of one - qubit rotations and cnots . more examples of the application of these transformations will be given below . recall from ref.@xcite our definition of a general multiplexor with @xmath189 control qubits @xmath190 and @xmath191 target qubits @xmath192 : @xmath193 . in a multiplexor whose matrices @xmath194 are independent of the @xmath180 component @xmath195 of @xmath14 , we can sum @xmath196 over @xmath197 to get 1 . such a multiplexor acts as the identity on qubit @xmath180 . when representing such a multiplexor in a circuit diagram , we can omit its half - moon node on qubit line @xmath180 . fig.[fig - free - half - moon ] shows a very special case of this principle , a special case that will be used in the circuit diagrams below . fig.[fig-1control ] shows two possible ways of decomposing an @xmath172-multiplexor with one control . the decomposition ( a ) in fig.[fig-1control ] is equivalent to : ( i_bbool _ bp_b)= e^i_0(1 ) ( 1)^n(0 ) e^i_1(1 ) ( 1)^n(0 ) . [ eq - decomp-1control ] let lhs and rhs stand for the left and right hand sides of eq.([eq - decomp-1control ] ) . recall that @xmath198 and @xmath199 . eq.([eq - decomp-1control ] ) can be proven as follows : [ eq - decomp-1control - proof ] @xmath200\}}\\ & = & lhs \label{eq - step - subst - theta } \;.\end{aligned}\ ] ] to arrive at eq.([eq - step - subst - theta ] ) , we expressed @xmath201 in terms of @xmath202 using = . if we take the hermitian conjugate of both sides of eq.([eq - decomp-1control ] ) , and then we replace the angles @xmath182 and @xmath203 by their negatives , we get ( i_bbool _ bp_b)= ( 1)^n(0 ) e^i_1(1 ) ( 1)^n(0 ) e^i_0(1 ) . [ eq - decomp-1control - rev ] eq.([eq - decomp-1control - rev ] ) is equivalent to decomposition ( b ) in fig.[fig-1control ] . thus , decompositions ( a ) and ( b ) in fig.[fig-1control ] transform into each other under time reversal . fig.[fig-2controls ] shows four possible ways of decomposing an @xmath172-multiplexor with two controls . fig.[fig-2controls ] was obtained by applying the results of figs.[fig - free - half - moon ] and [ fig-1control ] . in fig.[fig-2controls ] , note that decompositions ( a ) and ( b ) transform into each other under time reversal . decompositions ( c ) and ( d ) do too . furthermore , decompositions ( b ) and ( c ) transform into each other under bit permutation . the decompositions exhibited in fig.[fig-2controls ] can also be expressed analytically . for example , decomposition ( b ) is equivalent to : @xmath204 eq.([eq - decomp-2-controls ] ) can be proven using the same techniques that were employed in eqs.([eq - decomp-1control - proof ] ) to prove eq.([eq - decomp-1control ] ) . the proof requires that we assume : = . [ fig-3controls ] ( ditto , [ fig-4controls ] ) shows one of several possible decompositions of an @xmath172-multiplexor with 3 ( ditto , 4 ) controls . in general , decompositions for multiplexors with @xmath189 controls can be obtained starting from decompositions for multiplexors with @xmath205 controls . in this section , we finally define our approximation of multiplexors . we give the number of cnots required to express the approximant , and an upper bound to the error incurred by using it . so far we have used @xmath1 to denote a number of bits , and @xmath206 to denote the corresponding number of states . below , we will use two other numbers of bits , @xmath207 and @xmath208 , where @xmath209 and @xmath210 . their corresponding numbers of states will be denoted by @xmath211 and @xmath212 . define an @xmath213 dimensional matrix @xmath214 by v = h _ b _ g , [ eq - v - fun - h ] where @xmath99 is an arbitrary bit permutation on @xmath207 bits . eq.([eq - v - fun - h ] ) is a generalization of eq.([eq - w - fun - h ] ) . both equations define a new matrix ( either @xmath214 or the walsh matrix @xmath149 ) by permuting the columns of the hadamard matrix @xmath152 . @xmath214 becomes @xmath149 if we specialize the bit permutation @xmath99 to @xmath105 . if we denote the columns of @xmath214 by @xmath215 for @xmath216 , then _ j=__b_g(j ) , [ eq - vi - def ] which is the counterpart of eq.([eq - wi - def ] ) . in ref.@xcite , the decomposition of an @xmath172 multiplexor starts by taking the following hadamard transform : = h _ , [ eq - theta - fun - phi ] where @xmath217 and @xmath218 . the vectors @xmath219 constitute an orthonormal basis for the space @xmath220 in which @xmath221 lives , so @xmath221 can always be expanded in terms of them : = _ i=0 ^ -1 _ i ( _ i^ ) . now suppose that we truncate this expansion , keeping only the first @xmath222 terms , where @xmath223 and @xmath224 . let us call @xmath225 the resulting approximation to @xmath221 : = _ i=0^-1 _ i ( _ i^ ) . [ eq - def - phi - prime ] define @xmath226 , an approximation to @xmath227 , as follows : = h _ . [ eq - def - theta - prime ] if we let @xmath228 denote the standard basis vectors , then h__i= _ _ b_g(i ) = _ _ b_g(i ) . therefore , = _ i=0^-1 _ _ b_g(i ) ( _ i^ ) . [ eq - theta - prime ] by virtue of eq.([eq - theta - prime ] ) , if we list the components @xmath229 of @xmath226 in the grayish code order specified by the map @xmath230 , then the items in the list at positions from @xmath222 to the end of the list are zero . consider , for example , fig.[fig-2controls ] , which gives the exact decompositions for a multiplexor with 2 controls . suppose that in one of those decompositions , the angles @xmath203 s in the second half ( i.e. , the half that does not contain @xmath231 ) of the decomposition are all zero . then the one - qubit rotations in the second half of the decomposition become the identity . then the three cnots in the second half of the decomposition cancel each other in pairs except for one cnot that survives . the net effect is that the decomposition for a multiplexor with 2 controls degenerates into a decomposition for a multiplexor with only 1 control . the number of control bits is reduced by one in this example . in general , we can approximate a @xmath3-multiplexor by another @xmath3-multiplexor ( the approximant " ) that has fewer controls , and , therefore , is expressible with fewer cnots . we will call the reduction in the number of control bits the * bit deficit * @xmath10 . hence , @xmath232 . if @xmath4 denotes the number of cnots in an approximant with bit deficit @xmath10 , then it is clear from figs.[fig-1control ] , [ fig-2controls ] , [ fig-3controls ] and [ fig-4controls ] that : n_cnot= & 2 ^ -1 & 2 ^ -2 & 2 ^ -3 & & 2 & 0 + . [ eq - n - cnots ] hence , for @xmath233 , @xmath234 , but for @xmath235 , @xmath236 . the bit permutation @xmath99 on which the approximation of a multiplexor depends can be chosen according to various criteria . if we choose @xmath237 , then our approximation will keep only the higher constancy components of @xmath221 . such a smoothing , * high constancies approximation * might be useful for some tasks . similarly , if we choose @xmath238 , then our approximation will keep only the lower constancy components of @xmath221 , giving a * low constancies approximation*. alternatively , we could use for @xmath99 a bit permutation , out of all possible bit permutations on @xmath207 bits , that minimizes the distance between the original multiplexor and its approximant . such a * dominant constancies approximation * is useful if our goal is to minimize the error incurred by the approximation . the error incurred by approximating a multiplexor can be bounded above as follows . let @xmath239 denote the @xmath172-subset of an @xmath172-multiplexor @xmath240 and @xmath241 that of its approximant @xmath242 . call @xmath243 the error of approximating @xmath240 by @xmath242 . note that [ eq - error ] @xmath244 to arrive at step eq.([eq - step - ineq ] ) , we used the results of appendix [ ap - ineq ] . we will sometimes refer to @xmath245 as the linearized error , to distinguish it from the error @xmath243 . a simple picture emerges from all this . the error @xmath246 and the number of cnots @xmath4 are two costs that we would like to minimize . these two costs are fungible to a certain extent . given a multiplexor @xmath240 , and an upper bound @xmath247 on @xmath246 , we can use eqs.([eq - n - cnots ] ) and ( [ eq - error ] ) to find the approximant @xmath242 with the smallest @xmath4 . similarly , given a multiplexor @xmath240 , and an upper bound @xmath248 on @xmath4 , we can use eqs.([eq - n - cnots ] ) and ( [ eq - error ] ) to find the approximant @xmath242 with the smallest @xmath246 . at this point we encourage the reader to read appendix [ ap - computer - results ] . it discusses the output of a computer program that calculates @xmath225 from @xmath221 via eq.([eq - def - phi - prime ] ) . next we will show that eq.([eq - def - phi - prime ] ) can be simplified considerably by taking into account the explicit values of the column vectors @xmath215 . to get a quick glimpse of the simplification we seek , consider first the special case @xmath249 . we have h_2 = & & & + & & & + & & & + & & & + , w_2 = & & & + & & & + & & & + & & & + . define a matrix @xmath250 by = ( cc 1 & 1 + 1 & 1 ) . for any matrix @xmath40 , let @xmath251 be the submatrix of @xmath40 obtained by keeping only its columns from @xmath252 to @xmath159 . it is easy to check that w_2(:,0:3 ) w_2(:,0:3)^t = diag(1,1,1,1 ) , w_2(:,0:1 ) w_2(:,0:1)^t = ( cc & 0 + 0 & ) = ( ) , [ eq - w2-two - cols ] and w_2(:,0:0 ) w_2(:,0:0)^t = ( cc & + & ) = ( ) . in each case , we formed a decimated matrix " @xmath253 from @xmath149 , where @xmath223 . then we showed that the projection operator @xmath254 onto the column space of the decimated matrix , is a matrix whose entries are all either 0 or @xmath255 , and these entries sum to one along each row ( or column ) . given a set @xmath23 of real numbers , and given @xmath256 , call the average of the elements of @xmath257 a partial average " of the elements of @xmath23 . for example , if @xmath258 and @xmath237 , then @xmath259 and @xmath260 . from eq.([eq - w2-two - cols ] ) , the entries of @xmath225 are partial averages of the entries of @xmath221 . next we show how to simplify eq.([eq - def - phi - prime ] ) for arbitrary @xmath207 , not just for @xmath249 . define an @xmath213 dimensional matrix @xmath261 by ^ ( , )_qs= _ r=0^-1 ( w_)_qr ( w_^t)_rs . [ eq - gamma - fun - w ] below , we will show that @xmath261 reduces to : ^ ( , )_qs= ( = ) . [ eq - gamma - fun - floor ] for example , when @xmath262 , eq.([eq - gamma - fun - floor ] ) becomes = = c|cccc & 00&01 & 10&11 + & 1&1&0&0 + 01&1&1&0&0 + 10&0&0&1&1 + 11&0&0&1&1 . eqs.([eq - gamma - fun - w ] ) and ( [ eq - gamma - fun - floor ] ) are given in component form . the identical statements written in matrix form are , respectively , ^ ( , ) & = & w_(:,0:-1 ) w_(:,0:-1)^t + & = & w_w_^t , and ^ ( , ) = ( ^_b)^ . eq.([eq - gamma - fun - floor ] ) can be proven from eq.([eq - gamma - fun - w ] ) as follows : ^ ( , )_qs&= & _ r=0^-1 ( -1)^ _ = 0 ^ -1 _ = 0 ^ -1 ( q_r_+ r_s _ ) ( _ + ^ -1 + _ + ^ ) + & = & ( _ r_-1=0 ^ 1 _ r_1=0 ^ 1 _ r_0=0 ^ 1 ) ( r_-1 = r_-2 = = r_-_b=0 ) + & & ( -1)^ _ = 0 ^ -1 _ = 0 ^ -1 r_t _ ( _ + ^ -1 + _ + ^ ) ( where t = qs ) + & = & ( _ r_-1=0 ^ 1 _ r_1=0 ^ 1 _ r_0=0 ^ 1 ) ( r_-1 = r_-2 = = r_-_b=0 ) + & & ( -1)^ r_-1(t_0 + t_1 ) + r_-2(t_1 + t_2 ) + r_1(t_-2 + t_-1 ) + r_0(t_-1 ) + & = & 2 ^ -_b _ q_-1^s_-1 _ q_-2^s_-2 _ q__b^s__b [ eq - delta - fun - step ] + & = & ( = ) . to arrive at eq.([eq - delta - fun - step ] ) , we used eq.([eq - bool - delta - fun ] ) . recall that @xmath149 and @xmath214 can both we obtained by permuting the columns of @xmath152 : w = h _ r _ g , and v = h _ b _ g . from these two equations and from the fact , proven earlier , that @xmath99 and @xmath105 commute with @xmath152 , we get v = _ b _ r h _ r _ g = _ b _ r w . thus , & = & v_v_^t + & = & _ b _ r w_w_^t _ r _ b^t . hence , if we define @xmath263 and @xmath264 by = _ r _ b^t , = _ r _ b^t , then = ^ ( , ) . [ eq - phi - pr - fun - phi ] we see from eq.([eq - phi - pr - fun - phi ] ) that even when @xmath265 , the entries of @xmath264 ( which are the same as the entries of @xmath225 but in a different order ) are partial averages of the entries of @xmath263 ( which are the same as those of @xmath221 but in a different order ) . thus _ k(_k ) _ j _ k ( _ k ) , for all @xmath216 . this last equation implies as we mentioned before , the quantum compiling algorithm of ref.@xcite uses @xmath3-multiplexors that are more general than the @xmath172-multiplexors considered above . luckily , the above results for @xmath172-multiplexors are still valid , with minor modifications , for the more general ones . indeed , the @xmath3-subset of the multiplexors used in ref.@xcite is of the form @xmath266 \sigma_w^{f(b)}\}_{\forall b}$ ] . ref.@xcite defines vectors @xmath267 and @xmath268 from the parameters @xmath269 and @xmath270 , respectively . it then defines @xmath271 and @xmath272 as hadamard transforms of @xmath273 and @xmath274 , respectively , just as eq.([eq - theta - fun - phi ] ) defines @xmath227 as a hadamard transform of @xmath221 . we can define approximations @xmath275 and @xmath276 by replacing @xmath221 , @xmath227 , @xmath225 , @xmath226 by @xmath273 , @xmath271 , @xmath275 , @xmath276 , respectively , within eqs.([eq - def - phi - prime ] ) and ( [ eq - def - theta - prime ] ) . we can define approximations @xmath277 and @xmath278 analogously . the expansions of @xmath279 and @xmath280 in the @xmath281 basis can be truncated at the same @xmath222 . the table given in eq.([eq - n - cnots ] ) for the number of cnots still applies , except that @xmath4 may change by 1 if we eliminate the @xmath282 gate as in ref.@xcite . when a @xmath3-multiplexor @xmath240 with @xmath3-subset @xmath266 \sigma_w^{f(b)}\}_{\forall b}$ ] is approximated by a @xmath3-multiplexor @xmath242 with @xmath3-subset @xmath283 \sigma_w^{f(b)}\}_{\forall b}$ ] , one can show , using the results of appendix [ ap - ineq ] , that -_2 _ b , which is a generalization of eq.([eq - error ] ) . in this appendix , we establish a well known(see ref.@xcite , page 574 ) upper bound for the distance ( measured in either the 2-norm or the frobenius norm ) between two @xmath284 matrices . let @xmath285 be 3d real vectors . define @xmath286 . if @xmath287 , then e^i -e^i_2&= & e^i e^-i -1_2 + & & i_2 + & = & || . next , we will show that this approximation can be turned into an inequality . consider first the special case where @xmath288 and @xmath289 both point in the y direction . then e^i - e^i_2 & = & e^i-1_2 + & = & ( cc ( ) -1 & ( ) + - ( ) & ( ) -1 ) _ 2 + & = & 2 ( ) ( cc - ( ) & ( ) + - ( ) & - ( ) ) _ 2 + & = & 2 |()| + & & || . to find an upper bound for @xmath290 when either @xmath288 or @xmath289 does not point in the y direction , we will use the following identity . for @xmath291 and @xmath292 , e^(a+e)t - e^at = _ 0^t ds e^a(t - s ) e e^(a+e)s . [ eq - diff - eq - sol ] to prove eq.([eq - diff - eq - sol ] ) , let @xmath293 and @xmath294 stand for the left and right hand sides of eq.([eq - diff - eq - sol ] ) . it is easy to verify that ( l - r)(0)=0 , = a(l - r ) . this initial value problem has the unique solution @xmath295 . in eq.([eq - diff - eq - sol ] ) , set @xmath296 , and @xmath297 , and take the 2-norm of both sides . this yields e^i ( + ) - e^i _ 2 & & _ 0 ^ 1 ds e^i ( 1-s)_2 i _ 2 e^i ( + ) s _ 2 + & = & || . one can also find an upper bound for the distance , in the frobenius norm , between two @xmath284 matrices . if @xmath298 and @xmath299 , then the eigenvalues of @xmath300 are @xmath301 , for some real number @xmath173 . likewise , the eigenvalues of @xmath302 are @xmath303 . thus @xmath304 is real . if we denote the eigenvalues of @xmath305 by @xmath306 with @xmath307 , then @xmath308 has a single eigenvalue @xmath309 with algebraic multiplicity 2 . thus e^i -e^i_f = e^i -e^i_2 . [ eq-2-f - norms ] but we ve already proven that @xmath290 is bounded above by @xmath310 so e^i -e^i_f | | . in this appendix , we discuss a simple computer program that verifies and illustrates many of the results of this paper . our program is written in the octave language . octave is a gratis , open - source interpreter that understands a subset of the matlab language . hence , our program should also run in a matlab environment with few or no modifications . .... phi(1)= 0.133765891 phi(2)= 0.270447403 phi(3)= 0.307625920 phi(4)= 0.311291575 phi(5)= 0.452735037 phi(6)= 0.569045961 phi(7)= 0.653136015 phi(8)= 0.867156088 ----------------------- permutation 1 = ( 1,2,3 ) delta_b , phi_prime= 0 0.134 0.270 0.308 0.311 0.453 0.569 0.653 0.867 1 0.293 0.420 0.480 0.589 0.293 0.420 0.480 0.589 2 0.387 0.504 0.387 0.504 0.387 0.504 0.387 0.504 3 0.446 0.446 0.446 0.446 0.446 0.446 0.446 0.446 ----------------------- permutation 2 = ( 1,3,2 ) delta_b , phi_prime= 0 0.134 0.270 0.308 0.311 0.453 0.569 0.653 0.867 1 0.293 0.420 0.480 0.589 0.293 0.420 0.480 0.589 2 0.356 0.356 0.535 0.535 0.356 0.356 0.535 0.535 3 0.446 0.446 0.446 0.446 0.446 0.446 0.446 0.446 ----------------------- permutation 3 = ( 2,1,3 ) delta_b , phi_prime= 0 0.134 0.270 0.308 0.311 0.453 0.569 0.653 0.867 1 0.221 0.291 0.221 0.291 0.553 0.718 0.553 0.718 2 0.387 0.504 0.387 0.504 0.387 0.504 0.387 0.504 3 0.446 0.446 0.446 0.446 0.446 0.446 0.446 0.446 ----------------------- permutation 4 = ( 2,3,1 ) delta_b , phi_prime= 0 0.134 0.270 0.308 0.311 0.453 0.569 0.653 0.867 1 0.202 0.202 0.309 0.309 0.511 0.511 0.760 0.760 2 0.356 0.356 0.535 0.535 0.356 0.356 0.535 0.535 3 0.446 0.446 0.446 0.446 0.446 0.446 0.446 0.446 ----------------------- permutation 5 = ( 3,1,2 ) delta_b , phi_prime= 0 0.134 0.270 0.308 0.311 0.453 0.569 0.653 0.867 1 0.221 0.291 0.221 0.291 0.553 0.718 0.553 0.718 2 0.256 0.256 0.256 0.256 0.636 0.636 0.636 0.636 3 0.446 0.446 0.446 0.446 0.446 0.446 0.446 0.446 ----------------------- permutation 6 = ( 3,2,1 ) delta_b , phi_prime= 0 0.134 0.270 0.308 0.311 0.453 0.569 0.653 0.867 1 0.202 0.202 0.309 0.309 0.511 0.511 0.760 0.760 2 0.256 0.256 0.256 0.256 0.636 0.636 0.636 0.636 3 0.446 0.446 0.446 0.446 0.446 0.446 0.446 0.446 .... .... error as function of ( permutation\delta_b ) 0 1 2 3 1 1.110e-16 2.779e-01 3.627e-01 4.215e-01 2 1.110e-16 2.779e-01 3.324e-01 4.215e-01 3 1.110e-16 1.491e-01 3.627e-01 4.215e-01 4 1.110e-16 1.070e-01 3.324e-01 4.215e-01 5 5.551e-17 1.491e-01 2.316e-01 4.215e-01 6 5.551e-17 1.070e-01 2.316e-01 4.215e-01 .... in this example , @xmath311 so @xmath312 and @xmath313 . the first 8 lines of ` out_phis.txt ` give the components of @xmath221 . in this case , the computer picked 8 independent random numbers from the unit interval , and then it sorted them in non - decreasing order . ` my_moo.m ` can be easily modified so as to allow the user himself to supply the components of @xmath221 . after listing @xmath221 , ` out_phis.txt ` lists the @xmath314 permutations @xmath99 of @xmath207 bits . for each @xmath99 , it prints the components of @xmath225 , listed as a row , for each value of @xmath10(=row label ) . note that for @xmath315 , @xmath316 , and for @xmath317 , all @xmath318 are equal to the average of all the components of @xmath221 . note also that for all values of @xmath10 and @xmath159 , @xmath319 $ ] . the second output file , ` out_error.txt ` , gives a table of the linearized error @xmath245 as a function of permutation number(=row label ) and @xmath10(=column label ) . as expected , the error is zero when @xmath10 is zero , and it is independent of the permutation @xmath99 when @xmath10 is maximum ( when the bit deficit @xmath10 is maximum , the approximant has no control bits , so permuting bits at positions @xmath320 does not affect the error . ) note that in the above example , the last permutation minimizes the error for all @xmath10 . this last permutation is @xmath321 ( bit - reversal ) , and it gives a high constancies expansion . recall that for this example , ` my_moo.m ` generated iid ( independent , identically distributed ) numbers for the components of @xmath221 , and then it rearranged them in monotonic order . when @xmath221 is chosen in this way , the graph @xmath322 has a high probability of lying close to a straight line , and a high constancy staircase is the best fit for a straight line . for this reason , almost every time that ` my_moo.m ` is operated in the mode which generates iid numbers for the components of @xmath221 , the high constancies expansion minimizes the error for all @xmath10 . however , this need not always occur , as the following counterexample shows . try running ` my_moo.m ` for @xmath323 , and for @xmath221 with its first 7 components equal to 0 and its 9 subsequent components equal to 1 . for this @xmath221 , and for @xmath324 , the high constancies expansion yields an error of 7/8 while some of the other expansions yield errors as low as 5/8 . note that although ` my_moo.m ` visits all @xmath314 permutations of the control bits , visiting all permutations is a very inefficient way of finding the minimum error . in fact , the @xmath314 control bit permutations can be grouped into equivalence classes , such that all permutations in a class give the same error . it s clear from fig.[fig - approx ] that we only have to visit @xmath325 ( recall @xmath326 ) equivalence classes of permutations . whereas @xmath327 is exponential in @xmath207 , @xmath328 is polynomial in @xmath207 for two very important extremes . namely , when @xmath10 or @xmath208 is of order one whereas @xmath207 is very large . indeed , if @xmath329 or @xmath330 , then @xmath331 ; if @xmath332 or @xmath333 , then @xmath334 , etc .

what is the simplest hamiltonian which can implement quantum computation without requiring any control operations during the computation process ? in a previous paper we have constructed a 10-local finite - range interaction among qubits on a 2d lattice having this property . here we show that pair - interactions among qutrits on a 2d lattice are sufficient , too , and can also implement an _ ergodic computer _ where the result can be read out from the time average state after some post - selection with high success probability . two of the 3 qutrit states are given by the two levels of a spin-1/2 particle located at a specific lattice site , the third state is its absence . usual hopping terms together with an attractive force among adjacent particles induce a coupled quantum walk where the particle spins are subjected to spatially inhomogeneous interactions implementing holonomic quantum computing . the holonomic method ensures that the implemented circuit does not depend on the time needed for the walk . even though the implementation of the required type of spin - spin interactions is currently unclear , the model shows that quite simple hamiltonians are powerful enough to allow for universal quantum computing in a closed physical system .

quant-ph0506270

to understand which sets of quantum control operations are sufficient for quantum computing is still an important issue of research . whereas the standard model of the quantum computer is based on one- and two - qubit unitaries there is meanwhile a large number of alternative proposals , e.g. computing by measurements only @xcite or adiabatic computing @xcite . the latter model encodes a computational problem into an interaction such that the ground state of the hamiltonian indicates the solution . in order to understand the power of adiabatic computing it is particularly interesting to know to what extent the transition into the ground state of simple interactions can already be a non - trivial computation process . since it has been shown that every problem in the complexity class qma can be encoded in a nearest - neighbor interaction of qubits located on a 2d lattice @xcite it is clear that , indeed , relatively simple hamiltonians are sufficient for adiabatic computing . this highlights the computational power of simple hamiltonians from one point of view . another aspect of the `` computational power of hamiltonians '' has been studied in models where the natural time evolution is interspersed by external control operations ( see e.g. @xcite ) . the intention of the present paper is to understand to what extent models with simple interactions can already be autonomous devices with full quantum computing power in the sense that quantum computation reduces to the following protocol : ( 1 ) prepare an initial state in the computational basis which contains the data , ( 2 ) wait for a sufficiently long time , and ( 3 ) measure a sufficiently large set of qubits in the computational basis . in @xcite we have constructed a hamiltonian satisfying these conditions with the additional feature that the readout need not necessarily be performed within a specific time interval ; in our `` ergodic quantum computer '' the result is also present in the time average of the dynamics after some simple post - selection with high success probability . even though this property may be of minor practical relevance , we considered it as a necessary feature of an _ autonomous _ computer since one would otherwise require a clock as an additional device . here we construct an interaction that is even simpler than the one constructed in @xcite since the latter requires 10-qubit interactions . the motivation to consider autonomous quantum computers is twofold . first , it could trigger new ideas how to reduce the set of necessary control operations in current implementation proposals by using the `` natural power '' of the interactions . in addressing this issue , the ergodic model defines only an extreme case ; realistic perspectives for quantum computing could arise by combining it with more conventional approaches . second it is an interesting fundamental issue in the thermodynamics of computation how to realize computation in closed physical systems . benioff , feynman , and margolus have already presented such hamiltonian computers @xcite . margolus asynchronous hamiltonian cellular automaton ( ca ) has the appealing feature to be lattice - symmetric . in other words , it is driven by a finite - range interaction among qubits located on a 2d lattice such that the total hamiltonian is invariant under discrete translations . the synchronization of the ca is realized by a kind of spin wave propagating along the lattice and triggering the update of the cells according to some computationally universal update rules which are not specified any further in @xcite . however , its clock wave has to start in an uncertain _ position _ in order to obtain a well localized _ momentum _ distribution with mainly positive momenta . a localized wave front would also propagate backwards and would therefore not trigger a correct computation . in @xcite we have chosen the same synchronization scheme but we start with a localized wave front since we allow only for preparations of _ basis _ states . the fact that the dispersion of the wave front leads to completely undefined computation steps is irrelevant since the _ time average _ of the dynamics encodes the correct result . the feature of our ergodic model to show the correct computation result also in the time average was hence only a nice byproduct of the fact that we must use a concept which works with a strange kind of clock : the latter starts with a well - defined time but then it counts backwards and forward with completely undefined counting speed . note that the time average can also be considered as the `` generalized final state '' of our computer since a final output state in the usual sense can not exist for finite dimensional hamiltonian models . to ask for a simple finite range hamiltonian that is universal for quantum computation is in some sense similar to asking for a simple computationally universal quantum cellular automaton ( for some proposals see @xcite ) with the only difference of considering update rules which change the state only in an `` infinitesimal '' way . however , local interactions in lattices have typically the property to spread the information into increasingly large regions , whereas it is possible to construct update rules for cellular automata that work by propagating the information forward column by column @xcite . this apparent difference between discrete and continuous time dynamics can already be explained with the translation operator on @xmath0 qubits : the cyclic shift is a unitary that can be achieved by local update rules @xcite , whereas the hamiltonian obtained from the logarithm of the shift contains interactions between _ distant _ qubits . hamiltonians with finite interaction length will typically spread the information over the lattice . therefore , one has to be more modest and demand that the correct result can only be present with high probability . the paper is organized as follows . in section [ syn ] we explain the hardware of our model consisting of a chain of atoms on a 2d square lattice . we construct a nearest - neighbor interaction which leads via a simple effective hamiltonian to a coupled random walk that will later trigger the implementation of gates acting on the atom spins . in section [ holo ] we explain how to arrange spin - spin interactions among the atoms such that they implement logical gates based on holonomic quantum computing . in section [ complete ] we describe the complete hamiltonian of our computing device . in section [ sec : diag ] we show that the time evolution of the effective hamiltonian is solvable since it can be transformed into a quasifree evolution of fermions . based on this solution , we estimate the time required for the computation process . in section [ impl ] we briefly sketch some ideas on the realization . section [ sec : ca ] describes how to construct the hamiltonian such that it implements a universal quantum cellular automaton . this is to obtain programmable hardware . in the appendix we present an alternative model for the atom propagation which is caused by an even simpler hamiltonian since the nearest - neighbor interactions in the model presented in the main part involves also _ diagonal neighbors _ in the lattice . in contrast , the model in the appendix uses only _ nearest _ neighbors in a strict sense . however , the corresponding hamiltonian dynamics will not be solvable . we will therefore consider an incoherent analogue and show that the corresponding classical random walk would trigger the implementation of gates in the desired way . for the coherent model , we can only conjecture that an appropriate propagation is achieved . now we present a model of a 2d lattice with a coupled propagation of a chain of atoms which will later be the synchronization mechanism of the computer . while propagating along the horizontal direction ( see fig . [ diagonalpositionen ] ) the spin of the atoms ( which represent the logical states ) will be subjected to spatially inhomogeneous interactions that implement the desired gates . it is important that the chain does not tear since the spin - spin interactions to be described later will only be active between adjacent atoms . furthermore , it is important for the synchronization that the chain remains connected . this will become clear in section [ holo ] . we will now describe the synchronization hamiltonian in detail and show that it generates a diagonalizable quantum walk . the lattice and the initial atom configuration are shown in fig . [ diagonalpositionen ] . the lattice sites are given by @xmath1 black fields of the modified chess - board in fig . [ diagonalpositionen ] . the @xmath2 atoms are only allowed to move forward or backward along the horizontal direction , i.e. , along the rows . each atom can only hop from a black site to either the next or the previous black site in the row . furthermore , the atoms can only move such that the chain does not tear , i.e. , atoms in adjacent rows are always diagonal neighbors . [ possible < ] shows a possible configuration . now we describe the hamiltonian that allows only collective motions of the atoms respecting these rules . the grid of black fields is labeled by @xmath3 with @xmath4 . the sites @xmath5 with @xmath6 indicate the diagonal line leading from the leftmost field to the uppermost one and sites @xmath7 with @xmath8 indicate the diagonal line from the leftmost field to the lower - most one . we consider the hilbert space of all atom configurations as a subspace of an @xmath9-qubit register of the form @xmath10 where qubit @xmath3 corresponds to lattice site @xmath3 and the basis states @xmath11 and @xmath12 indicate that there is no atom or that there is an atom , respectively , at position @xmath3 . a subspace of @xmath13 is the clock hilbert space @xmath14 spanned by all _ allowed _ configurations . they correspond to those basis states in @xmath13 for which there is exactly one atom in each row and for which the atom chain is connected . first we introduce interactions between site @xmath3 and @xmath15 which generates independent hopping of atoms along the rows : @xmath16 to achieve that the atom configuration remains a connected chain we introduce a strong attractive force between diagonal neighbors : @xmath17 where @xmath18 is assumed to be much larger than @xmath19 and @xmath20 is the projection on the states with an atom in position @xmath21 . the sum runs over all unordered pairs of sites @xmath22 with @xmath23 . the physically irrelevant term @xmath24 is ( for purely technical reasons ) chosen such that the initial configuration has zero energy . the energy is minimal if all attractive interactions are active . otherwise , if the chain is not connected , the potential energy is at least @xmath25 . we define the `` synchronization hamiltonian '' , which leads to the desired coupled motion of atoms , by @xmath26 in order to analyze the dynamics generated by @xmath27 we first argue that it can be replaced with an effective clock hamiltonian @xmath28 having @xmath14 as an invariant subspace . the initial atom configuration defines a ground state of @xmath29 . due to @xmath30 the comparably small perturbation @xmath31 can not circumvent the gap to one of the first excited states of @xmath29 . if @xmath32 denotes the spectral projection of @xmath29 corresponding to the ground state energy @xmath33 we will therefore expect a dynamical evolution according to @xmath34 note that not all states in the image of @xmath32 are allowed clock states since the image of @xmath32 contains also states with more than one atom in a row . however , these ground states could only be reached from an allowed state if @xmath31 contained hopping terms along columns . the following lemma ( proven in the appendix ) justifies more formally that we may analyze the dynamics according to @xmath28 instead of @xmath27 : [ approx ] for @xmath35 and @xmath36 the norm distance between the effective hamiltonian in equation ( [ effdef ] ) and the restriction @xmath37 of @xmath27 to the eigenspace corresponding to eigenvalues smaller than @xmath25 satisfies @xmath38 if we increase the attractive part of the interaction proportional to @xmath39 if @xmath0 increases the time evolution induced by @xmath28 is a good approximation for the true one . if the intention of this article was to propose a physical implementation scheme this would be a rather bad scaling , leading possibly to hard practical problems . however , from the computer science point of view , it is essential that the increase of energy is only polynomial in @xmath0 and we will thus obtain an _ efficient _ model of computation . we will now give a more explicit form of @xmath28 . to be more precise , we will describe an operator whose restriction to @xmath14 coincides with @xmath28 . let @xmath40 denote some basis state in @xmath14 , i.e. , @xmath41 describes an allowed atom configuration . then @xmath42 is a superposition of all atom configurations with connected chain that can be reached from @xmath41 by moving one atom forward or backwards in its diagonal row . in other words , each hopping term @xmath43 is only active if the sites @xmath44 and @xmath45 are occupied . we have therefore a conditional hopping given by @xmath46 where the symbol @xmath14 at the right indicates the restriction to @xmath14 . using a slightly more intuitive notation , which respects the orientation of fig . [ diagonalpositionen ] , the summands of @xmath28 can be denoted by @xmath47 where this @xmath48-local operator acts on @xmath48 black sites enclosing a common white site . we are interested in the dynamical evolution of the quantum state defined by the initial atom configuration . since it is a ground state of the dominating hamiltonian @xmath49 the major part of the state vector is contained in that spectral subspace of @xmath27 which corresponds to energy values not greater than @xmath50 . using the bound of lemma [ approx ] we may then estimate the error in the dynamical evolution caused by replacing the true evolution with the evolution according to @xmath28 . let @xmath51 be an arbitrary allowed atom configuration . then we have @xmath52 where @xmath53 as in lemma [ approx ] . due to @xmath54 the average energy of @xmath55 satisfies @xmath56 where the last inequality is given by counting the number of terms in eq . ( [ vham ] ) . let @xmath57 be the spectral projection of @xmath27 corresponding to all eigenvalues less than @xmath25 . then we have @xmath58 using @xmath59 and @xmath60 we conclude @xmath61 we have @xmath62 where we have used @xmath63 for two self - adjoint matrices @xmath64 . the statement follows then by replacing the last term in eq . ( [ arrayeq ] ) with the right hand side of ineq . ( [ rem ] ) . @xmath65 one could use the condition @xmath66 of theorem [ fehler ] to determine the time scale for which @xmath28 is a good approximation when @xmath25 is given . we can also assume that the time scale is given and we have to determine the interaction strength @xmath25 . after all , we have to ensure that the approximation is valid for all @xmath67 where @xmath68 is some a priori given upper bound on the time required by the clock wave to pass the relevant region on the lattice , i.e. , the region containing the spin - spin interactions which will be described in the next section . the effective hamiltonian leads to a quantum walk on the space of atom configurations which will be diagonalized in section [ sec : diag ] . in the next section we will explain how this walk may trigger the implementation of gates . now we consider not only the positions of the atoms but also their inner degree of freedom , e.g. their spin . we replace the qubit at each lattice site with a qutrit and extend hence the space @xmath13 to @xmath69 the basis states @xmath70 indicate the absence of the atom or its two possible spin states , respectively . before we describe how to add a spin - spin interaction hamiltonian to @xmath27 which leads to the implementation of gates when the atoms are moving along the rows , we should first mention two obvious approaches to imprint gates by interactions and why they are not suitable for our goal . in @xcite and @xcite the gates are directly coupled ( as additional _ tensor product _ components ) to the synchronization hamiltonian . certainly we could define the hopping term in eq . ( [ vham ] ) between column @xmath71 and @xmath72 in any desired row such that the spin is not conserved during the atom propagation but subjected to some rotation . this would provide us with one - qubit gates . such a one - qubit gate would clearly be inverted when the atom moves back to column @xmath71 again . the conditional state of the considered qubit , given that the atom is found on the right of column @xmath71 would hence be subjected to the desired transformation , no matter whether the atom has traveled back and forth several times as is is typical for a random walk . to imprint two - qubit gates into the hamiltonian is more difficult . the obvious method to couple their implementation with some coupled motion of two adjacent atoms would require more non - local interactions than only two - particle terms . the second obvious method would be to complete the clock hamiltonian by an _ additive _ spin - spin interaction term between some adjacent sites in the same column . this leads to the following problems . first the atoms do not stay there for a well - defined time . some part of the wave packet moves already and one part stays . second the atoms travel back and forth and pass the interaction region several times . both effects would in general entangle atom position and logical spin states in an uncontrollable way . a solution to the latter problem is holonomic quantum computing @xcite . the idea of this approach , in the usual setting , is that some time - dependent hamiltonian @xmath73 is adiabatically changed along a closed loop such that , on an appropriate degenerate subspace , the overall effect is a unitary which depends only on the loop ( described by representing @xmath73 in some parameter space ) and not on the speed of the change of @xmath73 . the unitary in the end remains the same even if the vector @xmath73 has been moving back and forth on this loop in the parameter space . we apply this concept to our model where we achieve the time - dependence of the spin - spin interaction by the motion of the atoms when they pass spatially inhomogeneous interactions . we imprint a spin - spin interaction ( varying slowly along the rows ) such that is describes a closed loop after the atoms have passed a certain region . then the spin state of the atoms does not depend on the time required for the passing , neither does it depend on the number of times the atoms were traveling back and forth before the whole region has been passed . this ensures that the entanglement between the position degree of freedom and the spin is lost as soon as the region has been passed by the whole atom chain . before we explain how to make use of this idea in detail , we rephrase a result in section ii of ref . @xcite , which is useful for us , as a lemma : * ( holonomic gates ) * + [ adia ] given a family of hamiltonians @xmath73 on a finite dimensional hilbert space @xmath74 by @xmath75\,,\ ] ] where @xmath76 is a self - adjoint operator on @xmath74 such that the family @xmath73 is a closed loop , i.e. , @xmath77 . let the change of @xmath73 be sufficiently slowly to consider it as an adiabatic change , i.e. , for each entry @xmath78 of @xmath73 we have @xmath79 where @xmath80 is the smallest spectral gap of @xmath81 . let @xmath82 be the time evolution generated by the family @xmath73 after the closed loop . let @xmath83 be an eigenspace of @xmath84 with eigenvalue @xmath85 . then @xmath82 keeps @xmath83 invariant and its restriction is given by @xmath86 where @xmath57 is the projection onto @xmath83 . the unitary ( [ geoph ] ) is also called a non - abelian geometric phase . since the holonomic concept implements the well defined unitary @xmath87 only on a _ subspace _ @xmath83 of the full hilbert space our @xmath2 spin particles can not be used for @xmath2 logical qubits . instead , the atom in row @xmath88 will encode a logical qubit together with the atom in row @xmath89 . moreover , we can only use rows being not too far from the row in the middle since the others are too short to imprint interactions that implement gates . two adjacent spin particles encode one qubit with logical states @xmath90 via @xmath91 we will call their span @xmath92 the `` code space '' . to explain how to implement one - qubit gates we first restrict our attention to two adjacent rows . we add interactions as follows . the whole region which carries the gate consists of a stripe of width @xmath93 as shown in fig . [ fig2erwechs ] . we call this region `` interaction stripe '' . inside this stripe , we add a `` gate hamiltonian '' to @xmath27 which consists of spin - spin interactions @xmath94 between certain diagonal neighbors in adjacent rows ( as indicated by the edges labeled with @xmath95 in fig . [ fig2erwechs ] ) . each @xmath94 is a pair - interaction between two sites . we choose the following interactions : @xmath96 with @xmath97 . here @xmath76 is some self - adjoint operator on @xmath98 specified later and @xmath99 in eq . ( [ attract ] ) . the pauli matrix @xmath100 has to be read as formally acting on the space @xmath101 of the corresponding qutrit , even though it is zero for the state @xmath11 and acts therefore only on the spin states . it is important to note that the degenerate eigenspace of @xmath102 and @xmath103 coincides with the code space @xmath92 . if the spins are initialized such that their state is in @xmath92 before they pass the interaction region they will at every time instant remain in the degenerate subspace of the interaction that is active ( provided that the change between different @xmath94 is adiabatic ) . this will implement the `` non - abelian phase '' @xmath104 of lemma [ adia ] . the length @xmath105 of the interaction stripe is chosen such that the change of interactions from @xmath71 to @xmath72 can be considered as approximating a continuous change and furthermore as adiabatic when the atom chain is propagating in horizontal direction . to sketch how to choose @xmath105 we define a typical time scale @xmath106 on which an atom jumps from one site to its neighbor . then we have to ensure that each entry @xmath107 of @xmath94 satisfies @xmath108 where @xmath80 is the smallest energy gap of @xmath109 . here , the time @xmath106 is defined as a dimensionless quantity of the order @xmath19 since the hopping terms @xmath110 in eq . ( [ vham ] ) appears with constant one . physical dimensions are irrelevant since inequality ( [ adiaab ] ) is invariant with respect to a simultaneous rescaling of all @xmath94 and the hopping terms in eq . ( [ vham ] ) . we obtain therefore the dimensionless condition @xmath111 . now we may set @xmath112 with arbitrary angle @xmath113 , or @xmath114 where we have again slightly abused notation since we did not explicitly indicate the embedding of the 2-qubit operators in eqs . ( [ xdef1 ] ) and ( [ xdef2 ] ) ( acting on the 4 possible spin configurations ) into the two qutrit space . the following lemma shows which gates are implemented by the above choices for @xmath76 . [ onegate ] let two atoms pass a region where they are subjected to the hamiltonians @xmath115 in eq . ( [ vdef ] ) with @xmath76 as in eq . ( [ xdef1 ] ) or eq . ( [ xdef2 ] ) . assume furthermore that @xmath105 is sufficiently large to consider the change of interactions as adiabatic . then their encoded qubit is , up to an irrelevant global phase , subjected to the transformation @xmath116 or @xmath117 respectively . on the space @xmath118 spanned by the @xmath48 spin states the interaction changes according to @xmath119 with @xmath97 . due to lemma [ adia ] the code space @xmath92 is then subjected to @xmath120 with @xmath121 , where @xmath122 is the projection onto @xmath92 . for both choices for @xmath76 , the relevant operator @xmath123 consists then only of transitions between @xmath124 and @xmath125 , which corresponds to a @xmath126 if the transition amplitudes are @xmath19 ( as it is the case given by eq . ( [ xdef1 ] ) ) and to @xmath127 if they are @xmath128 and @xmath129 ( as in the case of eq . [ xdef2 ] ) . @xmath65 lemma [ onegate ] shows that we can generate arbitrary one - qubit transformations by concatenating interactions with @xmath76 as in eq . ( [ xdef1 ] ) and ( [ xdef2 ] ) using appropriate angles @xmath113 . for the implementation of logical two - qubit gates we consider interaction stripes that consist of @xmath130 adjacent rows ( see fig . [ fig3erwechs ] ) , where row @xmath19 and @xmath93 belong to the _ first _ logical qubit and row @xmath130 is part of the _ second _ logical qubit . we define interactions @xmath131 and @xmath132 such that @xmath131 connects row @xmath19 and @xmath93 and @xmath132 connects row @xmath93 and @xmath130 as shown in fig . [ fig3erwechs ] . we will refer to the atoms in row @xmath133 as atoms @xmath133 , respectively . the interaction between row 1 and the row 2 is always @xmath131 in the whole interaction stripe . row 2 interacts with row 3 according to @xmath134 . as soon as atom @xmath93 has entered the interaction stripe ( shown in fig . [ fig3erwechs ] for @xmath135 ) it is simultaneously subjected to interactions @xmath131 and @xmath136 . moreover , @xmath131 and @xmath137 are turned off simultaneously as soon as atom @xmath93 leaves the interaction stripe . the total interaction that is active on the system consisting of the spins of atoms 1,2,3 is therefore either @xmath138 when atom @xmath93 is inside the interaction stripe , or @xmath33 otherwise . we choose @xmath139 where the first tensor component refers to row 1 in fig . [ fig3erwechs ] and the second to row 2 . the operator @xmath140 has no effect on the spin of the atom in row @xmath93 . it only ensures that the @xmath100-hamiltonian is not switched on before the atom in row @xmath93 has entered the left - most black field in row 3 in fig . [ fig3erwechs ] . the interaction @xmath132 changes according to @xmath141 where @xmath142 is an operator that acts on the one - qutrit space . it will be specified later . on the space @xmath143 spanned by all @xmath144 possible spin states , we have @xmath145 the idea is that the adiabatic change of the degenerate hamiltonian on row @xmath19 and @xmath93 is controlled by the spin of the atom in row @xmath130 , i.e. , by the logical state of the second qubit . whenever the state in row @xmath130 is @xmath146 , rows @xmath19 and @xmath93 are subjected to the hamiltonians @xmath147 otherwise they are subjected to a constant hamiltonian . we find : let @xmath131 and @xmath132 be as in eq . ( [ v3def ] ) and eq . ( [ udef ] ) , respectively , such that they connect row @xmath19 and @xmath93 as well as @xmath93 and @xmath130 , respectively . choose the corresponding operator @xmath142 as @xmath148 with arbitrary angle @xmath113 . consider the logical 2-qubit space @xmath149 , where the first tensor component refers to the atoms in rows @xmath19 and @xmath93 and the second to those in @xmath130 and the additional row @xmath48 ( which is not depicted in fig . [ fig3erwechs ] ) . then the adiabatic change of the interaction @xmath150 from @xmath151 to @xmath152 implements a controlled-@xmath153 transformation on @xmath149 with the lower qubit as control qubit and the upper as target ( in the tensor product in eq . ( [ udef ] ) this correponds to the right and the left side , respectively ) . _ proof : _ let the second qubit be in the logical @xmath33 state @xmath154 . this means that the spins of atom @xmath130 and @xmath48 are in the state @xmath125 . however , for the interaction only the state of atom @xmath130 matters . since it is @xmath155 , atoms @xmath19 and @xmath93 are constantly subjected to the spin hamiltonian @xmath156 since the kernel of this operator coincides with the code space the logical state is not affected at all . assume now that the second qubit is in the state @xmath12 and atom @xmath130 is therefore in the state @xmath157 . then atom @xmath19 and @xmath93 are subjected to the hamiltonians @xmath158 with @xmath97 . due to lemma [ adia ] the unitary that is implemented on the code space after such an adiabatic change is given by @xmath159 with @xmath160 describing this operator with respect to the logical basis @xmath161 . this shows that the unitary @xmath162 is implemented after the atoms have passed the interaction stripe . since we have also shown that this unitary is only implemented if the state of the second qubit is @xmath163 , we have know shown that the net effect is a controlled-@xmath162 gate . @xmath65 . we can imprint interactions like those above between any adjacent logical qubits as well as we may obtain one - qubit rotations as in lemma [ onegate ] by appropriate interactions . it is known that conditional phase gates together with the set of one - qubit rotations allow for universal quantum computing @xcite . hence we are able to design interactions for arbitrary quantum circuits . we obtain the complete hamiltonian @xmath27 of our autonomous computing device as a sum of the synchronization hamiltonian and the spatially inhomogeneous spin - spin interaction that implement holonomic transformations . it reads : @xmath164 where @xmath29 is the strong attractive force defined in eq . [ attract ] and @xmath31 describes the hopping terms @xmath165 where we have , in slight abuse of notation , adapted the definition of eq . ( [ vham ] ) for spin - less creation and annihilation operators into those for spin 1/2 particles . for instance , @xmath166 creates a particle with spin down . similarly , the term @xmath29 is adapted to spin 1/2 particles in the sense that the particle number operator @xmath167 is given by @xmath168 the spin - spin interactions @xmath169 are only non - zero if site @xmath3 is adjacent to site @xmath170 and if the pair of sites belongs to some interaction stripe . inside these stripes , they are given by the interactions @xmath171 described in the preceding section . we assume that all these rectangles ( `` interaction stripes '' ) are confined to a square of length @xmath172 , i.e. , the spin - spin interaction is only non - zero for @xmath173 for some @xmath174 . we call this region the _ circuit region _ ( see fig . [ region ] ) . as soon as all atoms have passed it , the whole circuit is implemented . the complement of the interaction region will be called _ output region _ since we will read out the result of the computation there . now we show that the hamiltonian @xmath175 leads to a propagation of atoms that implements the gates in correct order . first of all , we observe that the gates are irrelevant for the quantum walk . different particle configurations correspond to mutually orthogonal vectors in the hilbert space . whether or not the particles are subjected to an additional change of their inner degree of freedom is irrelevant for the dynamics as long as we consider the adiabatic limit where the effect of the spin - spin interaction is only that it implements _ unitary gates _ on the spin states . hence we will analyze the dynamics generated by @xmath28 instead of the dynamics generated by @xmath175 to derive the relevant probabilities for finding the atom chain at certain positions . after recalling that the adiabatic change of the spin - spin interactions has implemented the desired gates given that the atom chain has passed the interaction stripes we have then shown that the circuit has been implemented . to diagonalize the quantum walk we show that it is isomorphic to a walk of free fermions propagating on a one - dimensional chain . on the left of figs . [ diagonalpositionen ] and [ possible < ] we have characterized the configurations by binary words of length @xmath176 . the symbols @xmath177 as @xmath71th digit indicate whether the atom in row @xmath72 is _ behind _ the atom on row @xmath71 or _ in front _ of it , respectively . the initial configuration is hence characterized by @xmath178 symbols @xmath33 followed by @xmath179 symbols @xmath19 . since the atom on the very top and that one on the very bottom are fixed , the number of symbols @xmath19 remains constant and the vector space spanned by the possible atom configurations is therefore the subspace @xmath180 of @xmath181 spanned by words with hamming weight @xmath179 . with respect to such a representation , the effective clock hamiltonian @xmath28 , when restricted to @xmath14 , consists of operators which replace some pattern @xmath182 by @xmath183 or vice versa . it can be written as @xmath184 where @xmath185 and @xmath186 are fermion annihilation and creation operators , respectively . they are defined @xcite by @xmath187 and satisfy the canonical anti - commutation relation @xmath188 where we have chosen the convention @xmath189 and @xmath190 . observe that @xmath191 is the xy - hamiltonian which is well - known in solid states physics @xcite and generates a quasi - free evolution of fermions . since the so - called jordan wigner transformation , describing the fermion interpretation formally , is standard @xcite , we rephrase it only briefly . the subspace @xmath180 can be reinterpreted as the space of @xmath179 fermions moving without interactions in a @xmath192-dimensional state space . then @xmath180 is simply the antisymmetric tensor product @xmath193 the correspondence between these two pictures can be described easily : given a basis state in the qubit register by a binary word with symbols @xmath19 at the positions @xmath194 . this state corresponds in the fermionic picture to @xmath195 where @xmath196 denotes here the @xmath71th canonical basis vector in @xmath197 . the restriction of @xmath191 to @xmath180 is in the fermionic picture given by @xmath198 where @xmath199 is the linear ( non - unitary ) shift acting on the basis states of @xmath197 via @xmath200 for @xmath201 and @xmath202 . this is because the term @xmath203 shifts the fermion by one site . in other words , the time evolution in each tensor component in eq . ( [ antisymm ] ) is generated by the hamiltonian @xmath204 , i.e. , the adjacency matrix of the linear chain . this implies that the time evolution transfers annihilators and creators to linear combinations of annihilators and creators , respectively , when considered in the heisenberg picture @xcite . let @xmath205 be the time evolution of one particle on a quantum walk on a chain and @xmath206 its entries . then the time evolution of the creation operator on site @xmath71 is @xmath207 i.e. , it evolves into an operator creating a fermion which is in a superposition of different sites . for the annihilation operator at site @xmath71 we obtain @xmath208 to analyze the time evolution of relevant observables we will only make use of this formulation of the dynamics . the running time of our computer is given by the length of the time one has to wait until all atoms will have passed the interaction region @xmath209 , i.e. , a square of length @xmath174 , almost with certainty . the dynamical evolution @xmath210 is clearly quasiperiodic because the hilbert space @xmath180 is finite dimensional . therefore the atoms will always return to @xmath209 . however , the essential idea of the ergodic quantum computer is that the probability of finding it outside of @xmath209 is high if one measures at a random time instant . later , we will therefore also consider the time average , before we estimate the time needed by the atoms to pass @xmath209 for the first time . we found : [ passingtime ] let the circuit region be a square of length @xmath172 . then there is a time @xmath211 such that the probability of finding all atoms outside the circuit region @xmath209 is at least @xmath212 given that the size @xmath213 of the whole lattice is sufficiently large . the proof can be found in the appendix . note that the probability of finding some atom in @xmath209 could even be made smaller if the factor @xmath48 in eq . ( [ jint ] ) in the proof would be replaced with some larger number . we conclude that the implementation time for our imprinted circuit is in @xmath214 . now we want to show that also at a random time instant the probability of finding all atoms in the output region ( i.e. , the probability of finding at least @xmath172 fermions on the left half ) is close to @xmath19 . we found : [ passingprobability ] let the circuit region be a square of length @xmath215 . then the probability of finding all atoms outside the circuit region tends to @xmath19 for @xmath216 . this theorem is also proven in the appendix . one should maybe mention the following limitation of our model : in a lattice with finite length @xmath0 the adiabatic approximation underlying the holonomic implementation is only true up to some error . therefore the implemented transformation , given that the atom chain has passed the circuit region , does not exactly coincide with the desired quantum circuit . instead , it depends to some extent on the time needed to pass the region . in the limit of averaging the time evolution over an infinite time interval this error will increase . roughly speaking ( and expressed in a too classical language ) , the atom chain has then traveled back and forth many times and the error can increase since passing the circuit region backward would not necessarily invert the implemented circuit . however , it is a matter of the time scale on which the time average is taken whether this error is relevant , i.e. , the degree to which the adiabatic approximation is justified determines the time scale on which the time average is described by our analysis . to judge whether it is likely that systems with hamiltonians as above could be found in real systems would go beyond the scope of this article . maybe one should rather ask which modifications are possible for our models that would make them more feasible . so far , the required spin - spin interactions are quite specific . to show that the required diagonal hopping is not a priori unphysical one could think of electrons on quantum dot arrays arranged as in fig . [ figdiadots ] . the hopping along the rows does not require tunneling between distant dots even though it is the _ diagonal _ direction with respect to the square lattice . spin - spin interactions would then only be needed in the direction of the dashed square lattice . [ dots ] attractive interactions between particles on adjacent dots could , for instance , be achieved if the particles alternate between electrons and holes from row to row . certainly , it will be difficult to find a system that combines the attractive interaction and the hopping with the spatially inhomogeneous spin - spin interactions . we have shown that nearest - neighbor interactions among qutrits located on a 2d square lattice can be designed in such a way that their autonomous time evolution simulates any desired quantum circuit . the hamiltonian given here is significantly simpler than that one given in our previous paper @xcite since the latter contained @xmath182-qubit interactions . however , the advantage of the hamiltonian in @xcite is that it is programmable . the hamiltonian given here contains the gates to be implemented already as hardware . in order to make a programmable quantum computer one could proceed as follows . consider a universal quantum cellular automaton in one dimension @xcite consisting of cells @xmath217 with @xmath218 . each cell contains a quantum system with hilbert space @xmath219 . using the so - called margolus partition scheme @xcite , one time step of the ca consists of two parts . the first one is given by identical copies of a unitary @xmath220 acting on all pairs @xmath221 , the second one by identical copies one could use a @xmath222-algebraic description and work with automorphisms on so - called quasilocal algebras @xcite . ] of @xmath131 acting on all pairs @xmath223 . in our scheme , we may represent each cell by the subspace of an appropriate number of adjacent logical qubits . then we cover some part of our 2d lattice with a pattern of interaction regions that implement unitaries @xmath220 and @xmath131 as shown in fig . [ margo ] . the number of pairs of columns in this pattern determines the number of time steps of the simulated ca . the translation symmetry of such a crystal would then , however , be described by huge unit cells . nevertheless , our construction shows that translation invariant finite - range interactions among finite dimensional quantum systems can in principle implement a universally programmable quantum computer . we have shown that nearest - neighbor interactions on 2-dimensional lattices can induce dynamical evolutions that are powerful enough to represent a programmable quantum computer . even though we have not described a realistic implementation scheme we have given interactions which are not a priori unphysical . the main reason why it may be difficult to construct systems having exactly the hamiltonians considered here is that the spatial homogeneity of the interactions follow a rather sophisticated law . to find even simpler hamiltonians which can perform quantum computing in a closed physical system is an interesting challenge for further research . part of this work was done during a visit of hans briegel s group at iqoqi , whose hospitality is gratefully acknowledged . the model was inspired by discussions with robert raussendorf about cellular automata and with several people at iqoqi about optical lattices and quantum dots , especially gregor thalhammer and alessio recati . thanks also to pawel wocjan for encouraging remarks and helpful discussions . this work was partly funded by the landesstiftung baden - wrttemberg , project az1.1422.01 . the clocking scheme of our autonomous computer presented so far is somewhat sophisticated in the sense that the hopping terms connect lattice sites lying diagonal with respect to the grid given by the possible atom positions . in other words , the rows defining the direction of atom propagation were _ diagonal _ with respect to the square lattice structure . it would be more natural to have tunneling along rows or columns of the square lattice itself . we will therefore describe an alternative model for the propagation of an atom chain which has , however , the disadvantage , that we can only conjecture that the atoms move in forward direction with sufficient speed and will appear outside the interaction region with reasonable probability . we can , however , prove , that the corresponding classical random walk would really implement the computation . the whole lattice consists of @xmath224 sites . for reasons that will become clear later , we will color the sites such that we obtain a pattern similar to a chess board . as indicated in fig . [ positionen ] the @xmath225 atoms start in the first column . in contrast to our first model , the white sites are used , too . the rules for the propagation of atoms along the rows are as follows . an atom on a white site is allowed to move forward or backwards by one column if the atoms in the two adjacent rows are currently in the same column . an atom on a black site is only allowed to move one column forward or backwards if it is finally located in between two adjacent atoms in the same column . in the first column , only the atoms on the white sites are allowed to move forward . a possible configuration obtained by applying these rules is shown on the left of fig . [ positionen ] . by induction , we will argue that the configuration of atoms has always the following properties : 1 . atoms located in adjacent rows are either in the same column or in adjacent columns . 2 . for every atom located on a white site there are always atoms on the two adjacent black sites in the same column . the only exception are particles on white sites in the upper - most and lower - most rows which have certainly only one black neighbor . it is clear that these conditions are satisfied by the initial state . in order to prove that they are preserved we observe that an atom that is located on a black site can only move to a white site if the latter is adjacent to two occupied black sites . but then it will satisfy condition ( 2 ) . when an atom on a white site moves to a black site , condition ( 1 ) is certainly preserved since ( 2 ) was true for its initial position . the atoms being on black sites on the boundary can only move to column @xmath71 when the atom in the row below is already in column @xmath71 . now we present a hamiltonian that induces these propagation rules . we start with @xmath226 which is a usual hopping term ( as it appears in hubbard models @xcite ) annihilating the atom at a certain position and creating it at the right or left neighboring site in the same row . physically , one could think of an optical lattice with periodic potentials generated by standing waves which result from the superposition of counter - propagating waves @xcite . to allow tunneling in horizontal direction and to avoid it along the vertical axis one could use high potential walls between rows and low potential walls between columns . this can be achieved by a superposition of high amplitude waves in vertical direction with low amplitude waves in horizontal direction . now we modify the lattice such that there are two types of minima in the potential . the lower potential corresponds to the black sites and the higher potential to the white sites . by superposition of two lattices with wave length @xmath85 and @xmath227 one could also generate a chess - board like alternating potential ( this is called a `` superlattice '' in @xcite ) . we describe the alternating potential by introducing an additional term @xmath228 where @xmath229 denotes the set of black sites . furthermore we introduce an attractive interaction between atoms in adjacent rows which is only active between pairs of atoms that are in the same column . the strength of the attractive interaction has strength @xmath25 , i.e. , we add a term @xmath230 this ensures that all the allowed atom configurations have the same energy : when an atom sitting on a white site moves to a black one the attractive interaction is switched off , this compensates the energy gap between the black and the white site . the same is true for an atom on a black field that moves to a white field since the attractive interaction is then switched on . for the upper - most and the lower - most rows we need to decrease the energy gap between black and white sites because there is only one attractive interaction compensating it . we achieve this by adding a potential @xmath231 given the hamiltonian @xmath232 , one checks easily that an allowed forward or backward motion of an atom leads to a configuration with the same energy . moreover , one can see that every step of an atom that violates the propagation rules would lead to a state with different energy . we observe furthermore that forbidden motions would always lead to a configuration with higher energy and never to lower energy . this additional feature provides the scheme with some thermodynamical stability . now we add the small perturbation @xmath31 ( @xmath25 is again thought to be large ) to the potentials and obtain @xmath233 in analogy to the discussion of our first model we obtain an effective hamiltonian that is given by @xmath234 where @xmath32 is projecting onto the subspace of states having the same energy as the initial atom configuration . the effective hamiltonian modifies the hopping terms such that they are multiplied by an additional number operator checking the position of the atoms in adjacent rows . in other words , each term of the form @xmath235 as well as those of the form @xmath236 are only active if the two black fields are occupied which are either the current two neighbors of the considered atom ( if the latter is located on a white site ) or the future two neighbors ( if it moves to a white site ) . explicitly , we get therefore @xmath237 the operators @xmath140 of these @xmath48-local terms act on black sites in the same column . to implement one - qubit gates in this scheme , we imprint interactions as shown in fig . [ 2erfig ] . the interactions @xmath115 are chosen as in section [ holo ] . a decisive difference to our first model is that the hopping of the atoms does not turn @xmath94 on immediately after @xmath238 was turned off . instead , the atoms pass always a configuration where all interactions are turned off . however , this is irrelevant for the holonomic scheme because the statement of lemma [ adia ] remains certainly true if the evolution is interspersed by time intervals where @xmath73 is completely switched off . to implement two - qubit gates , we design the spin - spin interactions as depicted in fig . [ 3erfig ] . now we have described how interactions can be imprinted that would implement the desired gates given that the whole atom chain passes indeed the circuit region . however , as already stated , it seems to be difficult to derive explicit formulas for the dynamical evolution . we will therefore only consider the corresponding classical random walk and argue that one will have at least probability @xmath239 to find all particles outside the circuit region in the time average . here the circuit region is defined by the first @xmath172 columns of the chess - board where @xmath172 is chosen such that the region contains all interaction stripes . the computation result is then obtained by measuring the spin of the particles after they have left the circuit region . the classical random walk would describe the physical dynamics in the limit of strongly incoherent atom hopping . the incoherent model is , of course inconsistent with our intention to study computation in a closed physical system . the motivation to study the classical walk is that the result gives us some _ small hope _ that the coherent walk behaves also as it is required for computation even though it may clearly be very different from the classical random walk . let @xmath240 be the set of allowed atom configurations on the @xmath224-chess - board and @xmath241 be the graph with nodes @xmath240 . two nodes @xmath242 are adjacent if @xmath243 can be reached from @xmath244 by an allowed step of an atom . the probability distribution on the set of possible atom configurations at some time instant @xmath245 is described by a vector @xmath246 having the @xmath105 nodes of @xmath241 as indices . then a continuous classical random walk on @xmath241 is described by @xcite @xmath247 where @xmath248 is the graph laplacian . its entries are @xmath249 for @xmath250 if @xmath3 is adjacent , and @xmath251 where @xmath252 is the degree of node @xmath71 , i.e. , its number of neighbors . since @xmath241 is connected , @xmath253 is a non - degenerate eigenvalue of @xmath248 @xcite and there is hence a unique stationary distribution . the latter is clearly invariant if we replace each atom configuration with the configuration obtained by reflecting it at the vertical symmetry axis of the chess - board . the latter is defined by a line between column @xmath254 and @xmath255 if @xmath179 , where we have assumed that @xmath179 is even for simplicity . therefore the probability of finding at least one atom of the chain in the right half is at least @xmath239 . since the chain does not tear this implies that the column index of every atom is at least @xmath256 . provided that @xmath257 we find with probability at least @xmath239 all atoms outside the circuit region . we have a dominating hamiltonian @xmath29 and a weak perturbation @xmath31 . in order to use a perturbation theorem of @xcite we first have to rephrase some notations used in @xcite . given a hamiltonian @xmath29 and a small perturbation @xmath31 , both acting on some finite dimensional hilbert space @xmath74 . let @xmath29 have a spectral gap @xmath80 such that no eigenvalues lie between @xmath258 and @xmath259 for some @xmath260 . let @xmath261 be the eigenspace of @xmath29 corresponding to the eigenvalues below @xmath262 and denote its complement by @xmath263 . denote the corresponding spectral projections by @xmath264 . for an arbitrary operator @xmath76 , write @xmath265 for @xmath266 and @xmath267 for @xmath268 . we define the self - energy operator @xmath269 for real - valued @xmath270 by @xmath271 where @xmath272 , called the unperturbed greens function ( resolvent ) in the physics literature , is defined by @xmath273 using the above notations , we rephrase the following theorem which can be found as theorem 8 in @xcite . [ resth ] set @xmath274 with @xmath275 . let there be an effective hamiltonian @xmath28 with spectral width @xmath276 . we assume that @xmath277 . let @xmath278 and furthermore 1 . there is an @xmath279 such that @xmath280 2 . for all @xmath281 such that @xmath282 , @xmath283 . then the restriction @xmath284 of @xmath27 to the eigenspaces with eigenvalues smaller than @xmath262 satisfies @xmath285 we set @xmath286 and @xmath287 . then the condition @xmath288 is true since the expression in eq . ( [ vham ] ) is a sum of less than @xmath9 terms of norm @xmath19 and we have hence @xmath289 due to the assumption @xmath290 . for the same reason , the spectral width @xmath276 of @xmath28 is less than @xmath9 . the operator @xmath291 projects onto @xmath14 and condition @xmath292 required by the theorem is satisfied by definition ( see eq . ( [ effdef ] ) ) . we will now derive a bound on @xmath293 in order to define an appropriate @xmath294 and a corresponding constant @xmath295 . due to @xmath296 and @xmath297 we have @xmath298 the eigenvalues of @xmath299 are @xmath300 according to the number of inactive interactions . therefore the norm of @xmath301 is at most @xmath302 for all @xmath270 with @xmath303 . with @xmath304 we have @xmath305 hence we get @xmath306 for @xmath307 we obtain with @xmath308 @xmath309 we may choose @xmath310 in order to fulfill @xmath311 . we may now use inequality ( [ entsch ] ) and obtain @xmath312 where we have inserted the following results and definitions from above : @xmath313 , @xmath314 , @xmath315 , @xmath316 . we have inserted @xmath317 only in the numerator of the second fraction and replaced it with @xmath19 ( as an upper bound ) at the other places . using @xmath318 for @xmath319 and @xmath36 and @xmath320 we have @xmath321 where we have used @xmath322 . thus , we have obtained the desired upper bound on the norm distance between @xmath37 and @xmath28 . let @xmath323 , where we have described the correspondence between clock states and binary words ( as indicated by the sequence of symbols @xmath19 and @xmath33 at the left of figure [ diagonalpositionen ] ) . then all atoms are outside of @xmath209 if and only if at least @xmath172 symbols @xmath19 have traveled from the right half of the chain to the left half , i.e. , at least @xmath172 fermions are contained in the left half of the interval of length @xmath192 . we have therefore to solve a mixing problem of a `` discrete free fermion gas '' where all particles start in the right interval . first we show that after the time @xmath214 it is likely that at least @xmath172 symbols @xmath19 can be found on the left side . we define the observable @xmath324 where @xmath325 is the projector on the upper state of qubit @xmath71 . the observable @xmath326 counts the number of symbols @xmath19 on the left side . in figure [ possible < ] this corresponds to the number of symbols @xmath19 above the row in the middle . we will estimate the probability that less than @xmath172 symbols @xmath19 have moved to the left by the chebyshev inequality . it states that for any random variable @xmath76 we have @xmath327 where @xmath328 and @xmath329 denote the expectation value and the variance of @xmath76 , respectively . in the sequel we will consider the probability distribution of @xmath326 as the distribution of such a classical random variable . let @xmath330 be the time evolved state after time @xmath245 . then the expectation value of @xmath326 after the time @xmath245 is given by @xmath331 using the dynamics ( [ heis ] ) and ( [ heis2 ] ) we get @xmath332 where we have used that @xmath333 is @xmath19 or @xmath33 depending on whether there the state @xmath334 contains the symbol @xmath19 at this position . each term @xmath335 is the probability for a particle starting at site @xmath105 to be found at site @xmath71 when measured after the time @xmath245 in a single particle quantum walk on a linear chain of length @xmath192 . since the time evolution of such a walk has been discussed in detail in the literature @xcite , we will only describe the implications for our model . consider a particle starting at site @xmath105 with @xmath336 in the notation of eq . ( [ idef ] ) such a fermion corresponds to one of the @xmath337 leftmost symbols @xmath19 . the assumption @xmath338 and hence @xmath339 ensures that the time interval considered in the sequel is sufficiently small compared to the time to reach the boundaries of the lattice . then reflections at the boundaries can be neglected . we have to wait only the time @xmath214 in order to achieve that the width of the wave function of a particle starting at a definite position is much larger than @xmath340 ( see @xcite ) but still smaller than the size @xmath0 of the whole computer . then the probability of finding it on the left half is larger than @xmath341 . recalling that this holds true for each @xmath105 satisfying eq . ( [ jint ] ) and that the number of expected fermions ( @xmath342 symbols `` @xmath19 '' ) in the left half is given by the sum in eq . ( [ exp ] ) , we may choose a time instant @xmath211 such that the expectation value satisfies exactly @xmath343 in order to estimate the variance @xmath344 we observe that eq . ( [ exp ] ) implies @xmath345 moreover , we have @xmath346 where we have used @xmath347 due to eq . ( [ exp ] ) . we rewrite one summand of the first term on the right of eq . ( [ anni ] ) as @xmath348 the inner product can only be nonzero if annihilators meet creators , i.e. , if either @xmath349 and @xmath350 or @xmath351 and @xmath352 or if all indices coincide . in the first case ( including the third ) the term is only non - vanishing for @xmath353 and @xmath354 since @xmath355 is the projection @xmath356 on qubit @xmath105 . in the second case we must have @xmath357 and @xmath358 since @xmath359 is the projection @xmath360 on qubit @xmath361 . hence eq . ( [ anni2 ] ) becomes @xmath362 the first term coincides with @xmath363 by eq . ( [ u_ij ] ) . we rewrite the second term as @xmath364 where the last equality is due to @xmath365 because @xmath366 is unitary ( see eq . . hence we have found @xmath367 combining ( [ vardef ] ) and ( [ kleiner ] ) with ( [ e2 t ] ) straightforward computation shows @xmath368 recall that we have chosen the time instant such that @xmath369 and hence @xmath370 by ineq . ( [ vung ] ) . assume we would find less than @xmath172 symbols @xmath19 on the left half . then the random variable defined by @xmath326-measurements would deviate at least @xmath371 from its expectation value . hence we can apply eq . ( [ tsche ] ) with @xmath370 and @xmath372 and find that the probability of finding less than @xmath172 symbols in the left half is at most @xmath373 . this completes the proof . in analogy to the proof of theorem [ passingtime ] we will compute the expectation value and the variance of @xmath326 in the time average state and then use the chebyshev inequality ( [ tsche ] ) . the time average expectation value of the fermion number on the left half is given by averaging eq . ( [ exp ] ) over all @xmath245 : @xmath374 where the last equality follows from eq . ( [ u_ij ] ) . recall that we interpret each summand for @xmath375 as the probability of finding a particle ( that has started at site @xmath105 ) in the left half of the chain when measured at a random time instant . we will show that it is close to @xmath239 up to an error in @xmath376 . it is known @xcite that the hamiltonian @xmath204 on @xmath197 generating the walk ( i.e. , the adjacency matrix of the `` path graph @xmath377 '' ) has @xmath192 different eigenvalues @xmath378 the eigenspaces of @xmath204 , are therefore one - dimensional and the @xmath295th eigenvector is given by @xcite @xmath379 where the normalization factor is @xmath380 for @xmath381 and @xmath382 for @xmath383 . using these eigenvectors , we may write the time average density matrix of a particle that has started at position @xmath105 as @xmath384 by evaluating the probability to be at position @xmath71 using this state we may compute the time average in eq . ( [ tav ] ) and obtain @xmath385 each eigenvector defines a probability distribution @xmath386 on @xmath387 by @xmath388\,.\ ] ] the spatial probability oscillations are given by waves with frequencies @xmath389 . we will refer to those frequencies @xmath390 that satisfy @xmath391 as _ high _ frequencies ( they are neither close to @xmath33 nor close to @xmath392 ) and to the others as _ low _ frequencies . for a high frequency wave , the probability distribution @xmath393 leads almost to equal probabilities for both halfs . this is seen from @xmath394 the sum over the oscillating terms can be bounded from above by observing @xmath395 where we have used that @xmath390 differs from both @xmath33 and @xmath392 by @xmath396 and the last inequality follows directly from the geometric sum formula @xmath397 using eq . ( [ limqr ] ) and eq . ( [ pdef ] ) , the term in the bracket in eq . ( [ tav ] ) can be written as @xmath398 neglecting the low frequency terms @xmath386 in this sum can only cause an error in @xmath376 due to @xmath399 taking into account that there are @xmath400 such terms . hence the probability of finding a particle that has started at any site @xmath105 in the left half is close to @xmath239 up to an error in @xmath376 . by summation over all @xmath375 we find @xmath401 . to derive bounds on the variance of @xmath326 in the time average state we use eq . ( [ anni ] ) and observe @xmath402 where the last inequality is due to ineq . ( [ kleiner ] ) . we may rewrite the term in the big bracket in equation ( [ secondline ] ) using eq . ( [ u_ij ] ) into @xmath403 recall that the coefficients @xmath404 are the matrix entries of a unitary that describes a random walk on a linear chain . note furthermore that the corresponding unitary group @xmath366 is generated by a real - symmetric hamiltonian and we have therefore @xmath405 . thus , we are allowed to interpret @xmath406 as the probability for finding a particle at position @xmath105 that has started at position @xmath128 , which interchanges the original roles of @xmath128 and @xmath105 . this will be convenient in the sequel . the product of the sums on the right hand side of eq . ( [ rein ] ) can either be interpreted as arising from two independent walks of two particles or the walk of one particle in two dimensions . explicitly , we consider a dynamical evolution in @xmath407 generated by the hamiltonian @xmath408 since the hamiltonian ( [ 2hamil ] ) is the adjacency graph of the square lattice @xmath409 we can consider the right hand of eq . ( [ rein ] ) as the probability of finding a particle starting at @xmath410 in a quantum walk on the square lattice in the `` target quadrant '' @xmath411 . we want to prove that we have sufficient mixing in order to obtain @xmath412 for the time average of term ( [ rein ] ) up to an error in @xmath376 . we proceed similarly as for the one - dimensional walk with the essential difference that the spectrum of the hamiltonian ( [ 2hamil ] ) is degenerate since @xmath413 and @xmath414 have the same eigenvalues . we denote the projection onto their span by @xmath415 . the rank of @xmath415 is @xmath93 for @xmath416 and @xmath19 for @xmath350 . using these projections , we may compute the time average of ( [ rein ] ) for fixed @xmath21 by @xmath417 where we have defined the coefficients @xmath418 and probability measures @xmath419 on the square lattice by @xmath420 and @xmath421 , the state vector given by normalizing @xmath422 for a given @xmath423 is a superposition of cosine wave functions with frequency vector @xmath424 with another wave having the vector @xmath425 . the corresponding probability distributions @xmath419 contains then terms with frequencies @xmath426 . we refer to a term as low frequency term whenever at least one of these frequencies is low . for each `` high frequency distribution '' @xmath419 the probability for the target quadrant is @xmath412 up to an error in @xmath376 . this is seen in straightforward analogy to the corresponding argument for the one - dimensional walk . moreover , the contribution of the low frequency terms in ( [ 2dimtav ] ) can be bounded from above by the number of low frequency terms ( which is in @xmath427 ) times the maximal coefficient @xmath428 of each term ( which is in @xmath429 due to @xmath430 ) . we conclude that the total contribution of all low frequency terms to ( [ 2dimtav ] ) is in @xmath376 . hence the total sum in eq . ( [ 2dimtav ] ) is @xmath412 up to an error in @xmath376 . we conclude @xmath431 using eq . ( [ secondline ] ) we conclude @xmath432 and hence @xmath433 hence the variance @xmath434 is in @xmath435 and the probability that a measurement of @xmath326 leads to a result with less than @xmath436 converges to zero with @xmath437 by the chebyshev inequality . therefore , the probability of finding less than @xmath436 fermions on the left half ( i.e. , the atom chain has left the circuit region ) tends to zero , too .

simultaneous wireless information and power transfer ( swipt ) has attracted significant attention in the communication community . the problem of waveform design has however never been addressed so far . in this paper , we first investigate how a communication waveform ( ofdm ) and a power waveform ( multisine ) compare with each other in terms of harvested energy . we show that due to the non - linearity of the rectifier and the randomness of the information symbols , the ofdm waveform is less efficient than the multisine waveform for wireless power transfer . this observation motivates the design of a novel swipt transceiver architecture relying on the superposition of multisine and ofdm waveforms at the transmitter and a power - splitter receiver equipped with an energy harvester and an information decoder . the superposed swipt waveform is optimized so as to maximize the rate - energy region of the whole system . its design is adaptive to the channel state information and result from a posynomial maximization problem that originates from the non - linearity of the energy harvester . numerical results illustrate the performance of the derived waveforms and swipt architecture . key ( and refreshing ) observations are that 1 ) a power waveform ( superposed to a communication waveform ) is useful to enlarge the rate - energy region of swipt , 2 ) a combination of power splitting and time sharing is in general the best strategy , 3 ) exploiting the nonlinearity of the rectifier is essential to design efficient swipt architecture , 4 ) a non - zero mean gaussian input distribution outperforms the conventional capacity - achieving zero - mean gaussian input distribution .

1607.05602

simultaneous wireless information and power transfer ( swipt ) has recently attracted significant attention in academia , with works addressing many scenarios , a.o . mimo broadcasting @xcite , architecture @xcite , interference channel @xcite , broadband system @xcite , relaying @xcite . wireless power transfer ( wpt ) is a fundamental building block of swipt and the design of an efficient swipt architecture fundamentally relies on the ability to design efficient wpt . the major challenge with wpt , and therefore swipt , is to find ways to increase the dc power level at the output of the rectenna without increasing the transmit power . to that end , the vast majority of the technical efforts in the literature have been devoted to the design of efficient rectennas . the rectenna is made of a non - linear device followed by a low - pass filter to extract a dc power out of an rf input signal . the amount of dc power collected is a function of the input power level and the rf - to - dc conversion efficiency . interestingly , the rf - to - dc conversion efficiency is not only a function of the rectenna design but also of its input waveform @xcite . this observation has triggered very recent interests on wireless power waveform design in the signal processing literature @xcite . the objective is to understand how to make the best use of a given rf spectrum in order to deliver a maximum amount of dc power at the output of a rectenna . this problem can be formulated as a link optimization where transmit waveforms are adaptively designed as a function of the channel state information ( csi ) so as to maximize the dc power at the output of the rectifier . in @xcite , the waveform design problem for wpt has been tackled by introducing a simple and tractable analytical model of the non - linearity of the diode through the second and higher order terms in the taylor expansion of the diode characteristics . comparisons were also made with a linear model of the rectifier , that only accounts for the second order term . assuming perfect channel state information at the transmitter ( csit ) can be attained , relying on both the linear and non - linear models , an optimization problem was formulated to adaptively change on each transmit antenna a multisine waveform as a function of the csi so as to maximize the output dc current at the energy harvester . important conclusions of @xcite are that 1 ) multisine waveforms designed accounting for nonlinearity are spectrally more efficient that those designed based on a linear model of the rectifier , 2 ) the linear model does not characterize correctly the rectenna behaviour and leads to inefficient multisine waveform design , 3 ) rectifier nonlinearity is key to design efficient wireless powered systems . interestingly , the swipt literature has so far entirely relied on the aforementioned linear model of the rectifier , e.g. @xcite and subsequent works . the problem of swipt waveform and transceiver design that accounts for the nonlinearity of the rectifier has indeed never been addressed so far . in view of the recent results in @xcite , it is expected that accounting for the rectifier nonlinearity is key to efficient swipt design . in this paper , we address the important problem of waveform and transceiver design for swipt . the contributions of the paper are summarized as follows . first , we leverage the analytical model of the rectenna non - linearity introduced in @xcite and investigate how a communication waveform ( ofdm ) and a power waveform ( multisine ) compare with each other in terms of harvested energy . comparison is also made with the linear model commonly used in the swipt literature @xcite . scaling laws of the harvested energy with multisine and ofdm waveforms are analytically derived as a function of the number of sinewaves and the propagation conditions . we show that with the nonlinear model of the rectifier , there is a clear benefit of using a deterministic multisine over an ofdm waveform for wpt . on the other hand , the benefits vanish when we consider the linear model . second , we introduce a novel swipt transceiver architecture relying on the superposition of multisine and ofdm waveforms at the transmitter and a power - splitter receiver equipped with an energy harvester and an information decoder . both cases where the receiver does and does not have the capability of cancelling the multisine waveform are considered . the swipt multisine / ofdm waveforms and the power splitter are jointly optimized so as to maximize the rate - energy region of the whole system . the superposed swipt waveforms are adaptive to the channel state information and result from a posynomial maximization problem that originates from the non - linearity of the energy harvester . third , numerical results illustrate the performance of the derived waveforms and swipt architecture . key observations are that 1 ) a power waveform ( superposed to a communication waveform ) is always useful to enlarge the rate - energy region of swipt , 2 ) a combination of power splitting and time sharing is in general the best strategy , 3 ) a non - zero mean gaussian input distribution outperforms the conventional capacity - achieving zero - mean gaussian input distribution . those observations are consequences of the non - linearity of the rectifier and highlights that exploiting the nonlinearity of the rectifier is essential to design an efficient swipt architecture . _ organization : _ section [ ms_vs_ofdm ] introduces the multisine and ofdm waveforms for wpt , section [ section_eh ] introduces the rectenna model , section [ wpt_design ] optimizes multisine and ofdm waveforms for wpt and discusses the impact of the waveform type ( multisine vs ofdm ) on the harvested energy . section [ swipt_section ] introduces the swipt architecture , section [ section_swipt_waveform ] addresses the swipt waveform design , section [ simulations ] evaluates the performance and section [ conclusions ] concludes the work . _ notations : _ bold lower case and upper case letters stand for vectors and matrices respectively whereas a symbol not in bold font represents a scalar . @xmath0 refers to the frobenius norm a matrix . @xmath1 refers to the dc component of a signal . @xmath2 refers to the expectation operator taken over the distribution of the random variable @xmath3 ( @xmath3 may be omitted for readability if the context is clear ) . @xmath4 refers to the conjugate of a scalar . @xmath5 and @xmath6 represent the transpose and conjugate transpose of a matrix or vector respectively . we consider two types of waveform : multisine and ofdm . consider the transmitter made of @xmath7 antennas and @xmath8 sinewaves whose multisine transmit signal at time @xmath9 on transmit antenna @xmath10 is given by @xmath11 with @xmath12 where @xmath13 and @xmath14 refer to the amplitude and phase of the @xmath15 sinewave at frequency @xmath16 on transmit antenna @xmath17 , respectively . we assume for simplicity that the frequencies are evenly spaced , i.e. @xmath18 with @xmath19 the frequency spacing . the magnitudes and phases of the sinewaves can be collected into @xmath20 matrices @xmath21 and @xmath22 . the @xmath23 entry of @xmath21 and @xmath22 write as @xmath13 and @xmath14 , respectively . the transmitter is subject to an average transmit power constraint @xmath24 . stacking up all transmit signals , we can write the transmit signal vector as @xmath25 where @xmath26^t$ ] . the multi - antenna transmitted sinewaves propagate through a multipath channel , characterized by @xmath27 paths whose delay , amplitude , phase and direction of departure ( chosen with respect to the array axis ) are respectively denoted as @xmath28 , @xmath29 , @xmath30 and @xmath31 , @xmath32 . we assume transmit antennas are closely located so that @xmath28 , @xmath29 and @xmath30 are the same for all transmit antennas ( assumption of a narrowband balanced array ) @xcite . denoting @xmath33 with @xmath34 the phase shift between the @xmath35 transmit antenna and the first one where @xmath36 is the inter - element spacing , @xmath37 the wavelength of the @xmath15 sinewave . ] , the signal transmitted by antenna @xmath17 and received at the single - antenna receiver after multipath propagation can be written as @xmath38 where the amplitude @xmath39 and the phase @xmath40 are such that @xmath41 with @xmath42 the frequency response of the channel of antenna @xmath17 at @xmath16 . the vector channel is defined as @xmath43 $ ] . the total received signal comprises the sum of over all transmit antennas , namely @xmath44 where @xmath45 . the baseband ofdm signal over one symbol duration @xmath46 can be written as @xmath47 , @xmath48 , where @xmath49 refers to the precoded input symbol on frequency tone @xmath50 and antenna @xmath17 . we assume a capacity achieving gaussian input distribution such that @xmath51 are i.i.d . circularly symmetric complex gaussian distributed . we further write the precoder @xmath52 and the input symbol @xmath53 . after adding the cyclic prefix over duration @xmath54 , it comes to @xmath47 , @xmath55 . vector - wise , the baseband ofdm signal vector on tone @xmath50 writes as @xmath56^t=\sum_{n=0}^{n-1 } \mathbf{x}_{n } e^{j\frac{2\pi t}{t}n}$ ] with @xmath57 and @xmath58^t$ ] is the precoder . after upconversion , the transmitted ofdm signal on antenna @xmath17 is written as @xmath59 where @xmath60 with @xmath61 and @xmath62 . we also define @xmath63 and @xmath64 where @xmath65 . vector - wise , @xmath66^t$ ] . we define @xmath20 matrices such that the @xmath23 entry of matrix @xmath67 , @xmath68 , @xmath69 , @xmath70 write as @xmath71 , @xmath72 , @xmath73 , @xmath74 , respectively . the ofdm waveform is subject to the power constraint @xmath75 . after multipath , the received signal can be written as @xmath76 where @xmath77 and @xmath78 . let us also define @xmath79 such that @xmath80 . in @xcite , a simple and tractable model of the rectifier nonlinearity in the presence of multisine excitation was derived and its validity verified through circuit simulations . in this paper , we reuse the same model and further expand it to ofdm excitation . a multisine waveform is determinsitic while an ofdm waveform exhibits randomness due to information symbols @xmath81 . this randomness has an impact on the amount of harvested energy and needs to be captured in the rectenna model . in the next two subsections , we introduce the antenna equivalent circuit and rectifier non - linearity . the antenna model reflects the power transfer from the antenna to the rectifier through the matching network . the signal impinging on the antenna is @xmath82 for a multisine or @xmath83 for an ofdm waveform . for readability , let us drop the index for a moment and simply refer to a received signal of the form @xmath84 , i.e. applicable to both multisine and ofdm . @xmath85 has an average power @xmath86 . a lossless antenna can be modelled as a voltage source @xmath87 followed by a series resistance @xmath88 ( see fig [ antenna_model ] ) . let @xmath89 denote the input impedance of the rectifier with the matching network . assuming perfect matching ( @xmath90 , @xmath91 ) , all the available rf power @xmath92 is transferred to the rectifier and absorbed by @xmath93 , so that @xmath94 and @xmath95 . since @xmath96 , @xmath87 can be formed as @xmath97 . we also assume that the antenna noise is too small to be harvested so as no antenna noise term is added and @xmath98 is delivered as such to the rectifier . consider a rectifier composed of a single series diode followed by a low - pass filter with load ( see fig [ antenna_model ] ) . denoting the voltage drop across the diode as @xmath99 where @xmath98 is the input voltage to the diode and @xmath100 is the output voltage across the load resistor , a tractable behavioural diode model is obtained by taylor series expansion of the diode characteristic equation @xmath101 ( with @xmath102 the reverse bias saturation current , @xmath103 the thermal voltage , @xmath50 the ideality factor assumed equal to @xmath104 ) around a quiescent operating point @xmath105 . the diode current is then written as @xmath106 where @xmath107 and @xmath108 , @xmath109 . assume a steady - state response and an ideal low pass filter such @xmath100 is at constant dc level . if @xmath110 is chosen as @xmath111 , can be simplified as @xmath112 the problem at hand will be the design of @xmath113 or @xmath114 such that the output dc current is maximized . under the ideal rectifier assumption and a deterministic incoming waveform @xmath85 ( i.e. with fixed weights @xmath115 and phases @xmath116 ) , the current delivered to the load in a steady - state response is constant and given by @xmath117 . in order to make the optimization tractable , we truncate the taylor expansion to the @xmath118 order . we consider two models : a non - linear model that truncates the taylor expansion to the @xmath118 order but retains the fundamental non - linear behaviour of the diode and a linear model that truncates to the second order term and ignores the non - linearity . after truncation , the output dc current approximates as @xmath119 let us first consider a _ multisine waveform_. applying a received signal of the form @xmath84 to and averaging over time , we get an approximation of the dc component of the current at the output of the rectifier ( and the low - pass filter ) with a multisine excitation over a multipath channel as @xmath120 where @xmath121 and @xmath122 are detailed in and , respectively ( at the top of next page ) . there is no odd ( first , third , fifth , etc ) order terms since @xmath123 for @xmath124 odd . we note that the second order term is linear , with the dc power being the sum of the power harvested on each frequency . on the other hand , even terms with @xmath125 such as are responsible for the non - linear behavior of the diode since they are function of terms expressed as the product of contributions from different frequencies . note that the non - linear model in with @xmath126 has been validated in @xcite using a circuit simulator for multisine excitation . @xmath127=\frac{1}{2}\left[\sum_{n=0}^{n-1 } \left|\mathbf{h}_n\mathbf{w}_{p , n } \right|^2\right]=\frac{1}{2}\left[\sum_{n=0}^{n-1 } \sum_{m_0,m_1 } s_{p , n , m_0}s_{p , n , m_1}a_{n , m_0}a_{n , m_1}\cos\left(\psi_{p , n , m_0}-\psi_{p , n , m_1}\right)\right],\label{y_dc_2_2}\\ \mathcal{a}\left\{y_p(t)^4\right\ } & = \frac{3}{8}\re\left\{\sum_{{\genfrac{}{}{0pt}{}{n_0,n_1,n_2,n_3}{n_0+n_1=n_2+n_3}}}\mathbf{h}_{n_0}\mathbf{w}_{p , n_0}\mathbf{h}_{n_1}\mathbf{w}_{p , n_1}\left(\mathbf{h}_{n_2}\mathbf{w}_{p , n_2}\right)^*\left(\mathbf{h}_{n_3}\mathbf{w}_{p , n_3}\right)^*\right\},\label{y_dc_4_1b}\\ & = \frac{3}{8}\left[\sum_{{\genfrac{}{}{0pt}{}{n_0,n_1,n_2,n_3}{n_0+n_1=n_2+n_3}}}\sum_{{\genfrac{}{}{0pt}{}{m_0,m_1,}{m_2,m_3}}}\bigg[\prod_{j=0}^3s_{p , n_j , m_j}a_{n_j , m_j}\bigg]\cos(\psi_{p , n_0,m_0}+\psi_{p , n_1,m_1}-\psi_{p , n_2,m_2}-\psi_{p , n_3,m_3})\right].\label{y_dc_4_3 } \ ] ] let us now consider the _ ofdm waveform_. it can be viewed as a multisine waveform for a fixed set of input symbols @xmath128 . hence , we can also write the dc component of the current at the output of the rectifier ( and the low - pass filter ) with an ofdm excitation and fixed set of input symbols over a multipath channel as @xmath129 . similar expressions as and can be written for @xmath130 and @xmath131 for a fixed set of input symbols @xmath128 . however , contrary to the multisine waveform , the input symbols @xmath128 of the ofdm waveform change randomly every symbol duration @xmath132 . for a given channel impulse response , the proposed model for the dc current with an ofdm waveform is obtained by averaging out over the distribution of the input symbols @xmath128 such that @xmath133 for @xmath134 with @xmath124 even , the dc component is first extracted for a given set of amplitudes @xmath135 and phases @xmath136 and then expectation is taken over the randomness of the input symbols @xmath81 . due to the i.i.d . circularly symmetric complex gaussian distribution of the input symbols , @xmath137 is exponentially distributed with @xmath138 and @xmath139 is uniformly distributed . from the moments of an exponential distribution , we also have that @xmath140 . this enables to express and as a function of @xmath141 . @xmath142= \frac{1}{2}\left[\sum_{n=0}^{n-1 } \left|\mathbf{h}_n\mathbf{w}_{i , n } \right|^2 \right]\label{e_y_i_2}\\ \mathcal{e}\left\{\mathcal{a}\left\{y_i(t)^4\right\}\right\ } & = \frac{3}{8}\mathcal{e}\left\{\sum_{{\genfrac{}{}{0pt}{}{n_0,n_1,n_2,n_3}{n_0+n_1=n_2+n_3}}}\sum_{{\genfrac{}{}{0pt}{}{m_0,m_1,}{m_2,m_3}}}\bigg[\prod_{j=0}^3\tilde{s}_{i , n_j , m_j}a_{n_j , m_j}\bigg]\cos(\tilde{\psi}_{i , n_0,m_0}+\tilde{\psi}_{i , n_1,m_1}-\tilde{\psi}_{i , n_2,m_2}-\tilde{\psi}_{i , n_3,m_3})\right\}\nonumber\\ & = \frac{6}{8}\bigg[\sum_{n_0,n_1}\sum_{{\genfrac{}{}{0pt}{}{m_0,m_1,}{m_2,m_3}}}\bigg[\prod_{j=0,2}s_{i , n_0,m_j}a_{n_0,m_j}\bigg]\bigg[\prod_{j=1,3}s_{i , n_1,m_j}a_{n_1,m_j}\bigg ] \cos(\psi_{i , n_0,m_0}+\psi_{i , n_1,m_1}-\psi_{i , n_0,m_2}-\psi_{i , n_1,m_3})\bigg]\label{e_y_i_40}\\ & = \frac{6}{8}\left[\sum_{n=0}^{n-1 } \left|\mathbf{h}_n\mathbf{w}_{i , n } \right|^2 \right]^2\label{e_y_i_4}\end{aligned}\ ] ] the linear model was orginally introduced a few decades ago in @xcite and is nowadays extensively used throughout the swipt literature , e.g. @xcite . it could be argued that if @xmath143 is very small ( i.e. for a very low input power ) , the higher order terms would not contribute much to @xmath144 . hence , the linear model truncates the taylor expansion to @xmath145 such that @xmath146 for multisine and @xmath147 for ofdm . it therefore completely omits the non - linearity behavior of the rectifier . the linear model is motivated by its simplicity rather than its accuracy . in @xcite , the linear model was shown to be inaccurate in predicting multisine waveform performance . nevertheless , the loss incurred by using a linear vs a non - linear model for the design of wpt based on other types of waveforms ( i.e. others than deterministic multisine ) and the design of swipt has never been addressed so far . in the next section , we study the design of multisine and ofdm waveforms under the assumption of a linear and non - linear model . assuming the csi ( in the form of frequency response @xmath148 ) is known to the transmitter , we aim at finding the optimal set of amplitudes and phases ( @xmath149 for multisine and @xmath150 for ofdm ) that maximizes @xmath144 , i.e.@xmath151 for @xmath152 . following @xcite , problem can equivalently be written as @xmath153 where @xmath154 for the multisine waveform , and @xmath155 for the ofdm waveform . in and , we define @xmath156 ( with a slight abuse of notation ) . assuming @xmath157 , a diode ideality factor @xmath158 and @xmath159 , typical values of those parameters for second and fourth order are given by @xmath160 and @xmath161 ( and will be used as such in any evaluation in the sequel ) . the problem of ( deterministic ) multisine waveform design with a linear and non - linear rectenna model has been addressed in @xcite . the linear model leads to the equivalent problem @xmath162 subject to @xmath163 whose solution is the adaptive single - sinewave ( ass ) strategy @xmath164 the ass performs a matched beamformer on a single sinewave , namely the one corresponding to the strongest channel @xmath165 . on the other hand , the non - linear model leads to a posynomial maximization problem that can be formulated as a reverse geometric program and solved iteratively . interestingly , for multisine waveforms , the linear and nonlinear models lead to radically different strategies . the former favours transmission on a single frequency while the latter favours transmission over multiple frequencies . design based on the linear model was shown to be inefficient and lead to significant loss over the nonlinear - based design . the design of ofdm waveform is rather different . from and , both the second and fourth order terms are exclusively function of @xmath166 . this shows that both the linear and nonlinear model - based design of ofdm waveforms for wpt lead to the ass strategy and the optimum @xmath167 should be designed according to . this is in sharp contrast with the deterministic multisine waveform design and originates from the fact that the ofdm waveform is subject to randomness due to the presence of input symbols @xmath81 . note that this ass strategy has already appeared in the swipt literature with ofdm transmission , e.g. @xcite . in order to assess the performance benefits of a multisine waveform over an ofdm waveform for wpt , we quantify how @xmath168 scales as a function of @xmath8 . for simplicity we truncate the taylor expansion to the fourth order ( @xmath126 ) . we consider frequency - flat and frequency - selective channels . we assume that the complex channel gains @xmath169 are modeled as independent circularly symmetric complex gaussian random variables . @xmath29 are therefore independent rayleigh distributed such that @xmath170 with @xmath171 . the impulse responses have a constant average received power normalized to 1 such that @xmath172 . in the frequency flat channel , @xmath173 and @xmath174 @xmath175 this is met when the bandwidth of the multisine waveform @xmath176 is much smaller than the channel coherence bandwidth . in the frequency selective channel , we assume that @xmath177 and frequencies @xmath16 are far apart from each other such that the frequency domain circularly symmetric complex gaussian random channel gains @xmath148 fade independently ( phase and amplitude - wise ) across frequencies and antennas . taking the expectation over the distribution of the channel , we denote @xmath178 . leveraging the derivation of the scaling laws for multisine waveforms designed based on the ass in @xcite , we can easily compute the scaling laws of ofdm waveform designed based on ass . table [ scaling_law_summary ] summarizes the scaling laws for both multisine and ofdm waveforms based on the ass strategy . it also compares with the upmf strategy for multisine ( introduced in @xcite ) that consists in uniformly allocating power to all sinewaves and matching the waveform phase to the channel phase . such a upmf srategy is suboptimal for multisine excitation and its scaling law is therefore a lower bound on what can be achieved with the optimal strategy of . recall that the ass strategy for ofdm is optimal for the maximization of @xmath168 with the linear and nonlinear model . the ass for multisine is only optimal for the linear model . as it can be seen from the scaling laws , the upmf strategy leads to a linear increase of @xmath179 with @xmath8 while the ass strategy for multisine and ofdm only lead to at most a logarithmic increase with @xmath8 . [ cols="<,<,<,<",options="header " , ] [ scaling_law_summary ] [ obs_main ] the linear model highlights that there is no difference in using a deterministic multisine waveform or an ofdm waveform for wpt , since according to this model the harvested energy is a function of @xmath180 . hence ofdm and multisine waveforms are equally suitable . on the other hand , the nonlinear model highlights that there is a clear benefit of using a deterministic multisine over a ( randomized ) ofdm waveform in wpt , with the scaling law of multisine significantly outperforming that of ofdm . while @xmath179 scales linearly with @xmath8 with a multisine waveform , it scales at most logarithmically with @xmath8 with an ofdm waveform . this loss in scaling law is inherently due to the randomness of information symbols . observation [ obs_main ] motivates a novel swipt architecture that is based on a superposition of multisine waveform for efficient wpt and ofdm waveform for efficient wireless information transfer ( wit ) . in figure [ wipt_transceiver ] , we introduce the proposed swipt architecture where power and information are transmitted simultaneously from one transmitter to one receiver equipped with a power splitter . + + the swipt waveform on antenna @xmath17 , @xmath181 , consists in the superposition of one multisine waveform @xmath182 at frequencies @xmath18 , @xmath183 for wpt and one ofdm waveform @xmath184 at the same frequencies for wit . the total transmitted swipt waveform on antenna @xmath17 writes as @xmath185 with @xmath186 and @xmath187 . this is illustrated in figure [ wipt_transceiver](a ) . due to the superposition of the two waveforms , the total average transmit power constraint now writes as @xmath188 . at the receiver , we can write the received signal as @xmath189 , i.e. the sum of two contributions at the output of the channel , namely one from wpt @xmath190 and the other from wit @xmath191 , expressed as in and , respectively . using a power splitter with a power splitting ratio @xmath192 and assuming perfect matching ( as in section [ antenna_eq_circuit ] ) , the input voltage signal @xmath193 is conveyed to the input of the energy harvester ( eh ) while @xmath194 is conveyed to the information decoder ( i d ) . [ remark_distribution ] it is worth noting the effect of the deterministic multisine waveform on the input distribution in . recall that @xmath195 . hence @xmath196 and the effective input distribution on a given frequency and antenna is not zero mean . the magnitude @xmath197 is ricean distributed with a k - factor on frequency @xmath50 and antenna @xmath17 given by @xmath198 . let us consider two types of information decoders : the perfect cancellation ( pc ) receiver and the no - cancellation ( nc ) receiver . : since @xmath182 does not contain any information , it is deterministic and can be cancelled at the i d receiver . therefore , after down - conversion and adc , the contribution of the wpt waveform is subtracted from the received signal ( fig [ wipt_transceiver](b ) ) . conventional ofdm processing is then conducted , namely removing the cyclic prefix and performing fft . we can write the equivalent baseband system model of the i d receiver as @xmath199 where @xmath200 is the awgn noise on tone @xmath50 ( with variance @xmath201 ) originating from the antenna and the rf to baseband down - conversion . assuming perfect cancellation ( pc ) , the rate writes as @xmath202 naturally , @xmath203 can never be larger than the maximum rate achievable when @xmath204 , i.e. @xmath205 , which is obtained by performing matched filtering on each subcarrier and water - filling power allocation across subcarrier . : in the event that the receiver is not equipped with a wpt multisine waveform canceller ( nc ) , the equivalent baseband system model of the i d receiver is @xmath206 the i d receiver treats the contribution from the wpt as noise and the rate writes as @xmath207 contrary to wpt , in swipt , @xmath208 is conveyed to the input of the energy harvester . hence both wpt @xmath190 and wit @xmath191 now contribute to the dc component @xmath209 assuming @xmath126 , we can compute @xmath168 as in , where we use the fact that @xmath210 , @xmath211 , @xmath212 and @xmath213 . we can now define the achievable rate - harvested energy ( or more accurately rate - dc current ) region as @xmath215\leq p \big\}.\end{gathered}\ ] ] optimal values @xmath216,@xmath217,@xmath218,@xmath219 are to be found in order to enlarge as much as possible the rate - harvested energy region . we derive a methodology that is general to cope with any taylor expansion order @xmath220 but the derived algorithm works for any @xmath220 . ] . looking at and , it is easy to conclude that matched filtering w.r.t . the phases of the channel is optimal from both rate and harvested energy maximization perspective . this leads to the same phase decisions as for wpt in @xcite , namely @xmath221 and guarantees all arguments of the cosine functions in @xmath222 ( expressions and ) and in @xmath223 ( expressions and ) to be equal to 0 . @xmath218 and @xmath224 are obtained by collecting @xmath225 and @xmath226 @xmath227 into a matrix , respectively . @xmath228\right ] + \frac{3k_4\rho^2}{8}r_{ant}^2\left[\sum_{{\genfrac{}{}{0pt}{}{n_0,n_1,n_2,n_3}{n_0+n_1=n_2+n_3}}}\sum_{{\genfrac{}{}{0pt}{}{m_0,m_1,}{m_2,m_3}}}\bigg[\prod_{j=0}^3s_{p , n_j , m_j}a_{n_j , m_j}\bigg]\right]\nonumber\\ & \hspace{0.6cm}+\frac{k_2 \rho}{2}r_{ant}\left[\sum_{n=0}^{n-1 } \sum_{m_0,m_1 } \bigg[\prod_{j=0}^1 s_{i , n , m_j}a_{n , m_j}\bigg]\right]+ \frac{3 k_4\rho^2}{4}r_{ant}^2\left[\sum_{n=0}^{n-1 } \sum_{m_0,m_1 } \bigg[\prod_{j=0}^1 s_{i , n , m_j}a_{n , m_j}\bigg]\right]^2 \nonumber\\ & \hspace{0.6cm}+\frac{3k_4\rho^2}{2 } r_{ant}^2\left[\sum_{n=0}^{n-1 } \sum_{m_0,m_1 } \bigg[\prod_{j=0}^1 s_{p , n , m_j}a_{n , m_j}\bigg]\right]\left[\sum_{n=0}^{n-1 } \sum_{m_0,m_1 } \bigg[\prod_{j=0}^1 s_{i , n , m_j}a_{n , m_j}\bigg]\right]\end{aligned}\ ] ] with such phases @xmath218 and @xmath224 , @xmath229 can be finally written as . similarly we can write @xmath230 where @xmath231 . recall from @xcite that a monomial is defined as the function @xmath232 where @xmath233 and @xmath234 . a sum of @xmath235 monomials is called a posynomial and can be written as @xmath236 with @xmath237 where @xmath238 . as we can see from , @xmath229 is a posynomial . in order to identify the achievable rate - energy region , we formulate the optimization problem as an energy maximization problem subject to transmit power and rate constraints @xmath239\leq p,\\ & i_{pc}(\mathbf{s}_i,\mathbf{\phi}_i^{\star},\rho)\geq \bar{r}.\end{aligned}\ ] ] it therefore consists in maximizing a posynomial subject to constraints . unfortunately this problem is not a standard geometric program ( gp ) but it can be transformed to an equivalent problem by introducing an auxiliary variable @xmath240 @xmath241\leq p,\\ & t_0/z_{dc}(\mathbf{s}_p,\mathbf{s}_i,\mathbf{\phi}_p^{\star},\mathbf{\phi}_i^{\star},\rho)\leq1,\\&2^{\bar{r}}/\left[\prod_{n=0}^{n-1}\left(1+\frac{(1-\rho)}{\sigma_n^2 } c_n\right)\right]\leq 1.\label{swipt_opt_problem_eq_4}\end{aligned}\ ] ] this is known as a reverse geometric program @xcite . a similar problem also appeared in the wpt waveform optimization @xcite . note that @xmath242 and @xmath243 $ ] are not posynomials , therefore preventing the use of standard gp tools . the idea is to replace the last two inequalities ( in a conservative way ) by making use of the arithmetic mean - geometric mean ( am - gm ) inequality . let @xmath244 be the monomial terms in the posynomial @xmath245 . similarly we define @xmath246 as the set of monomials of the posynomial @xmath247 with @xmath248 . for a given choice of @xmath249 and @xmath250 with @xmath251 and @xmath252 , we perform single condensations and write the standard gp as @xmath253\leq p,\\ & t_0\prod_{k=1}^k\left(\frac{g_k(\mathbf{s}_p,\mathbf{s}_i,\mathbf{\phi}_p^{\star},\mathbf{\phi}_i^{\star},\rho)}{\gamma_k}\right)^{-\gamma_k}\leq1,\\ & 2^{\bar{r}}\prod_{n=0}^{n-1}\prod_{k=1}^{k_n}\left(\frac{g_{nk}(\mathbf{s}_i,\bar{\rho})}{\gamma_{nk}}\right)^{-\gamma_{nk}}\leq 1,\\ & \rho+\bar{\rho}\leq 1.\label{standard_gp_swipt_4}\end{aligned}\ ] ] it is important to note that the choice of @xmath254 plays a great role in the tightness of the am - gm inequality . an iterative procedure can be used where at each iteration the standard gp - is solved for an updated set of @xmath254 . assuming a feasible set of magnitude @xmath255 and @xmath256 and power splitting ratio @xmath257 at iteration @xmath258 , compute at iteration @xmath124 @xmath259 @xmath260 and @xmath261 , @xmath183 , @xmath262 and then solve problem - to obtain @xmath263 , @xmath264 and @xmath265 . repeat the iterations till convergence . the whole optimization procedure for pc - id receiver is summarized in algorithm [ algthm_opt_wipt ] . * initialize * : @xmath266 , @xmath267 , @xmath218 and @xmath224 , @xmath21 , @xmath68 , @xmath192 , @xmath248 , @xmath268 [ algthm_opt_wipt_step_initialize ] @xmath269 , @xmath270 , @xmath271 , @xmath272 , @xmath273 @xmath274 , @xmath260 [ algthm_opt_wipt_step_gamma ] @xmath275 , @xmath183 , @xmath262 [ algthm_opt_wipt_step_gamma_2 ] @xmath276 [ algthm_opt_wipt_step_opt ] @xmath277 note that the successive approximation method used in the algorithm 1 is also known as a sequential convex optimization or inner approximation method @xcite . it can not guarantee to converge to the global solution of the original problem , but only to yield a point fulfilling the kkt conditions @xcite . when @xmath278 , previous section derives a general methodology to optimize the superposed waveform weights jointly across space and frequency . it is worth wondering whether we can decouple the design of the spatial and frequency domain weights without impacting performance . the optimal phases in are those of a matched beamformer . looking at , , , and , the optimum weight vectors @xmath279 and @xmath280 that maximize the @xmath281 and @xmath282 order terms and the rate , respectively , are matched beamformers of the form @xmath283 such that , from , @xmath284 and from , @xmath285 and @xmath286 . hence , with , the multi - antenna multisine swipt weight optimization is converted into an effective single antenna multisine swipt optimization with the effective channel gain on frequency @xmath50 given by @xmath287 and the amplitude of the @xmath15 sinewave given by @xmath288 and @xmath289 for the multisine and ofdm waveform , respectively ( subject to @xmath290 ) . the optimum magnitude @xmath288 and @xmath289 in can now be obtained by using the posynomial maximization methodology of section [ pc_id ] . namely , focusing on @xmath126 for simplicity , plugging into , , and , we get and . @xmath291,\mathcal{a}\left\{y_p(t)^4\right\}=\frac{3}{8}\left[\sum_{{\genfrac{}{}{0pt}{}{n_0,n_1,n_2,n_3}{n_0+n_1=n_2+n_3}}}\bigg[\prod_{j=0}^3s_{p , n_j}\left\|\mathbf{h}_{n_j}\right\|\bigg]\right],\label{y_dc_w_3}\\ \mathcal{e}\left\{\mathcal{a}\left\{y_i(t)^2\right\}\right\}&=\frac{1}{2}\left[\sum_{n=0}^{n-1 } \left\|\mathbf{h}_n\right\|^2 s_{i , n}^2\right],\mathcal{e}\left\{\mathcal{a}\left\{y_i(t)^4\right\}\right\}=\frac{6}{8}\left[\sum_{n=0}^{n-1 } \left\|\mathbf{h}_n\right\|^2 s_{i , n}^2\right]^2.\label{y_dc_w_4}\end{aligned}\ ] ] the dc component @xmath168 as defined in can simply be written as @xmath292 , expressing that it is now only a function of the @xmath8-dimensional vectors @xmath293 $ ] and @xmath294 $ ] . similarly , @xmath295 writes as a function of @xmath296 . following the posynomial maximization methodology , we can write @xmath297 and @xmath298 with @xmath299 , apply the am - gm inequality and write the standard gp problem @xmath300\leq p,\\ & t_0\prod_{k=1}^k\left(\frac{g_k(\mathbf{s}_p,\mathbf{s}_i,\rho)}{\gamma_k}\right)^{-\gamma_k}\leq1,\\ & 2^{\bar{r}}\prod_{n=0}^{n-1}\prod_{k=1}^{k_n}\left(\frac{g_{nk}(\mathbf{s}_i,\bar{\rho})}{\gamma_{nk}}\right)^{-\gamma_{nk}}\leq 1,\label{standard_gp_constraint_3_simple}\\ & \rho+\bar{\rho}\leq 1.\label{standard_gp_swipt_4_simple}\end{aligned}\ ] ] algorithm [ algthm_opt_simple ] summarizes the design methodology with spatial and frequency domain decoupling . * initialize * : @xmath266 , @xmath267 , @xmath301 and in , @xmath302 , @xmath296 , @xmath192 , @xmath248 , @xmath268 [ algthm_opt_wipt_step_initialize_simple ] @xmath269 , @xmath303 , @xmath304 , @xmath272 , @xmath273 @xmath305 , @xmath260 [ algthm_opt_wipt_step_gamma_simple ] @xmath306 , @xmath183 , @xmath262 [ algthm_opt_wipt_step_gamma_2_simple ] @xmath307 [ algthm_opt_wipt_step_opt_simple ] @xmath308 such an approach would lead to the same performance as the joint space - frequency design of algorithm [ algthm_opt_wipt ] but would significantly reduce the computational complexity since only @xmath8-dimensional vectors @xmath302 and @xmath296 are to be optimized numerically , compared to the @xmath309 matrices @xmath21 and @xmath68 of algorithm [ algthm_opt_simple ] . with nc - id receiver , the multisine creates interference to the information decoder , which is reflected in the denominator of the sinr in the rate expression . this implies that decoupling the space frequency design by choosing the weights vectors as in is not guaranteed to be optimal . we will therefore resort to a more general approach where space and frequency domain weights are jointly designed , similarly to the one used in section [ pc_id ] . let us assume the same phases @xmath218 and @xmath224 as in the pc - id receiver in . such a choice is optimal for @xmath310 ( even though there is no claim of optimality for @xmath278 ) . with such a choice of phases , @xmath229 also writes as , while the achievable rate is written as @xmath311 where @xmath231 and @xmath312 . the optimization problem can now be written as - with replaced by @xmath313 defining the set of monomials of the posynomial @xmath314 . for a given choice of @xmath315 with @xmath316 and @xmath317 , we perform single condensations and write the standard gp for nc - id receiver as @xmath318\leq p,\\ & t_0\prod_{k=1}^k\left(\frac{g_k(\mathbf{s}_p,\mathbf{s}_i,\mathbf{\phi}_p^{\star},\mathbf{\phi}_i^{\star},\rho)}{\gamma_k}\right)^{-\gamma_k}\leq1,\\ & 2^{\bar{r}}\prod_{n=0}^{n-1}\left(1+\frac{\bar{\rho}}{\sigma_n^2}d_n(\mathbf{s}_p)\right)\nonumber\\ & \hspace{1cm}\prod_{j=1}^{j_n}\left(\frac{f_{nj}(\mathbf{s}_p,\mathbf{s}_i,\rho)}{\gamma_{nj}}\right)^{-\gamma_{nj}}\leq 1,\\ & \rho+\bar{\rho}\leq 1.\label{standard_gp_swipt_4_nc}\end{aligned}\ ] ] the whole optimization procedure for nc - id receiver is summarized in algorithm [ algthm_opt_wipt_nc ] . * initialize * : @xmath266 , @xmath267 , @xmath218 and @xmath224 , @xmath21 , @xmath68 , @xmath192 , @xmath248 , @xmath268 [ algthm_opt_wipt_step_initialize_nc ] @xmath269 , @xmath270 , @xmath271 , @xmath272 , @xmath273 @xmath274 , @xmath260 [ algthm_opt_wipt_step_gamma_nc ] @xmath319 , @xmath183 , @xmath320 [ algthm_opt_wipt_step_gamma_2_nc ] @xmath321 [ algthm_opt_wipt_step_opt_nc ] @xmath277 note that algorithm [ algthm_opt_wipt_nc ] boils down to algorithm [ algthm_opt_wipt ] for @xmath322 @xmath175 . 2 we now illustrate the performance of the optimized swipt architecture . parameters are taken as @xmath160 , @xmath161 and @xmath323 . we assume a wifi - like environment at a center frequency of 5.18ghz with a 36dbm eirp , 2dbi receive antenna gain and 58db path loss . this leads to an average received power of about -20dbm . we assume a large open space environment with a nlos channel power delay profile with 18 taps obtained from model b @xcite . taps are modeled as i.i.d . circularly symmetric complex gaussian random variables , each with an average power @xmath324 . the multipath response is normalized such that @xmath325 . equivalently , this system model can be viewed as a transmission over the aforementioned normalized multipath channel with an average transmit power fixed to -20dbm . [ z_dc_results_swipt_n](a ) illustrates the frequency response of one realization of the multipath channel . the superposed ofdm and multisine waveforms are designed assuming a frequency gap @xmath19 with @xmath326 with @xmath327 . the @xmath8 sinewaves are centered around 5.18ghz . fig [ z_dc_results_swipt_n](b),(c),(d ) illustrate the rate - energy region obtained for pc ( with algorithm [ algthm_opt_wipt ] ) , nc ( with algorithm [ algthm_opt_wipt_nc ] ) and no wpt ( multisine ) waveform ( with algorithm [ algthm_opt_wipt ] by allocating zero power to the wpt multisine waveform ) . @xmath310 and @xmath328 are considered and transmission is performed over the frequency response of fig [ z_dc_results_swipt_n](a ) . the noise power @xmath201 is fixed at -40dbm , corresponding to a snr ( defined as @xmath329 ) of 20db . the rate is normalized w.r.t . @xmath8 . for all three strategies ( pc , nc , no wpt waveform ) , the extreme right point on the x - axis refers to the rate achieved by the water - filling solution ( with all transmit power allocated to the ofdm waveform ) with @xmath204 . with pc and nc , points on the y - axis refer to the @xmath168 achieved with the multisine waveform ( wpt only ) of @xcite with @xmath330 . a first observation is that for the first two strategies , namely pc and nc , increasing @xmath8 helps increasing the harvested energy . on the other hand , in the absence of multisine waveform , the harvested energy does not change much . this was predicted from the scaling laws in table [ scaling_law_summary ] that showed that the average @xmath168 with ofdm waveform is independent of @xmath8 in frequency flat channel and only increases logarithmically with @xmath8 in frequency selective channel , and is lower than that achieved with the multisine waveform ( which scaled linearly with @xmath8 ) . a second observation is that without wpt waveform , the rate - energy region appears convex such that power - splitting ( ps ) outperforms time sharing ( ts ) , irrespectively of @xmath8 . on the other hand , with a wpt waveform , for both pc and nc , the rate - energy region with ps is convex for low @xmath8 ( @xmath331 ) but concave - convex for larger @xmath8 ( @xmath332 ) . this has for consequences that ps - only is suboptimal . looking at fig . [ z_dc_results_swipt_n](b ) for @xmath333 , the convex hull is obtained by ts between wpt ( multisine waveform only and @xmath330 ) and swipt with @xmath334 ( illustrated by point a ) . going for @xmath335 , the convex hull is obtained by ts between the two extreme points : wpt ( multisine waveform only and @xmath330 ) and ofdm with water - filling ( with @xmath204 ) . this shows that the convex hull with pc and nc can in general be obtained by a combination of ts and ps . this behaviour is observed because of the concacity - convexity of the rate - energy region achieved by pc , and fundamentally originates from the non - linearity of the rectifier . it is important to note that those observations are in sharp contrast with the existing swipt literature ( relying on the linear modeling of the rectifier ) . it is indeed shown in @xcite that ps always outperforms ts . 2 fig [ z_dc_results_swipt_snr](a),(b),(c),(d ) further studies the performance of the aforementioned three strategies for @xmath335 and four different snrs ( 10,20,30,40 db ) under the same channel response as in fig [ z_dc_results_swipt_n](a ) . a first observation is that the rate - energy region is augmented by having a superposition of two ( power and communication ) waveforms in both cases where the communication receiver does and does not cancel the contribution of the power waveform , i.e. @xmath336 and @xmath337 respectively . this can be seen by noting that the rate - energy region with pc and nc is larger than that obtained with no multisine waveform ( no wpt waveform ) . this key observation fundamentally comes from the non - linear behaviour of the wireless power channel . the usefulness of the multisine originates from the fact that the rectifier is a non - linear device and that the multisine is deterministic . on the contrary , the ofdm waveform exhibits randomness due to the information symbols that leads to some uncontrollable fluctuation of the waveform at the input of the rectifier and therefore some loss in terms of harvested energy . interestingly , this observation is in sharp contrast with the swipt literature relying on the conventional linear model . it was indeed shown in @xcite that , in the event where the power waveform is not eliminated by the communication receiver , the power waveform is useless and does not help enlarging the rate - energy region . as we can see from fig . [ z_dc_results_swipt_snr ] and observation [ obs_main ] , the wpt multisine waveform is actually useful when the rectifier non - linearity is properly taken into account in the design of swipt . a second observation is that the gain of pc over nc increases as the snr increases . at low snr , the interference from the multisine waveform in nc is drawn within the noise and therefore does not impact much the rate - energy region . pc and nc therefore have similar regions at very low snr . on the other hand , as the snr increases , in order for nc not to become interference limited , the nc architecture needs to allocate a very small power to the multisine waveform over a wide range of rates . this leads to a loss in terms of @xmath168 compared to the pc architecture that fully benefits from the presence of the multisine waveforms ( and its superior efficiency in increasing @xmath168 ) . a third observation is that at lower snrs ( 10 and 20db ) , the rate - energy achieved by pc with ps is actually outperformed by a ts between wpt ( multisine waveform only and @xmath330 ) and wit ( ofdm waveform only and @xmath204 ) . at higher snr ( 30 and 40db snr ) , ps performs better than ts between those two extreme points . this behaviour is observed because of the concacity - convexity of the rate - energy region achieved by pc . on the other hand , in the absence of the wpt waveform ( ofdm - only transmission ) , the rate - energy region appears convex because of the inefficiency of the ofdm waveform to boost @xmath168 . at low snr ( as on fig [ z_dc_results_swipt_snr](a ) ) , the wf strategy for rate maximization coincides with the ass strategy for energy maximization . hence the ass strategy is used across the boundary and only changing the power splitting ratio enables to draw the region boundary . as the snr increases and wf allocates power over an increasing number of frequencies , the region boundary with `` no wpt waveform '' is getting rounded because of the log behaviour of the rate expression . in summary , we draw the following conclusions from fig [ z_dc_results_swipt_n ] and [ z_dc_results_swipt_snr ] : 1 ) a power ( multisine ) waveform ( superposed to a communication ( ofdm ) waveform ) is useful to enlarge the rate - energy region of swipt , for @xmath8 sufficiently large ( @xmath338 ) . 2 ) without wpt multisine waveform , ps is preferred over ts . 3 ) with pc and nc , ps is favoured for low @xmath8 , a combination of ps and ts for medium @xmath8 and ts for large @xmath8 . 4 ) with pc and sufficiently large @xmath8 , ts is favoured at low snr and ps at high snr . it is important to note that 1 ) , 3 ) and 4 ) are consequences of the non - linearity of the rectifier . it is also noteworthy to recall that if we had used the linear model for the design and evaluation of the rate - energy region , following observation [ obs_main ] , there would not be any benefit of using a multisine waveform on top of the ofdm since both are equally suitable from an energy harvesting perspective . moreover , according to section [ ofdm_design ] , the optimum design of an ofdm waveform for wpt is the same for both the linear and non - linear model , namely based on the ass strategy . these facts imply that the rate - energy region achieved with a design of swipt based on the linear model is the same as the one achieved by `` no wpt waveform '' in fig [ z_dc_results_swipt_n ] and [ z_dc_results_swipt_snr ] . moreover , since the `` no wpt waveform '' strategy is always significantly outperformed by both pc and nc strategies in fig [ z_dc_results_swipt_n ] and [ z_dc_results_swipt_snr ] , this also implies that the design of swipt based on the non - linear rectifier model is more spectrally efficient than a design based on the linear model . the importance of accounting for the non - linearity of the rectifier in the design and evaluations of wpt was highlighted in @xcite . evaluations in this paper shows that this is also true for swipt . as a final but very interesting observation , we note that the input distribution with `` no wpt waveform '' or with a linear model - based design is gaussian with zero mean . the variance is frequency - dependent . this is the classical capacity achieving input distribution over a gaussian channel with average transmit power constraint . following remark [ remark_distribution ] , we note however that with `` pc '' and `` nc '' , the input distribution is gaussian but can have a non - zero mean . it is zero - mean as long as the ofdm waveform - only is used , but as we reduce the threshold @xmath267 and aim at a higher harvested energy , the waveform design algorithm starts allocating power over the multisine waveform . this has the effect of changing the input distribution from zero mean to non - zero mean . concurrently the variance of the distribution decreases since the transmit power allocated to the ofdm waveform is decreased , which leads to increasing k - factors ( defined in remark [ remark_distribution ] ) . this also suggests an alternative way of interpreting the transmit waveform . rather than viewing it as the superposition of a deterministic multisine and an ofdm waveform with zero - mean gaussian input distribution , we can view it as an ofdm waveform with non - zero mean gaussian input distribution . fig [ z_dc_results_swipt_n ] and [ z_dc_results_swipt_snr ] effectively show that designing an ofdm waveform for swipt with non - zero mean gaussian inputs leads to enlarged rate - energy regions compared to zero - mean gaussian inputs . an interpretation of fig [ z_dc_results_swipt_snr](b ) in terms of k - factor is provided in fig [ k_factor ] . for @xmath335 and @xmath310 over the multipath channel of fig . [ z_dc_results_swipt_n](a ) with 1mhz bandwidth and snr=20db . evolution fo the inpput distribution and k - factor . ] the paper derived a methodology to design waveforms for miso swipt . contrary to the existing swipt literature , the non - linearity of the rectifier is modeled and taken into account in the swipt waveform and transceiver optimization . motivated by the fact that a mutlsine waveform outperforms an ofdm waveform from a wireless power transfer persepctive , a swipt waveform is introduced as the superposition of a wpt waveform ( multisine ) and a wit waveform ( ofdm ) . the waveforms are adaptive to the csi ( assumed available to the transmitter ) and result from a non - convex posynomial maximization problem . the algorithm allows to draw the fundamental limits of swipt in terms of rate - energy region . results highlight that accounting for the nonlinearity of the rectifier radically changes the design of swipt . it favours a different waveform and transceiver architecture and a different use of the rf spectrum . exploiting the nonlinearity of the rectifier in the swipt design also makes a more efficient use of the resources by enabling enlarged rate - energy regions compared to those obtained by ignoring the nonlinearity in the system design . a lot of interesting questions arise from this work . to name a few , a fundamental question is , how to make the best use of the rf spectrum for swipt ? results here highlight that the conventional capacity - achieving zero - mean gaussian input distribution is suboptimal for swipt . what is the optimal input distribution and how to build corresponding swipt architecture ? results also highlight that the design of swipt differs from conventional communication due to the inherent non - linearity of the wireless power channel . in contrast , the wireless communication channel is commonly assumed linear . interesting works consist in re - thinking swipt architecture in light of non - linearity and designing swipt transceivers for broadcast , multiple access , interference and relay channels accounting for the non - linearity of the rectifier . b. clerckx , `` waveform optimization for swipt with nonlinear energy harvester modeling , '' 20th international itg workshop on smart antennas , march 2016 , munich . visser , r.j.m . vullers , `` rf energy harvesting and transport for wireless sensor network applications : principles and requirements , '' proceedings of the ieee | vol . 101 , no . 6 , june 2013 . j. park and b. clerckx , `` joint wireless information and energy transfer in a k - user mimo interference channel , '' ieee trans . wireless commun . 5781 - 5796 , oct . j. park and b. clerckx , `` joint wireless information and energy transfer with reduced feedback in mimo interference channels , '' ieee jsac , vol . 33 , no . 8 , pp . 1563 - 1577 , aug 2015 . k. huang and e , g. larsson , `` simultaneous information and power transfer for broadband wireless systems , '' ieee trans . sig . process . 5972 - 5986 , dec . 2013 . x. zhou , r. zhang and c. k. ho , `` wireless information and power transfer in multiuser ofdm systems , '' ieee trans . wireless commun . 2282 - 2294 , apr . 2014 . d. w. k. ng , e. s. lo and r. schober , `` wireless information and power transfer : energy efficiency optimization in ofdma systems , '' ieee trans . wireless commun . 6352 - 6370 , dec . 2013 . a. a. nasir , x. zhou , s. durrani , and r. a. kennedy , `` relaying protocols for wireless energy harvesting and information processing , '' ieee trans . wireless commun . 12 , no . 7 , pp . 3622 - 3636 , jul . y. huang and b. clerckx , `` joint wireless information and power transfer for an autonomous multiple - antenna relay system , '' ieee comm . letters , vol . 19 , no . 7 , pp . 1113 - 1116 , july 2015 . trotter , j.d . griffin and g.d . durgin , `` power - optimized waveforms for improving the range and reliability of rfid systems , '' 2009 ieee international conference on rfid . s. boaventura and n. b. carvalho , `` maximizing dc power in energy harvesting circuits using multisine excitation , '' 2011 ieee mtt - s international microwave symposium digest ( mtt ) . a. collado and a. georgiadis , `` optimal waveforms for efficient wireless power transmission , '' ieee microwave and wireless components letters , vol . 24 , no.5 , may 2014 . b. clerckx , e. bayguzina , d. yates , and p.d . mitcheson , `` waveform optimization for wireless power transfer with nonlinear energy harvester modeling , '' ieee iswcs 2015 , august 2015 , brussels , arxiv:1506.08879 . b. clerckx and e. bayguzina , `` waveform design for wireless power transfer , '' submitted to ieee tsp , arxiv : 1604.00074 . y. huang and b. clerckx , `` waveform optimization for large - scale multi - antenna multi - sine wireless power transfer , '' ieee spawc 2016 , arxiv:1605.01191 . s. wetenkamp , `` comparison of single diode vs. dual diode detectors for microwave power detection , '' in ieee mtt - s int . boston , ma , usa , 1983 , pp . 361 - 363 . b. clerckx and c. oestges , `` mimo wireless networks : channels , techniques and standards for multi - antenna , multi - user and multi - cell systems , '' academic press ( elsevier ) , oxford , uk , jan 2013 . z.b . zawawi , j. park and b. clerckx , `` simultaneous wireless information and power transfer in a two - user ofdm interference channel , '' ieee iswcs 2015 , august 2015 , brussels . m. chiang , `` geometric programming for communication systems , '' foundations and trends in communications and information theory , 2005 . duffin and e.l . peterson , `` geometric programming with signomials , '' journal of optimization theory and applications , vol . 11 , no . 1 , 1973 . m. grant , s. boyd , and y. ye , `` cvx : matlab software for disciplined convex programming [ online ] , '' available : http://cvxr.com/cvx/ , 2015 . marks and g.p . wright , `` a general inner approximation algorithm for nonconvex mathematical programs , '' operations research , vol . 681 - 683 , 1978 . m. chiang , c.w . tan , d.p . palomar , d. oneill , and d. julian , `` power control by geometric programming , '' ieee trans . on wireless comm . , vol . 6 , no . 7 , pp . 2640 - 2651 , july 2007 . j. medbo , p. schramm , `` channel models for hiperlan/2 in different indoor scenarios , '' 3eri085b , etsi ep bran , march 1998 . j. xu , l. liu , r. zhang , `` multiuser miso beamforming for simultaneous wireless information and power transfer , '' in ieee transactions on signal processing , vol.62 , no.18 , pp.4798 - 4810 , sept.15 , 2014 .

we study the effects of dissipation on electron transport in a semiconductor superlattice with an applied bias voltage and a magnetic field that is tilted relative to the superlattice axis . in previous work , we showed that although the applied fields are stationary , they act like a thz plane wave , which strongly couples the bloch and cyclotron motion of electrons within the lowest miniband . as a consequence , the electrons exhibit a unique type of hamiltonian chaos , which creates an intricate mesh of conduction channels ( a stochastic web ) in phase space , leading to a large resonant increase in the current flow at critical values of the applied voltage . this phase - space patterning provides a sensitive mechanism for controlling electrical resistance . in this paper , we investigate the effects of dissipation on the electron dynamics by modifying the semiclassical equations of motion to include a linear damping term . we demonstrate that even in the presence of dissipation , deterministic chaos plays an important role in the electron transport process . we identify mechanisms for the onset of chaos and explore the associated sequence of bifurcations in the electron trajectories . when the bloch and cyclotron frequencies are commensurate , complex multistability phenomena occur in the system . in particular , for fixed values of the control parameters several distinct stable regimes can coexist , each corresponding to different initial conditions . we show that this multistability has clear , experimentally - observable , signatures in the electron transport characteristics .

0706.2004

semiconductor superlattices ( sls ) are nanostructures made from alternating layers of two different semiconductor materials , usually with very similar lattice constants , for example gaas and ( alga)as @xcite . due to the different energy band gaps of the two materials , the conduction band edge of an ideal superlattice is periodically modulated . typically , a sl made by molecular beam epitaxy contains 10 - 100 quantum wells in series , which are coupled by tunnel barriers . this periodic potential leads to the formation of energy bands , known as `` minibands '' , for electron motion perpendicular to the layers @xcite . a voltage applied to ohmic contacts at the two ends of the sl generates an electric field , @xmath0 , perpendicular to the plane of the layers ( see fig . [ schema ] ) , which causes charge to flow through the device . the current - voltage characteristics of sls are usually highly nonlinear due to a variety of quantum mechanical effects including resonant tunneling , the formation of wannier - stark energy level ladders , and the occurrence of bloch oscillations @xcite , whose frequency is proportional to the spatial period , @xmath1 , of the sl and , also , to @xmath2 . in natural crystals , @xmath1 is so small ( @xmath3 nm ) that bloch oscillations do not occur because the bloch frequency is much less than the electron scattering rate . but in sls , @xmath1 and the corresponding bloch frequency can be large enough for the electrons to perform bloch oscillations , which then play a key role in both the dc and high - frequency charge transport processes . the onset of bloch oscillations localizes the electrons , thus causing their drift ( average ) velocity to decrease as the electric field increases . this negative differential velocity can induce high - frequency collective oscillations of the conduction electrons within the sl layers @xcite , making sls attractive for the generation and detection of electromagnetic radiation in the ghz to thz frequency range @xcite . , which lies in the @xmath4 plane at an angle @xmath5 to the sl ( @xmath6 ) axis . the lower arrow shows direction of the electric field , @xmath0 , applied perpendicular to the plane of the layers and anti - parallel to the @xmath6 axis . ] the occurrence of negative differential velocity for electrons in sls has also led to fundamentally new regimes of charge transport involving deterministic chaos , characterized by complex irregular electron dynamics @xcite . understanding the effects of chaos on the current - voltage characteristics , @xmath7 , and high - frequency electromagnetic properties of sls is an emerging research area at the interface between nonlinear dynamics and condensed matter physics . in most previous studies , chaos in sls was identified in the cooperative motion of interacting electrons in either periodically driven @xcite or undriven sls @xcite . recent theoretical and experimental work has revealed that the single - particle trajectories and collective behavior of electrons moving through the lowest miniband of a biased sl with a tilted magnetic field have unique properties @xcite . for example , the electrons exhibit an unusual type of hamiltonian chaos , known as `` non - kam '' chaos @xcite , which does not obey the kolmogorov - arnold - moser ( kam ) theorem and provides a sensitive new mechanism for controlling electron transport . remarkably , the effective classical hamiltonian for the electron motion in a sl has an intrinsically quantum - mechanical origin as it depends explicitly on the energy versus wavevector dispersion relation of the miniband @xcite . analysis of hamilton s equations reveals that the stochastic electron motion switches on abruptly when the field parameters satisfy certain resonance conditions , described in detail below . the onset of chaos delocalizes the electrons by imprinting an intricate mesh of conduction channels ( known as a `` stochastic web '' ) in phase space , which produces large resonant enhancement of the electron velocity and current flow measured in experiment @xcite . this phase - space patterning provides a fundamentally new concept for controlling electrical conductance @xcite and operates even at room temperature @xcite . in the absence of dissipation , the resonant delocalization of the conduction electrons would produce a series of delta - function peaks in the @xmath7 curves . but in real sls , the resonances are broadened because the electrons scatter both elastically @xcite , due mainly to roughness at the interfaces between the tunnel barriers and quantum wells and to ionized donor atoms , and inelastically via the emission and absorption of phonons . when calculating the transport characteristics of the sls , the effect of scattering on the electron dynamics must be taken into account in order to obtain @xmath7curves that agree with experiment . in our previous theoretical work @xcite , we incorporated electron scattering in the following phenomenological way . first , we calculated the electron trajectories by solving hamilton s equations for the system , assuming @xmath8 scattering . next , we used these trajectories to determine the average electron velocity from a simple kinetic formula in which scattering appears _ a posteriori _ as an exponential damping term whose decay rate depends on the elastic and inelastic relaxation times . in this paper , we investigate whether the striking electron resonance effects and phase space patterning predicted by our earlier collision - free hamiltonian model of the electron dynamics persist when scattering is included _ a priori _ in the equations of motion . we find that when the scattering is described in this way , it creates far richer electron dynamics than expected from our previous hamiltonian analysis . in particular , scattering makes the system highly sensitive to changes in the control parameters ( electric and magnetic fields ) , which induce complex bifurcation sequences characterized by alternating windows of stability and chaos . counter - intuitively , dissipation also enhances the resonant delocalization of the semiclassical trajectories by creating attractors that drive the electrons rapidly through the sl . the regimes of dissipative chaos that we have identified have a pronounced effect on the velocity - field characteristics of the miniband electrons and provide mechanisms for controlling sl magnetotransport by exploiting complex nonlinear dynamics . the structure of the paper is as follows . in section ii , we introduce the semiclassical equations of motion for a miniband electron in a tilted magnetic field and show how scattering is included _ a priori _ in those equations . in section iii , we show that scattering dramatically changes the electron trajectories and phase space structure of the system by creating attractors corresponding to periodic orbits that extend through the whole sl . in section iv , we present a detailed analysis of the stability of the electron orbits . section v explores the complex series of resonances that scattering induces in the electron dynamics and elucidates the effect of these resonances on the electron velocity and experimentally - measured electrical current . finally , in section vi , we summarize our results and draw conclusions . in a sl , quantum mechanical tunneling broadens the discrete quantized energy levels of each individual quantum well into sl minibands . the miniband states are delocalized bloch functions specified by the crystal momentum @xmath9 @xmath10=@xmath11 , where @xmath12 is the corresponding electron wavevector . within the tight - binding approximation , the energy versus crystal momentum dispersion relation for the lowest miniband is @xmath13/2+(p_y^2+p_z^2)/2m^{\ast}$ ] , where @xmath14 is the miniband width , @xmath1 is the sl period , and @xmath15 is the electron effective mass for motion in the @xmath16 plane . the crystal momentum component , @xmath17 , is taken to lie within the first minizone of the sl . throughout this paper , we consider electron motion in an electric field @xmath18 applied anti - parallel to the @xmath6 axis , and a tilted magnetic field @xmath19 ) ( fig . [ schema ] ) . in a semiclassical picture , which neglects inter - miniband tunneling , the force produced by the electric and magnetic fields changes the electron s crystal momentum at a rate @xmath20 where @xmath21 is the electronic charge . equation ( 1 ) can be written in the component form @xmath22 where the left hand terms are time derivatives of the crystal momentum components and @xmath23=@xmath24 is the cyclotron frequency corresponding to the magnetic field component along the @xmath6 axis . it follows from eqs . ( 2 - 4 ) that @xmath25 where @xmath26 , @xmath27 , and @xmath28 is the bloch frequency . the phase , @xmath29 , depends on the initial conditions and equals zero for electrons starting from rest @xcite . equation ( 5 ) describes completely the electron motion because its solution , @xmath30 , uniquely determines all of the other dynamical variables @xcite . consequently , to include the effects of dissipation _ ab initio _ in the equations of motion , we formally introduce a relaxation term , @xmath31 , into the left hand side of eq . ( 5 ) , which becomes @xmath32 where @xmath33 is the damping constant . in this approach , @xmath33 has the same meaning as a constant collision rate in the boltzmann equation . the dissipation term in eq . ( [ peneq ] ) directly affects the in - plane momentum component , @xmath34 . but when @xmath35 , electron motion in the @xmath6 , @xmath36 , and @xmath37 directions is coupled , which means that the dissipation term affects all of the dynamical variables , thus simulating the scattering processes that occur in a real sl device . recent experiments @xcite have shown that the presence of a high magnetic field strongly affects the inelastic scattering mechanisms required for current to flow through a biased sl @xcite . in particular , when @xmath38 and @xmath39 , the emission of longitudinal optical ( lo ) phonons is strongly suppressed if the phonon energy , @xmath40 = 36 mev in gaas , exceeds @xmath14 @xcite , which is the case for the sl that we studied in ref . @xcite . in this low - dissipation regime , the current falls almost to zero because the oscillatory bloch motion along the @xmath6 direction is not damped by lo phonon emission . our model equations ( 2 - 6 ) capture this magneto - suppression of the current , because eqs . ( 2 ) , ( 4 ) , and ( 6 ) decouple in the limit @xmath41 , which means that the @xmath6 motion is undamped , and so the mean electron velocity is zero . in the experiments @xcite , as @xmath5 increases from 0 , the current rises dramatically because the selection rules that prevent lo phonon emission when @xmath38 are broken . our model equations ( 2 - 6 ) describe this activation of lo phonon emission by increasing the coupling between motion parallel and perpendicular to @xmath6 as @xmath5 increases from 0 , thus enhancing scattering - assisted transport through the sl . consequently , the equations encapsulate the strongly @xmath5-dependent scattering rates observed experimentally in the high magnetic field regime that is the focus of this paper . as a consequence , however , they do not accurately describe the strongly damped bloch motion that occurs in the limit @xmath42 , which we do not consider here . we used eq . ( [ peneq ] ) to investigate the effect of dissipation on the motion of electrons in sls similar to those described and studied experimentally in @xcite , taking @xmath43 mev , @xmath44 nm , and @xmath45 , where @xmath46 is the mass of a free electron . in our analysis , we first solve eq . ( 6 ) numerically to obtain @xmath30 , which we then use to determine the other crystal momentum components @xmath47 and the electron velocity components @xmath48 eq . ( [ peneq ] ) describes the motion of a damped harmonic oscillator , whose natural frequency equals the cyclotron frequency @xmath23 , driven by a plane wave of wave number @xmath49 and frequency equal to the bloch frequency @xmath50 . the linear dissipation term describes the coupling of the oscillator to a reservoir " of many additional degrees of freedom , which , for the particular case of electrons in a sl , is provided by elastic and inelastic scattering processes . an _ undamped _ harmonic oscillator driven by a plane wave is one of few systems known to exhibit non - kam chaos @xcite . its rich dynamical properties have been studied by many authors and provide insights for understanding a wide range of problems in , for example , plasma physics , tokamak fusion , turbulent fluid dynamics , ion traps and quasi - crystals @xcite . some interesting effects of dissipation on non - kam chaos , including the identification of cantor sets in phase space , were studied in ref . @xcite . but , to our knowledge , the bifurcation phenomena considered in the present paper , and their relation to resonance effects in sls , have not been considered elsewhere . moreover , what has not been clear previously is the nature of the characteristic instabilities and dynamical regimes that can be induced by dissipation in systems that exhibit non - kam chaos , or how these regimes evolve with variation of the control parameters . our analysis of such dynamics for electrons in a sl with a tilted magnetic field is therefore also of general interest in nonlinear dynamics and in the diverse areas of physics mentioned above , which involve the behavior of a harmonic oscillator driven by a plane wave . in ref.@xcite , the dynamics of a miniband electron in an electric and tilted magnetic field were studied in the absence of dissipation by solving eq . ( [ peneq ] ) with @xmath51 . two distinct types of chaotic trajectories were identified from the different patterns that they produce in the poincar sections through phase space . the first type of pattern occurs only when the resonance condition @xmath52 is satisfied , where @xmath53 is a rational number . at such resonances , the chaotic orbits map out a stochastic web comprising an intricate mesh of filaments , which threads the phase space and is a unique feature of non - kam chaos . by contrast , the second type of pattern , a large continuous chaotic sea " in the poincar section , does not require the resonance condition to be satisfied . . this section reveals the inner filaments of a stochastic web . colored triangles numbered 1 - 6 : poincar sections calculated for six coexisting stable limit cycles formed when @xmath54 s@xmath55 . blue circle numbered 7 : poincar section calculated for the single stable limit cycle found when @xmath56 s@xmath55 . ( b ) electron trajectories corresponding , by number and color , to the labeled features in the poincar sections shown in ( a ) . all orbits are projected onto the @xmath4 plane ( axes inset ) , have a common spatial scale shown by the horizontal line whose length is 30 sl periods , and are plotted over a fixed time interval of @xmath57 ps . the field parameters @xmath58 t , @xmath59 and @xmath60 vm@xmath55 satisfy the resonance condition @xmath61 . ] we now illustrate these two types of chaos and investigate how the electron dynamics and associated phase space patterns change when dissipation is introduced by increasing @xmath33 from zero in eq . ( [ peneq ] ) . first , we consider the case when the phase space of the dissipation - free ( @xmath51 ) system contains a stochastic web corresponding to the @xmath62 resonance condition attained when @xmath60 vm@xmath55 , @xmath58 t , and @xmath59 . the black dots in fig . [ web](a ) , which merge to form a continuous pattern , show a stroboscopic poincar section constructed from the electron trajectories by plotting the momentum components @xmath63 in the plane of the sl layers at discrete times separated by the bloch period @xmath64 . the poincar section reveals a stochastic web , formed by the black dots and labeled 0 " in fig . [ web](a ) , which contains both ring - like and radial filaments . the radial filaments act as conduction channels , which enable the electron to diffuse rapidly outwards away from the web centre , thus increasing its momentum @xmath65 in the plane of the sl layers . in real space , the electron progresses rapidly along the sl axis @xcite , gaining kinetic energy from the electric field , which is transferred into the @xmath16 motion by the tilted magnetic field @xcite . to illustrate the resonant delocalization of the electron trajectories that results from stochastic web formation , the black curve labeled 0 " in fig . [ web](b ) shows an electron orbit within the stochastic web projected onto the @xmath4 plane in which the magnetic field lies ( axes inset ) . the electron starts from rest at the left - hand edge of the orbit , which corresponds the center point @xmath66 of the stochastic web in fig . [ web](a ) . a slow modulation of the density of orbital loops , which gradually increases with increasing @xmath6 , is the only indication of irregularity in the electron trajectory labeled 0 . for this reason , stochastic web chaos is often called weak chaos " @xcite , because the narrow width of the web filaments largely suppresses the highly erratic behavior usually found for nonintegrable systems . as the electron travels along the orbit from left to right in real space , it diffuses outwards along the almost vertical radial filament in the stochastic web towards higher @xmath34 values . the web filaments enmesh islands of stability , which are slices through invariant phase space tori generated by regular quasiperiodic trajectories @xcite . for clarity , in fig . [ web](a ) we show no tori corresponding to orbits within these islands of stability . the inclusion of even a very small dissipation term in eq . ( [ peneq ] ) destroys both the stochastic webs and islands of stability , replacing them with _ attracting _ limit cycles to which all orbits tend as time progresses . to illustrate this , we first set @xmath67 s@xmath55 , which is far less than the momentum relaxation rate @xmath68 s@xmath55 typical of electrons in semiconductor sl devices @xcite . for this value of @xmath33 , there are several distinct limit cycles , six of which are marked by the numbered and colored triangles in the stroboscopic poincar section shown in fig . [ web](a ) . each limit cycle appears within , and attracts trajectories from , one of the islands of stability found when @xmath51 . segments of the electron trajectories for each limit cycle are shown in fig . [ web](b ) . each orbit is shown over the same fixed time interval of 9.45 ps , chosen when @xmath69 is large enough for the orbit to approach very close to the limit cycle , and linked , by number and color , to the corresponding points in the phase space plot shown in fig . [ web](a ) . the orbital segments reveal that the inclusion of weak dissipation makes the electron trajectories more regular because each limit cycle creates a particular repeating loop pattern . as the level of dissipation increases , the different attracting limit cycles degenerate until only one remains in the phase space , whose position for @xmath70 s@xmath55 is marked by the blue circle labeled 7 " in fig . [ web](a ) . part of the corresponding electron orbit is shown by the blue curve labeled 7 " at the bottom of fig . [ web](b ) . the simple loop structure of this orbit transports the electron rapidly through the sl and is qualitatively similar to that of orbit 0 , which generates the stochastic web in the absence of dissipation . so , although dissipation completely changes the phase space structure , replacing the stochastic web with a small number of limit cycles , it preserves the overall form of the delocalized orbits found when @xmath71 and actually _ enhances _ resonant electron transport by preventing the spatial compression of the orbital loops , which slows the electron in the absence of dissipation ( see , for example , orbit 0 in fig . [ web](b ) ) . . colored dots : stroboscopic poincar sections calculated for the chaotic attractors of electrons starting from rest when @xmath54 s@xmath55 ( yellow dots ) , @xmath72 s@xmath55 , ( red dots ) , and @xmath73 s@xmath55 ( green dots ) . ( b ) electron trajectories corresponding , by specified value of @xmath33 and color , to the poincar sections shown in ( a ) . all orbits are projected onto the @xmath4 plane ( axes inset ) , have a common spatial scale shown by the horizontal line whose length is 10 sl periods , and are plotted over a fixed time interval of @xmath74 ps , which is long enough to reveal the form of the electron trajectories . @xmath75 t , @xmath76 , @xmath77 vm@xmath55 , @xmath78 . ] we now consider the second type of hamiltonian chaos , which , in the absence of dissipation , produces an extended chaotic sea in phase space even when @xmath79 is not a rational number . the poincar section formed by the black dots in fig . [ sea](a ) shows the chaotic sea generated by solving eq . ( [ peneq ] ) for an electron starting from rest in the sl with @xmath51 , @xmath77 vm@xmath55 , @xmath75 t , and @xmath76 , for which @xmath80 . introducing dissipation into the equation of motion affects this type of extended chaotic sea far less than stochastic webs . in particular , although dissipation changes the form of the chaotic sea and the orbits within it , the orbits remain unstable and spatially irregular even for fairly large values of @xmath33 . the reason is that dissipation transforms the chaotic sea into a @xmath81 attractor , rather than into a @xmath82 limit cycle like those labeled 1 - 6 in fig . 2 . in fig . [ sea](b ) , the yellow , red and green dots show the chaotic attractors generated by plotting @xmath63 at every bloch period for electron trajectories starting from rest with , respectively , @xmath67 s@xmath55 , @xmath72 s@xmath55 , and @xmath73 s@xmath55 . the irregular arrangement of points within these attractors demonstrates that they represent chaotic orbits , a conclusion that is confirmed by calculating their lyapunov exponents . however , with increasing @xmath33 , the dimension of the chaotic attractors diminishes . for the three attractors shown in fig.[sea](b ) , the lyapunov dimension calculated using the kaplan - yorke formula @xcite is @xmath83 , @xmath84 , and @xmath85 . chaotic electron orbits calculated for @xmath51 ( black curve ) , @xmath86 s@xmath55 ( yellow curve ) , @xmath72 s@xmath55 ( red curve ) , and @xmath73 s@xmath55 ( green curve ) are shown projected into the @xmath4 plane in fig . [ comparison of the trajectories and poincar sections shown in fig . [ sea ] ( a ) and ( b ) for different values of @xmath33 reveals that as the dissipation increases , the chaotic attractors gradually contract towards simple curves in phase space and , as a consequence , the orbits become spatially more regular . we emphasize that , in spite of the presence of energy dissipation , the electron trajectories remain chaotic , although their their topological structure in phase space changes as @xmath33 varies over a wide range of values . in contrast to the hamiltonian case , the phase space of the dissipative system contains unstable limit sets , which attract trajectories from certain regions of phase space known as basins of attraction . however , for very large @xmath87 s@xmath55 chaos completely disappears leaving only periodic orbits in the system . calculated as a function of @xmath2 and @xmath88 for the attractors approached by electron trajectories starting from rest . the color scale is shown on the right . ( b ) black ( green ) curves are loci of the primary period - doubling ( saddle - node ) bifurcations of limit cycles corresponding to resonances with rational @xmath53 values ( marked for @xmath89 ) . light green shaded areas are saddle - node tongues . ( c ) location of chaotic attractors ( red areas where @xmath90 ) and additional bifurcations within the region of @xmath91 space bounded by the @xmath92 primary period - doubling bifurcation ( outer black curve ) . black ( green ) curves are loci of period - doubling and saddle - node bifurcations corresponding to rational values of @xmath93 . light green shaded areas are saddle - node tongues . dashed line marks @xmath58 t , and blue arrow shows the path through @xmath91 space corresponding to the bifurcation diagram in fig . 5 . in all panels , crosses mark @xmath94 coordinate corresponding to the chaotic orbit shown ( green ) in fig . 3(b ) . @xmath76 and @xmath70 s@xmath55 . ] in this section , we investigate how the stability of the electron orbits changes with the field parameters @xmath2 , @xmath88 , and @xmath5 , and explore the bifurcation sequences that drive the transition between regular motion and dissipative deterministic chaos . we consider a fixed value of @xmath70 s@xmath55 . first , we identify different regimes of electron dynamics by calculating the lyapunov exponents for those limit sets that attract trajectories starting from rest . to characterize the stability of the electron motion , we calculate the largest lyapunov exponent , @xmath95 . the orbit is chaotic if @xmath90 , but otherwise is stable and regular . the color map in fig . [ bif1](a ) shows the value of @xmath95 calculated as a function of @xmath2 and @xmath88 for @xmath76 . for small @xmath2 or @xmath88 , the system exhibits stable periodic oscillations , of frequency @xmath50 , driven by the plane wave . we call such oscillations period-1 limit cycles " . as @xmath2 or @xmath88 grows , the system undergoes a sequence of bifurcations , which eventually leads to the appearance of chaos within the yellow and red areas in fig . [ bif1](a ) , where @xmath90 . the crosses in fig . [ bif1](a - c ) mark the coordinate in @xmath91 space for which an electron starting from rest eventually approaches the chaotic attractor shown green in fig.[sea](b ) . the location and shape of the various islands of stable and chaotic electron motion in the @xmath91 plane [ fig . [ bif1](a ) ] can be understood by considering the bifurcations that the limit cycles undergo as @xmath2 and/or @xmath88 increases . the black curves in fig . [ bif1](b ) show the loci of the primary sequence of period - doubling bifurcations , i.e. those that occur first as @xmath2 increases . each locus encloses a large v "- shaped region , which we call a period - doubling tongue " surrounding at least one area of chaos in the @xmath91 plane . when the field parameters enter one of these period - doubling tongues , the limit cycle electron trajectory undergoes a period - doubling bifurcation . as a result of this bifurcation , the initial limit cycle loses its stability and , in its vicinity , another stable limit cycle appears whose period is twice as long . a second type of bifurcation , namely the saddle - node bifurcation for limit cycles , also plays a key role in the electron dynamics . loci of the primary saddle - node bifurcations are shown by the green curves in fig . [ bif1](b ) , which enclose saddle - node " tongues indicated by the green shaded areas . the top edges of these tongues are close to those of the period - doubling bifurcation loci [ black curves in fig . [ bif1](b ) ] and lie near the lines @xmath96 ( @xmath97 ) along which @xmath71 . consequently , each of the period - doubling and saddle - node tongues can be associated with a particular value of @xmath53 [ those for @xmath97 are labeled in fig . [ bif1](b ) ] , which relates to a resonance of a period-1 cycle . note , however , that when @xmath98 , nonlinear resonances can occur near , but not exactly at , field values for which @xmath53 is a rational number . when the field coordinate @xmath94 enters a saddle - node tongue , a new stable limit cycle is created , which corresponds to an unbounded electron trajectory , like that labeled 7 " in fig . [ web](a ) for @xmath92 . this new limit cycle is twinned with an unstable saddle counterpart . as the field coordinate @xmath94 leaves the boundary of the saddle - node tongue , the stable and saddle cycles approach one another in phase space and eventually combine via the saddle - node bifurcation that occurs at the edge of the tongue ( we consider this process in more detail below ) . at this bifurcation point , the stable and unstable saddle cycles both disappear , and the electron trajectory jumps abruptly onto another stable orbit located in a different part of phase space . within the area of @xmath91 space bounded by each of the primary period - doubling tongues in fig . [ bif1](b ) , the system undergoes a cascade of additional bifurcations . in fig . [ bif1](c ) , we illustrate this for the region of @xmath91 space enclosed by the @xmath92 primary period - doubling bifurcation ( outer black curve ) . within this region , further period - doubling and saddle - node bifurcations occur along the black and green curves respectively . for example , the sharp saddle - node and period - doubling tongues whose apexes are at @xmath99 vm@xmath55 , 2 t ) in fig . [ bif1](c ) are associated with the @xmath100 nonlinear resonance and produce orbits of period @xmath101 . due to the presence of multiple tongues , the bifurcation diagram exhibits a complex self - similar structure . it is characterized by a cascade of bifurcations , which drives into chaos the electron trajectories that start from rest . these chaotic trajectories lie within the yellow and red areas of fig . [ bif1](a ) and ( c ) , where @xmath90 . an interesting effect occurs within those regions of @xmath91 space where period - doubling and saddle - node tongues intersect , for example the area marked by the cross in fig . [ bif1](c ) . as noted above , each period - doubling tongue contains a family of attractors . but in addition , the corresponding saddle - node tongue contains a _ distinct _ stable limit cycle and a saddle ( unstable ) periodic orbit . consequently , in the overlap region , at least two different attracting limit sets can coexist , which means that multistability occurs in the system . the initial condition of the orbit determines which attractor it eventually approaches . whenever @xmath102 along limit cycles for electrons starting from rest ( black dots ) and for stable / unstable period-1 limit cycles ( red / green circles ) with @xmath103 vm@xmath55 , @xmath76 and @xmath70 s@xmath55 . increasing @xmath88 induces period - doubling ( pd1 , pd2 ) and saddle - node ( sn1 , sn2 ) bifurcations , which correspond to crossing the black and green curves along the blue arrow in fig . 4(c ) , and a boundary crisis ( bc ) . arrow marks @xmath88 value for which @xmath92 . ( b ) an enlargement of the poincar section in the vicinity of saddle - node bifurcation sn1 and boundary crisis bc . ] to reveal in detail the bifurcations that the system undergoes inside the tongues in fig . [ bif1](c ) , we now consider how the electron orbits change as @xmath88 increases along the blue arrow in fig . [ bif1](c ) with fixed @xmath103 v m@xmath55 . in particular , we construct the single - parameter bifurcation diagram shown in fig . [ 1dbif ] by finding the limit sets that attract electrons starting from rest for @xmath104 t and , for each @xmath88 , plotting the values that @xmath34 attains ( black dots in the figure ) whenever @xmath102 in the limit cycle . for @xmath105 t , this attractor is a period-1 cycle , which produces a single black dot in fig . [ 1dbif](a ) at a @xmath34 value that gradually increases with increasing @xmath88 . when @xmath106 t , the period-1 cycle undergoes a period - doubling bifurcation , labeled pd1 in fig . [ 1dbif](a ) , which corresponds to entering the @xmath92 primary period - doubling tongue enclosed by the outer black curve in fig . [ bif1](c ) . as a result , @xmath34 takes one of two values when @xmath102 , causing the black dots to split into two distinct branches in fig . [ 1dbif](a ) . as @xmath88 increases further , the limit cycle for electrons starting from rest undergoes a sequence of additional period - doubling bifurcations , leading to the onset of chaos when @xmath107 t. thereafter , chaotic motion alternates with stability windows " , which is a characteristic feature of chaos induced by a cascade of period - doubling bifurcations @xcite , and suddenly disappears when @xmath108 t , near the r=1 nonlinear resonance , which occurs at the @xmath88 value marked by the arrow in fig . [ 1dbif](a ) . this suppression of chaos occurs because the attractor for electrons starting from rest switches back to a stable period-1 cycle owing to a boundary crisis " , which we explain in detail below . at this transition from chaotic to stable motion , the distribution of black dots in fig . [ 1dbif](a ) , which is shown more clearly in the enlargement in fig . [ 1dbif](b ) , changes from an irregular scatter to an ordered set of points located along a single - valued curve . to gain further insights into the electron dynamics , we now consider the nature and evolution of period-1 cycles in the system over the entire range of @xmath88 values shown in fig . [ 1dbif](a ) . to do this , we plot colored circles in fig . [ 1dbif ] showing the values of @xmath34 whenever @xmath102 in the period-1 cycle . if the period-1 cycle is stable ( unstable ) we plot a red ( green ) circle . for @xmath105 t , the period-1 orbit is stable and is the only attractor in the system , being the limit cycle that all trajectories approach irrespective of their initial conditions . at the first period - doubling bifurcation [ labeled pd1 in fig . [ 1dbif](a ) ] , the period-1 cycle becomes unstable ( and therefore no longer an attractor ) and co - exists with the period - doubled stable limit cycle produced by the bifurcation . when @xmath88 reaches @xmath109 t [ labelled sn1 in fig . [ 1dbif](a ) and ( b ) ] it enters the @xmath92 primary saddle - node tongue shown in fig . [ bif1](c ) . the associated saddle - node bifurcation creates _ new _ stable and unstable saddle period-1 cycles , which generate , respectively , the red and upper green circles in fig . [ 1dbif](a),(b ) . thus , the two new period-1 cycles formed by the saddle - node bifurcation coexist with the original period-1 cycle , which produces the lower colored circles in fig . [ this original period-1 cycle stabilizes ( circles change from green to red ) when @xmath88 reaches @xmath110 t and therefore crosses the upper edge of the @xmath92 primary period - doubling tongue shown in fig . [ bif1](c ) . at a slightly higher @xmath88 value of @xmath111 t , it coalesces with the unstable saddle period-1 cycle and thereby disappears as a result of saddle - node bifurcation sn2 . at this bifurcation , the field coordinate @xmath94 exits through the upper edge of the @xmath92 primary saddle - node tongue in fig . [ bif1](c ) . at higher @xmath88 values there is a single period-1 cycle , which is the only attractor in the system . note that the nonlinear @xmath92 resonance has a pronounced effect on the bifurcation diagram in fig . [ in particular , it gives rise to the characteristic loop formed by the different branches of the period-1 limit cycles that coexist between bifurcations sn1 and sn2 in fig . [ 1dbif](a ) and also causes the abrupt disappearance of the chaotic attractor at the boundary crisis , bc . we now consider the latter in more detail . @xmath112 when @xmath113 ( a ) 2.46 t just before boundary crisis , ( b ) 2.461 t at the boundary crisis . the sections are constructed by plotting @xmath114 whenever @xmath102 along the chaotic attractor for electrons starting from rest ( black dots ) , and for the stable period-1 limit cycle ( red circles ) and unstable saddle period-1 limit cycle ( green circle ) , which are formed by saddle - node bifurcation sn1 in fig . white ( gray ) regions are basins of attraction for the chaotic attractor and stable period-1 limit cycle respectively . arrows in ( a ) pointing towards ( away from ) the green circle represent stable ( unstable ) manifolds of the unstable saddle period-1 limit cycle . arrow and dotted curve in ( b ) show the direction along which the electron trajectory eventually leaves the chaotic attractor and subsequently approaches the stable period-1 limit cycle . vm@xmath55 , @xmath76 and @xmath70 s@xmath55 . ] as explained above , fig . [ 1dbif ] reveals a small field range between @xmath115 t and @xmath116 t ( i.e. between the points labeled sn1 and bc in the figure ) where two distinct attractors coexist , a chaotic one and the stable period-1 limit cycle . figure [ collision](a ) shows a stroboscopic poincar section calculated for the chaotic attractor ( black dots ) , stable period-1 limit cycle ( red circle ) , and unstable saddle period-1 limit cycle ( green circle ) constructed by plotting the momentum components @xmath63 at times @xmath117 @xmath112 when @xmath118 t , i.e. just before the boundary crisis that destroys the chaotic attractor . each attractor has a distinct basin of attraction comprising the set of initial conditions for which the trajectory will eventually approach the attractor . the basins of attraction for the chaotic set and period-1 limit cycles are shown respectively by the white and gray areas of the @xmath119 plane in fig . [ collision](a ) . note that the boundary between these two basins of attraction is the stable manifold of the unstable saddle periodic orbit ( green circle ) , which originates from the saddle - node bifurcation labeled sn1 in fig . [ 1dbif ] . trajectories with initial conditions exactly on this stable manifold approach the unstable saddle periodic orbit along the directions shown by the arrows pointing towards the green circle in fig . [ collision](a ) . conversely , the unstable manifold of the saddle , represented schematically by the two arrows pointing away from the green circle in fig . [ collision](a ) repels electron trajectories away from the vicinity of the saddle orbit . note that the lower right - hand part of the chaotic attractor in fig . [ collision](a ) approaches very close to the boundary of its ( white ) basin of attraction . this suggests that if changing @xmath88 increases the size of the chaotic attractor , there will be a boundary crisis , which occurs when an attractor hits the boundary of its basin of attraction . to investigate this possibility , we studied how the phase space structure shown in fig . [ collision](a ) responds to small changes of @xmath88 . our calculations confirm that increasing @xmath88 to @xmath120 t does indeed trigger a boundary crisis , which causes the abrupt disappearance of the chaotic attractor at the field value labeled bc in fig . [ 1dbif ] . to illustrate this , fig . [ collision](b ) shows the stroboscopic poincar section calculated for the chaotic set ( black dots ) , stable period-1 limit cycle ( red circle ) and unstable saddle period-1 limit cycle ( green circle ) at the boundary crisis . the lower right - hand edge of the chaotic attractor touches the saddle ( green circle ) , which , for lower @xmath88 , was located on the boundary between the basins of attraction of the chaotic and period-1 attractors [ fig . [ collision](a ) ] . when the chaotic attractor and saddle point touch at the boundary crisis , all trajectories that approach the chaotic attractor eventually leave it at the bottom right - hand tip in fig . [ collision](b ) , then cross the curve in the @xmath119 plane that used to separate the two basins of attraction , and thereafter move in the direction of the arrow along the dotted curve in fig . [ collision](b ) towards the stable period-1 limit cycle ( red circle ) . consequently , for _ all _ initial conditions , the electron trajectory eventually approaches the stable period-1 limit cycle , either directly or via the tip of the former chaotic attractor , which now allows the electron trajectories to escape . over the field range between the boundary crisis and the second period - doubling transition ( i.e. between the fields labeled bc and pd2 in fig . [ 1dbif ] ) the period-1 limit cycle is therefore the only attractor in the system and its basin of attraction [ gray in fig . [ collision](b ) ] fills the entire phase space . we now consider how the attractors evolve as @xmath88 sweeps either up or down through the boundary crisis . as noted above , for @xmath88 values in the range @xmath121 to @xmath122 t ( i.e. between points pd1 and sn1 in fig . [ 1dbif ] ) , the chaotic set ( black dots in fig . [ 1dbif ] ) is the only attractor in the system . as @xmath88 increases above @xmath123 t , the state of the system remains pinned to the chaotic attractor even though a second attractor ( red circles in upper part of fig . [ 1dbif ] ) emerges from the saddle - node bifurcation labeled sn1 in fig . [ 1dbif ] . when @xmath88 reaches the value at which the boundary crisis occurs , the system state changes abruptly from chaotic to the period-1 attractor , as discussed above . conversely , if @xmath88 now decreases , the state remains pinned to the period-1 attractor until the saddle - node bifurcation sn1 [ where the red and green circles meet in fig . [ 1dbif](b ) ] removes this attractor when @xmath124 t , i.e. at a field value _ below _ that corresponding to the boundary crisis . consequently , the transition between chaotic attractors and regular limit cycles occurs at a higher @xmath88 value when the field sweeps up through the boundary crisis than when it sweeps down , and the system exhibits _ hysteresis_. for given @xmath2 , the hysteresis occurs over the magnetic field range for which the @xmath92 saddle - node tongue , shown by the upper gray region in fig . [ bif1](c ) , overlaps with the upper red island in the figure , in which the attractor is chaotic . figure [ bif1](c ) shows that this overlap region broadens with increasing @xmath2 , meaning that the hysteresis becomes more pronounced , that is , it extends over a wider magnetic field range . the occurrence of hysteresis in bifurcation diagrams like fig . [ 1dbif ] is a common feature of resonances in non - linear systems @xcite . note that two distinct mechanisms cause the onset of chaotic limit cycles in the system . as @xmath88 increases from 0 , the transition to chaos occurs via the cascade of period - doubling bifurcations shown in fig . [ 1dbif ] . conversely , as @xmath88 decreases from 3 t , chaos switches on abruptly at the saddle - node bifurcation ( point sn1 in fig . [ 1dbif ] ) that removes the stable period-1 limit cycle . calculated as a function of @xmath2 and @xmath5 for the attractors approached by electron trajectories starting from rest when @xmath58 t and @xmath70 s@xmath55 . color scale is shown on the right . yellow ( green ) curves are loci of period - doubling ( saddle - node ) bifurcations of electron limit cycles . cross marks @xmath125 coordinate corresponding to limit cycle 7 in fig . the two main mechanisms responsible for the onset of chaos ( the period - doubling sequence and saddle - node bifurcation ) occur for a wide range of @xmath5 values . this can be seen from fig . [ bif2 ] , which shows the loci of the main period - doubling ( yellow curves ) and saddle - node ( green curves ) bifurcations in the @xmath126 plane calculated for fixed @xmath127 . these loci are overlaid on a color map showing the value of the largest lyapunov exponent for the limit cycles approached by electron trajectories starting from rest . note that , as in fig . [ bif1 ] , the limit cycles become chaotic ( yellow or red in the color map ) within the regions of @xmath126 space bounded by the period - doubling bifurcations ( yellow lines ) , which partially overlap the areas enclosed by the saddle - node bifurcations ( green lines ) . as a consequence of this overlap , multistability , and the associated hysteretic evolution of limit cycles , occurs for a range of @xmath2 and @xmath5 in the same way that hysteresis occurs within islands in the @xmath91 plane ( see fig . [ bif1 ] ) . note , however , that for the particular field value @xmath58 t used to calculate fig . [ bif2 ] , chaos only occurs for @xmath128 . for smaller values of @xmath5 , electrons starting from rest approach only stable limit cycles , for example that labeled 7 in fig . 2 , which is calculated for the @xmath125 coordinate marked by the cross in fig . [ bif2 ] . , of @xmath129 oscillations calculated versus @xmath2 for a stable / unstable period-1 limit cycle ( red / green circles ) and a stable / unstable period-2 limit cycle ( red / green triangles ) . arrows mark positions of @xmath53 = 1 and 2 resonances . ( b ) @xmath130 calculated as a function of @xmath2 for the attractor approached by electrons starting from rest . peaks labeled r1(r2 ) occur at @xmath2 values near the @xmath131 nonlinear resonances . labels o1(o2 ) mark off - resonance field values for which @xmath132 . ( c ) orbits approached by electrons starting from rest at the resonant ( r1,r2 ) and off - resonant ( o1,o2 ) @xmath2 values marked in ( b ) . all attractors are projected onto the @xmath4 plane ( axes inset ) , have a common spatial scale shown by the horizontal line whose length is 30 sl periods , and are plotted over the same time interval . @xmath127 , @xmath76 , @xmath70 s@xmath55 . ] in section ii , we showed that electron motion in the sl can be described by eq . ( [ peneq ] ) , corresponding to a damped harmonic oscillator driven by a plane wave . when the driving frequency , @xmath133 , is commensurate with the natural frequency , @xmath134 , of the harmonic oscillator nonlinear resonances occur in the system . the effect of such resonances on the electron dynamics is complex . for example , in sections iii and iv we saw that the @xmath92 resonance creates a stable limit cycle , which corresponds to a spatially unbounded electron trajectory , via a saddle - node bifurcation , and also triggers a boundary crisis . in this section , we investigate how resonances affect the momentum and drift velocity of the electrons , both parallel and perpendicular to the sl layers . we focus on these dynamical variables because they strongly influence the transport characteristics , such as @xmath7 curves , measured in experiments @xcite . first , we consider how resonances affect the variation of the in - plane momentum component , @xmath34 , as the electron performs a limit cycle . this momentum component is the most natural parameter to study because it is the dependent variable in eq . ( [ peneq ] ) . for each limit cycle , we define the amplitude of the @xmath34 oscillations to be @xmath135 , where @xmath136 and @xmath137 are , respectively , the minimum and maximum values of @xmath34 attained during the limit cycle . we then investigate how @xmath138 varies with @xmath2 for several different limit cycles , including those produced by the bifurcations induced by changing @xmath2 . in fig . [ res](a ) , we show @xmath138 calculated as a function of @xmath2 for period-1 ( circles ) and period-2 ( triangles ) limit cycles when @xmath58 t and @xmath76 . the @xmath139 plots , analogous to the frequency response curves familiar to engineers , exhibit the resonant features that characterize nonlinear dissipative systems . the resonances are clearly seen as sharp increases of @xmath138 when @xmath53 is close to rational @xcite , with the largest peaks [ arrowed in fig . [ res](a ) ] occurring near the @xmath2 values for which @xmath92 and 2 . the smaller peaks seen at lower @xmath2 correspond to rational @xmath140 . a period-1 limit cycle exists over the entire range of @xmath2 shown in fig . [ res](a ) , but undergoes bifurcations as @xmath2 increases , which changes it between stable ( red circles ) and unstable ( green circles ) . the location of the primary bifurcations for @xmath58 t and @xmath76 can be seen from fig . [ bif2 ] . for @xmath2 values close to @xmath141 vm@xmath55 , when @xmath142 , there is a saddle - node tongue ( enclosed by the green curve in fig . [ bif2 ] ) , which means that up to three period-1 orbits coexist ( as in the loop region of fig . [ 1dbif ] ) . consequently , the @xmath139 curve for the period-1 limit cycle contains a loop when @xmath142 . within the loop region , the stable period-1 limit cycle undergoes a period - doubling bifurcation ( along the right - hand yellow curve in fig . [ bif2 ] ) , which produces the period-2 limit cycle whose @xmath139 curve [ triangles in fig . [ res](a ) ] reveals large resonant peaks when @xmath92 and 2 . curves ( with gray underfill ) calculated for electron trajectories labeled in fig . [ res ] ( b ) and ( c ) as : ( a ) r1 , ( b ) o1 , ( c ) r2 , ( d ) o2 . vertical dashed lines mark extremal velocities @xmath143 . @xmath58 t , @xmath76 , @xmath70 s@xmath55 . ] the electron velocity component , @xmath144 , along the sl axis is a particularly important dynamical variable because it determines the electrical current that flows through the device and the power dissipation rate per particle , @xmath145 . it follows from the equations of motion ( [ peneq ] ) and ( [ veleq ] ) that for each attractor , @xmath144 oscillates as a function of time . to determine the drift component of the electron velocity , which determines the current flow @xcite , we therefore calculate the time average @xmath146 of the velocity throughout the attractor . figure [ res](b ) shows @xmath130 calculated as a function of @xmath2 for the attractor approached by electrons starting from rest with @xmath58 t and @xmath76 . the shape of this graph is similar to the @xmath139 data shown in fig . [ res](a ) . in particular , there is strong resonant enhancement of @xmath130 near rational @xmath53 values where maxima also occur in the @xmath139 curves calculated for the period-1 and period-2 limit cycles . the peak labeled r1 in the @xmath130 versus @xmath2 curve corresponds to the @xmath92 resonance of the period-1 limit cycle , whereas peak r2 originates from the @xmath100 resonance of the period-2 limit cycle . those peaks to the left of r1 arise from resonances of the period-1 limit cycle . to relate the resonant peaks in the @xmath130 versus @xmath2 curve directly to the electron motion , fig . [ res](c ) shows segments of period-1 ( lower and two upper trajectories ) and period-2 ( third orbit down ) limit cycles calculated at the field values labeled r1 , r2 , o1 , and o2 in fig . [ res](b ) over a fixed time interval of 12.38 ps . during this time interval , the electron travels much further along in the two orbits on resonance ( r1 and r2 ) than in the two off - resonance orbits ( o1 and o2 ) . calculated as a function of increasing ( black curve ) and decreasing ( under - filled gray curve ) @xmath88 ( upper axis ) . lower axis shows values of @xmath147 . peaks in @xmath130 originate from @xmath148 and 2 resonances ( located by arrows ) . the difference between the two curves originates from the hysteretic evolution of the electron limit cycle discussed in section iv . dashed curve shows @xmath149 ( with _ no _ vertical offset ) calculated as a function of @xmath88 ( or @xmath53 ) using eq . ( [ eq : vd ] ) . @xmath150 vm@xmath55 , @xmath76 , @xmath70 s@xmath55 . ] to further investigate how resonances affect the electron transport , we calculated the probability distribution , @xmath151 , of all the @xmath144 values attained during one period of the limit cycles approached by electrons starting from rest both on and off resonance . figure [ vx ] shows normalized @xmath151 curves calculated for the orbits labeled ( a ) r1 , ( b ) o1 , ( c ) r2 , ( d ) o2 in fig . [ res ] ( b ) and ( c ) . all attainable @xmath144 values lie between the two vertical dashed lines located at @xmath143 , where @xmath152 is the maximum velocity of an electron in the lowest miniband . comparison of the velocity distributions on ( a),(c ) and off ( b),(d ) resonance reveals that on resonance , the most probable @xmath144 value approaches @xmath153 , thus producing the large peak @xmath130 values shown in fig . [ res](b ) . the hysteretic response exhibited by the limit cycles as @xmath88 is swept up and down through a boundary crisis has a clear manifestation in the @xmath130 versus @xmath88 curves calculated for large @xmath2 values where the hysteresis is most pronounced , as discussed in section iv . to illustrate this , fig . [ hyst ] shows @xmath130 calculated as a function of increasing ( black curve ) and decreasing ( under - filled gray curve ) @xmath88 ( upper axis ) when @xmath150 vm@xmath55 and @xmath76 . note that the lower axis of the graph shows the ratio , @xmath147 , of the bloch and cyclotron frequencies . near the @xmath154 and @xmath155 resonances ( marked by arrows in fig . [ hyst ] ) , saddle - node tongues coexist with chaotic attractors . this can be seen , for example , from fig . [ bif1](c ) , which , as discussed in section iv , reveals a large overlap between the @xmath92 saddle - node tongue and upper red island of chaos when @xmath150 vm@xmath55 . as a consequence , the limit cycle evolves in a different way when @xmath88 increases or decreases , and so the corresponding @xmath130 versus @xmath88 curves exhibit hysteresis . this hysteresis is particularly pronounced near the @xmath92 and @xmath100 resonances in fig . [ hyst ] , which reveals that for some @xmath88 values @xmath130 differs by a factor of two between up and down sweeps of @xmath88 . since the current flow depends critically on the electron drift velocity , we predict that measured @xmath156 curves should also reveal strong resonant peaks that exhibit clear hysteresis . in fig . [ hyst ] , we compare the @xmath130 versus @xmath88 curves obtained by including dissipation _ a priori _ in the equations of motion , with drift velocities , @xmath157 , calculated using the traditional esaki - tsu approach @xcite , based on non - dissipative electron trajectories . within this approach , which we used in our previous studies of electron transport in sls with a tilted magnetic field @xcite , @xmath158 where @xmath159 is determined from eqs . ( [ peneq ] ) and ( 8) , setting @xmath51 , and the electron scattering time , @xmath160 , includes contributions from both elastic and inelastic scattering processes @xcite . the @xmath157 variation obtained from eq . ( [ eq : vd ] ) is shown by the dashed curve in fig . comparison with the @xmath130 versus @xmath88 curves ( black and under - filled gray ) reveals that both approaches give qualitatively similar results , with prominent maxima at @xmath88 values close to @xmath161 1 , 3/2 , and 2 . however , the positions of these maxima do not always coincide exactly , in particular for @xmath92 , mainly because including dissipation _ a priori _ in the equations of motion shifts the resonances slightly away from rational values of @xmath53 , as discussed in section iv . although the resonant peaks in the @xmath130 versus @xmath88 and @xmath157 curves have very similar heights , especially for @xmath53 = 2 and 3/2 , away from the resonances @xmath130 is significantly lower than @xmath149 . the reason for this is that off resonance , @xmath34 increases less rapidly with increasing @xmath69 , which means that the damping term , @xmath162 , in eq . ( [ peneq ] ) remains small and therefore causes less scattering - induced transport . figure [ hyst ] reveals that , for the sl considered here , the most pronounced resonant enhancement of @xmath130 occurs near @xmath100 [ as also shown in fig . [ res](b ) ] . the two other resonances shown in fig . [ hyst ] near @xmath161 1 and 3/2 weaken as @xmath53 decreases because the corresponding increase of @xmath88 reduces the distance that the electrons travel along the @xmath6 axis in a given time interval [ compare , for example , orbits r2 and r1 calculated for @xmath53 = 2 and 1 respectively in fig . [ res](c ) ] . to investigate how such resonances manifest themselves in the electrical properties of a real device , we measured @xmath7 characteristics for the sl studied in ref . @xcite , whose parameters are similar to those described in section [ sec : model ] . figure [ exp ] shows the differential conductance @xmath163=@xmath164 measured as function of @xmath53 at @xmath165 = 0.25 v and @xmath76 . in the experiments , we varied @xmath88 and , for each value , measured @xmath166 over a very small range of @xmath165 near 0.25 v , taking the derivative to find @xmath163 . for each @xmath88 value , we determined the corresponding @xmath167 by assuming that @xmath168 and is uniform throughout the sl . the dominant feature in fig . [ exp ] is the pronounced conductance peak at the @xmath100 resonance ( arrowed ) . additional , far weaker , features can be seen for smaller @xmath53 , but are not strong enough for us to link them definitively to particular @xmath53 values . the overall shape of the experimental @xmath169 curve in fig . [ exp ] is broadly similar to the @xmath130 versus @xmath53 plots shown in fig . [ hyst ] . however , only the @xmath53 = 2 conductance peak is clearly visible in the experimental data . this difference between experiment and theory occurs because in the actual device , @xmath2 is not constant throughout the device , as assumed in our present calculations , but varies with @xmath6 due to the formation of charge domains @xcite . the spatial variation of @xmath2 means that the resonance condition is only satisfied in a small region of the sl , which weakens and blurs the resonances in @xmath169 @xcite . this effect is less pronounced for stronger resonances , at which the electric field is more uniform through the sl @xcite . consequently , the @xmath53 = 2 resonance is clearly revealed and correctly positioned in fig . , versus @xmath170 , measured for different values of @xmath88 at fixed @xmath165 = 0.25 v and @xmath76 for the sl described in ref . @xcite , whose parameters are similar to those specified in section [ sec : model ] . arrow marks @xmath100 resonance . ] we have shown that both hamiltonian and dissipative chaos strongly affect the transport of miniband electrons in biased sls with a tilted magnetic field . chaos originates from a complex nonlinear interaction between bloch and cyclotron oscillations associated , respectively , with the electric field and magnetic field component , @xmath171 , along the sl axis . in the absence of dissipation , the system exhibits non - kam chaos , characterized by the formation of intricate stochastic web patterns when @xmath50 and @xmath23 are commensurate . when weak dissipation is included _ a priori _ in the equations of motion , limit cycles replace the stochastic web filaments and stable orbits that they enclose . as the dissipation strength increases , the number of limit cycles and the volume of phase space occupied by chaotic trajectories both decrease however , for the electron momentum relaxation rates measured for sls used in recent experiments @xcite , chaos occurs for a wide range of @xmath2 and @xmath88 values , usually when @xmath172 ( see figs . [ bif1 ] , [ bif2 ] ) . two distinct bifurcation mechanisms drive the transition to chaos : a period - doubling cascade , where the onset of chaos is gradual , and a boundary crisis , which switches chaos on abruptly . both mechanisms operate for all tilt angles in the range @xmath173 . for both hamiltonian and dissipative models of the electron dynamics , the onset of chaos occurs near resonances when @xmath50 and @xmath23 are commensurate . such resonances cause an abrupt delocalization of the electron trajectories , due either to stochastic web formation in a hamiltonian picture , or to the creation of attractors corresponding to unbounded stable limit cycles when dissipation is included _ ab initio _ in the equations of motion . this delocalization produces large resonant peaks in the electron drift velocity , which , in turn , generate the strong resonant enhancement of the current flow observed in our recent experiments @xcite . in the case of dissipative dynamics , the drift velocity increases near resonances because the most probable value of @xmath144 that the electron attains during its unbounded limit cycle is very close to the maximal possible value , determined by the sl parameters ( fig . resonances between @xmath50 and @xmath23 also give rise to multistability phenomena when two or more different attractors coexist in phase space . multistability makes limit cycles , and the corresponding @xmath130 values , evolve differently when @xmath2 or @xmath88 sweeps up or down through a resonance . we expect that this hysteresis will also have striking experimental signatures . in particular , our calculations suggest that experimental @xmath156 or @xmath7 curves will exhibit pronounced hysteresis , reflecting that shown in fig . [ hyst ] for electron drift velocities , especially at high field values . finally , we emphasize that the driven harmonic oscillator equation ( [ peneq ] ) , which describes the motion of miniband electrons in a tilted magnetic field , is also of fundamental importance in many other physical systems , for example plasma @xcite , ultra - cold atoms in optical lattices @xcite , the transmission of light through photonic crystals @xcite , turbulent flow and in understanding the patterns of quasi - crystals @xcite . although the motion of the _ dissipation - free _ harmonic oscillator driven by a plane wave is well understood , to our knowledge the effects of dissipation have not previously been considered . our study of the rich dynamics of the _ dissipative _ driven harmonic oscillator is relevant to a wide range of topics in physics , engineering , and applied mathematics . the work was supported by the uk engineering and physical sciences research council . we thank m. henini ( university of nottingham ) for mbe growth of the sl samples used in our experiments , r. airey ( university of sheffield ) for processing the samples , and d.k . maude ( grenoble high magnetic field laboratory ) for assistance with the high magnetic field measurements . e. schomburg , j. grenzer , k. hofbeck , t. blomeier , s. winnerl , s. brandl , a.a . ignatov , k.f . renk , d.g . pavelev , y. koschurinov , v. ustinov , a. zhukov , a. kovsch , s. ivanov , p.s . kopev , solid - state electronics * 42 * 1495 ( 1998 ) . a.a . vasiliev , g.m . zaslavsky , m.y . natenzon , a.i . neishtadt , b.a . petrovichev , r.z . sagdeev , a.a . chernikov , sov . jetp * 67 * , 2053 ( 1989 ) [ zh . teor . fiz . * 94 * , 170 ( 1988 ) ] s.p . stapleton , s. bujkiewicz , t.m . fromhold , p.b . wilkinson , a. patan , l. eaves , a.a . krokhin , m. henini , n.s . sankeshwar , and f.w . sheard , physica d * 199 * , 166 ( 2004 ) .

we have previously identified 10 m31 black hole candidates ( bhcs ) in m31 , from their x - ray properties alone . they exhibit `` hard state '' emission spectra that are seen at luminosities% eddington in x - ray binaries ( xbs ) containing a neutron star ( ns ) or black hole ( bh ) , at luminosities that significantly exceed the ns threshold . nine of these are associated with globular clusters ( gcs ) ; hence , these are most likely low mass x - ray binaries ( lmxbs ) ; eight are included in this survey . we have recently discovered that analysis of the long term 0.54.5 kev variability of xbs via structure functions allows us to separate xbs from agn , even though the emission spectra are often similar ; this has enabled us to search for bhcs outside of gcs . we have identified 26 new bhcs ( 12 strong , 14 plausible ) within 20 of the m31 nucleus ( m31 * ) , using 152 chandra observations spaced over years ; some of our classifications were enhanced with xmm - newton observations . of these , 7 appear within 100 of m31 * ; this supports the theory suggesting that this region experiences enhanced xb production via dynamical processes similar to those seen in gcs . we have found a parameter space where our black hole candidates are separated from galactic neutron star binaries : we show that modelling a simulated hard state spectrum with a disk blackbody + blackbody model yields parameters that lie outside the space occupied by neutron star binaries that are modeled this way . the probability that our bhcs all lie within the ns parameter space is .

1304.7780

we know of @xmath220 x - ray binaries with dynamically confirmed black hole ( bh ) accretors ; these include 15 low mass x - ray binaries ( lmxbs ) and 3 high mass x - ray binaries ( hmxbs ) in the milky way and magellanic clouds ( see e.g. * ? ? ? * and references within ) , as well as bh + wolf - rayet binaries in ic10 @xcite and ngc300 @xcite . these systems were identified using x - ray and optical observations ; the mass function is calculated from periodic radial velocity shifts in emission lines from the optical counterpart . all of the bh lmxbs identified this way are necessarily transient , because the optical spectra of bright lmxbs are dominated by the accretion disc ( see e.g. * ? ? ? we have established a method for identifying bhcs from their x - ray properties alone . this makes use of the `` low / hard '' emission state seen in bh and neutron star ( ns ) xbs @xcite , that is only seen at 0.011000 kev luminosities @xmath010% eddington in ns xbs @xcite ; @xcite recently found that the low / hard state is limited to luminosities @xmath010% eddington in bh xbs also . bh xbs may exhibit low / hard emission states at considerably higher luminosities than ns xbs , due to the higher accretor mass . however , it is necessary to differentiate between our bhcs and distant active galactic nuclei ( agn ) , since agn and xb emission spectra are often similar . we have identified 10 bhcs from their high luminosity low states to date . of these , 9 are associated with m31 gcs @xcite , and are therefore likely lmxbs . only one of these is transient ; however , persistently bright gc bh xbs are consistent with tidal capture of a main sequence donor ( * ? ? ? * although the donor may be disrupted in the process ) , or with an ultra - compact system with a degenerate donor @xcite . we identified our first bhc outside of a gc from its x - ray spectra , long term ( @xmath212 year ) behaviour , and a serendipitous hst observation @xcite ; the faint optical counterpart ( @xmath5 @xmath6 @xmath70.4 ) suggests a low mass donor for this system also . we cover 8 of the 9 gc bhcs in our survey : bhcs 1 , 2 , 20 , 25 , 28 , 31 , 32 , and 34 . the field bhc described in @xcite is bhc3 . we obtained accurate positions for our x - ray sources by registering 27 x - ray sources associated with m31 gcs to the m31 field 5 b band image provided by @xcite . the r.m.s . offsets were 0.11@xmath8 in r.a . and 0.09@xmath8 in dec @xcite ; this would be extremely unlikely unless the 27 x - ray sources used for calibration were indeed associated with the gcs . the x - ray spectra of bh binaries are usually described with two components : a thermal component ( often modelled as a multi - temperature disk blackbody ) , and a power law component to represent unsaturated , inverse - compton scattering of cool photons on hot electrons @xcite . the hard state is classified by a power law component with photon index ( @xmath9 ) = 1.42.1 , and a thermal component that contributes @xmath10 of the 220 kev flux @xcite . our very first gc bhc ( xb045 , xbo 45 in * ? ? ? * ) was associated with the m31 gc b045 , named following the revised bologna catalogue v.3.4 ( rbc , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? its @xmath217,000 count xmm - newton / pn spectrum was well described by an absorbed power law with line - of - sight absorption ( @xmath11 ) = 1.41@xmath120.11@xmath13 atom @xmath14 , and @xmath9 = 1.45@xmath120.04 ; @xmath15/dof = 517/487 ( good fit probability , gfp , = 0.17 ) . adding a blackbody component improved the fit somewhat , but the thermal contribution to the flux was too small to be constrained ; we therefore considered xbo 45 to be in its hard state . since its unabsorbed 0.310 kev luminosity was 2.5@xmath120.2@xmath16 erg s@xmath17 , 140% eddington for a 1.4 @xmath18 ns , we considered it a bhc . the 8 remaining gcs bhcs are included in this survey , unlike xbo 45 , so we do nt describe each spectrum in detail here . however , we note that fitting a two component model to xb144 ( xbo 144 in * ? ? ? * ) yielded k@xmath19 = 0.0082@xmath120.0016 kev , indicating a complete lack of thermal component in the spectrum . we also note that even though xb082 was best described by @xmath9 = 1.20@xmath120.09 , the @xmath15/dof for that fit was @xmath20.9 , and we were able to obtain fits where @xmath9=1.4 and @xmath15/dof @xmath201 @xcite . the unabsorbed 0.310 kev luminosities of our gc bhcs exhibiting hard state spectra range over @xmath2545@xmath21 erg s@xmath17 @xcite . comparison with the 0.510 kev agn flux distribution obtained by @xcite yields a 2.4@xmath22 probability that our gc bhcs are coincident agn . the probability that our brightest gc bhc , xb135 , is a coincident agn is 1.2@xmath23 . we will present evidence in a separate paper that xb135 may contain the most massive stellar mass bh known to date ; it may have been formed by direct collapse of a high mass , metal poor star ( r. barnard et al . , 2013 , in prep ) . we found that the gcs hosting these very bright x - ray sources were significantly more massive and/or metal rich than the other gcs in m31 , agreeing with previous work . however , two gcs had particularly low metalicities ; one of these is b135 , consistent with the direct collapse formation scenario for xb135 ( see * ? ? ? * and references within ) . @xcite have shown that while bh masses are limited to @xmath215 @xmath18 for solar metalicties , they could theoretically reach @xmath280 @xmath18 for metalicities @xmath20.01 solar . high accretion rate ns xbs exhibit multi - component emission that may appear to be hard state spectra in extragalactic x - ray sources , where the spectra have relatively few photons . to compare the emission spectra of our bhcs with ns xrb , we use the double thermal ( disk blackbody + blackbody ) model used by lin et al . ( 2007 , 2009 , 2012 ) to describe hundreds of rxte observations of ns xbs in all their varied emission states . we note that the physical interpretation of their model is contradicted for persistent xbs by a substantial body of work ; however , lin et al . ( 2007 , 2009 , 2012 ) have sampled spectra from the full range of ns lmxb emission states in a consistent manner , allowing us to compare our bhcs with ns xbs in a single parameter space . for a long time , ns xbs were divided into two types , z - sources ( high lumonisity ) and atoll sources ( low luminosity ) , based on their luminosities and color - color diagrams ( cds ) ; the cds of z - sources had three branches ( horizontal , normal , and flaring ) , and evolved along these branches without ever jumping from one branch to the other ; the cds of the atoll sources were more fragmented @xcite . furthermore , the z - sources were split into those like cygnus x-2 , and those like scorpius x-1 @xcite . it was believed that the differences were due to more than just the accretion rate , because the z - sources varied by a factor of a few when tracing their z - shaped cds , while atoll source intensities varied by 12 orders of magnitude @xcite . however , we now know of two transient ns systems that exhibited both types of z - source behavior before evolving to atoll source behavior during decay @xcite . @xcite examined the spectral evolution of two galactic x - ray transients , aql x-1 and 4u1608@xmath752 , over many rxte observations covering @xmath620 outbursts . they devised a new double - thermal model ( disk blackbody + blackbody ) to describe a ns transient soft state that is analagous to the bh soft state described by @xcite . they have since applied their model to rxte observations of xtej1701@xmath7462 , one of the transients that exhibits cyg - like and sco - like z - source behavior as well as atoll behavior @xcite , and also to the sco - like z - source gc 17 + 2 @xcite . they have applied their model to hundreds of rxte spectra from galactic ns binaries including the full gamut of ns spectral behavior . they found that their disk blackbody + blackbody was unsuccessful in two situations . firstly , they found the hard state spectra to be power law dominated , as expected @xcite . we therefore expect our bhcs to inhabit a separate parameter space to the ns xbs because fitting the double thermal model to hard state spectra will yield unphysical results . secondly , they found that z - sources required a three - component spectrum ( disk blackbody + blackbody + power law ) on the horizontal branch @xcite . @xcite also performed broadband analysis of suzaku and bepposax observations of the persistently bright galactic ns xb 4u1705@xmath744 , with energy ranges 0.1600 kev and 0.1300 kev respectively . when using a disk blackbody + blackbody model to the soft state spectra , they found temperatures that were similar to those they observed in the rxte observations @xcite . fitting an additional comptonization component , via either a power law or simpl convolution model ( following * ? ? ? * ) resulted only in small changes in the thermal components ; this is because the comptonized component only contributed @xmath010% of the total flux in the soft state ( see fig . 6 in * ? ? ? hence the parametric differences between our bhcs and the ns xbs should not be due to differences in energy bands used in the observations , or the lack of a third ( power law ) component . evidence against the double thermal model indicates an extended corona that contributes a substantial portion of the x - ray flux . this evidence includes the ingress times of periodic absorption dips in the x - ray lightcurves of the high inclination xbs ( the dipping sources , see e.g. * ? ? ? * ; * ? ? ? * and references within ) , and also broadened emission lines in a chandra grating spectroscopy of cyg x-2 @xcite . furthermore , compact corona models where the inner disk temperature is tied to the seed photon energy for comptonization are rejected for ulxs in ngc253 , and the confirmed bh+wolf - rayet binary ic10 x-1 @xcite , as well as bhc3 in the steep power law state @xcite . in @xcite , we applied structure function analysis to xbs for the first time , following @xcite , who created an ensemble structure function for agn . they used a structure function ( sf ) to estimate the mean intensity deviation for data separated by time @xmath24 : @xmath25 where @xmath26 is the photon noise and @xmath27 is the x - ray flux . they grouped the sf into logarithmic bins with width 0.5 ; each bin in the range log(@xmath24 ) = 0.03.0 contained more than 100 measurements . our sample consisted of 37 x - ray sources associated with objects in the rbc ; these were classified by @xcite as 30 confirmed gcs , 4 gc candidates , 1 star , and 2 agn . the sfs of gc xbs with 0.310 kev luminosities @xmath2250@xmath28 erg s@xmath17 tended to show significantly more variability than agn over a wide range of time - scales . the sfs of brighter xbs generally showed comparable or less variability than agn , despite their high signal to noise ; however , their high fluxes made them unlikely agn . sfs provide an effective mechanism for distinguishing between xbs and agn @xcite . in this work , we combine these techniques to search for bhcs in the whole region covered by 152 chandra observations from our monitoring programme ; the roll angle is unrestricted , resulting an approximately circular field of view with radius @xmath29 . we surveyed 530 x - ray sources in this region . we also obtain spectra from archival xmm - newton observations , to strengthen the cases for several bhcs . the 100@xmath3 region surrounding m31 * is particularly interesting , because @xcite found an excess of x - ray binaries over the radial distribution expected from k band light ( tracing stellar mass ) ; this excess had the distribution expected of dynamically formed xbs @xcite . dynamical xb formation requires stellar densities rarely seen outside gcs , and @xcite proposed that the m31 bulge is sufficiently large and dense to form a significant number of xbs dynamically . however , since the stellar velocities in the m31 bulge are @xmath2510 times higher than in gcs , @xcite expect only short period binaries to survive , with most of these containing bh accretors . evidence for bhcs in this region would therefore provide strong support for their theory . in the next section , we discuss the observations and data analysis ; this is followed by our results , then by our discussion . the central region of m31 has been observed with chandra on a @xmath2 monthly basis for the last @xmath213 years in order to monitor transients ; we exclude periods when m31 can not be observed due to satellite constraints ( @xmath2 march april ) . we have analyzed 98 acis observations and 54 hrc observations . we determined the position of each source from a merged 0.37.0 kev acis image , using the iraf tool imcentroid . the x - ray positions were registered to the field 5 b band image of m31 from the local group galaxy survey ( lgs , see * ? ? ? * ) , using 27 gc x - ray sources and following the procedure described in @xcite . the r.m.s uncertainties in registration were 0.11@xmath8 in r.a . and 0.09@xmath8 in dec . we obtained 0.37.0 kev lightcurves and spectra from circular source and background regions for each source . the background region was the same size as the source region , and at a similar off - axis angle . the extraction radius varied between sources , because larger off - axis angles resulted in larger point spread functions . we used the ciao v4.5 software suite , with corresponding caldb to reduce the data , and xspec v12.7 to analyze the spectra . for acis observations , we obtained corresponding response matrices and ancillary response files . we initially estimated the conversion from flux to luminosity by assuming a power law emission spectrum with photon index 1.7 , with @xmath11 = 7@xmath30 atom cm@xmath17 , then determining the unabsorbed 0.310 kev luminosity equivalent to 1 count s@xmath17 at the location of the source . after correcting for the exposure , vignetting and background , multiplying the source intensity by this conversion factor gave the source luminosity . source spectra with @xmath6200 net counts were freely fitted . when good spectral fits were found for a particular observation of a source , the parameters of these fits replaced the default parameters when estimating the source luminosity for observations of the source with @xmath20200 photons . cccccccccccc bhc & position & @xmath32 & @xmath33 & @xmath34 & @xmath15/dof & @xmath35 + 1 & 00:42:15.786 + 41:01:14.24 & 18 & 48 & 21.3@xmath120.4 & 191/65 & 2145 + 2 & 00:42:18.648 + 41:14:01.85 & 83 & 54 & 5.54@xmath120.06 & 1479/136 & 9316 + 3 & 00:42:22.954 + 41:15:35.23 & 91 & 54 & 7.57@xmath120.06 & 23850/144 & 10440 + 4 & 00:42:26.070 + 41:19:15.03 & 93 & 54 & 2.53@xmath120.04 & 3576/146 & 10731 + 5 & 00:42:28.193 + 41:10:00.30 & 85 & 54 & 3.85@xmath120.05 & 486/138 & 9591 + 6 & 00:42:28.285 + 41:12:22.95 & 89 & 54 & 4.73@xmath120.05 & 6113/142 & 10153 + 7 & 00:42:31.147 + 41:16:21.67 & 93 & 54 & 9.32@xmath120.08 & 1123/146 & 10585 + 8 & 00:42:34.669 + 40:57:14.20 & 13 & 12 & 0.64@xmath120.08 & 696/24 & 300 + 9 & 00:42:39.585 + 41:16:14.30 & 70 & 21 & 0.063@xmath120.006 & 3480/90 & 4095 + 10 & 00:42:40.654 + 41:13:27.32 & 93 & 53 & 1.40@xmath120.03 & 4724/145 & 10585 + 11 & 00:42:42.177 + 41:16:08.23 & 74 & 34 & 0.057@xmath120.006 & 2331/107 & 5778 + 12 & 00:42:44.831 + 41:11:37.89 & 92 & 54 & 3.6@xmath120.04 & 3220/145 & 10585 + 13 & 00:42:45.122 + 41:16:21.68 & 96 & 54 & 2.3@xmath120.03 & 3291/149 & 11175 + 14 & 00:42:45.946 + 41:10:36.53 & 54 & 27 & 0.035@xmath120.007 & 1759/80 & 3240 + 15 & 00:42:46.969 + 41:16:15.58 & 95 & 54 & 3.1@xmath120.04 & 3202/148 & 11026 + 16 & 00:42:47.176 + 41:16:28.41 & 92 & 54 & 0.294@xmath120.014 & 40403/145 & 10585 + 17 & 00:42:47.893 + 41:15:32.87 & 95 & 54 & 2.6@xmath120.03 & 3278/148 & 11026 + 18 & 00:42:48.529 + 41:15:21.12 & 95 & 54 & 11.6@xmath120.09 & 2369/148 & 11026 + 19 & 00:42:48.546 + 41:25:22.12 & 60 & 51 & 1.6@xmath120.05 & 8701/110 & 6105 + 20 & 00:42:52.030 + 41:31:07.87 & 11 & 54 & 49.4@xmath120.4 & 343/64 & 2080 + 21 & 00:42:52.534 + 41:18:54.46 & 95 & 54 & 11.79@xmath120.09 & 1271/148 & 11026 + 22 & 00:42:54.935 + 41:16:03.19 & 95 & 54 & 4.8@xmath120.05 & 18769/148 & 11026 + 23 & 00:42:57.900 + 41:11:04.65 & 93 & 54 & 5.3@xmath120.05 & 826/146 & 10731 + 24 & 00:42:59.675 + 41:19:19.35 & 93 & 54 & 5.4@xmath120.05 & 887/146 & 10731 + 25 & 00:42:59.872 + 41:16:05.72 & 95 & 54 & 5.6@xmath120.06 & 1074/148 & 11026 + 26 & 00:43:02.937 + 41:15:22.53 & 95 & 54 & 3.3@xmath120.04 & 4838/148 & 11026 + 27 & 00:43:03.220 + 41:15:27.69 & 95 & 54 & 1.5@xmath120.03 & 11960/148 & 11026 + 28 & 00:43:03.876 + 41:18:04.91 & 94 & 54 & 4.1@xmath120.05 & 2920/147 & 10878 + 29 & 00:43:05.667 + 41:17:02.43 & 77 & 29 & 0.053@xmath120.005 & 10972/105 & 4851 + 30 & 00:43:09.866 + 41:19:00.76 & 85 & 48 & 0.127@xmath120.010 & 2675/132 & 8646 + 31 & 00:43:10.622 + 41:14:51.30 & 90 & 54 & 13.4@xmath120.11 & 915/143 & 10296 + 32 & 00:43:17.593 + 41:27:44.87 & 17 & 0 & 0.11@xmath120.02 & 409/16 & 136 + 33 & 00:43:34.334 + 41:13:23.08 & 45 & 54 & 3.8@xmath120.06 & 1380/98 & 4851 + 34 & 00:43:37.320 + 41:14:43.09 & 34 & 54 & 6.7@xmath120.09 & 209/87 & 3828 + 35 & 00:43:44.593 + 41:24:10.25 & 32 & 39 & 0.5@xmath120.03 & 2296/71 & 2346 + for hrc observations , we we included only pi channels 48293 , thereby reducing the instrumental background . we used the webpimms tool to find the unabsorbed luminosity equivalent to 1 count s@xmath17 , assuming the same emission model as for the acis observations with @xmath20200 photons . we created a 1 kev exposure map for each observation , and compared the exposure within the source region with that of an identical on - axis region , in order to estimate the estimate the necessary exposure correction . we multiplied the background subtracted source intensity by the correction factor to get the 0.310 kev luminosity . we created long term 0.310 kev lightcurves for each source , using the luminosities obtained from each observation as described above ; all luminosities assume a distance of 780 kpc @xcite . we only included observations with net source counts @xmath36 0 after background subtraction . we fitted each long term lightcurve with a line of constant intensity , in order to ascertain the source variability . we note that the ratio of hrc to acis luminosity depends strongly on the spectral model , and should be 1.0 if the model is correct . differences between actual and assumed emission spectra during hrc and faint acis observations may lead to systematic offsets between luminosities from adjacent hrc and acis observations . we derived sfs from the 0.54.5 kev fluxes of each observation of every target by assuming a power law spectrum with the same photon index as for the hrc and faint acis observations ; an m31 x - ray source with a 0.310 kev unabsorbed luminosity of 1.0@xmath21 erg s@xmath17 , @xmath11 = 7@xmath30 atom @xmath14 and @xmath9 = 1.7 has a 0.54.5 kev flux of 0.80@xmath37 erg @xmath14 s@xmath17 . @xcite calculated the noise component from @xmath38 assuming that @xmath39 = @xmath40 , and @xmath41 is the number of photons . our lightcurves are background subtracted , and arf - corrected ; furthermore , uncertainties in the luminosities of bright sources include uncertainties in the spectral parameters . as a result , our uncertainties are not simply due to photon counting noise . hence in our case , @xmath42 ^ 2 + \left [ \frac{\sigma f_{\rm x}\left(t\right)}{f_{\rm x}\left(t\right ) } \right]^2\right>,\ ] ] where @xmath43 is the uncertainty in x - ray flux and often significantly larger than @xmath44 . in the simplest case where only 1 pair has a particular separation , then @xmath45,\ ] ] where @xmath46 and @xmath47 are the fluxes of the two observations . in addition to our chandra observations , we also obtained spectra for some of our bhcs from archival xmm - newton observations . their higher sensitivity and longer exposure times often yielded superior spectra for bright sources . we used only the pn instrument , because the mos detectors are more prone to pile up . we used sas v10 for the data reduction . we filtered out background flares by creating a lightcurve with the selection `` ( pattern==0)&&(flag==0)&&(pi in [ 10000:12000 ] ) '' and 100 s binning , then excluding intervals with @xmath60.4 count s@xmath17 . source and background spectra were obtained from circular regions with 2040@xmath8 radius , along with corresponding response files ; the filter `` ( pattern@xmath20=4)&&(flag==0 ) '' was applied along with the good time filter and extraction region . our bhc identification process involved three steps . first , we had to establish that the x - ray source was an xb rather than an agn . then we checked the acis observations of each xb , to see if it exhibited a high luminosity hard state ( @xmath9 @xmath48 2.1 at a 0.310 kev unabsorbed luminosity @xmath63@xmath21 erg s@xmath17 , i.e. 10% eddington for a 2 @xmath49 ns ) . finally we found the highest quality observation that exhibited a high luminosity hard state , fitted it with the double - thermal model , and compared the results with the known range of parameters exhibited by the ns systems studied by lin et al . ( 2007 , 2009 , 2012 ) . likely xbs were identified from their associations with gcs , variable sfs or relatively high fluxes . there are @xmath2170 gcs within our field @xcite ; using the 0.510 kev agn flux distribution @xcite , we expect to find 0.0005 agn with flux @xmath61.4@xmath37 erg @xmath14 s@xmath17 ( 10@xmath50 erg s@xmath17 ) within 1@xmath8 of any gc . hence any bright x - ray source associated with a gc is a likely xb . we also identify xbs using sfs that are strikingly more variable than the ensemble agn sf . furthermore , only @xmath21 agn per square degree is expected with a flux equivalent to @xmath51 erg s@xmath17 ; hence brighter x - ray sources are likely xbs too . lin et al . ( 2007 , 2009 ) found that they were unable to successfully fit hard state spectra with their double thermal model . this is because the disk blackbody component must account for the low - energy flux , resulting in an unphysically low temperature . to demonstrate this , we created a fake hard state spectrum using the xspec command fakeit ; we used a power law with @xmath9 = 1.7 , with absorption equivalent to 7@xmath30 atom @xmath14 . we present the best fit double thermal model for this spectrum in fig . [ example ] , which consists of a 0.67 kev disk blackbody and a 1.76 kev blackbody ; the blackbody component contributes @xmath280% of the 210 kev flux . by contrast , the soft states of ns xbs studied by lin et al . ( 2007 , 2009 , 2012 ) yielded disk blackbody temperatures @xmath521 kev , with the blackbody component contributing less than 50% of the 210 kev flux . since lin et al . ( 2007 , 2009 , 2012 ) failed to fit hard state spectra with the double thermal model , we expect disk blackbody + blackbody fits to our bhcs to inhabit a different parameter space from the ns binaries if they really are in the hard state . if a bhc exhibits a blackbody temperature @xmath201.5 kev , a disk blackbody temperature @xmath201.0 kev ( @xmath201.2 kev for disk blackbody 210 kev luminosity @xmath62@xmath21 erg s@xmath17 ) , and/or a blackbody contribution to the total 210 kev flux @xmath645% at a 3@xmath53 level , then we consider it a strong bhc ; otherwise , we label it a plausible bhc . these criteria are drawn from the hundreds of rxte spectra of aql x-1 , 4u1608@xmath752 , xtej1701@xmath7462 , and gx 17 + 2 analyzed by @xcite . table [ props1 ] gives the location of each of our bhcs , followed by the number of observations with acis and hrc . next we provide the best fit line of constant 0.310 kev luminosity over @xmath213 years , with the corresponding @xmath15 and number of degrees of freedom ( dof ) . finally , we show the number of observation pairs used to create the structure function . usually this is @xmath54 , where @xmath55 is the total number of observations ; however , bhcs 7 , 29 , 30 , and 35 included observations with zero luminosity that could not be included in the sf . all uncertainties in this work are quoted at the 1@xmath53 level unless specified otherwise . for nearly all bhcs , closely - spaced acis and hrc observations give consistent luminosities , although some small systematic offsets occur due to differences between assumed and actual spectra when fitting is not possible . however , bhc32 appears to have substantial signal in the hrc observations even when this transient source is in quiescence ; the hrc is considerably more sensitive than acis to low energy photons , hence the hrc obsevations of bhc32 must be contaminated by soft x - rays . bhc32 is located near 2 variable stars ( 16@xmath8 and 17@xmath8 distant ) ; since the extraction radius for this souce is 15@xmath8 due to the high off - axis angle , our hrc observations are likely contaminated by one or both foreground stars . as a result , we excluded all hrc observations of bhc32 from our analysis . . figure [ lgsbh ] shows a detail of the b band image of m31 field 5 from the lgs ( massey et al . , 2006 ) , superposed with the positions of our bhcs ; north is up , east is left . the circle encloses the region within 100@xmath3 ( @xmath2380 pc ) of m31 * ; this region is enlarged in the top right portion of the figure . bhcs 1 and 8 lie below the southern edge of field 5 . we find 7 bhcs within 100@xmath3 of m31@xmath56 , 20% of our total sample : bhcs 9 , 11 , 13 , 15 , 16 , 17 , and 18 . this result appears to be consistent with the predictions of @xcite . bhc9 is a recurrent transient with a peak 0.310 kev luminosity of @xmath28@xmath21 erg s@xmath17 . bhc11 was bright for @xmath212 years , then turned off . bhc18 appears to be persistently bright ( @xmath57 erg s@xmath17 ) , and may vary rapidly if the hrc luminosities are correct ; however , the emission from bhc18 is variable ( @xmath9 @xmath21.42.4 for bright acis observations ) , meaning that the hrc luminosities may be systematically offset due to differences between assumed and true spectra . bhcs 13 , 15 , 16 , and 17 are extremely variable , with bhcs 13 , 15 , and 17 luminosities ranging over @xmath216@xmath21 erg s@xmath17 , and bhc16 exhibiting luminosities from @xmath25@xmath28 to @xmath22@xmath58 erg s@xmath17 . we observed similar behavior in the x - ray source associated with b158 @xcite , a high inclination lmxb with a @xmath210000 s period @xcite , and an asymmetric , precessing disc @xcite . such behavior is seen in low mass ratio ( short period ) systems where the outer disc reaches the 3:1 resonance with the donor star , causing additional tidal torques that lead to elongation and precession of the disc ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? b158 exhibited rapid variation over 440@xmath21 erg s@xmath17 , which we speculated might be due to varying accretion rates over the disc precession cycle ; we suggested that other xbs exhibiting such behaviour may also be in precessing ( i.e. short period ) systems @xcite . therefore we propose that bhcs 13 , 15 , 16 , and 17 are short period bh xbs , as predicted by @xcite . two of our new bhcs are associated with confirmed old gcs ; bhc4 is associated with b096 ( following the rbc naming convention ) , and bhc24 with b143 . @xcite rank b096 as the 36th most massive and 34th most metal rich out of 379 gcs ( [ fe / h ] = @xmath70.28 ) ; b143 is ranked 52nd most massive and 17th most metal rich , with [ fe / h ] = @xmath70.07 . we have found that the gcs hosting our previous bhcs were particularly massive , or metal rich , or both ( barnard et al . , 2011a , 2012 ) ; hence it is encouraging that the new gc bhcs are associated with gcs that are amongst the most massive and most metal rich of the m31 gcs studied by caldwell et al . ( 2009 , 2011 ) . in @xcite we classified the semi - regular transient associated with hii rich gc candidate sk059a as a possible hmxb associated with the hii , because the intervals between its outbursts were consistent with @xmath59 cycles of 120 days ; hence the accretor could make an eccentric , @xmath2120 day orbit around the high mass donor , with mass transfer restricted to times near periastron . this system is bhc30 . cccccccccccc bhc & obs & cnts & @xmath11/10@xmath60 @xmath14 & @xmath9 & @xmath15/dof & @xmath61 /10@xmath50 + 1 & 0109270101 & 3229 & 0.40@xmath120.05 & 1.24@xmath120.09 & 133/155 [ 0.89 ] & 27.1@xmath121.1 + 2 & 303 & 704 & 0.15@xmath120.10 & 1.5@xmath120.2 & 33/29 [ 0.29 ] & 6.6@xmath120.6 + 3 & 13827 & 5597 & 0.07 f & 1.50@xmath120.04 & 161/182 [ 0.87 ] & 14.4@xmath120.4 + 4 & 0690600401 & 2370 & 0.26@xmath120.04 & 1.67@xmath120.14 & 153/146 [ 0.32 ] & 6.6@xmath120.3 + 5 & 13825 & 1425 & 0.21@xmath120.09 & 1.68@xmath120.18 & 8/11 [ 0.72 ] & 4.2@xmath120.3 + 6 & 0112570101 & 10384 & 0.112@xmath120.012 & 1.72@xmath120.06 & 428/411 [ 0.26 ] & 7.30@xmath120.17 + 7 & 0112570101 & 14523 & 0.110@xmath120.010 & 1.76@xmath120.04 & 519/490 [ 0.18 ] & 8.79@xmath120.17 + 8 & 9524 & 352 & 0.07 f & 1.29@xmath120.17 & 16/15 [ 0.37 ] & 17@xmath123 + 9 & 1575 & 3097 & 0.10@xmath120.03 & 1.73@xmath120.09 & 106/109 [ 0.55 ] & 5.0@xmath120.2 + 10 & 13827 & 1679 & 0.13@xmath120.06 & 1.56@xmath120.13 & 72/67 [ 0.32 ] & 3.6@xmath120.2 + 11 & 303 & 704 & 0.10@xmath120.09 & 1.43@xmath120.18 & 58/45 [ 0.10 ] & 9.7@xmath120.8 + 12 & 13825 & 2620 & 0.15@xmath120.04 & 2.00@xmath120.10 & 90/98 [ 0.71 ] & 5.2@xmath120.2 + 13 & 14197 & 1593 & 0.12@xmath120.05 & 1.63@xmath120.14 & 72/64 [ 0.24 ] & 3.2@xmath120.2 + 14 & 9521 & 183 & 0.07 f & 1.5@xmath120.4 & 6/7 [ 0.56 ] & 10@xmath123 + 15 & 14198 & 1860 & 0.13@xmath120.05 & 1.56@xmath120.12 & 75/74 [ 0.47 ] & 3.9@xmath120.2 + 16 & 13825 & 4003 & 0.12@xmath120.03 & 1.40@xmath120.07 & 161/165 & 10.2@xmath120.4 + 17 & 14197 & 1954 & 0.14@xmath120.04 & 2.05@xmath120.13 & 94/76 & 3.25@xmath120.18 + 18 & 13825 & 4207 & 0.09@xmath120.03 & 1.45@xmath120.08 & 140/154 [ 0.78 ] & 9.0@xmath120.4 + 19 & 0112570101 & 5032 & 0.41@xmath120.03 & 1.89@xmath120.08 & 210/226 [ 0.77 ] & 9.74@xmath120.4 + 20 & 0402560901 & 20025 & 0.275@xmath120.009 & 1.60@xmath120.02 & 771/769 [ 0.44 ] & 45.0@xmath120.4 + 21 & 13827 & 3847 & 0.22@xmath120.05 & 1.73@xmath120.09 & 127/142 [ 0.82 ] & 10.0@xmath120.4 + 22 & 0112570101 & 10145 & 0.134@xmath120.014 & 1.81@xmath120.05 & 490/407 [ 3 e-3 ] & 8.2@xmath120.2 + 23 & 14198 & 2508 & 0.13@xmath120.04 & 1.64@xmath120.10 & 111/95 [ 0.12 ] & 5.1@xmath120.2 + 24 & 0112570101 & 6804 & 0.120@xmath120.015 & 1.91@xmath120.06 & 265/294 [ 0.89 ] & 5.13@xmath120.15 + 25 & 0112570101 & 5116 & 0.108@xmath120.019 & 1.50@xmath120.06 & 254/236 [ 0.20 ] & 5.3@xmath120.17 + 26 & 13825 & 1425 & 0.16@xmath120.07 & 1.71@xmath120.17 & 54/57 [ 0.60 ] & 3.3@xmath120.2 + 27 & 14197 & 2099 & 0.14@xmath120.04 & 1.83@xmath120.12 & 69/75 [ 0.67 ] & 4.1@xmath120.2 + 28 & 13826 & 635 & 0.07 f & 1.73@xmath120.13 & 23/27 [ 0.69 ] & 5.3@xmath120.4 + 29 & 1577 & 1375 & 0.07 f & 1.33@xmath120.08 & 79/58 [ 0.04 ] & 31@xmath122 + 30 & 13299 & 274 & 0.07 f & 1.2@xmath120.2 & 10/10 [ 0.42 ] & 9.9@xmath121.6 + 31 & 0112570101 & 14338 & 0.081@xmath120.009 & 1.64@xmath120.04 & 450/488 [ 0.89 ] & 10.4@xmath120.2 + 32 & 8184 & 521 & 0.07 f & 1.43@xmath120.13 & 27/23 [ 0.25 ] & 13.4@xmath121.5 + 33 & 13825 & 1411 & 0.14@xmath120.08 & 1.81@xmath120.17 & 57/60 [ 0.57 ] & 4.4@xmath120.3 + 34 & 0112570101 & 5181 & 0.124@xmath120.019 & 1.64@xmath120.07 & 246/230 [ 0.22 ] & 7.0@xmath120.2 + 35 & 14198 & 1646 & 0.18@xmath120.08 & 1.75@xmath120.16 & 85/69 [ 0.10 ] & 5.7@xmath120.4 + we initially identified most of our bhcs from their long term ( @xmath213 year ) light curves and spectral histories . they exhibited spectra consistent with the hard state ( @xmath9 @xmath202.1 ) at 0.310 kev luminosities @xmath63@xmath21 erg s@xmath17 in at least one observation . for such spectra , we expect the 0.011000 kev luminosities to be @xmath623@xmath21 erg s@xmath17 , our limit for neutron star hard states following @xcite . we present the @xmath213 year , corrected , 0.310 kev luminosity lightcurve for each bhc in fig . [ lcs ] , along with its spectral history . the top panel in each case shows the lightcurve , with acis and hrc observations represented by circles and crosses respectively . the bottom panel shows @xmath9 for each acis observation ; sufficiently bright observations have freely fitted @xmath9 , while faint observations have @xmath9 fixed to the mean value ; this approach should be valid because we expect the faint observations to be in the hard state also , with similar a emission spectrum . figure [ hist ] shows histograms of the number of observation pairs with separation @xmath24 for the most frequently observed bhc , bhc13 . the histogram with filled circles includes acis and hrc observations , while the open circle histogram contains only acis observations . the observation pairs were logarithmically binned by @xmath24 , with a bin width of 0.2 dex . we present the 0.54.5 kev structure function for each of our bhcs in fig . [ sfs ] , with same scheme as for fig . the ensemble sf for typical agn obtained by vagnetti et al . ( 2011 ) is represented by a dashed line in each panel : sf(@xmath24 ) = 0.11@xmath63 . we included an acis only sf for each source in addition to an acis + hrc sf because faint hrc observations add extremely large noise components for some bhcs , suppressing the variation ; bhcs 8 and 35 only show significant excess variability over typical agn in their acis only sfs . we find that 28 out of 35 bhcs exhibit sfs with significantly more variation than the vagnetti et al . ( 2011 ) ensemble agn sf ( @xmath64@xmath53 excess in at least one @xmath24 channel , @xmath63@xmath53 excess in at least 2 channels , or @xmath62@xmath53 excess in at least 3 channels over the 3@xmath53 limit for the agn sf , assumed to be 0.11@xmath64 ) . this distinction is important , because agn and xbs often have very similar spectra . the x - ray sources with sfs exhibiting similar or less variation than the ensemble agn sf ( bhcs 2 , 5 , 7 , 21 , 23 , 24 , 25 , and 34 ) had mean 0.310 kev fluxes @xmath525.5@xmath37 erg cm @xmath65 s@xmath17 . we have found 24 x - ray sources in our observations that exhibited mean 0.310 kev fluxes @xmath525.5@xmath6610@xmath67 erg @xmath14 s@xmath17 over the @xmath213 years of chandra monitoring ; this includes 8 gc xbs and 1 possible supernova remnant . the 0.510 kev flux distribution of agn found by @xcite predicts 0.6 agn at this flux level within our observed region ( approximated by a circle with radius @xmath220@xmath1 or @xmath24 kpc ) . since the observed number of unidentified x - ray sources at these flux levels is larger than the predicted agn number by a factor @xmath225 , we conclude that they are likely xbs . after the initial identification of candidates , we sought the observations that best represented the bhc case for each object ; these were the highest quality spectra that appeared to exhibit high luminosity hard states . table [ bestspec ] provides the details of these observations . for each bhc we give the observation number , and net sources counts . we then give the absorption ( @xmath11 ) , photon index ( @xmath9 ) , @xmath15/dof and 0.310 kev luminosity for the best fit absorbed power law model . observations 30313827 are chandra acis observations , while 0109270101 , 0112570101 , and 0402560901 are xmm - newton observations . we note that the spectrum of bhc 22 rejected simple absorbed power law models . it is well described by an absorbed disk blackbody + power law model with @xmath68 = 1.97@xmath69 kev , and @xmath9 = 2.16@xmath70 , with the power law contributing 61@xmath1212% of the 8.0@xmath121.1 @xmath21 erg s@xmath17 0.310 kev luminosity ; it is consistent with either of two canonical black hole states : the hard state , or a steep power law state similar to grs 1915 + 105 ( mcclintock & remillard , 2006 ) . we applied a disk blackbody + blackbody emission model to our bhc spectra . the disk blackbody components were generally better constrained than the blackbody components , particularly for spectra with relatively few net source counts ; this is simply because there are fewer high energy photons than low energy photons and the disk blackbody is cooler than the blackbody . our best fits yielded blackbody contributions of 6599% to the 210 kev flux , in stark contrast to the neutron star systems where the disk blackbody dominates . the blackbody component dominates the 210 kev flux most when the disk blackbody temperature is lowest . four of our bhcs only exhibited apparent high luminosity hard states during 5 ks acis observations , and we were unable to constrain double thermal models for the resulting spectra . however , future deep observations may strengthen their bhc cases . the 31 bhcs that we did fit with the double thermal model yielded k@xmath68 0.628@xmath53 outside the ns parameter space ; the combined @xmath15 was 212 , for 30 dof ( probability @xmath23@xmath71 ) . clearly our bhcs are inconsistent with being soft ns xbs . ccccccccccccc bhc & @xmath11 / 10@xmath60 @xmath14 & k@xmath72 / kev & @xmath73/10@xmath50 & k@xmath74 / kev & @xmath75/10@xmath50 & @xmath76 & bh + 1 & 0.34@xmath120.07 & 2.00@xmath77 & 21@xmath122 & 0.88@xmath78 & 7.6@xmath120.5 & 0.73@xmath120.09 & p + 2 & 0.07f & 1.12@xmath79 & 3.0@xmath120.5 & 0.50@xmath80 & 0.8@xmath120.4 & 0.8@xmath120.2 & p + 3 & 0.07f & 1.18@xmath81 & 7.3@xmath120.3 & 0.49@xmath82 & 0.8@xmath120.3 & 0.90@xmath120.06 & s + 4 & 0.19@xmath120.06 & 1.50@xmath83 & 3.9@xmath120.8 & 0.62@xmath84 & 0.8@xmath120.4 & 0.8@xmath120.2 & p + 5 & 0.07f & 1.3@xmath85 & 2.0@xmath120.7 & 0.7@xmath86 & 0.6@xmath120.3 & 0.8@xmath120.4 & p + 6 & 0.07f & 1.34@xmath87 & 3.67@xmath120.15 & 0.510@xmath88 & 0.42@xmath120.05 & 0.90@xmath120.05 & s + 7 & 0.07f & 1.47@xmath89 & 4.59@xmath120.13 & 0.51@xmath90 & 0.54@xmath120.06 & 0.89@xmath120.04 & s + 8 & & & & & & & p + 9 & 0.07f & 0.93@xmath82 & 2.14@xmath120.12 & 0.38@xmath91 & 0.14@xmath120.05 & 0.94@xmath120.08 & s + 10 & 0.07f & 2.4@xmath92 & 2.3@xmath120.6 & 0.86@xmath93 & 0.7@xmath120.3 & 0.8@xmath120.3 & p + 11 & 0.07f & 1.11@xmath79 & 3.0@xmath120.5 & 0.50@xmath80 & 0.8@xmath120.2 & 0.80@xmath120.18 & p + 12 & 0.07f & 1.24@xmath94 & 1.74@xmath120.17 & 0.59@xmath95 & 0.7@xmath120.3 & 0.72@xmath120.12 & s + 13 & 0.07f & 1.5@xmath96 & 1.8@xmath120.4 & 0.67@xmath97 & 0.5@xmath120.3 & 0.8@xmath120.2 & p + 14 & & & & & & & p + 15 & 0.07f & 1.12@xmath79 & 1.6@xmath120.2 & 0.56@xmath98 & 0.6@xmath120.3 & 0.74@xmath120.17 & p + 16 & 0.07f & 2.2@xmath99 & 4.5@xmath120.5 & 1.03@xmath93 & 2.4@xmath120.8 & 0.65@xmath120.11 & p + 17 & 0.07f & 0.96@xmath100 & 1.17@xmath120.09 & 0.43@xmath87 & 0.17@xmath120.07 & 0.87@xmath120.10 & s + 18 & 0.07f & 1.05@xmath101 & 4.44@xmath120.19 & 0.41@xmath102 & 0.22@xmath120.08 & 0.95@xmath120.06 & s + 19 & 0.26@xmath120.04 & 1.7@xmath103 & 3.6@xmath120.5 & 0.83@xmath93 & 1.7@xmath120.5 & 0.68@xmath120.13 & p + 20 & 0.173@xmath120.010 & 1.61@xmath104 & 18.1@xmath121.7 & 0.83@xmath105 & 6.9@xmath121.9 & 0.7@xmath120.2 & s + 21 & 0.1@xmath120.05 & 1.7@xmath106 & 4.1@xmath120.5 & 0.77@xmath107 & 1.9@xmath120.7 & 0.68@xmath120.12 & p + 22 & 0.07f & 1.31@xmath87 & 3.71@xmath120.13 & 0.55@xmath108 & 0.57@xmath120.08 & 0.87@xmath120.04 & s + 23 & 0.07f & 0.99@xmath109 & 2.6@xmath120.5 & 0.45@xmath110 & 0.25@xmath120.13 & 0.9@xmath120.2 & s + 24 & 0.07f & 1.12@xmath87 & 2.12@xmath120.09 & 0.43@xmath90 & 0.19@xmath120.04 & 0.92@xmath120.05 & s + 25 & 0.07f & 1.56@xmath95 & 3.22@xmath120.13 & 0.61@xmath111 & 0.39@xmath120.07 & 0.89@xmath120.05 & s + 26 & 0.07f & 2.3@xmath112 & 2.1@xmath120.6 & 0.78@xmath113 & 0.5@xmath120.2 & 0.8@xmath120.3 & p + 27 & 0.07f & 0.94@xmath114 & 1.56@xmath120.14 & 0.46@xmath81 & 0.28@xmath120.15 & 0.85@xmath120.12 & s + 28 & 0.07f & 1.4@xmath115 & 2.3@xmath120.5 & 0.63@xmath97 & 0.9@xmath120.4 & 0.71@xmath120.2 & p + 29 & 0.07f & 1.00@xmath116 & 16.4@xmath120.9 & 0.29@xmath102 & 0.22@xmath120.11 & 0.99@xmath120.08 & s + 30 & & & & & & & p + 31 & 0.07f & 1.36@xmath117 & 5.72@xmath120.15 & 0.45@xmath118 & 0.4@xmath120.05 & 0.93@xmath120.03 & s + 32 & & & & & & & p + 33 & 0.07f & 1.15@xmath119 & 1.87@xmath120.17 & 0.52@xmath120 & 0.38@xmath120.17 & 0.83@xmath120.12 & s + 34 & 0.07f & 1.56@xmath121 & 3.7@xmath120.3 & 0.63@xmath111 & 0.67@xmath120.12 & 0.85@xmath120.08 & s + 35 & 0.07f & 1.44@xmath122 & 2.4@xmath120.4 & 0.68@xmath123 & 0.8@xmath120.3 & 0.75@xmath120.17 & p + in figs . [ dbbvl ] and [ bbvl ] we plot the 2.010 kev luminosity vs. temperature for the disk blackbody and blackbody components of the best fit double thermal models for each of our bhcs . we overlay the parameter space inhabited by the neutron star systems observed by lin et al . ( 2007 , 2009 , 2012 ) for spectral states where comptonization is not required in their models , i.e. the atoll soft states and z - source normal and flaring branches . we note that lin et al . quote luminosities from broader energy ranges ; however , our disk blackbody components make a negligible contribution above 10 kev , while the luminosities of our blackbody components are typically @xmath210% higher in the 2.020 kev band than the 2.010 kev band ; hence our observed 2.010 kev luminosity for our bhcs provide a fair comparison with the neutron star systems observed by lin et al . ( 2007 , 2009 , 2012 ) . as predicted , our bhcs inhabit a separate region of the disk blackbody parameter space to the ns systems studied by lin et al . ( 2007 , 2009 , 2012 ) , although some sources are consistent with ns values for @xmath68 within 3@xmath53 . we also see a correlation between the 210 kev luminosity and temperature of the disk blackbody component ; this is due to lower disk temperatures resulting in lesser contributions to the 210 kev flux . the bhcs are scattered more widely in the blackbody space , but are also systematically offset from the ns systems . remembering that lin et al . ( 2009 , 2012 ) failed to fit the horizontal branches of xte j1701@xmath7462 and gx 17 + 2 with their double thermal model , we also modeled our brightest bhcs ( @xmath124 @xmath6 10@xmath125 erg s@xmath17 ) with the disk blacbkody + blackbody + comptonization emission model that they employed for the horizontal branch spectra . @xcite used a cut - off power law to model the comptonization in gx 17 + 2 over the 2.960 kev band ; they required simultaneous fitting of several spectra to constrain the parameters of the cut - off power law , which yielded @xmath9 = 1.40@xmath120.14 , and a cut - off energy of 9.9@xmath121.0 kev . we therefore modeled the 0.310 kev xmm - netwon pn spectra , or 0.37.0 kev chandra acis spectra of our brightest bhcs with a blackbody , disk blackbody and a power law with @xmath9 fixed to 1.4 . @xcite found that the compotonized component contributed @xmath21050% of the total flux on the horizontal branch of gx 17 + 2 ; we fixed the flux of the power law component in our fits to @xmath230% of the total 210 kev flux . we fitted the spectra of bhcs 1 , 3 , 16 , 20 , 21 , and 31 in this way ; the results are summarized in table [ bdp ] . we find that this three component model results in lower temperatures for both thermal components than the double thermal model fits to the same spectra . i.e. , the addition of the power law component drives our bhcs away from the ns parameter space . increasing the power law contribution to 50% of the flux decreased the temperatures further . since none of the spectra from our brightest bhcs have three component fits that are consistent with the fits that successfully described the horizontal branch ( lin et al . , 2009 , 2012 ) , we conclude that none of our bhcs are z - sources . we have identified 26 new black hole candidates in the central region of m31 , using their structure functions or luminosities to identify them as x - ray binaries , and their high luminosity hard state spectra to classify them as bhcs . of these , 12 are strong candidates , and 14 are plausible candidates that may benefit from further observations . we were previously limited to identifying bhcs in globular clusters , due to the similarities between xb and agn spectra . this brings the total number of bhcs within 20@xmath1 of m31@xmath56 identified by their high luminosity hard states to 35 . ccccc bhc & @xmath11 / 10@xmath60 @xmath14 & k@xmath72 / kev & k@xmath74 / kev + 1 & 0.35 f & 1.99@xmath126 & 1@xmath103 + 3 & 0.07 f & 1.07@xmath127 & 0.42@xmath128 + 16 & 0.07 f & 1.44@xmath129 & 0.8@xmath106 + 20 & 0.197@xmath120.011 & 1.29@xmath107 & 0.73@xmath130 + 21 & 0.11@xmath120.07 & 1.6@xmath131 & 0.7@xmath132 + 31 & 0.07 f & 1.27@xmath133 & 0.43@xmath108 + the structure functions of most of our bhcs reveal them to be substantially more variable than typical agn ( as measured by vagnetti et al . , 2011 ) over a wide range of time scales . those bhcs with comparable or less variability than the agn have 0.510 kev luminosities matched by @xmath20.6 agn within the observed region , according to @xcite . it is therefore unlikely that our new bhcs are agn . we have found that our bhc spectra exist in a separate parameter space to galactic neutron star systems when we compare our best fits for absorbed blackbody + disk blackbody emission models with the systems studied by lin et al . ( 2007 , 2009 , 2012 ) . this is expected because this double thermal model fails to fit hard state spectra ; the disk blackbody component is forced low temperatures in order to provide the low energy flux . indeed , the probability that our bhcs all lie in the ns parameter space when fitted with the double - thermal model is just @xmath23@xmath71 . seven of our bhcs ( 20% of our total sample ) were found within 100@xmath3 of the m31 nucleus , lending support to the hypothesis that the m31 bulge is sufficiently dense to form a significant number of xbs dynamically , as seen in globular clusters ; since the stellar velocities in the m31 bulge are considerably higher than in gcs , surviving xbs are expected to have short periods and are likely to contain black hole accretors @xcite . we thank the referee for their thoughtful comments . we thank z. li for merging the chandra data . this research has made use of data obtained from the chandra data archive , and software provided by the chandra x - ray center ( cxc ) . we also include analysis of data from xmm - newton , an esa science mission with instruments and and contributions directly funded by esa member states and the us ( nasa ) . is funded by chandra grants go0 - 11106x and go1 - 12109x , along with hst grants go-11833 and go-12014 . m.r.g . and s.s.m are partially supported by nasa grant nas-03060 .

high energy jet distributions measured since 1992 at the fermilab tevatron proton antiproton collider are presented and compared to theoretical predictions . the statistical uncertainties on these measurements are significantly reduced relative to previous results . the systematic uncertainties are comparable in size to the uncertainty in the theoretical predictions . although some discrepancies between theory and measurements are noted , the inclusive jet and dijet cross sections can be described by quantum chromodynamics . prospects for reducing the uncertainty in the theoretical predictions by incorporating tevatron measurements into the proton parton distributions are discussed . dijet distributions , in excellent agreement with quantum chromodynamics , set a 2.5 tev limit on the mass of quark constituents . epsf 0d cross sections , mass and angular distributions , compositeness , qcd , scaling , parton distribution functions

hep-ex9903058

in the past seven years , quantum chromodynamics ( qcd ) , the accepted theory of quark and gluon interactions , has been confronted with a set of precise and varied measurements of jet production from the fermi national accelerator laboratory proton antiproton collider . in @xmath0 collisions , jet production can be understood as a point like collision of a quark or gluon from the proton and a quark or gluon from the antiproton . after colliding , these scattered partons manifest themselves as sprays of particles or `` jets '' . the extremely high energy of these interactions provides an excellent opportunity to test our understanding of perturbative qcd ( pqcd ) . in particular , differences between qcd calculations and measured jet processes reveal information about both the partonic content of the proton and the nature of the strong interaction . further , unexpected deviations may signal the existence of new particles or interactions down to distance scales of @xmath1 meters or less . historically , jet measurements have involved tests of qcd at the highest energies and searches for new physics . prior to 1992 , the inclusive jet cross section , one of the most important hadron collider results , had been measured at the isr @xmath2 collider at a center of mass energy of @xmath3 gev @xcite , the cern @xmath0 collider at 546 and 630 gev @xcite , and the fermilab tevatron @xmath0 collider at 1800 gev @xcite . these data span a factor of twenty in beam energy and a factor of two hundred in jet energy transverse to the proton beam ( @xmath4 ) and , in general , are reasonably well described by qcd . studies of the correlations between the leading two jets of an interaction also test qcd and provide opportunities to search for new physics . measurements of dijet distributions prior to 1992 , such as the invariant mass of the leading two jets or the angular separation between the jets , are also in good agreement with qcd @xcite . using the model of eichten , lane , and peskin @xcite these data are used to search for composite quarks @xcite , with the best limits coming from an inclusive jet cross section measurement @xcite . in this model excess jet production at large relative to qcd is interpreted as the product of quark constituent scattering . because these data sets are relatively small all searches are limited by statistical precision rather than systematic uncertainties at the highest @xmath4 . a summary of these inclusive jet and dijet results can be found in the review of qcd by huth and mangano @xcite . in 1992 , with the advent of high luminosity data collection at the tevatron , a new era of precise @xmath0 jet physics began . from 19921993 ( run 1a ) each of the two collider experiments accumulated @xmath5 20 @xmath6 of data and from 19941996 ( run 1b ) each accumulated @xmath5 100 @xmath6 . both data sets were taken at @xmath7 1800 gev . at the end of run 1b a small portion of data ( @xmath5 600 @xmath8 ) was taken at @xmath7 630 gev . the two runs together represent a data set @xmath5 50 times larger than any preceding data set . in contrast to previous tests of qcd , the run 1 jet measurements have statistical uncertainties significantly smaller than experimental or theoretical systematic uncertainties . in fact , the inclusive jet analysis from the run 1a sample yielded a significant discrepancy between data and contemporaneous theoretical predictions as well as good agreement with previous measurements @xcite . in addition to stimulating great excitement over the possibility of departures from qcd , the result motivated a re evaluation of the uncertainties associated with inclusive jet cross section calculations @xcite . later measurements of the inclusive jet cross section using the run 1b samples stimulated further discussions since one confirmed @xcite the run 1a measurement while the other was well described by qcd . in addition , subsequent studies have indicated that qcd can describe the observed run 1a cross section through adjustments to the parton distribution functions ( pdfs ) , which represent the fraction of proton momentum carried by the constituent quarks and gluons . independent of the actual high @xmath4 behavior of the inclusive jet cross section , the run 1a and 1b results have revealed large uncertainties in the theoretical predictions . in particular , the pdfs are derived from fits to data from many different experiments , all of which are collected at low energy and extrapolated to tevatron jet energies . representation of the uncertainties in the resulting pdfs is a complex issue and has never been precisely resolved . consequently , the tevatron collaborations have taken an interest in derivation of the pdfs from information collected at the tevatron . measurements of dijet mass distributions and dijet cross sections at a variety of scattering angles provide information on the momentum distributions of the partons . by comparing the different measurements , constraints on the pdfs can be established , and sensitivity to the presence of new physics should ultimately be improved . this review opens with a description of the theoretical framework behind perturbative qcd and a discussion of the uncertainties associated with predictions . descriptions of the two detectors in which these measurements were performed and jet reconstruction algorithms are also provided . the review then devotes a section each to the run 1 measurements of the inclusive jet cross section , the ratio of inclusive cross sections at different beam energies , dijet differential cross sections , and dijet mass and angular distributions . as will be seen , these measurements offer a detailed look into the composition of the proton and the nature of the strong interaction . a conclusion summarizes the results and suggests future avenues of research . the proton antiproton interaction , a fairly general scattering process , nicely introduces the concepts of leading order and next to leading order pqcd . as illustrated in fig . [ order ] , inelastic scattering between a proton and an antiproton can be described as an elastic collision between a single proton constituent and single antiproton constituent . these constituents are collectively referred to as partons and in qcd are quarks and gluons . the non colliding constituents of the incoming proton and anti proton are called beam fragments or spectators . predictions for jet production are given by folding experimentally determined parton distribution functions @xmath9 with perturbatively calculated two body scattering cross sections @xmath10 . see , for example , reference @xcite for a detailed discussion . the two ingredients can be formally combined to calculate any cross section of interest : @xmath11 $ ] . the parton distribution functions @xmath12 describe the momentum fraction @xmath13 of the incident hadron momentum @xmath14 carried by a parton of type @xmath15 ( gluons or quarks ) . the pdf is defined in terms of the factorization scale @xmath16 . the hard two body cross section is a function of the momentum carried by each of the incident partons @xmath17 , the strong coupling parameter @xmath18 , the scale @xmath19 characterizing the energy of the hard interaction , and the renormalization scale @xmath20 . final state partons manifest themselves as collimated streams or `` jets '' of particles . this formal description includes no explicit hadronization or fragmentation functions to describe the transition from partons to jets . for most high @xmath4 measurements these effects are small and jets are identified as partons . the two body scattering has been illustrated in fig . [ order ] with a leading order ( lo ) graph . the cross section for this process is proportional to two powers of the strong coupling parameter , @xmath18 , which come from the two vertices . although useful , the lo picture is too simple and has large normalization uncertainties . next to leading order ( nlo ) or o@xmath21 calculations include one additional parton emission . a schematic example is shown in the second illustration of fig . [ order ] . here a final state quark has radiated an additional gluon and the entire scattering process is proportional to @xmath22 . depending on the proximity of the other partons , a jet " could result from one or two ( combined ) partons . this in turn results in parton level predictions for the shape of jets , and for the effects of clustering parameters . = 5.0 in = 2.5 in a complete theoretical prediction of jet production should not depend on internal calculational details ; however , this is not the case for fixed order perturbative qcd , which depends on the choice of factorization and renormalization scales . the factorization scale , a free parameter , determines how the contributions of initial state radiation are factorized between the pdfs and @xmath10 . the renormalization scale is also arbitrary and is related to choices in the theoretical calculations designed to control or renormalize ultraviolet singularities . typically , as in this report , @xmath20 and @xmath16 are set equal to each other , @xmath23 , and @xmath24 is chosen to be of the same order as @xmath19 . the sensitivity of the theoretical predictions to @xmath24 is often taken as a measure of the uncertainty from the contributions of higher order terms . predictions for jet production at nlo have been derived by ellis , kunszt , and soper @xcite ; aversa _ et al . _ @xcite ; and giele , glover , and kosower @xcite . the nlo predictions are much less sensitive to the choice of @xmath24 @xcite . the `` eks '' program by ellis , kunszt , and soper @xcite can generate analytic predictions for jet cross sections as a function of final state parameters . the `` jetrad '' program by giele , glover , and kosower @xcite generates weighted `` events '' with final state partons . cross sections are calculated by generating a large number of events as a function of final state parameters . all predictions in this document have been generated either with or jetrad . the two programs agree within a few percent@xcite . a nlo qcd calculation requires selection of several input parameters including specification of a parton clustering algorithm , a perturbative scale , and pdfs . taken together these choices can result in approximately 30% variations in the theoretical predictions . the single largest uncertainty is due to the pdfs . these input choices and the resultant changes in the predictions are discussed in the following paragraphs . the variations are illustrated by comparing predictions for the inclusive jet cross section . as in the experimental results to follow , the cross section is reported as a function of jet and pseudorapidity @xmath25 where @xmath26 is the angle between the jet and the proton beam , and is summed over @xmath27 , the azimuthal angle around the beam . for reference , @xmath28 for jet production at 90 degrees relative to the proton beam . the nlo predictions may include two or three final state partons . to convert the partons to the equivalent of jets measured in a detector , a clustering algorithm is employed . the snowmass cone algorithm @xcite was proposed as a way to minimize the difference between theoretical predictions and experimental measurements . in theory calculations , two partons which fall within a cone of radius r in @xmath29@xmath27 space ( r = @xmath30 and @xmath31 and @xmath32 are the separation of the partons in pseudorapidity and azimuthal angle ) are combined into a jet " . a standard cone radius of @xmath33 0.7 is used by both collider experiments for most measurements . a consequence of this algorithm is that partons have to be at least a distance of 2r apart to be considered as separate jets . subsequent studies indicated that the experimental clustering algorithms ( described in later sections ) were more efficient at separating nearby jets @xcite than the idealized snowmass algorithm . in other words , two jets would be identified even though they were separated by less than 2r . an additional parameter , @xmath34 , was introduced in the qcd predictions to mimic the experimental effects of cluster merging and separation . partons within @xmath35 were merged into a jet , while partons separated by more than @xmath35 were identified as two individual jets . as described in detail in the references , a value of @xmath34 = 1.3 was found to give the best agreement with the data for cross section and jet shape measurements @xcite . this corresponds in the data to the 50% efficiency point in jet separation @xcite , i.e. two jets within @xmath34=1.3 of one another are merged 50% of the time and identified as two individual jets 50% of the time . at separations of 1.0r the algorithms nearly always merge the two jets and at values of 1.6r they nearly always identify two separate jets . the top panel of figure [ theory_err ] shows the change in the prediction for the inclusive jet cross section using the eks program when @xmath34 is decreased from 2 to 1.3 . the result is primarily a normalization change of 5 to 7% . = 4.0 in = 2.5 in because a nlo calculation is truncated at order @xmath36 there is some residual dependence on the scale @xmath24 at which the calculation is performed . the scale is usually taken to be proportional to @xmath37 ( or just @xmath4 ) , maximum transverse jet energy @xmath38 in a given event , or the total center of mass energy . other choices for the scale are possible . to study the scale dependence of the predictions , the magnitude of the scale is varied by a multiplicative coefficient , common choices are @xmath39 and @xmath40 . the middle panel of fig . [ theory_err ] shows a mild dependence at the 29% level on the definition of the scale . the lowest panel shows that typical variations of the multiplicative coefficient lead to 520% shifts in the cross section with only small changes in shape above 100 gev . parton distribution functions are derived from global fits to data primarily from deep inelastic scattering ( dis ) fixed target and electron proton collider experiments . in dis , a lepton probe is used to sample the partonic structure of the target hadrons . for use at the tevatron , the resulting distributions must be evolved from the low energy dis results to the high energy range of jet measurements . as the data from dis experiments increases and improves , new parton distribution functions are derived . some recent pdfs also incorporate data from the tevatron . the result is a plethora of pdfs , each with its own specific list of data included in the fit , assumptions as to the value of @xmath41 , functional forms for the quark and gluon momentum distributions , and assumptions concerning the contributions of the gluons , which are not constrained by the dis results . the most recent pdfs include the most precise data and combined knowledge , and supplant the previous pdfs . several groups have analyzed the available data and produced families of candidate pdfs . within a family , the individual pdfs represent the range of predictions resulting from changes in one of the input assumptions such as the value of @xmath41 . additional variations come from differences in the input data sets or the relative weighting between the data sets . the cteq2 m @xcite , cteq3 m @xcite and mrsa@xmath42 @xcite families of pdfs incorporate data published before 1994 and do not include tevatron jet data . the cteq4 m pdfs @xcite include data published before 1996 ; cteq4hj @xcite additionally has a high @xmath13 gluon adjustment designed to accommodate the run 1a high jet cross section measurements . the mrst @xcite pdf family utilizes data published before 1998 and adds a contribution for a putative initial transverse momentum of the partons . comparison between observed cross sections and nlo predictions with alternate pdfs provide some insight into the quark and gluon composition of the proton . as shown in the fig . [ pdf_err ] the pdfs can result in @xmath5 20% variations in jet cross sections ; typical variations within a set are on the order of 510% . although the quark distributions were thought to be well known and the contribution of the gluons was expected to be small , investigations @xcite have shown that there are uncertainties at high @xmath13 , which were ignored in the derivation of early pdfs . in addition , studies@xcite revealed that the gluon distribution could be adjusted to give a significant increase in the jet cross section at high @xmath43 , while maintaining reasonable agreement with the low energy data sets . = 3.5 in = 2.5 in the collider detector at fermilab ( cdf ) @xcite and the d0 detector @xcite at the fermi national accelerator laboratory tevatron @xmath0 collider are complementary general purpose detectors designed to study a broad range of particle physics topics . the tevatron is designed with six locations at which the counter rotating proton and antiproton beams can collide . the cdf and d0 detectors each occupy one of these interaction regions ; d0 takes its name from the alphanumeric designation of its interaction region . each detector is comprised of a series of concentric sub systems , which surround the @xmath0 interaction region . immediately outside the tevatron beam pipe each detector has a tracking system designed to detect charged particles . the tracking systems are surrounded by calorimeters , which measure the energy and direction of electrons , photons , and jets . finally , the calorimeters are covered with muon tracking systems . luminosity monitors at each detector measure the beam exposure . these monitors are located near the beam lines and detect particles from the @xmath0 collisions . the luminosity at a given site is given by dividing the luminosity monitor event rate by that portion of the total @xmath0 cross section to which the luminosity monitors have acceptance . a direct comparison of jet cross sections at the two experiments is complicated by the fact that the collaborations adopt slightly different values for the total @xmath0 cross section @xcite . as a result , the cdf jet cross section measurements are 2.7% higher than the corresponding d0 jet cross section measurements @xcite . in the following sections we mention primarily the central tracking systems and the calorimeters since these systems are the most important for jet identification and reconstruction . as will be seen , cdf has a high resolution particle tracking system , which makes a crucial contribution to jet energy calibration . on the other hand , and in a complementary fashion , d0 has highly segmented , uniform , and thick calorimetry well suited to _ in situ _ jet energy measurements . both calorimeters are segmented into projective towers . each tower points back to the center of the nominal interaction region and is identified by its pseudorapidity and azimuth . the polar angle @xmath26 in spherical coordinates is measured from the proton beam axis , and the azimuthal angle @xmath27 from the plane of the tevatron . the towers are further segmented longitudinally into individual readout cells . the energy of a calorimeter tower is obtained by summing the energy in all the cells of the same pseudorapidity and azimuth . the transverse energy or @xmath4 of a tower is the sum of the components of the cells . these components are calculated by assigning each cell of a tower a massless four vector with magnitude equal to the energy deposited and with the direction defined by the unit vector pointing from the event origin to the center of the tower segment . the central tracking volume ( @xmath44 ) of d0 includes the vertex chamber , transition radiation detector , and central drift chamber , arranged in three cylinders concentric with the beamline , and two forward drift chambers . the non magnetic tracker is compact , with an outer radius of 75 cm and an overall length of 270 cm centered on @xmath45 . without the need to measure momenta of charged particles , the prime considerations for tracking were good two track resolving power , high efficiency , and good ionization energy measurement . for jet physics the central tracker is used to find event vertices . jet detection in the d0 detector primarily utilizes liquid argon uranium + calorimeters which are hermetic , finely segmented , thick , uniform and have unit gain . these sampling calorimeters are composed of alternating layers of liquid argon and absorber . the particles in a jet interact with the absorber , and the resulting particle shower ionizes the liquid argon . this ionization is detected as a measure of the jet energy . the calorimeter is enclosed in three cryostats : the central calorimeter ( cc ) covers @xmath46 , and the end calorimeters ( ecs ) extend the coverage to @xmath47 . the calorimeters have complete @xmath48 azimuthal coverage . between the cryostats the calorimeter sampling is augmented by scintillator tiles with segmentation matching the argon calorimeters . [ twojet ] is a graphical representation of the calorimeter modules described below . = 3.0 in = 2.5 in the cc includes three concentric rings of calorimeter modules . there are 32 electromagnetic calorimeter modules ( ccem ) in the inner ring , 16 fine hadronic modules ( ccfh ) in the middle ring , and 16 coarse hadronic modules ( ccch ) in the outer ring . the ccem and ccfh calorimeters have uranium absorber plates , and the ccch has copper absorber plates . longitudinal segmentation includes seven samples : four in the ccem , three in the ccfh , and one in the ccch . at @xmath28 the cc has a thickness of 7.2 nuclear absorption lengths . the calorimeter cells are segmented into @xmath49 except at shower maximum in the third layer of the ccem where the segmentation is @xmath50 . the calorimeter segmentation is designed to form projective towers of size @xmath51 geometry which point to @xmath45 . in the cc because of the finer resolution in the third layer of the ccem each tower is comprised of eleven cells . the two mirror image ecs contain four module types . an electromagnetic module ( ecem ) surrounding the beam pipe is backed by an inner hadronic module ( ecih ) which also surrounds the pipe . outside the ecem and ecih are concentric rings of 16 middle and outer hadronic modules ( ecmh and ecoh ) . the ecem and ecih have uranium absorber plates and four longitudinal segments . the ecmh has four uranium absorber segments nearest the interaction point and one stainless steel absorber segment behind . the ecoh employs stainless steel plates and includes three longitudinal segments . at = 2.0 a jet encounters 89 longitudinal segments and a nuclear interaction thickness of @xmath52 . the transverse segmentation is similar to that of the cc . the cc(ec ) electron energy resolution is @xmath53 plus 0.3(0.3)% added in quadrature and for hadrons @xmath54 plus 4.5(3.9)% added in quadrature . the calorimeter response to the different types of particles is the most important aspect of the d0 jet energy calibration . electromagnetically interacting particles , like photons ( @xmath55 ) and electrons deposit most of their energy in the electromagnetic sections of the calorimeters . hadrons , by contrast , lose energy primarily through nuclear interactions and extend over the full interaction length of the calorimeter . in general , the calorimeter response to the electromagnetic ( @xmath56 ) and nuclear or hadronic components ( @xmath57 ) of hadron showers is not the same . non compensating calorimeters have a response ratio @xmath58 greater than one and suffer from non - gaussian event to event fluctuations in the fraction of energy lost through electromagnetic production . such calorimeters give a non - gaussian signal distribution for hadrons and jets . the d0 calorimeter is nearly compensating and the hadronic and jet response are well described by a gaussian distribution @xcite . in fact , single jet resolution as measured with dijet events has a gaussian line shape and is approximately 7% at 100 gev and 5% at 300 gev @xcite . the d0 jet calibration requires correction for the hadronic response of the jet ; showering of energy outside the cone ; and subtraction of an offset which can be attributed to instrumental effects , pile up from previous beam beam crossings , additional interactions , and spectator energy . these corrections are derived primarily _ in situ _ the correction for hadronic response begins with the electromagnetic calibration of the calorimeter which is performed with dielectron and diphoton decays of the @xmath59 and @xmath60 resonances since electrons and photons deposit their energy in the electromagnetic calorimeters . the hadronic response for centrally produced @xmath55jet events is derived from data using balance between the photon and jet . because of the rapidly falling @xmath55jet cross section the central calorimeter balance technique is limited to jet energies below 150 gev . balancing central , well measured photons against high rapidity jets permits the energy calibration to exceed 300 gev . the response calibration is extended above 300 gev using simulated @xmath55jets events . for a r=0.7 , 100 gev jet at @xmath28 the hadronic response is about @xmath61 . the showering correction for this jet is about @xmath62% and is derived from a study of jet profiles . the total offset correction for a r=0.7 jet at @xmath28 as determined from a study of data sets with various triggering and luminosity conditions is about @xmath63 gev . the contribution from the underlying event or spectator energy alone is roughly @xmath64 gev . at @xmath28 the mean total energy correction for a r=0.7 , 100 gev jet is @xmath65% and decreases slowly with energy . the cdf central ( @xmath66 ) tracking system has three sub systems located within a 1.5 t magnetic field , that is provided by a superconducting solenoid coaxial with the beam . nearest the beam , a four layer silicon microstrip vertex ( svx ) detector @xcite occupies the radial region between 3.0 and 7.9 cm from the beamline and provides precision @xmath67 measurements . outside the svx a vertex drift chamber ( vtx ) provides @xmath68@xmath69 tracking information and is used to locate the position of the @xmath0 interaction ( event vertex ) in @xmath69 , along the beam line . both the svx and the vtx are mounted inside a 3.2 m long drift chamber called the central tracking chamber ( ctc ) . the radial coverage of the ctc is from 31 to 132 cm . the momentum resolution @xcite of the svx ctc system is @xmath70^{1/2}$ ] where @xmath71 has units of gev / c . the ctc provides _ in situ _ measurement of the calibration and response of the calorimeter to low energy particles ( where test beam information is not available ) along with measurements of jet fragmentation properties . the solenoid and tracking volumes of cdf are surrounded by calorimeters , which cover @xmath48 in azimuth and @xmath72 . the central electromagnetic ( cem ) calorimeter covers @xmath73 and is followed at larger radius by the central hadronic calorimeters ( cha and wha ) which cover @xmath74 . these calorimeters use scintillator as the active medium . the cem absorber is lead and the cha / wha absorber is iron . the calorimeters are segmented into units of 15 degrees in azimuth and 0.1 pseudorapidity . two phototubes bracket each tower in @xmath27 and the average of the energy in the two tubes is used to determine the @xmath27 position of energy deposited in a tower . the interaction length of both the cha and wha is @xmath75 . electron energy resolution in the cem is @xmath76 plus 2% added in quadrature . for hadrons the single particle resolution depends on angle and varies from roughly @xmath77 plus 3% added in quadrature in the cha to @xmath78 plus 4% added in quadrature in the wha . in the forward regions ( @xmath79 ) calorimetric coverage is provided by gas proportional chambers . the plug electromagnetic ( pem ) and hadronic calorimeters ( pha ) cover the region from @xmath80 . the forward electromagnetic ( fem ) and hadronic calorimeters ( fha ) cover the region from @xmath81 . the segmentation of these detectors is roughly 0.1 in @xmath29 and 5 degrees in @xmath27 . figure [ cdflego ] shows a 3jet event in the cdf calorimeter . in this `` lego '' plot the calorimeter is `` rolled out '' onto the @xmath29@xmath27 plane . the tower height is proportional to the @xmath4 deposited in the tower . the darker and lighter shading of each tower corresponds to the @xmath4 of the electromagnetic and hadronic cells of the tower respectively . the oval around each clump of energy indicates the jet clustering cone described in the next section . = 4.0 in = 4.0 in observed jet energies are corrected for a number of effects including the calibration and response of the detectors and the background energy from the remnants of the @xmath0 interaction . in the central detectors corrections associated with detector response are obtained from a monte carlo program , which is tuned to give good agreement with the data from electron and hadron test beams ( @xmath82 ranged from 10 to 227 gev ) , and to data from an _ in situ _ study using ctc track momenta and isolated hadrons with 400 mev / c@xmath83 10 gev / c . in addition , the monte carlo jet fragmentation parameters were tuned to agree with both the number and momenta of particles observed in jets . the resulting simulation was then used to determine `` response functions '' , which represent the response of the calorimeter to jets as a function of jet @xmath4 . the width of the response functions represents the jet energy resolution and can be expressed as @xmath84 = 0.1@xmath4 + 1 gev @xcite . calibration of the plug and forward detectors is achieved through a dijet balancing technique @xcite . the @xmath4 of jets in the plug and forward detectors is balanced against jets in the central region whose calibration is pinned by the tracking information . the energy left in a jet cone from the hard interaction spectators is called the underlying event energy " . a precise theoretical definition of this quantity does not exist . experimentally , it has been estimated from events in which there is no hard scattering , and from the energy in cones perpendicular to the jet axis in dijet events . for a cone of radius 0.7 the contribution to the jet @xmath4 is of order 1 gev . an uncertainty of 30% is assigned to this quantity to cover reasonable variations in the definition of this quantity . the magnitude of the total jet energy corrections and the corresponding uncertainty depend on the @xmath4 of the jets and their location in the detector , and on the specific analysis . typically , the jet correction factor is in the range of 1.0 to 1.2 . the fermilab tevatron is undergoing major improvements for run 2 , which is slated to begin in the year 2000 . the energy of the proton and antiproton beams will increase from 900 to 1000 gev and the instantaneous luminosity will increase by at least a factor of 5 . the expected run 2 data sample is 2@xmath85 in the first 2 years of operation . both the cdf and d0 collaborations are improving their detectors for run 2 operations . among other things , cdf is replacing the gas calorimeters with scintillating tiles and is closing gaps between calorimeters . cdf is also extending the tracking such that calibration of the new plug detector can be accomplished using tracks , as in the central detector . d0 is also upgrading a number of detector components . for jet measurements the most important upgrade is the replacement of the nonmagnetic tracking system with a magnetic tracker . the tracker ( including a high resolution silicon strip detector , a scintillating fiber tracker , and electron / photon preshower detectors ) will enable calibration of the calorimeter with single particles . thus , in run 2 the detectors will be more similar than in run 1 : cdf will have more uniform calorimetry and d0 will have magnetic tracking . the result of these improvements should be an overall reduction of jet measurement uncertainties . from the experimental point of view , jets are the manifestation of partons as showers of electromagnetic and hadronic matter . jets are observed as clusters of energy located in adjacent detector towers . typically , a jet contains tens of neutral and charged pions ( and to a lesser extent kaons ) , each of which showers into multiple cells . a single jet illuminates roughly 20 towers ( this corresponds to approximately 40 calorimeter cells in the cdf detector and over one hundred detector cells in the d0 detector ) . [ twojet ] illustrates the energy distribution of a two jet event in the d0 detector . each rectangular outline represents the energy deposited in calorimeter cells at fixed @xmath29 and depth . the eye quite naturally clusters the energy into two jet like objects . however , for the purposes of jet cross section measurements a more quantitative definition of the jet is required . for high @xmath4 measurements at cdf and d0 , cone algorithms are used to identify jets in the calorimeters . cone algorithms operate on objects in pseudorapidity and azimuth space , such as particles , partons , calorimeter cells , or calorimeter towers . for the sake of simplicity towers will be used in the following discussions . the cdf and d0 algorithms are both based on the snowmass algorithm @xcite . this algorithm defines a jet as those towers within a cone of radius @xmath86 = 0.7 . the jet is the sum of the transverse energies of the towers in the cone , and the location of the jet is defined by the weighted @xmath29 and @xmath27 centroids : @xmath87 @xmath88 @xmath89 the sum over @xmath90 is over all towers that are within the jet radius @xmath91 . in figure [ cdflego ] the clustering cone is shown by the oval around each jet and is centered on the cluster centroid . the two overlapping cones in this event indicate that the two nearby clusters have been identified as two separate jets . the snowmass algorithm does nt specify cell thresholds or the handling of such overlapping jet cones . these details must be dealt with according to the needs of individual experiments , and led to the introduction of the @xmath34 parameter in the theoretical calculations . in the d0 experiment , jets are defined in two stages @xcite . in the first or clustering stage , all the energy that belongs to a jet is accumulated , and in the second stage , the @xmath92 @xmath27 and of the jet are defined . the clustering consists of the following steps : 1 ) calorimeter towers with @xmath93 1 gev are enumerated . starting with the highest tower , preclusters are formed by adding neighboring towers within a radius of r=0.3 . 2 ) the jet direction is calculated for each precluster using the sums defined for the snowmass algorithm . 3 ) all the energy in the towers in a cone of radius r=0.7 around each precluster is accumulated and used to recalculate @xmath29 and @xmath27 . 4 ) steps 2 and 3 are repeated until the jet direction is stable . finally , in the second stage , the jet energy and directions are calculated according to the equations : + @xmath94 @xmath95 @xmath96^{2}\ ] ] where @xmath97 , @xmath98 , and @xmath99 . studies have shown that for @xmath100 the final d0 jet directions are , within experimental errors , equal to the snowmass directions . overlapping jets are merged if more than 50% of the smaller jet is contained in the overlap region . if less than 50% of the energy is contained in the overlap region , the jets are split into two jets and the energy in the overlap region is assigned to the nearest jet . after merging or splitting the jet directions are recalculated . the cdf cluster algorithm @xcite has two stages of jet identification similar to the d0 algorithm . the first stage consists of the following steps : 1 ) a list of towers with @xmath1011.0 gev is created ; 2 ) preclusters are formed from an unbroken chain of contiguous seed towers with continuously decreasing tower @xmath4 ; if a tower is outside a window of @xmath102 towers surrounding the seed it is used to form a new precluster ; 3 ) the preclusters are grown into clusters by finding the weighted centroid and collecting the energy from all towers with more than 100 mev within r=0.7 of the centroid ; 4 ) a new centroid is calculated from the set of towers within the cone and a new cone drawn about this position ; steps 3 and 4 are repeated until the set of towers contributing to the jet remains unchanged ; 5 ) overlapping jets are merged if they share @xmath9375% of the smaller jet s energy ; if they share less , the towers in the overlap region are assigned to the nearest jet . in the second stage , the final jet parameters are computed . the angles and energy are calculated as in the d0 algorithm , however the jet is given by : @xmath103 one philosophical difference between the cdf and snowmass algorithms is that the cdf jets have mass ( @xmath104 ) . studies @xcite found that the cdf clustering algorithm and the snowmass algorithm were numerically very similar . both the cdf and d0 algorithms are based on the snowmass algorithm , but the reality of jets of hadrons measured by a finitely segmented calorimeter necessitates the introduction of additional steps and cuts . apart from small definitional details , the most significant difference between the snowmass algorithm and its experimental implementations is the handling of cluster merging and separation . to simulate these effects in the nlo calculations the @xmath34 parameter was introduced . considering the complexity of the hadronic showers is it remarkable that a nlo calculation , with only 2 or 3 partons in the final state and a single additional parameter can provide a good description of the observed jet shapes and cross sections @xcite . the inclusive jet cross section represents one of the most basic tests of qcd at a hadron hadron collider . it reflects the probability of observing a hadronic jet of a given and rapidity in a @xmath0 collision . the term inclusive indicates that all the jets in an event are included in the cross section measurement and that the presence of additional non jet objects ( for example electrons or muons ) is irrelevant . theoretical calculations are normally expressed in terms of the invariant cross section : @xmath105 where @xmath106 and @xmath107 are the jet energy and momentum . in terms of the natural experimental variables , the cross section is given by @xmath108 which is related to the first expression by , @xmath109 massless jets have been assumed ( @xmath71 = @xmath4 ) ; the @xmath48 comes from the integration over the azimuthal angle @xmath27 . the measured cross section is simply the number of jets @xmath110 observed in an @xmath29 and interval normalized by the total luminosity exposure , @xmath111 : @xmath112 . the cdf and d0 inclusive jet analyses place similar requirements on the events and jets selected for calculation of the cross sections . both collaborations eliminate poorly measured events by requiring the event vertices to be near the center of the detector . backgrounds ( such as cosmic rays ) are eliminated by rejecting events with large missing @xmath43 . spurious jets from cosmic ray or instrumental backgrounds are eliminated with quality cuts based on jet shapes . corrections are made to the measured cross sections to account for the event and jet detection inefficiencies , mismeasurement of the jet energies , and for energy falling in a jet cone from other sources ( such as remnants of the @xmath0 collision , or additional @xmath0 interactions ) . no corrections are made for partons showering outside the jet cone as this should be included in the nlo theoretical calculations . we begin our discussion of the inclusive jet cross section with measurements by the cdf collaboration . the run 1a measurement stimulated interest due to a discrepancy with qcd predictions at high @xmath4 . a subsequent measurement by the d0 collaboration with run 1b data is well described by theoretical predictions , while the preliminary cdf run 1b result is in good agreement with the 1a measurement . in the next sections , the cdf and d0 measurements are each described in some detail and then comparisons of the two results are presented . in 1996 cdf published the inclusive jet cross section measured from the run 1a data sample @xcite for jet @xmath4 from 15 to 440 gev in the central rapidity range 0.1@xmath113 0.7 . this analysis followed the same procedures as previous measurements @xcite for correcting the cross section and for estimating the systematic uncertainties . this process will be described briefly below . more detail is available in references @xcite . the measured jet @xmath4 spectrum requires corrections for energy mismeasurement and for smearing effects caused by finite @xmath4 resolution . this is accomplished with the unsmearing procedure " described in @xcite . as mentioned earlier , a monte carlo simulation program @xcite tuned to the cdf data is used to determine detector response functions . a trial true ( unsmeared ) spectrum is smeared using these response functions and compared to the raw data . the parameters of the trial spectrum are iterated to obtain the best match between the smeared trial spectrum and the raw data . the corresponding unsmeared curve is referred to as the `` standard curve '' , and is used to correct the measured spectrum . the simultaneous correction for response and resolution produces a result which is independent of the @xmath4 binning used in the measurement while preserving the statistical uncertainty on the measured cross section . for jet @xmath4 between 50 and about 300 gev these corrections increase both the @xmath4 and the cross section by @xmath510% . at lower and higher @xmath4 the cross section corrections are larger due to the steepening of the spectrum . figure [ cdfinc1a ] shows the cdf measurement of the inclusive jet cross section from the 20 run 1a data sample compared to a nlo prediction from the eks program with @xmath114 , @xmath34 = 2.0 and mrsd0 pdfs . the inset shows the cross section on a logarithmic scale . there is good agreement between data and theory over eleven orders of magnitude . the main figure shows the percentage difference between the data ( points ) and theory . the bars on the points represent the statistical uncertainty . while excellent agreement is observed below 200 gev , an excess of events over these theoretical predictions is observed at high @xmath4 . predictions with other pdfs which were available at the time are shown by the additional curves . for these predictions the percentage from the default theory ( mrsd0 ) is shown . as can be seen from the figure , the best agreement below 200 gev is with the mrsd0 pdf . other pdfs agree less well , and none show the rise at high @xmath4 observed in the data . = 3.5 in = 2.0 in the systematic uncertainties on the run 1a cross section were evaluated using the procedures in reference @xcite . in short , new parameter sets are derived for @xmath115 standard deviation shifts in the unsmearing function for each source of systematic uncertainty . the parameters for the run 1a cross section are given in reference @xcite . figures [ cdfincsys1a](a h ) show , for the run 1a data sample , the percentage change from the standard curve as a function of @xmath4 for the seven largest systematic uncertainties : ( a ) charged hadron response at high @xmath116 ; ( b ) the calorimeter response to low@xmath116 hadrons ; ( c ) @xmath117 on the jet energy for the stability of the calibration of the calorimeter ; ( d ) jet fragmentation functions used in the simulation ; ( e ) @xmath118 on the underlying event energy in a jet cone ; ( f ) detector response to electrons and photons and ( g ) modeling of the detector jet energy resolution . an eighth uncertainty , an overall normalization uncertainty of @xmath1193.8% , was derived from the uncertainty in the luminosity measurement@xcite ( @xmath1193.5% ) and the efficiency of the acceptance cuts ( @xmath1191.5% ) . the eight uncertainties arise from different sources and are not correlated with each other . the 1@xmath84 shifts are evaluated by changing only one item at a time in the monte carlo simulation , such as high @xmath120 hadron response . the resulting uncertainty in the unsmeared cross section is thus 100% correlated from bin to bin , but independent of the other seven uncertainties . = 3.5 in = 2.5 in to analyze the significance of the run 1a result , the cdf collaboration used four normalization independent , shape dependent statistical tests @xcite . the eight sources of systematic uncertainty are treated individually to include the @xmath4 dependence of each uncertainty . the effect of finite binning and systematic uncertainties are modeled by a monte carlo calculation . between 40 and 160 gev , the agreement between data and theory is @xmath12180% for all four tests . above 160 gev , however , each of the four methods yields a probability of 1% that the excess is due to a fluctuation . if the test is performed for other pdfs , agreement at low @xmath4 is reduced , as is the significance of the excess at high @xmath4 . the best agreement at high @xmath4 for the curves shown is with cteq2m@xcite which gives 8% , but the low @xmath4 agreement is reduced to 23% . the excess of events observed at high @xmath4 initiated a re evaluation of the uncertainties in the pdfs , particularly at high @xmath13 . one outcome of this re examination was cteq4hj . this pdf incorporates the same low energy dis data as in cteq4 m , but the high jet data is weighted to accentuate its contribution to the global parton distribution fit in the high @xmath19 and @xmath13 region . figure [ cdfinc1b ] shows the preliminary run 1b result @xcite compared to the run 1a results and to cteq4 m ( a more modern pdf than the ones in figure [ cdfinc1a ] ) . the run 1a and 1b data sets are in excellent agreement . only statistical uncertainties are shown . the greatly reduced statistical error bars on the run 1b data are due to the five fold increase of luminosity between the two runs . analysis of the run 1b data follows exactly the sequence of the run 1a data with some additional corrections specific to the run 1b running conditions . the run 1b systematic uncertainties are very similar to the uncertainties derived for the run 1a sample . figure [ cdfinc1bpdf ] shows the run 1b result compared to predictions with other recent pdfs . although cteq4hj is a bit contrived the good agreement with the data demonstrates the flexibility of the pdfs and the ability of the qcd calculations to describe these run 1a and run 1b inclusive jet cross sections . quantitative comparisons between the run 1b data and theoretical predictions are underway . = 3.5 in = 2.5 in = 3.5 in = 2.5 in in 1998 the d0 collaboration finalized a 92 measurement of the inclusive jet cross section @xcite . the d0 analysis differs from the cdf analysis in that the spectrum is corrected independently for energy calibration and then for distortion by finite jet energy resolution . after passing the various jet and event selection criteria each jet is corrected individually for the average response of the calorimeter . as mentioned in section 3 the jet response was determined with well measured photon jet events . however , the background free , energy corrected @xmath122 spectrum still remains distorted by jet energy resolution . the distortion is corrected by first assuming that an ansatz function @xmath123 will describe the actual spectrum , then smearing it with the measured resolution , and finally comparing the smeared result with the measured cross section . the procedure is repeated by varying parameters @xmath124 and @xmath125 until the best fit is found between the observed cross section and the smeared trial spectrum . at all @xmath122 , the resolution ( measured by balancing @xmath122 in jet events ) is well described by a gaussian distribution ; at the standard deviation is 7 gev . the ratio of the initial ansatz to the smeared ansatz is used to correct the cross section on a bin by bin basis @xcite . the resolution correction reduces the observed cross section by ( 13@xmath1193)% [ ( 8@xmath1192)% ] at [ ] . fig . [ d0xsec ] shows the final inclusive jet cross section as measured by the do collaboration in the rapidity region @xmath126 @xcite . this rapidity interval was chosen since the detector is uniformly thick ( seven or more interaction lengths with no gaps ) and both jet resolution and calibration are precise . the figure shows a theoretical prediction for the cross section from . there is good agreement over seven orders of magnitude . for the calculation shown here @xmath127 , the pdf is cteq3 m , and @xmath128=@xmath129 . [ d0unc ] shows the cross section uncertainties . each curve represents the average of nearly symmetric upper and lower uncertainties . the energy scale uncertainty which varies from 8% at low @xmath122 to 22% at 400 gev dominates all other sources of uncertainty , except at low @xmath122 , where the 6.1% luminosity uncertainty is of comparable magnitude . the total systematic error is 10% at 100 gev and 23% at 400 gev . although the individual errors are independent of one another each error is either 100% or nearly 100% correlated point to point , and the overall systematic uncertainty is highly correlated . table [ tab : d0correls ] shows that the bin to bin correlations in the full uncertainty for representative @xmath122 bins are greater than 40% and positive . the high degree of correlation will prove a powerful constraint on data theory comparisons . .d0 inclusive jet cross section total uncertainty correlations . [ cols="^,^,^,^,^,^",options="header " , ] the d0 analysis follows the inclusive analysis very closely . jet and event selection , energy correction , and resolution unsmearing are all quite similar . an additional correction for vertex resolution is also important in the very forward backward rapidity bins . the cross section uncertainties are quite similar to the inclusive cross section for the four rapidity bins limited by @xmath130 . in the high @xmath29 bins the systematic errors are approximately doubled [ d0triple1 ] and [ d0triple2 ] show the fractional difference between data and theory for all eight bins @xcite . the theoretical prediction is from with cteq3 m and @xmath131 . the bars on the data represent the statistical uncertainty and the outer symbols the total uncertainty . note there is good agreement over all rapidity for @xmath132 to @xmath133 . significantly , this pdf includes no collider jet data . although not shown , the agreement with cteq4 m is also reasonable . = 4.0 in = 2.5 in = 4.0 in = 2.5 in the differential cross sections show great promise for constraining the pdfs . in all rapidity bins the cdf data appears to prefer cteq4hj over cteq4 m or mrst , while the d0 data seems to be good agreement with the predictions using cteq3 m . this apparent disagreement mirrors the situation in the inclusive cross sections . however , a firm statement on the agreement or disagreement of the two data sets is obscured by the different techniques and the current lack of quantitative comparisons between data and theory . the lo cross section for @xmath134 @xmath135 + @xmath136 + @xmath137 events ( where @xmath135 and @xmath136 are the leading two jets ) can be completely described in terms of three orthogonal center of mass variables . these are cos @xmath138 , where @xmath138 is the center of mass scattering angle between the two leading jets , the boost of the dijet system and the dijet mass , @xmath139 as follows @xcite : @xmath140 is the hard scattering matrix element , @xmath141 ( @xmath142 ) is the fraction of the proton ( antiproton ) momentum carried by the parton and @xmath143 is the parton momentum distribution . typically the dijet mass is derived from measured variables such as , @xmath144 , and @xmath145 . in the case of higher order processes , where more than two jets are produced , the mass is calculated using the two highest jets in the event and additional jets are ignored . integration of the general dijet cross section over boost and production angle leads to the dijet mass spectrum . the spectrum is a useful test of qcd sensitive to the pdfs . on the other hand integration over mass and boost leads to the dijet angular distribution a marvelous test of the hard scattering matrix elements almost totally insensitive to the pdfs . comparisons of the angular distributions and certain mass spectra ratios to theoretical predictions can establish stringent limits on the presence of conjectured quark constituents . we turn now to a discussion of the these spectra and compositeness limits . at lo , where only two jets are produced , the dijet invariant mass is given by @xmath146 where @xmath147 is the center of mass energy of the interacting partons and @xmath148 the total center of mass energy . since the dijet mass represents the center of mass energy of the interaction it directly probes the parton distribution functions . experimentally , the dijet mass cross section is given by @xmath149 where @xmath150 and @xmath151 are the pseudorapidities of the jets . as with the inclusive jet cross section the dijet mass cross section is integrated over a range of pseudorapidity . for example , fig . [ mass_theory_uncertainties ] shows nlo qcd dijet predictions for @xmath152 relative to a reference prediction . there are 2030% variations due to the scale and to the pdfs . = 4.0 in = 2.5 in the d0 and cdf collaborations have both measured the dijet mass spectrum with run 1b data samples . the cdf measurement uses the four vector definition for the dijet mass @xmath153 where @xmath106 and @xmath14 are the energy and momentum of a jet and allows the rapidity of the jets to extend to @xmath154 . to ensure good acceptance , the cdf analysis also imposes a cut on @xmath155 , where @xmath138 is the scattering angle in the center of mass frame . figure [ cdfdjmlog ] shows the cdf preliminary data compared to predictions @xcite . the error bars represent the quadrature sum of the statistical and systematic uncertainties . the data and predictions using cteq4 m , @xmath156 and @xmath157 are in good agreement . figure [ cdfdjmlin ] shows the percent difference between the data and the theoretical prediction . the shaded band shows the preliminary estimate of the systematic uncertainties . these uncertainties are derived in a manner similar to the inclusive jet cross section measurement with additional contributions for jets outside the central region . the percent difference between the default prediction and those using other pdfs are also shown and are all consistent with the data given the systematic errors . however , as with the inclusive and dijet cross section measurements , cteq4hj seems to provide the best agreement with the data in the high @xmath4 region . = 4.0 in = 2.5 in = 4.0 in = 2.5 in d0 has performed the dijet mass measurement in two rapidity regions : central , @xmath158 , and more forward , @xmath159 , @xmath160 @xcite . the dijet mass at d0 is defined assuming massless jets : @xmath161 , where @xmath162 and @xmath163 are the rapidity and azimuthal separation of the two jets . figure [ d0dijeteta ] shows the difference between the data and theory divided by the theory . as with the d0 inclusive jet measurement there is good agreement with theory , well within systematic uncertainties . at larger rapidity the data shows a tendency to be slightly above the predictions at high @xmath4 . theory @xmath164 comparisons similar to those described in section 4 yield large probabilities . as fig . [ d0dijetpdf ] illustrates there is sensitivity to the pdf choice but unfortunately its significance is obscured by the systematic uncertainties . as with the inclusive jet cross section these uncertainties are highly correlated . in fig . [ djmcdfd0 ] , the d0 @xmath165 result is compared to the cdf result and to predictions with cteq4 m . the kinematic cuts of the two analyses have a significant overlap ; 59% of the cdf sample has the two leading jets within @xmath165 . there is remarkable agreement between these data samples over the full @xmath4 range . = 4.0 in = 4.0 in = 4.0 in = 2.5 in the dijet angular distribution measured in the center of mass is sensitive primarily to the hard scattering matrix elements . in fact , the distribution is unique among high @xmath4 measurements in that it is almost independent of the pdfs . the shape of the angular distribution is dominated by t channel exchange and is nearly identical for all dominant scattering subprocesses ( e.g. @xmath166 , @xmath167 , and @xmath168 ) . the dijet angular distribution predicted by lo qcd ( two jets only ) is proportional to the rutherford cross section @xmath169 to flatten out the distribution and facilitate the comparison to theory , the variable transformation @xmath170 is used giving @xmath171 for measurement of the dijet angular spectra this quantity is integrated over pseudorapidity and mass regions and normalized to the total number of events n in those regions . the normalization reduces both experimental and theoretical uncertainties . figure [ d0dijang1 ] shows the @xmath172 distributions from the d0 collaboration for four mass ranges @xcite . in all cases the jets are limited to regions of full acceptance . the nlo predictions from with @xmath173 provide the best agreement with the shape of the data . the large @xmath174 acceptance of the measurement allows discrimination between lo and nlo predictions . figure [ cdfdijang ] shows cdf data compared to qcd predictions ( ) for different mass bins @xcite . in this case both lo and nlo qcd are in good agreement with the distribution . = 4.0 in = 2.5 in = 4.0 in = 2.5 in the dijet mass and angular distributions are sensitive to new physics such as quark compositeness . in qcd parton parton scattering , the dominant exchange involves the t channel . this produces distributions peaked at small center of mass scattering angles ( near the beam axis in the lab ) e.g. large @xmath29 and @xmath174 . in contrast , the compositeness model @xcite predicts a more isotropic angular distribution . thus , relative to qcd predictions , the contributions of composite quarks would be most noticeable in the central region , near @xmath28 and @xmath174 = 1 . compositeness signals may be parameterized by a mass scale @xmath175 which characterizes the quark substructure coupling . limits on @xmath175 are set assuming that @xmath176 , such that the dominant force is still qcd . the substructure coupling is approximated by a four fermi contact interaction giving rise to an effective lagrangian @xcite . the lagrangian contains eight terms describing the coupling of left and right handed quarks and antiquarks . currently only the term describing the left handed coupling of quarks and anti quarks has been calculated , and this term has an unknown phase . limits are reported for the case where specific quarks or all quarks are composite with either constructive interference ( @xmath177 ) or destructive interference ( @xmath178 ) . to set compositeness limits cdf and d0 both use the nlo prediction times a `` k factor '' from lo qcd + compositeness @xcite . nlo calculations with compositeness are not available . figure [ cdfdijang ] includes a curve with a compositeness signal added . the additional contribution at low @xmath174 is most pronounced in the highest dijet mass bins . both cdf and d0 set compositeness limits based on the ratio of number of events at low @xmath174 to those in the high @xmath174 region . figure [ cdfang ] shows the cdf result for the ratio of the number of events below @xmath174=2.5 and between @xmath174= 2.5 and 5.0 as a function of the dijet mass , along with curves which correspond to the different values of @xmath175 . using this ratio , the cdf data excludes at the 95% c.l . a contact interaction scale of @xmath179 tev and @xmath180 tev . for a model where all quarks are composite @xmath181 tev and @xmath182 tev . figure [ d0dijang2 ] shows the angular distribution as measured by d0 for very high dijet masses ( greater than 635 gev / c@xmath183 ) compared to predictions which include composite quarks . compositeness limits in this analysis are derived from the ratio of the number of events above and below @xmath174=4 . the data excludes ( at the 95% cl ) contact interaction scales with @xmath184 : @xmath185 tev , @xmath186 tev , @xmath187 tev , and with @xmath188 : @xmath189 tev , @xmath190 tev , @xmath191 tev . = 3.5 in = 2.5 in = 3.5 in = 2.5 in the best limits on compositeness now come from recent dijet mass measurements from the d0 collaboration @xcite which combine the sensitivity of the dijet mass distributions with the pdf independence of the angular distributions . in much the same way jet production from a composite interaction increases the @xmath174 spectrum at low values of @xmath174 , the dijet mass spectrum will increase at central rapidities relative to forward rapidities . thus the ratio of the central and forward dijet mass spectra will be sensitive to @xmath175 . in addition , both theoretical and experimental uncertainty are reduced in the ratio . [ d0masscomp ] shows the ratio of cross sections for @xmath192 and @xmath193 as a function of dijet mass . as indicated by the family of curves , the compositeness model predicts changes in shape to this ratio at high mass . the spectrum rules out quark compositeness at the 95% confidence level for @xmath194 below @xmath195 tev and @xmath196 below @xmath197 tev . = 4.0 in = 2.5 in the inclusive jet and dijet cross sections provide fundamental tests of qcd predictions at the highest jet @xmath4 and thus are the deepest probes into the structure of the proton . with the increased luminosity performance of the tevatron , measurements of these cross sections are no longer limited by statistical uncertainties . in fact , the systematic uncertainties from the experimental measurements and from the theoretical predictions are comparable in size and are significantly larger than the statistical uncertainty on all but the highest @xmath4 data points . uncertainties in the pdfs dominate the theoretical uncertainty , while uncertainties in the jet energy scale dominate the experimental measurements . the inclusive jet cross sections from the tevatron have proved a particularly interesting test of qcd . the run 1a inclusive jet measurement showed disagreement with concurrent pqcd predictions at the highest @xmath13 and @xmath19 . two subsequent measurements , each using five times the data , show mixed agreement between the data and theory at high @xmath4 . one measurement is consistent with the run 1a measurement , and can be described by qcd if the pdfs are suitably modified . the second run 1b measurement is well described by qcd with pdfs that either do or do not include high jet data . further , within statistical and systematic errors , the three measurements are compatible . the preliminary measurements of the inclusive jet cross section at @xmath198=630 gev and the ratio of the cross sections show good agreement with previous results and marginal agreement with qcd predictions . derivation of quantitative results is in progress . unfortunately this sample is too statistically limited to provide a constraint on the high @xmath4 behavior of the cross section at 1800 gev . the apparent excess of events at high @xmath4 observed in the inclusive jet cross section from run 1a created intense scrutiny of the theoretical predictions . as a result , there is now a better understanding of the uncertainty in the theoretical predictions , particularly for the parton distribution functions . the dijet mass distributions and differential cross sections reflect a story similar to the inclusive jet measurements . the results are generally consistent with each other and with qcd , but with hints that moderately forward regions indicate some high @xmath13 modifications of the pdfs might be appropriate . incorporation of this information into the global fitting procedures used to derive the pdfs should be particularly helpful in reducing the uncertainty associated with extrapolation from low energy dis data to the kinematic region covered by the tevatron jet measurements . until the uncertainty in the pdfs is reduced , or a method of analytically evaluating their uncertainties is available , the best limits on compositeness will come from measurements insensitive to the pdfs , and in particular , measurements that rely on angular information . these angular distributions show the partonic hard scattering to be well described by nlo theoretical calculations . comparisons of the angular and mass data to jet production models augmented by quark compositeness show no preference for deviations from standard qcd predictions . in fact , the analyses now show that the compositeness scale must be greater than 2.5 tev if it exists at all . the high precision tevatron jet data has fostered a period of great progress in qcd . due to the flexibility of theoretical predictions , pqcd can describe nearly all inclusive jet and dijet observations . limits on quark substructure from run 1 have nearly doubled from the previous measurements . however , more stringent tests will require improved pdfs and reduced theoretical uncertainties . looking forward , expectations are high that the experimental measurements of jets and their properties will continue to improve in run 2 with a 20fold increase in data sample size , increased beam energy , and reduced systematic uncertainties from the upgraded detectors . we wish to thank the members of the cdf and d0 collaborations for permitting us to include their results ; in particular we would like to thank iain bertrum , anwar bhatti , frank chlebana , bjoern hinrichsen , bob hirosky and john krane for helpful discussions during preparation of this review and for providing excellent illustrations . see for example j. huston ( for the cdf collaboration ) fermilab conf99006e , to be published in the proceedings of 29th international conference on high energy physics ( ichep 98 ) , vancouver , canada , 2329 jul 1998 . b. abbott _ et al . _ , ( d0 collaboration ) , accepted by phys . rev . lett . preprint hep ph/9807018 . r. k. ellis , w. j. stirling , and b. r. webber , qcd and collider physics ; cambridge univ . press ( 1996 ) ; p238 . ellis , z. kunszt , d. soper , phys . 69 * 3615 ( 1992 ) and s. ellis cern th686193 , jun 1993 . 13pp . presented at 28th rencontres de moriond : qcd and high energy hadronic interactions , les arcs , france , 2027 mar 1993 . published in moriond 1993 : hadronic:235244 ; hep ph/9306280 . g. blazey in the proceedings of the 29th international conference on high energy physics ( ichep 98 ) , vancouver , canada , 2329 july 1998 . j. krane , unpublished theses university of nebraska , lincoln ( 1998 ) .

we present a series of numerical simulations that explore how the ` x - factor ' , the conversion factor between the observed integrated co emission and the column density of molecular hydrogen varies with the environmental conditions in which a molecular cloud is placed . our investigation is centred around two environmental conditions in particular : the cosmic ray ionisation rate ( crir ) and the strength of the interstellar radiation field ( isrf ) . since both these properties of the interstellar medium have their origins in massive stars , we make the assumption in this paper that both the strength of the isrf and the crir scale linearly with the local star formation rate ( sfr ) . the cloud modelling in this study first involves running numerical simulations that capture the cloud dynamics , as well as the time - dependent chemistry , and ism heating and cooling . these simulations are then post - processed with a line radiative transfer code to create syntheticco ( 1 - 0 ) emission maps from which can be calculated . we find that for virialised clouds with mean density 100 , is only weakly dependent on the local sfr , varying by a factor of a few over two orders of magnitude in sfr . in contrast , we find that for similar clouds but with masses of , the x - factor will vary by an order of magnitude over the same range in sfr , implying that extra - galactic star formation laws should be viewed with caution . however , for denser ( ) , super - virial clouds such as those found at the centre of the milky way , the x - factor is once again independent of the local sfr . galaxies : ism ism : clouds ism : molecules molecular processes

1506.06503

in both the milky way , and in other local galaxies , star formation is known to take place within large clouds of molecular gas , the so - called giant molecular clouds ( gmcs ) . understanding the properties of these gmcs is important for the light it sheds on the process of star formation . unfortunately , the main chemical constituent of the gas in gmcs , molecular hydrogen ( h@xmath6 ) , is extremely difficult to observe _ in situ _ , owing to the fact that the characteristic temperature of the gas in typical gmcs ( 1020 k ; @xcite ) is much smaller than the temperature required to excite even the lowest rotational transition of the h@xmath6 molecule . therefore , most observational studies of gmcs rely on carbon monoxide ( co ) as a proxy for molecular hydrogen . observations of nearby gmcs find a surprisingly tight correlation between the integrated intensity coming from the clouds in the @xmath7 rotational transition line of @xmath1co and the molecular hydrogen column density of the clouds ( see e.g. @xcite ; or the recent review by @xcite ) , and yield a conversion factor between co intensityco . ] and h@xmath6 column density given approximately by @xcite @xmath8 where @xmath9 is the velocity - integrated intensity of the co @xmath10 emission line , averaged over the projected area of the gmc , and @xmath11 is the mean h@xmath6 column density of the gmc , averaged over the same area . this latter quantity can not be directly determined from observations of h@xmath6 emission , but can be inferred from measurements of the total column density if the hi column density has also been mapped . the total column density of gas in the cloud can itself be determined for nearby clouds using extinction mapping ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) or measurements of the diffuse @xmath12-ray flux produced by interactions between high energy cosmic rays and atomic hydrogen , atomic helium and h@xmath6 ( e.g. * ? ? ? the fact that the values for @xmath0 derived for nearby clouds show very little variation has led to its widespread use as a general co - to - h@xmath6 conversion factor ( commonly referred to as the ` x - factor ' ) , even in environments very different from the galactic interstellar medium ( ism ) . it is therefore important to understand when the adoption of a constant value for @xmath0 is appropriate , and when it may be seriously misleading . in clouds with a fixed h@xmath6 column density , any variation in @xmath0 that occurs from cloud to cloud must be due solely to a variation in @xmath9 . numerical simulations show that this is sensitive to three main parameters : the velocity dispersion of the gas in the cloud , the temperature of the co - emitting gas , and the filling factor of bright co emission @xcite . increasing the gas temperature increases the typical brightness temperature of the emitting gas , thereby increasing @xmath9 and decreasing @xmath0 . however , as a change in the temperature also affects the excitation of the co molecules , the scaling of @xmath9 with temperature is not linear , but instead is closer to @xmath13 @xcite . increasing the velocity dispersion also increases @xmath9 by increasing the co linewidth , but the increased velocity dispersion can also lead to lower co line opacities and hence lower brightness temperatures , so again the dependence is sub - linear @xcite . finally , @xmath9 is sensitive to the fraction of the cloud traced by bright co emission , as in cases where this is small , the _ mean _ integrated intensity is much smaller than the _ peak _ integrated intensity , owing to the effects of beam dilution . in solar metallicity clouds situated in a standard galactic radiation field , the filling factor of the co emission is of order unity ( see e.g. * ? ? ? * or section 3 below ) , and so small variations in its value from cloud to cloud have little effect on @xmath0 . at lower metallicities , however , the co filling factor can become very small @xcite and hence @xmath0 can increase dramatically . if we also allow the h@xmath6 column density to vary , then this adds an additional parameter to the problem . however , in real clouds , it is unlikely that all of these parameters vary independently . in particular , we expect the filling factor of co - bright gas to be sensitive to the mean extinction of the cloud @xcite , which at fixed metallicity is directly related to @xmath11 . nevertheless , in order to understand why @xmath0 is close to constant in local gmcs , we need to understand why these four parameters do not all vary by large amounts . if gmcs are in virial equilibrium with linewidths that satisfy the standard galactic size - linewidth relationship @xcite @xmath14 where @xmath15 is the linewidth and @xmath16 is a measure of the characteristic size of the cloud , then we would expect that @xmath11 would be approximately constant from cloud to cloud @xcite and that the co linewidth would scale only weakly with the cloud mass as @xmath17 @xcite , thereby explaining much of the constancy in @xmath0 . however , it has become clear over the last few years that this is not the only possible explanation and that virial equilibrium is not required in order to explain a constant @xmath0 . the reason is that even if gmcs are not virialized , their mean h@xmath6 column densities are unlikely to vary by a large amount . clouds with low values of @xmath11 will provide insufficient dust shielding to allow widespread co formation and hence will not be identified as `` molecular '' clouds @xcite , while clouds with high @xmath11 will form stars rapidly , and the resulting stellar feedback will prevent their column densities from growing too large @xcite . the small variation that we find in the mean co brightness temperature of local gmcs can also be easily understood . since @xmath1co emission from gmcs is usually optically thick , the observed brightness temperature is determined by the density and temperature of the gas at the point in the cloud where @xmath18 @xcite . these values vary much less from cloud to cloud than do the mean values for the cloud as a whole @xcite , and so there is little variation in the resulting mean brightness temperatures @xcite . putting all of this together , we see that we can understand fairly well why @xmath0 is approximately constant for local gmcs . however , the behaviour of @xmath0 as we move to environments with much stronger uv radiation fields and higher cosmic ray fluxes is less easy to predict . we would expect that as we increase the strength of the ambient radiation field , more photodissociation of co will occur . this will decrease the filling factor of the co emission and may also decrease the co linewidth if the co - emitting gas is no longer well - distributed throughout the volume of the cloud . these effects will tend to decrease @xmath9 and hence increase @xmath0 . at the same time , we also expect a stronger uv field to lead to greater dissociation of h@xmath6 , and hence a decrease in @xmath11 . in addition , we would expect the cloud to be systematically warmer , owing to the higher photoelectric heating rate in low @xmath19 regions , and the greater cosmic ray heating rate in high @xmath19 regions . this means that the mean brightness temperature of the co emission may also be larger . together , these two effects will tend to decrease @xmath0 . it is not immediately clear which set of effects will dominate , and hence whether one would expect @xmath0 to increase or decrease as the strength of the uv radiation field and the cosmic ray flux increase . in this paper , we investigate this issue with the help of numerical simulations . we simulate the coupled chemical , thermal and dynamical evolution of several model clouds in a variety of different environments , spanning a range of 100 in terms of uv field strength and cosmic ray ionisation rate . we model and analyze the resulting emission in the 1 - 0 line of @xmath1co , and thereby are able to explore how @xmath0 varies as a function of the external environment in clouds with a fixed mass and mean volume density . in section 2 , we briefly describe our numerical approach , and in section 3 , we explore how the changing environment affects the physical structure of the clouds . in section 4 , we examine how @xmath0 varies from cloud to cloud , and how this depends on the definition used for the cloud boundary . we discuss some implications of our results in section 5 and conclude in section 6 . the computations presented in this paper were performed using a modified version of the publicly available smoothed particle hydrodynamics ( sph ) code gadget-2 @xcite . a major modification to the original gadget-2 is the inclusion of time - dependent chemistry . the version we use here follows the formation and destruction of h@xmath20 as introduced by @xcite in addition to the simplified treatment of the co chemistry that was proposed by @xcite . in @xcite , it was shown that this co network has a similar accuracy to the more exhaustive treatment of @xcite , but incurs around one third of the computational cost . we do not include freeze - out of co onto dust grains in this model , but as this generally occurs only in regions where the @xmath1co emission is already optically thick @xcite , we do not expect this omission to significantly affect our results . as well as following the chemical evolution of the gas , we also model its thermal evolution . we account for dynamical heating due to shocks and adiabatic compression and cooling due to adiabatic rarefaction in the same fashion as in the unmodified version of gadget-2 . in addition , we also account for the main radiative and chemical heating and cooling processes occurring in the ism . these include fine structure cooling from c@xmath21 , c and o , molecular line cooling from h@xmath6 and co , photoelectric heating and cosmic ray heating . at high gas densities , collisions between gas particles and dust grains also play an important role in regulating the thermal energy balance , cooling the gas if @xmath22 and heating it if @xmath23 . full details of how we treat these processes and a number of other , less important , contributors to the overall thermal energy balance can be found in our previous papers @xcite , and a summary of the most important processes included can be found in figure 4 in @xcite . the attenuation of the interstellar radiation field ( isrf ) is treated using the treecol algorithm , introduced by @xcite . in this paper , the spectral shape of the isrf is based on the prescription of @xcite in the ultraviolet and @xcite at longer wavelengths . the strength of the isrf is varied in the different simulations as described in section [ sec : sfr - proxy ] below . the clouds are assumed to be bathed in a uniform isrf , and treecol is used to compute the attenuated spectrum that reaches each sph particle in the computational volume . since the above - mentioned papers , we have made two significant changes to the chemical model . the first was to update our treatment of the photodissociation of co from the prescription given in @xcite , to that described in the recent paper of @xcite . the second change that we have made to the chemical model is the inclusion of the effects of cosmic - ray ionisation of atomic carbon @xmath24 and cosmic - ray induced photodissociation of c and co @xmath25 these processes were not included in the chemical model described in @xcite . we assume that the rates of all three processes are proportional to @xmath26 , the cosmic ray ionisation rate of atomic hydrogen . for processes [ react1 ] and [ react2 ] , we use the rates given in @xcite as a basis , but rescale them by a factor @xmath27 , where @xmath28 is the value of the cosmic ray ionisation rate adopted by @xcite . similarly , to compute the rate of process [ react3 ] , we use the value given in @xcite as a basis , but rescale it to make it consistent with our choice of @xmath26 . the @xmath1co ( 1 - 0 ) emission maps that form the basis of the analysis in this paper were created using the radmc-3d radiative transfer codedullemond / software / radmc-3d/ ] . the level populations in the non - lte limit are computed using the the large velocity gradient approximation @xcite , as implemented in radmc-3d by @xcite . we assume that collisions with h@xmath6 dominate the excitation of the co rotational levels and adopt the values for the collisional excitation rate coefficients given in the leiden atomic and molecular database @xcite . all of the clouds in this study start as uniform spheres of gas , onto which a three - dimensional turbulent velocity field has been superimposed . two cloud masses are examined in this paper : a `` low - mass '' cloud , with a mass of @xmath29 , and a `` high - mass '' cloud of mass @xmath30 . in most of the runs , we take the initial hydrogen nuclei number density of the gas to be @xmath31 , yielding initial radii for the low- and high - mass clouds of 8.8 pc and 19 pc , respectively . in addition , we performed two simulations of high - mass clouds with a much higher initial density , @xmath32 ; these denser clouds had an initial radius of around 4 pc . the mass resolution in this study is kept fixed , with the mass of an sph particle being set at 0.005@xmath33 . the low - mass clouds therefore have @xmath34 sph particles , while the high - mass clouds have @xmath35 sph particles . the minimum resolvable self - gravitating mass - element in this calculation is therefore 0.5@xmath33 in both cases @xcite . in practice , this means that we can follow the collapse of the cloud until the gas number density reaches a value of around @xmath36 . at this point , we halt the collapse and perform the analysis that is presented in this paper . we note that we would not expect the results presented here to change if we were to follow the collapse to higher densities , as at these densities , the co ( 1 - 0 ) line will be highly optically thick ( see e.g. * ? ? ? * ) , and in any case much of the co will be frozen out onto dust grains . the velocity fields are generated with a ` natural ' mix of solenoidal and compressive modes ( i.e. a ratio of 2:1 ) . this is generated on a 128@xmath37 grid . this velocity field is left to freely decay in shocks , rather than being continuously driven during the course of the simulation . in most of the simulations , the initial energy in the velocity field , @xmath38 , was set to be half the initial gravitational potential energy of the cloud , @xmath39 , so that the clouds are initially in virial equilibrium . this means that the initial rms turbulent velocity @xmath40 is 2.4 kms@xmath41 for the low - mass clouds and 5.2 kms@xmath41 for the high - mass clouds . in terms of the virial parameter @xmath42 these clouds have @xmath43 . in addition , we also examine one high density cloud for which we set @xmath44 , so that the cloud is initially unbound ( see section [ sect : sfr ] ) . note that most of the simulations presented here adopt the same random seed for the turbulent velocity field . however we also run one of the simulations with a different seed , to gauge the sensitivity of our results to the underlying cloud structure that is created by the turbulent motions . from previous modelling , it has been found that in cold gas with densities of around @xmath45 , most of the hydrogen is in the form of h@xmath20 . in contrast , the carbon in this gas is still predominantly in the form of c@xmath46 ( see e.g. * ? ? ? * ; * ? ? ? * ) . the initial chemical state of the gas in the majority of the runs presented in this study is motivated by this previous work . we start most of our runs with all of the hydrogen already in the form of h@xmath6 , but assume that the carbon and oxygen are present in the form of c@xmath21 and o , respectively . we adopt total abundances ( relative to the number density of hydrogen nuclei ) for the carbon and oxygen nuclei of @xmath47 and @xmath48 respectively , consistent with the values measured in the local ism @xcite . similarly , we adopt a dust - to - gas mass ratio of 0.01 , consistent with the value in the local ism . in our runs with very high uv fields , it is unclear whether starting with all of the hydrogen in molecular form is a good approximation , as in this case , one would expect the equilibrium h@xmath6 fraction in low density gas to be much smaller than in the models of @xcite and @xcite . to address this uncertainty , we ran two additional high - mass models with the hydrogen initially in atomic form . the post - processing of the simulation data in radmc-3d first requires that the sph particle data is interpolated onto a regular cartesian grid . this is done using the standard sph smoothing formalism . to ensure that we catch small , high density pockets of gas , we employ a radmc-3d cell - size of 0.068pc , such that the high - mass cloud calculations have @xmath49 cells and the low - mass cloud calculations have @xmath50 cells . we find that such a resolution is sufficient to get converged values for the probability density function of co ( 1 - 0 ) emission , from which it follows that the various mean values that we examine later are also converged . once the position - position - velocity maps are obtained , we then integrate the emission along the z - axis to create maps of the integrated intensity , @xmath51 . in this study , we vary two of the environmental conditions that can affect the chemical balance of clouds : the strength of the isrf and the cosmic - ray ionisation rate . both of these are thought to vary with the local star formation rate , and so we assume here that the strength of these processes can be used as a proxy for the star formation rate . as mentioned above , the isrf used here is taken to have a shape described by a combination of the @xcite and @xcite radiation fields . in one set of runs those representing clouds in an environment with a star formation rate similar to that in the local ism , which we denote as sfr@xmath52 we adopt the same normalization for the isrf as in the papers of @xcite and @xcite . our fiducial isrf therefore has a strength @xmath53 in @xcite units . for this study , we are mainly interested in how the isrf heats the gas and affects its chemical state . photons with energies above 6 ev are responsible for the photoelectric heating ( the dominant heat source in low extinction regions of these clouds ) , and photons with energies above 11.2 ev and 11.5 ev are responsible for dissociating h@xmath20 and co respectively . since most of the photons in this part of the isrf come from massive , young stars , it is reasonable to assume that , to a first approximation , the strength of the relevant portion of the isrf scales linearly with the local star formation rate . at longer wavelengths , the isrf is dominated by older stellar populations and this assumption is less well - founded . however , the strength of the isrf at these wavelengths has little effect on the temperature or chemistry of the gas , and so for simplicity , we assume that in regions with higher star formation rates , we can simply scale the entire isrf upwards , rather than changing its spectral shape . we also assume that the cosmic ray ionisation rate scales linearly with the star formation rate . this assumption is reasonable : supernova remnants are the main source of cosmic rays in the galactic ism , and the lifetime of a typical cosmic ray within the galaxy is around 15 myr @xcite , so the cosmic ray energy density , and hence the cosmic ray ionisation rate , should track the star formation rate fairly closely . in this study , in addition to the runs representing the behaviour of clouds in the local ism , with @xmath54 , we also perform simulations where we increase the strength of the isrf and the cosmic ray ionisation rate by factors of 10 or 100 , corresponding to star formation rates of @xmath55 or @xmath56 , respectively . [ cols="^,^,^,^,^,^",options="header " , ] in the left - hand columns of figures [ fig : image_m4 ] and [ fig : image_m5 ] , we show column density images of the gas , taken at a point in the simulations just before the onset of star formation in our fiducial clouds i.e. those seeded with our standard turbulent velocity field , and which started life with all their hydrogen in the form of h@xmath20 . this particular stage in the cloud s evolution is when we perform the analysis that forms the basis of this paper . for most of our simulated clouds , this corresponds to a time @xmath572 myr after the beginning of the simulation , as shown in table [ tab : sims ] . in both the @xmath29 and @xmath30 clouds we see a similar change in behaviour of the strength of the isrf and cosmic ray ionisation rate are increased to mimic progressively higher rates of ambient star formation activity : the filamentary structure imposed by the cloud s turbulence and self - gravity is less pronounced as we move to higher sfrs . this is simply a result of gas temperatures rising due to the combined effects of the enhanced photoelectric emission and cosmic - ray ionisation rates . this systematic increase in the temperature as we increase the sfr is also illustrated in figure [ fig : phase ] , where we show temperature - density phase diagrams for four of our runs , colour - coded by the co abundance . in addition , we see from figures [ fig : image_m4 ] and [ fig : image_m5 ] that the overall morphology of the clouds changes as the sfr increases . in this respect , there are two trends . the first is that the dense , filamentary structure becomes confined to a progressively smaller region as the sfr increases . this makes sense , as the gas is more easily structured near the centre of the cloud where the extinction is higher and the heating effects of the photoelectric emission are less . as the sfr increases , the heating by photoelectric emission becomes progressively more important in shaping the gas , and so the regions structured by the turbulence and gravity are those towards the centre . the second effect is that the structure of the cloud envelope changes . the sph particles have open ( i.e. vacuum ) boundary conditions , and so the particles are free to expand into the surrounding space . the stronger the isrf , the more the surface of the cloud is heated , causing it to `` boil off '' into the void , creating the halo that we can see around the clouds . in the case of the 100 @xmath58 clouds ( i.e. those illuminated by radiation fields and cosmic ray fluxes that are 100 times greater than the values in the local ism ; see section [ sec : sfr - proxy ] ) , the isrf is so strong that this halo becomes hot enough to push back on the cloud , creating the sharp boundary that we see in figures [ fig : image_m4 ] and [ fig : image_m5 ] . although dramatic , this feature of the increasing sfr is actually of little interest to this current study , as both the h@xmath20 and co lie within the cloud boundary . the middle panels in figures [ fig : image_m4 ] and [ fig : image_m5 ] show the h@xmath20 column densities images at the same point in the cloud evolution . in general , we see that the h@xmath20 column density tracks the total column density extremely well for the bulk of the cloud . given that all the clouds in these two figures were initialised with all their hydrogen in molecular form , this is not surprising . the main difference between the left and middle panels in figures [ fig : image_m4 ] and [ fig : image_m5 ] is that the h@xmath20 in the outer envelope has been photodissociated by the isrf , and so does not trace the very low density ( both column and volume ) that surrounds the main cloud structure . we also see that the column density at which h@xmath20 starts to appear increases with increasing sfr , moving from around @xmath59 to around @xmath60 as the sfr increases from the solar neighbourhood value to a factor of 100 greater . this behaviour is consistent with the predictions of detailed models for the structure of photodissociation regions ( see e.g. * ? ? ? * ; * ? ? ? the fact that a large change in the strength of the isrf leads to a relatively small change in the column density of gas required to effectively shield h@xmath6 is a consequence of the fact that dust shielding is an exponential function of the column density , which implies that in the dust - dominated region , the critical column density scales only logarithmically with the sfr . the integrated emission in the 1 - 0 line of @xmath1co , @xmath9 , is shown in the right - hand column in figures [ fig : image_m4 ] and [ fig : image_m5 ] , and shows the most striking variation with increasing sfr . for the sfr@xmath61 clouds , we see that the surface filling factor of the bright co emission is very similar to that of the h@xmath6 gas . despite this , it is not a good tracer of the physical structure of the cloud : some features seen in the emission map are far more blobby than their h@xmath6 counterparts and others are missing entirely . this behaviour is expected ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and is simply a result of the large optical depth of the 1 - 0 line the features that we see are the @xmath62 surfaces , which are a function of co column density , velocity , and temperature , and so often bear little resemblance to the structure we see in the h@xmath20 column density images . as we move to higher sfrs , we see that the surface filling factor of the co emission progressively decreases , a feature we see in all of the cloud models run in this study . in the case of the @xmath63 clouds , the effect is extreme , with emission falling below @xmath64 for much of the cloud s interior the sensitivity limit in many large - scale surveys and even below @xmath65 along lines of sight with h@xmath20 column densities as high as a few @xmath60 . however , in the cases with the higher sfr , we see that the _ peak _ co emission becomes brighter towards regions of higher column density . this is a consequence of increased heating in the cloud , which results in a higher overall excitation temperature of the co. the broader line - width in the gas also helps to the lower the optical depth , allowing more of the emission from the line to escape @xcite . due to the higher spatial resolution in the reproduced image ( the true resolution is the same in all images ) , this effect is perhaps most easily seen in the @xmath29 clouds in figure [ fig : image_m4 ] . here we see the brightening of the clump towards the left of the @xmath66 image , which goes from having a peak brightness of around @xmath67 for sfr@xmath61 , to nearly @xmath68 in the case of 100sfr@xmath52 conversion factor , @xmath0 , in each of our clouds , as a function of our sfr proxy ( see section [ sec : sfr - proxy ] for details ) . red and blue circles denote the results from the @xmath69 clouds with fully molecular and fully atomic initial conditions respectively . the orange circles correspond to a cloud with @xmath69 run using a different random seed for the turbulent velocity field . the dark and light green circles denote the @xmath70 clouds with , respectively , @xmath43 and 2 . in the top two panels , we compute @xmath0 using integrated intensities and h@xmath6 column densities only for those lines - of - sight with an integrated intensity above the specified threshold , @xmath71 . in the bottom panel , we show the results that we obtain if we do not apply an integrated intensity threshold . in each case , we compute @xmath0 by dividing the mean h@xmath6 column density in the considered area by the mean value of @xmath9 from the same area . the horizontal dashed line shows the canonical local value of @xmath0 , and the shaded area indicates the typical scatter in the value of @xmath0 derived observationally for the milky way and other nearby spiral galaxies ( see * ? ? ? * ) .,width=316 ] conversion factor , @xmath0 , in each of our clouds , as a function of our sfr proxy ( see section [ sec : sfr - proxy ] for details ) . red and blue circles denote the results from the @xmath69 clouds with fully molecular and fully atomic initial conditions respectively . the orange circles correspond to a cloud with @xmath69 run using a different random seed for the turbulent velocity field . the dark and light green circles denote the @xmath70 clouds with , respectively , @xmath43 and 2 . in the top two panels , we compute @xmath0 using integrated intensities and h@xmath6 column densities only for those lines - of - sight with an integrated intensity above the specified threshold , @xmath71 . in the bottom panel , we show the results that we obtain if we do not apply an integrated intensity threshold . in each case , we compute @xmath0 by dividing the mean h@xmath6 column density in the considered area by the mean value of @xmath9 from the same area . the horizontal dashed line shows the canonical local value of @xmath0 , and the shaded area indicates the typical scatter in the value of @xmath0 derived observationally for the milky way and other nearby spiral galaxies ( see * ? ? ? * ) .,width=316 ] conversion factor , @xmath0 , in each of our clouds , as a function of our sfr proxy ( see section [ sec : sfr - proxy ] for details ) . red and blue circles denote the results from the @xmath69 clouds with fully molecular and fully atomic initial conditions respectively . the orange circles correspond to a cloud with @xmath69 run using a different random seed for the turbulent velocity field . the dark and light green circles denote the @xmath70 clouds with , respectively , @xmath43 and 2 . in the top two panels , we compute @xmath0 using integrated intensities and h@xmath6 column densities only for those lines - of - sight with an integrated intensity above the specified threshold , @xmath71 . in the bottom panel , we show the results that we obtain if we do not apply an integrated intensity threshold . in each case , we compute @xmath0 by dividing the mean h@xmath6 column density in the considered area by the mean value of @xmath9 from the same area . the horizontal dashed line shows the canonical local value of @xmath0 , and the shaded area indicates the typical scatter in the value of @xmath0 derived observationally for the milky way and other nearby spiral galaxies ( see * ? ? ? * ) .,width=316 ] in the previous section , we saw that as the sfr in the ambient cloud environment is increased , the co emission traces progressively less of the h@xmath20 distribution . however we have also seen that the peak emission in the cloud is higher in the case of high sfr . @xcite have suggested that these two effects may effectively cancel out , resulting in an x - factor that does not depend on the cloud s environment . in this scenario , the greater emission that comes from the dense bright peaks in the high sfr case makes up for the fact that much of the cloud is not emitting in co at all , since the molecule is easily photodissociated . using our self - consistent models for n@xmath72 and @xmath9 , we can now explore this idea . in what follows , we compare how @xmath0 varies as we alter the environmental conditions in our suite of turbulent clouds . we calculate @xmath0 for each cloud via @xmath73 where the averages are taken over a specified set of the pixels in the images from figures [ fig : image_m4 ] and [ fig : image_m5 ] . in our evaluation of equation [ eq : xcopix ] , we take two approaches . first , we examine the values of x@xmath74 obtained when we average over all of the pixels in the images . in our second approach , we instead restrict the calculation to include only those pixels with co 1 - 0 integrated intensities above some threshold value , @xmath71 . these two approaches mimic different types of observations . the first approach is more appropriate when considering unresolved molecular clouds that could inhabit the beam in an extragalactic co observation . in this case , the co emission may not trace all of the h@xmath20 gas in the beam , i.e. there is a ` dark ' molecular component surrounding the co - bright portion of the cloud , as recently highlighted by @xcite . the second approach , in which we only examine pixels with emission above @xmath71 , is more applicable to studies of nearby , resolved molecular clouds , where it is common to define the extent of a cloud in terms of its co emission , a definition which inevitably leads to one ignoring those regions without detectable co emission ( see e.g. * ? ? ? * ; * ? ? ? the mean values of @xmath0 that we derive from the simulations , and how they scale with the sfr , are shown in figure [ fig : xco_sfr ] . the top two panels show the results that we obtain for two different co detection thresholds , @xmath75 and @xmath76 , while the bottom panel shows the results in the case where we do not apply a detection threshold . we also indicate the canonical local value of the x - factor , @xmath77 ( the horizontal dashed line ) and illustrate the typical scatter in the value of @xmath0 derived for the disks of nearby spiral galaxies ( the shaded region ) , as summarized in @xcite . we begin our study of the effects of the local environment on @xmath0 by looking at the behaviour of clouds located in an environment with a star formation rate similar to the local ism , i.e. with @xmath54 ( see section [ sec : sfr - proxy ] ) . the first point to note is that when we consider clouds in this environment that have properties similar to those of local molecular clouds ( i.e. virial parameters close to unity and mean initial densities around 100 @xmath3 ) , we find that @xmath0 does not appear to be sensitive to mass . both our low - mass and our high - mass clouds yield a mean x - factor of between 4 - 5 @xmath78 , consistent with the large body of literature suggesting that the x - factor is not strongly dependent on the cloud mass ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? also , we find that the x - factor is relatively insensitive to the value of @xmath71 , indicating that the co emission is a good proxy for the total h@xmath20 present in such clouds , even though the emission is clearly not tracing all of the molecular gas . while this is encouraging , it should also be noted that the values we recover for @xmath0 in this case are roughly twice as large as the canonical galactic value , @xmath77 , and on the edge of the range of values found for the disks of nearby spiral galaxies @xcite . the reason for this discrepancy is not completely clear , but there are several factors that may contribute to our recovering a systematically higher @xmath0 . the first of these stems from the way in which we treat the effects of self - shielding in our simulations . when calculating each particle s sky maps of h@xmath20 and co column densities , the treecol algorithm includes contributions from all the molecules present in the cloud , regardless of the velocity at which they are moving with respect to the target particle . in reality , much of the cloud will not contribute to the self - shielding of a given location , since the cloud s supersonic motions will doppler shift the molecules out of the line - profile . neglecting this effect means that we overestimate the effectiveness of h@xmath6 and co self - shielding , and hence overproduce the abundances of both molecules . we expect the error to be larger for h@xmath6 than for co , since self - shielding plays a much greater role in determining the h@xmath6 abundance than the co abundance , and therefore it will lead to a systematic increase in our values of @xmath0 compared to those for real clouds . the importance of this effect will depend on the size of the velocity dispersion in the cloud , but should not depend strongly on our choice of sfr , and it therefore represents a systematic error that biases all of our derived values of @xmath0 to slightly higher values . another factor that may contribute to the discrepancy between our results and @xmath77 is our choice to start the simulations with the hydrogen in fully molecular form . although this approximates the behaviour that we find in previous studies of cloud formation @xcite , in practice the hydrogen in the assembling clouds is never _ completely _ molecular there is always some atomic component associated with the cloud . in equilibrium , we expect the surface density of atomic hydrogen associated with the clouds to be of order @xmath79 @xcite , but the time taken to reach this equilibrium can be longer than the lifetime of the cloud . in our simulations with fully molecular initial conditions , we will therefore overestimate the h@xmath6 content of the clouds by around 10 - 20% , and hence will overestimate @xmath0 by a similar factor . in order to verify that the effect on @xmath0 is not greater than this , we ran two simulations of our high - mass cloud in which we started with all of the hydrogen in atomic form . in one of these simulations , we adopted the solar neighbourhood value for sfr ( i.e. @xmath54 ) , while in the other we set @xmath80 . the values of @xmath0 that we derived from these two runs are indicated in figure [ fig : xco_sfr ] by the blue circles . we see that , as expected , the use of atomic rather than molecular initial conditions results in a decrease in @xmath0 , but that the effect is not large and seems to be independent of our choice of sfr . finally , our decision to focus on clouds with a virial parameter @xmath43 may also lead to us deriving a larger value of @xmath0 than that found for large samples of galactic gmcs . as discussed in the introduction , we know that the mean integrated co intensity of a gmc depends sensitively on the co line - width for that gmc ( see e.g. * ? ? ? * ; * ? ? ? increasing the line - width while keeping all other cloud properties the same leads to an increase in @xmath9 and hence a decrease in @xmath0 . similarly , decreasing the line - width decreases @xmath9 and increases @xmath0 . this implies that @xmath0 should depend to some extent on @xmath81 , in the sense that clouds with larger virial parameters that survive for long enough to form co should have smaller values of @xmath0 than similar clouds with smaller values of @xmath81 . in any large sample of gmcs , we would not expect to find many clouds with @xmath82 : any clouds that do initially have @xmath83 will undergo gravitational collapse , inducing motions that rapidly increase @xmath38 driving @xmath81 towards 1 if the clouds continue to collapse or 0.5 if they settle into virial equilibrium @xcite . on the other hand , there is both theoretical @xcite and observational @xcite evidence that many gmcs are gravitationally unbound , with @xmath84 . it is therefore plausible that the mean value of @xmath81 for the population of gmcs in the milky way is of order unity or larger . if so , then it follows that the clouds that we model in this study have systematically smaller co line - widths than the average galactic gmc , helping to explain why the values of @xmath0 that we derive for them are larger than @xmath77 . we now come to the question at the heart of this paper : does the value of @xmath0 vary with the local ambient sfr ? while the results displayed in figure [ fig : xco_sfr ] clearly show that @xmath0 is indeed a function of sfr , the extent to which the x - factor varies depends on several parameters , and we will now discuss these in turn . and below.,width=331 ] and below.,width=331 ] and below.,width=331 ] first , examining the red circles in figure [ fig : xco_sfr ] , we see that , in general , the mean x - factor derived from the cloud maps increases as sfr increases . given that co is more easily dissociated than h@xmath20 , this makes sense : the co is doing a progressively poorer job of tracing the true extent of the molecular gas as the ambient sfr increases . however we see that the effect is more pronounced for the high - mass clouds than for the low - mass clouds . indeed , there is a slight decrease in the x - factor of the 100 sfr@xmath61 case with respect to the 10 sfr@xmath61 case for the low - mass clouds . so while the x - factor at solar neighbourhood values of the sfr ( i.e. sfr@xmath61 ) is independent of mass , this is not true when the sfr is increased . the reason behind this behaviour in the low - mass clouds can be inferred from figure [ fig : cumu_wco ] , where we show the fraction of the cloud s total co emission that comes from pixels below a given @xmath9 . for progressively higher sfr , we see that fraction of the overall integrated emission coming from very bright lines of sight depend on the cloud mass . for the low - mass clouds at high sfr , the majority of the emission comes from regions with integrated intensities of 10 @xmath85 or greater , while the opposite is true for the high - mass clouds , in that they gain most of their emission from lines of sight with lower @xmath9 . the idea summarised by @xcite that a few bright lines of sight can compensate for the lack of co elsewhere in the cloud only appears to hold for our low - mass clouds . the reason why the clouds behave differently has to do with their column density distributions . in the low - mass clouds , the mean column density is low enough that co is efficiently photodissociated throughout much of their volume as the sfr is increased . the co that survives is confined to the highly overdense regions produced by turbulent compression and gravitational collapse , in which the shielding is more effective . because of the high density , the excitation temperature of the co in these regions is high , and so is the integrated intensity of the co emission . these clouds therefore have many regions with very low @xmath9 and a few with very high @xmath9 that dominate the total emission . in the high - mass clouds , on the other hand , the mean column density of the gas is larger and the co is therefore better able to resist photodissociation . because of this , co survives in lower density regions in these clouds even when the sfr is increased significantly . the excitation temperature of the co in these regions is lower than in the dense cores , and so the integrated co intensities coming from the lower density gas are smaller , but the much larger filling factor of these regions compensates for this , and leads to their contribution dominating the total emission from the cloud . it is plausible that molecular clouds in environments with higher sfrs will be denser , on account of the higher ambient gas pressure in these environments ( see e.g. * ? ? ? * ; * ? ? ? some support for this picture comes from observations of molecular clouds in the central molecular zone of the milky way . the star formation rate in this region is roughly 50 - 100 times that in the solar neighbourhood @xcite and the molecular clouds located there have high mean densities , of the order of @xmath86 or more ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? there is also evidence to suggest that these clouds have higher values of @xmath81 than local clouds @xcite . to investigate this scenario , we performed two additional simulations that probe this regime . we modelled two clouds with mass @xmath87 and initial number density @xmath88 , in which the hydrogen was taken to be initially in molecular form . one of the clouds was initialized with a turbulent field resulting in a virial parameter @xmath43 ( i.e. virial equilibrium ) , while the other had an @xmath44 ( i.e. the cloud was gravitationally unbound ) . these are shown as the green circles in figure [ fig : xco_sfr ] . we see that the x - factors for these clouds are now much closer to the standard milky way value , demonstrating that the physical properties of the clouds play a significant role in controlling the x - factor . we also see that , as expected , the x - factor is lower in the case where the amount of turbulent kinetic energy in the cloud is higher . another factor that strongly affects the mean x - factor derived from the cloud maps is the value used for deciding whether or not a pixel should be included in the averaging process , @xmath71 . if we set @xmath89 , and hence include all of the lines of sight in our synthetic emission maps when determining @xmath90 and @xmath91 , then we find that @xmath0 increases by between a factor of three ( for the low - mass clouds ) and an order of magnitude ( for the high - mass clouds ) as we increase the sfr from sfr@xmath52 to 100 sfr@xmath52 . however , if we restrict our calculation of the x - factor to only those lines of sight with detectable co emission , we find that the dependence of @xmath0 on the sfr becomes much weaker . for example , if we set @xmath92 , we find that the x - factor increases at most by a factor of two as we increase the sfr from sfr@xmath52 to 100 sfr@xmath52 . this suggests that the main effect driving the increase in @xmath0 with sfr in the @xmath89 case is a large increase in the number of lines of sight that have molecular hydrogen but no detectable co emission , or in other words that pass through `` co - dark '' molecular gas @xcite . in those regions of the clouds that remain co - bright , the relationship between co emission and h@xmath6 column density varies only weakly with the sfr . this behaviour is similar to what is seen in observational studies of the dependence of @xmath0 on metallicity : observations that focus only on co - bright regions find a weak dependence of @xmath0 on z ( e.g. * ? ? ? * ; * ? ? ? * ) , while those that sample both co - bright and co - dark gas find a much stronger dependence ( e.g. * ? * ; * ? ? ? . finally , it is also interesting to note that the x - factor for the high density , @xmath93 clouds is insensitive to the value of @xmath71 . this is because these clouds are so dense that the `` skin '' in @xmath19-space that contains h@xmath20 but not co is so thin that very little mass is contained within it . as such , the co typically always does a good job of tracing the molecular content of these clouds . the results presented in the previous section demonstrate that if we restrict our attention to those portions of molecular clouds that are traced by relatively bright co emission , then the x - factor that we have to adopt in order to correctly convert from the co intensity of these regions to their h@xmath6 column density does not vary strongly with the local environment . increasing the size of the assumed star formation rate ( and hence the strength of the isrf and the cosmic ray ionisation rate ) by a factor of 100 leads to an increase in @xmath0 of around a factor of two at most . this result is consistent with the finding by @xcite and @xcite that in the lmc @xmath0 does not vary strongly with the strength of the isrf , since these authors determine the masses of the clouds that they study using virial mass estimates that are sensitive only to the properties of the co - bright gas . unfortunately , the number of galaxies for which we can distinguish between co - bright and co - faint portions of gmcs is very small . this is routinely done in the milky way , has already been done to some extent in the magellanic clouds , and with the advent of alma becomes possible to do in more distant members of the local group . however , in most extragalactic observations , the best spatial resolution that we can achieve is comparable to or larger than the size of the individual clouds ( see e.g. * ? ? ? * ) . in this case , we can not easily distinguish between co - bright and co - faint portions of the clouds , and so the results that we present for the case when @xmath89 give the best guide to the behaviour of @xmath0 in these systems . for these unresolved clouds , our results show that @xmath0 increases substantially as we increase the sfr , provided that the other properties of the clouds remain unchanged ( i.e. provided that their characteristic densities or velocity dispersions do not also vary as functions of the sfr ) . one important implication of this is that in order to be able to use co emission as a reliable tracer of molecular mass , we need to understand the nature of the underlying cloud population and how this varies as a function of the star formation rate . if we make the simplifying assumption that cloud properties do not depend to any great degree on the local value of the sfr , then our results imply that the standard interpretation of the kennicutt - schmidt relation may need to be reassessed . if the observed co luminosity is dominated by high - mass clouds that have number densities similar to those of our fiducial cloud models , then our simulations suggest that @xmath0 should scale with the ambient rate of star formation as roughly @xmath94 provided that the sfr does not vary widely over the area of the observational beam , it then follows that @xmath95 measurements of the kennicutt - schmidt relation at gas surface densities where the molecular component dominates ( e.g. * ? ? ? * ; * ? ? ? ? * ; * ? ? ? * ) find a power - law relationship that we can write as @xmath96 where @xmath97 denotes the observed power - law index . in practice , however , @xmath98 is usually not measured directly ; instead , it is inferred from the co luminosity through the application of some constant co - to - h@xmath6 conversion factor , which in normal spiral galaxies is generally taken to have the canonical galactic value . therefore , we can more accurately write the above relationship as @xmath99 where @xmath100 is the surface density of co emission . now , if the actual co - to - h@xmath6 conversion factor is not fixed , but instead varies with @xmath101 according to equation [ xco_sfr ] , then it follows that the actual relationship between @xmath101 and @xmath98 can be written as @xmath102 rearranging this , we find that after correcting for the dependence of @xmath0 on the star formation rate , the true relationship between @xmath101 and @xmath98 can be written as @xmath103 where @xmath104 recent work by @xcite has suggested a value of @xmath105 from a sample of nearby galaxies . in this case , the ` true ' index of the kennicutt - schmidt relation would be around @xmath106 , implying that co is a progressively worse tracer of the star formation rate as one moves to more extreme environments ( see the discussion in @xcite ) . however , more recently , there have also been claims that @xmath107 ( i.e. that the kennicutt - schmidt relation is sub - linear ) , even when assuming a constant x - factor ( e.g. @xcite ) , with the most likely value lying at around 0.76 . in this case , we would predict a value of @xmath108 . it is important to note that this conclusion depends strongly on our assumption that the cloud properties do not vary strongly as a function of the local star formation rate . if this is not the case , and clouds in regions with higher values of @xmath101 are systematically denser and/or more turbulent than clouds in less actively star - forming regions , then it is plausible that the decrease in @xmath0 caused by the higher densities and turbulent velocities could be large enough to overwhelm the increase caused by the higher star formation rate . indeed , there is evidence that this is the case in the centres of many spiral galaxies : measurements of @xmath0 often find that it decreases close to the centre ( see e.g. * ? ? ? * ) , while we know from the study of the central molecular zone of our own galaxy that molecular clouds in this environment are very dense and highly turbulent ( see e.g. the summary in the review of * ? ? ? similarly , observational determinations of @xmath0 in nearby ultraluminous infra - red galaxies ( ulirgs ) or rapidly star - forming systems at high redshift typically find values that are smaller than the canonical local value ( see e.g. @xcite or the recent review by @xcite ) . however , these observations probe regions with much higher gas surface densities and turbulent velocities than are found in the local ism , once again suggesting that the decrease in @xmath0 caused by the increased densities and velocities is able to overcome the increase caused by the higher radiation field strength and cosmic ray ionization rate . ultimately , therefore , the real message to take away from this study is that it is necessary to understand the nature of the cloud population _ before _ one can properly interpret the co emission . unless one understands how the properties of the clouds depend on the local star formation rate , it is impossible to be certain whether @xmath0 increases with increasing @xmath101 ( as is the case in our simulations at fixed @xmath81 ) , decreases with increasing @xmath101 ( as appears to be the case in the centres of many spirals ) , or remains approximately constant . further compounding this problem is the fact that co measurements are one of the main observational tools used for determining the mass of molecular gas in clouds , potentially leaving us with a circular argument . supplementary information on cloud masses , from tracers such as dust emission , can probably help to break this circularity , but a detailed examination of how best to do this is outside of the scope of this paper . it is interesting to compare our results on the dependence of the x - factor on the sfr with those of previous numerical studies that have looked at this problem . @xcite use the time - dependent ucl_pdr code @xcite to examine how @xmath0 varies as a function of depth within a semi - infinite slab of gas as they vary a number of different physical parameters , including the radiation field strength . they find that at high @xmath19 , increasing the radiation field strength has little effect on @xmath0 : changing @xmath109 by a factor of @xmath110 leads to no more than a factor of three change in @xmath0 . at low @xmath19 , on the other hand , the dependence of @xmath0 on the radiation field is much stronger . at very low @xmath19 , the gas is co - faint and @xmath0 is very large , but as we move to higher @xmath19 , the co content of the gas grows , and the value of @xmath0 approaches the canonical galactic value . the depth into the cloud at which this transition occurs depends on the radiation field strength , with stronger fields implying that the transition occurs at higher @xmath19 . it is not straightforward to convert from these results to a cloud - averaged @xmath0 , but it is clear that they are at least qualitatively consistent with our findings that @xmath0 measured only for co - bright regions does not vary strongly with the sfr , while @xmath0 averaged over the whole cloud ( including the co - faint regions ) shows a much stronger dependence . @xcite have also explored the dependence of @xmath0 on the local radiation field . they performed a large hydrodynamical simulation of the formation of a typical @xmath111 galaxy ( see also * ? ? ? * ) and then post - processed the results of this simulation using a sub - grid prescription for @xmath0 that assumes that it is primarily determined by the mean visual extinction of the individual clouds and the strength of the isrf . for @xmath112 , @xcite use the results of @xcite to calibrate their subgrid model , while for @xmath113 and @xmath114 , they estimate the co abundance ( and hence @xmath0 ) using a simple remapping procedure that assumes that the clouds are in photodissociation equilibrium . they find that for clouds with mean extinctions @xmath115 and below , the value of @xmath0 depends strongly on the radiation field strength . for higher column density clouds , on the other hand , @xmath0 becomes largely independent of the radiation field strength . as most of the molecular clouds that form in their simulations have relatively large column densities , they find that when you average over the whole population of clouds , the dependence of @xmath0 on the strength of the isrf becomes very weak . at first sight , these results would appear to contradict our finding that @xmath0 can vary significantly as we change the sfr . however , there are a couple of important points that one should bear in mind . first , @xcite do not vary the cosmic ray ionisation rate , only the radiation field strength . therefore , their models do not account for the increased destruction of co by dissociative charge transfer with he@xmath21 that occurs when the cosmic ray ionisation rate is large . at high @xmath19 , this is the main co destruction mechanism , and its inclusion in our models is one of the main reasons why we see at least some dependence of @xmath0 on sfr even when we restrict our attention to portions of the cloud that are highly shielded and co - bright . second , the hydrodynamical simulations used as a basis for the @xcite analysis have a resolution of only 60 pc and hence are sensitive only to the largest , densest clouds , which are naturally the ones least affected by changes in the sfr . another large - scale numerical study of the dependence of @xmath0 on environment was presented by @xcite . they post - processed an extensive series of sph simulations of isolated and merging galaxies in order to determine the local chemical and thermal state of the gas and the consequent co emission ( see also @xcite and @xcite ) . they found that in their models , @xmath0 scales inversely with the molecular gas surface density as @xmath116 . in these models , @xmath117 by construction , and so the implication is that @xmath0 should decrease weakly with increasing star formation rate , scaling as @xmath118 . however , as with many of the observational studies of @xmath0 in extreme environments mentioned in the previous section , direct comparison of these results with our own is complicated by the fact that the mean properties of the clouds in regions with high sfr in these simulations are not the same as those of the clouds in regions with low sfr . @xcite also studied the effects of changes in the strength of the isrf and the cosmic ray ionization rate on the value of @xmath0 using a simple semi - analytical approach . they used the galform semi - analytical galaxy formation model @xcite to compute the atomic and molecular gas content of a series of model galaxies , and then used the ucl_pdr code to determine the co luminosity of the molecular gas , from which they could then infer @xmath0 . in their pdr calculations , it was necessary for them to assume some representative number density and visual extinction for the molecular gas . in most of their models , @xcite assumed a characteristic number density @xmath119 and a characteristic extinction @xmath120 . in these conditions , one would expect co photodissociation to be completely negligible even when the isrf is very strong , meaning that changes in the radiation field strength will only affect @xmath0 indirectly , through its effect on the thermal balance of the dust . indeed , @xcite find that in their model , even very large changes in the radiation field strength have only a small influence on @xmath0 . however , in practice , in realistic gmc models only a small fraction of the total gas mass is found in regions with a mean visual extinction as high as @xmath120 ( see e.g. * ? ? ? * ) , and so by focussing only on these conditions , @xcite ignore the large changes in the co content and co luminosity of the lower density , less shielded gas that drive much of the variation that we find in @xmath0 . @xcite also examine the effect of varying the cosmic ray ionization rate while keeping the radiation field strength fixed , and find that in this case , @xmath0 increases slowly with increasing ionization rate , in general agreement with our results . the dependence of @xmath0 on the uv field strength was also examined by @xcite . they computed @xmath0 for clouds illuminated by radiation fields with 1 and 10 times the strength of the @xcite field , using the 3d - pdr code @xcite . when computing @xmath0 , they only considered pixels in their synthetic emission maps with co integrated intensities @xmath121 . they find that @xmath122 for their @xmath53 run and @xmath123 for their @xmath124 run , a roughly 50% increase . this is somewhat smaller than the difference we find between our @xmath54 and @xmath125 runs , which for a similar integrated intensity cut show values of @xmath0 that differ by closer to a factor of two . however , it is important to note that @xcite do not vary the cosmic ray ionisation rate at the same time that they vary the radiation field strength , and so it is likely that there is less destruction of co in their @xmath124 run than in our @xmath125 run . finally , in a recent paper , @xcite have studied how the typical co abundance in a gmc varies as a function of the cosmic ray ionisation rate . they find that increasing the cosmic ray ionisation rate by a factor of ten or more leads to substantial destruction of co and argue that this will lead to clouds in regions with @xmath126 being co - faint . however , they do not compute @xmath0 for any of their clouds models and hence do not quantify its dependence on the sfr . there are two important methodological caveats that the reader should bear in mind when considering our results . first , as we have already discussed in section [ sfr0 ] , we do not currently account for the effects of line - of - sight velocity gradients when determining the effectiveness of h@xmath6 self - shielding . this means that we will tend to overestimate the effectiveness of self - shielding in gas with h@xmath6 column densities of around @xmath127 , the regime where doppler broadening of the lyman - werner lines dominates the uv absorption spectrum . consequently , we will overestimate the total h@xmath6 content of the cloud . however , as most of the h@xmath6 in the cloud is found in regions with much higher h@xmath6 column densities , we would not expect the omission of this effect to have a large impact on our results . second , we have assumed that the cosmic ray ionisation rate is uniform throughout the cloud , or in other words that the cosmic ray absorption spectrum is not significantly affected by absorption within the cloud . whether or not this is a good approximation depends on the details of the low energy portion of the cosmic ray energy spectrum , which is poorly constrained @xcite . there is some tentative evidence from astrochemical studies that the cosmic ray ionisation rate in dense cores may be significantly lower than in the diffuse ism ( see e.g. * ? ? ? * ) , but the issue is far from being settled . if there is indeed a significant fall - off in the cosmic ray ionisation rate as we move from diffuse to dense gas , then our models will tend to overestimate the heating rate in the dense gas , and also the rate at which co is destroyed there . the net effect of this on @xmath0 is difficult to predict without detailed modelling . however , we note that the destruction of the relatively diffuse inter - clump co component that occurs as we increase the sfr is a consequence of the increasing strength of the isrf , and not the increase in the cosmic ray ionisation rate . therefore , even if cosmic ray absorption within the cloud becomes significant , this behaviour will remain the same , and so we would expect to still see an increase in @xmath0 as we increase the sfr . in addition to these methodological caveats , we also take this opportunity to remind the reader that when we refer to changes in the `` sfr '' , what we mean are changes in the strength of the isrf and the size of the cosmic ray ionisation rate , since we assume that both scale linearly with the star formation rate . it could well be that this assumption is too simplistic , and that the functional dependence of the radiation field strength and the cosmic ray ionisation rate on the local sfr is more complicated than a simple linear scaling . if so , then the results of our study will still hold , but the actual star formation rates corresponding to our @xmath55 and @xmath56 models will differ from what we have assumed here . in this paper , we have performed a series of numerical simulations of molecular cloud evolution that explore how the distribution of co emission coming from the clouds changes as we change the strength of the interstellar radiation field and the size of the cosmic ray ionisation rate in the clouds . if one makes the reasonable assumption that these quantities scale linearly with the local star formation rate , then our results indicate how the co - to - h@xmath6 conversion factor for the clouds , @xmath0 , depends on the star formation rate . we find that as we increase the radiation field strength and the cosmic ray ionisation rate , the co content of our simulated clouds decreases . co survives more effectively in well - shielded clump and filaments than in the more diffuse inter - clump gas , leading to a significant decrease in the filling factor of bright co emission . the integrated intensity of the co emission from the brightest regions increases with increasing sfr , owing to the heating of the gas by the higher cosmic ray flux , but this increase is too small to compensate for the loss of co emission from elsewhere in the cloud , and so overall the co luminosity of the cloud decreases . the change in the chemical composition of the cloud as we increase the sfr leads to a change in @xmath0 . however , the size of this change depends on the method we use to compute @xmath0 . if we consider the whole area of the cloud when computing the mean h@xmath6 column density and co integrated intensity , then we find that @xmath0 changes substantially , scaling as roughly @xmath128 for our high - mass clouds . on the other hand , if we consider only regions with co integrated intensities exceeding an intensity threshold @xmath71 , then we find that the dependence of @xmath0 on sfr weakens substantially as we increase @xmath71 . the reason for this difference in behaviour is that the fraction of the cloud filled with `` dark '' molecular gas ( i.e. h@xmath6 without associated co ) increases as we increase the sfr . this dark gas contributes to our calculation of @xmath0 in the case where we do not apply an intensity threshold , but is ignored when we do apply a threshold . we have also explored whether increasing the density and turbulent velocity dispersion of our model clouds at the same time as we increase the sfr affects the relationship between @xmath0 and the sfr . we find that if we increase the initial number density to @xmath32 and raise the virial parameter to @xmath44 , then we can recover a value of @xmath0 close to the canonical galactic value even when the local sfr is 100 times our default value . we conclude from this that we will only recover a positive correlation between @xmath0 and the sfr if there is not a strong correlation between mean cloud densities and the sfr . the fact that observational determinations of @xmath0 in starbursting systems such as ulirgs find a value _ lower _ than the canonical galactic value therefore provides a strong indication that the properties of typical molecular clouds in these systems must differ significantly from those of molecular clouds in the local ism . consequently , we argue that the most important message to take away from this study is that it is necessary to understand the nature of the cloud population before one can properly interpret the co emission coming from the clouds . finally , we stress that if the co emission from nearby galaxies is dominated by virialised clouds with masses around @xmath129 @xmath130 and densities around a few 100 @xmath131 , then the x - factor may vary significantly as the local sfr changes . in these circumstances , our standard interpretation of the kennicutt - schmidt relation , in which @xmath0 is assumed to be constant throughout the galaxy , should be rethought . the authors would like to thank j. pety , r. feldmann , n. gnedin , and f. israel for interesting discussions concerning the physics of the co - to - h@xmath6 conversion factor in strong radiation fields . they would also like to thank the referee , d. narayanan , for a very helpful and constructive report that helped to improve the paper . scog acknowledges financial support from the deutsche forschungsgemeinschaft via sfb 881 , `` the milky way system '' ( sub - projects b1 , b2 and b8 ) and spp 1573 , `` physics of the interstellar medium '' ( grant number gl 668/2 - 1 ) . molinari , s. , bally , j. , glover , s. , moore , t. , noreiga - crespo , a. , plume , r. , testi , l. , vzquez - semadeni , e. , zavagno , a. , bernard , j .- p . , & martin p. 2014 , in protostars and planets vi , eds . h. beuther , r. s. klessen , c. p. dullemond and t. henning , ( tucson : university of arizona press ) , 125

the calculation of the dephasing time in electron systems is presented . by means of the keldysh formalism we discuss in a unifying way both weak localization and interaction effects in disordered systems . this allows us to show how dephasing arises both in the particle - particle channel ( weak localization ) and in the particle - hole channel ( interaction effect ) . first we discuss dephasing by an external field . besides reviewing previous work on how an external oscillating field suppresses the weak localization correction , we derive a new expression for the effect of a field on the interaction correction . we find that the latter may be suppressed by a static electric field , in contrast to weak localization . we then consider dephasing due to inelastic scattering . the ambiguities involved in the definition of the dephasing time are clarified by directly comparing the diagrammatic approach with the path - integral approach . we show that different dephasing times appear in the particle - particle and particle - hole channels . finally we comment on recent experiments .

cond-mat9903146

quantum corrections to the classical drude formula for the electrical conductivity give rise to singular terms in low - dimensional systems@xcite . there are two types of such terms . the first is known as the weak localization correction ( wl ) and arises as a result of the quantum interference of electron waves in disordered systems . the second is a consequence of the enhancement of the electron - electron interaction in a disordered system and is usually referred to as electron - electron interaction correction ( eei ) . in two dimensions , in particular , the correction to the conductivity is logarithmic for both the wl and eei correction . the argument of the logarithm contains the ratio of two length scales . the first is the mean free path , @xmath0 , which sets the microscopic scale , beyond which the system behaves diffusively . the second length scale differs in the two cases . for wl it is @xmath1 , the scale over which inelastic scattering starts to destroy the interference effects . for eei it is usually given by the thermal length @xmath2 . in a diffusive system , all length scales correspond to characteristic times : the elastic scattering time @xmath3 , the dephasing time @xmath4 , and the thermal time @xmath5 . in general , dephasing occurs either as a result of the interaction of the system with the external environment or as a consequence of the internal electron interactions . for example , it is known that an ac - field suppresses the wl . on the other hand , while it is clear that inelastic scattering contributes to dephasing , the exact way this happens is by far less obvious as witnessed by the recent hot debate in the literature @xcite . natural questions to ask are : how does the dephasing time @xmath6 that enters the wl correction and cuts off the logarithmic singularity depend on inelastic scattering and on the external environment ? does dephasing also affect the eei correction ? the two issues need to be addressed together if one wants to understand the experimental data in detail . although one can find in the literature discussions of both these issues , there seems to exist no unified discussion of them . it is one of the aims of the present paper to fill this gap . it was first noted by schmid@xcite that the inelastic quasiparticle scattering time is enhanced in the presence of disorder . the dephasing time , which controls the wl correction was initially assumed to be identical to the inelastic quasiparticle scattering time in a disordered fermi - liquid @xcite . according to this analysis , the inverse dephasing time was assumed to be @xmath7 in 2d , thus predicting a violation of the fermi - liquid behavior at low temperatures . however , altshuler , aronov , and khmelnitsky@xcite ( hereafter referred to as aak ) , by means of a semi - classical path - integral approach to the wl correction , were able to calculate the dephasing time directly and predicted an inverse dephasing time proportional to @xmath8 , in contrast to ref.@xcite . some time later fukuyama and abrahams@xcite ( hereafter indicated as fa ) , re - examined the problem in terms of standard diagrams and calculated the `` mass '' term that develops in the particle - particle propagator . they found the same result as ref.@xcite . the origin of the discrepancy of the aak and fa results lies in a genuine ambiguity in the definition of the dephasing time itself . a review of the attempts to clarify this issue made at the time@xcite may be found in @xcite . the issue was further complicated from the fact that a dephasing time also appears in the particle - hole channel@xcite . the mass term in the particle - hole propagator turned out to coincide with that found by fa in the particle - particle propagator . more recently inelastic scattering and dephasing have been addressed from a general point of view by stern et al @xcite and in connection with disordered mesoscopic systems by various groups@xcite , in all cases confirming the aak result . furthermore , during the last couple of years , various experimental groups have observed a saturation of the dephasing time at low temperatures @xcite . usually this saturation has been related to heating effects or to the presence of magnetic moments . in ref.@xcite , however , these possibilities have been excluded experimentally , and the observed saturation seems to be in contrast to the available theories . if the observed saturation of the dephasing time is due to an intrinsic mechanism of disordered electron systems , this will have dramatic consequences on the localization theory , such as preventing an insulating ground state . the above consideration led to a fresh re - examination of the theory whose historical development has been sketched above . in particular , such an intrinsic mechanism has been claimed to be discovered in ref.@xcite . however in refs.@xcite this new theoretical result has been questioned and at the same time in ref.@xcite a conventional mechanism has been suggested as a possible explanation of the experimentally observed saturation of the dephasing time . the authors of ref.@xcite concluded that there is no need to revise the existing theory of the dephasing time in wl . while we completely agree with ref.@xcite on this latter point , we think that the interpretation of the experiments is not yet settled and , especially , the effect of dephasing on the eei correction is not clear . in this paper we discuss dephasing in the particle - hole channel , and in parallel we clarify the reason for the discrepancy in the aak and fa calculations . the difference between the results may be traced back to a different definition of the dephasing time . in the fa case , the dephasing time is assumed to be the mass of the particle - particle propagator as due to the self - energy corrections . in this type of corrections , one does not include diagrams which connect the upper and the lower electron lines entering the particle - particle propagator . in the aak approach these diagrams , i.e. the vertex corrections , are taken into account . the inclusion of these vertex diagrams in the diagrammatic approach of fa leads to the aak result@xcite . in the particle - hole channel , on the other hand , we find that vertex corrections are negligible and the dephasing is entirely determined by the self - energy . to see this we will adopt a real time - formulation based on the keldysh technique which allows us to discuss wl and eei corrections in a unified way . furthermore , in the real - time formulation non - linear effects , for example those due to an external field , can be incorporated more easily . in order to address the issue of how dephasing arises in the particle - particle and in the particle - hole channels , it is instructive to start considering the situation of an external field . this analysis , besides providing a relatively simple physical situation to analyse has also a genuine experimental interest . for example , there exists no satisfactory theoretical explanation for the non - linear field effects observed in refs.@xcite . to this end we will devote a substantial part of our analysis to the nonlinear effects on wl and eei corrections for which we shall find an unexpected difference in their dependence on a static electric field . the plan of the paper is the following . in the next section , by means of the keldysh formalism , we derive an expression for the additional current in a disordered conductor due to both the wl and eei corrections . the formulae obtained are valid in the presence of an arbitrary time - dependent external field . this result allows us , in section iii , to discuss the wl and eei corrections in the presence of both a dc and an ac electric field . we show that , in the case of a dc field , the eei correction is suppressed , in contrast to what happens for the case of wl correction . an ac field , on the other hand , suppresses both the wl and eei correction , but with a different dephasing time . in section iv , we will review the fa calculation of the inelastic scattering time in two dimensions and show how the aak result may be obtained from it . in section v the path integral approach to the dephasing time of aak is briefly outlined and connection is made with the diagrammatic analysis . special attention is paid to how the cancellation of the infrared singularity occurs . we extend this approach to a calculation of the dephasing time in the particle - hole channel . finally , section vi is devoted to some discussion of the experimental results . in appendix a we give technical details concerning the calculation of the dephasing time in the particle - hole channel . in a weakly disordered metal , quantum corrections lead at low temperature to deviations from the drude - boltzmann theory of transport . the weak localization ( wl ) corrections are due to electrons diffusing along closed paths , where quantum interference causes an enhanced back - scattering probability . the correction to the current in the sample is given by@xcite @xmath9 where @xmath10 is the diffusion constant , @xmath11 the elastic scattering time and @xmath12 is the so - called cooperon or particle - particle propagator . its graphical definition is given in fig.[fig2 ] . in the presence of a vector potential the cooperon is given by the solution of the differential equation @xmath13 ^ 2 \right\ } c^t_{\eta \eta'}({\bf r } , { \bf r } ' ) & \nonumber\\ & = { 1\over \tau } \delta(\eta -\eta ' ) \delta({\bf r } -{\bf r } ' ) .&\end{aligned}\ ] ] here and below we use units where @xmath14 . in addition to wl there are quantum corrections to the conductivity due to the interplay of disorder and electron - electron interaction ( eei ) . in the long - wavelength , low - frequency limit the electron motion is diffusive , leading to an enhancement of the effective interaction , @xmath15 , and results in either an enhanced or suppressed conductivity in the case of the direct and exchange contribution of the interaction , respectively . in the following we derive these corrections to the conductivity in the presence of an external electromagnetic field . since one has to calculate a time - dependent , non - linear response function , we work within the keldysh formalism , using the notation of ref.@xcite . this has the advantage that no analytic continuation is necessary at @xmath16 in order to get the physical response functions . in particular the two - particle propagators like the cooperon are directly defined in terms of a retarded and an advanced green s function , and there is no restriction on the energies of these green s function in contrast to what is found within the matsubara formalism . the starting point is the general equation for the current in the presence of both interactions and an external field . for fixed disorder configuration , the current in @xmath17-dimensions is found from the keldysh component of the electron green s function , @xmath18 in order to simplify the notation , we only write down the time dependence of the green s function in the following . to first order in the interaction the correction to the green s function is given by @xmath19 where @xmath20 is the non - interacting green s function in the presence of both disorder and an external field , @xmath21 is the self energy , and one has to integrate over @xmath22 and @xmath23 . it turns out , that only the first two terms contribute to the quantum correction to the conductivity , so the term @xmath24 will be neglected from now on . the first two terms in eq.([eq4 ] ) correspond , upon impurity averaging , to the diagrams of fig.2 of the first of refs.@xcite . to lowest order in the interaction the retarded self - energy is given by @xmath25 where @xmath26 and @xmath27 denote the retarded and keldysh component of the interaction . singular corrections due to the interplay of disorder and interaction arise , after averaging over disorder , from singular vertex corrections , i.e. diffusons , which appear on the vertices that connect an advanced and a retarded green s function . corrections to @xmath28 where both vertices are singular are much larger than the contributions with only one singular vertex . as a consequence , these various , less singular , terms are neglected . in particular , the third term in eq.([eq4 ] ) , and the second term in eq.([eq5 ] ) give rise to only one singular vertex and are therefore neglected . the structure of the impurity averaged keldysh green s function is then @xmath29 where @xmath30 is the disorder average of @xmath20 and we have indicated by @xmath31 and @xmath32 the vertices dressed by disorder . notice that @xmath28 is purely imaginary , and the current in eq.([eq3 ] ) is real . in order to make the structure of this equation clear , we represent @xmath33 and the vertices @xmath31 , @xmath32 graphically in fig.[fig1 ] . an external electromagnetic field affects all the green s functions and all the vertex functions in this expression . however , it is known@xcite that the sum of the diagrams with two diffusons , i.e. the sum of the terms with an electromagnetic field insertion in the green s functions , cancel . diagrams which contribute to the quantum corrections to the conductivity must have at least three diffusons . in our formalism these diagrams are generated by inserting the electromagnetic field in the vertex corrections @xmath31 and @xmath32 . the singular part of the vertices is given by @xmath34 where the brackets denote the impurity average . @xmath35 is the single - particle density of states . @xmath32 can be easily expressed in terms of the diffuson , @xmath36 with @xmath37 ^ 2 \right\ } d^{\eta}_{t , t ' } & \nonumber \\ & = { 1\over \tau } \delta ( t - t ' ) \delta ( { \bf r } - { \bf r } ' ) .&\end{aligned}\ ] ] the graphical definition of the diffuson is given in fig.[fig2 ] . in comparing the diffuson and the cooperon in fig.[fig2 ] , it is worth recalling that the time evolution of the diffuson is associated with the center - of - mass time of the particle - hole pair , whereas the relative time controls the evolution of the particle - particle pair of the cooperon . it is also useful to compare the different way an external vector potential enters in the equation for the diffuson and the cooperon . in the case of the diffuson , the vector potential is felt only if the particle and hole are delayed with respect to one another , i.e. when @xmath38 . in this respect , processes , which are characterized by the simultaneous creation of a particle - hole pair in a given point in space , are not affected by an external vector potential . this is the case for the density correlation function which is relevant for screening the coulomb interaction and justifies the fact that the screened coulomb interaction , in eq.(15 ) below , does not depend on the external vector potential . also the particle - hole pair , which appears in the vertex @xmath32 , defined above , is composed of particles created simultaneously . @xmath32 therefore does not depend on the external vector potential . by contrast the singular part of the vertex @xmath31 also involves the keldysh component of the disorder averaged green s function , when it is expressed in terms of a diffusion propagator see fig.[fig2]c . the keldysh function decays only slowly in time and introduces a time delay between the creation of the particle and hole . we approximate the keldysh function by its equilibrium value @xmath39 , \ ] ] where @xmath8 is the electron temperature . notice that in strong electric fields where the electron temperature increases due to heating , the electron distribution function may differ from the fermi function , and eq.([eq11 ] ) may break down . such effects are neglected in the present work . going back to the real - time formalism , we fourier transform @xmath40 . the fourier transform of @xmath41 is @xmath42 , which has to be convoluted with the advanced green s function . the delta function @xmath43 can be neglected since it eventually leads to a term proportional to @xmath44 in the self - energy @xmath45 which is zero . @xmath31 is then found as @xmath46 and definitely depends on the external field due to the time delay @xmath47 in the particle - hole propagator . calculating the current requires also momentum integrations . if the interaction @xmath48 transfers a momentum @xmath49 with @xmath50 , the @xmath51-integration over the green s functions gives @xmath52 @xmath53 the temperature dependent factor is due to @xmath54 , with @xmath55 . note that the vertex corrections @xmath31 and @xmath32 do not depend explicitly on @xmath56 but only on the difference of momenta @xmath57 . finally , we obtain the correction to the current due to the interplay of interaction and disorder as @xmath58 here for simplicity we assumed a homogeneous external field . in an inhomogeneous field the diffuson is a function of two momenta , and the current is determined from @xmath59 . in the case of long range coulomb interactions , the expression for the current simplifies . in the relevant low momentum , low frequency region the screened coulomb interaction is given by @xmath60 the frequency dependent numerator cancels when multiplied with the diffuson , so that the product is frequency independent , and transforms into a delta function after fourier transformation . thus the convolution of the interaction with the first of the two diffusons appearing in eq.([eq14 ] ) gives @xmath61 and the expression for the current becomes @xmath62 this equation will be the starting point for our further investigations . we cut off the time integration with @xmath11 since the diffusive regime requires that all times are longer than @xmath11 . we see that as with @xmath63 , @xmath64 is also related to the propagation of two particles . however , whereas wl is related to the propagation around a closed loop , i.e. the probability of return @xmath65 , this is not the case for eei , due to the prefactor @xmath66 in the momentum integration . also the time dependencies are different . in the case of wl two particles start diffusing at @xmath67 at @xmath68 and return at @xmath69 at the same point . from the time dependence of the diffuson in eq.([eei ] ) , we conclude that , in the case of eei , the first particle starts at @xmath70 at @xmath68 and arrives at @xmath67 at @xmath71 . the second particle starts at @xmath67 at @xmath68 and arrives at @xmath69 at the final point @xmath71 . equation ( [ eei ] ) allows us to obtain the results known in the literature . for instance , the interaction correction to the conductivity is found by expanding eq.([eei ] ) to first order in the electric field . by using eq.([eq10 ] ) one finds the diffuson to linear order in the electric field as @xmath72 and one arrives at @xmath73 where @xmath17 is the dimension . the hyperbolic sine provides a cutoff for large times @xmath47 . by approximating the temperature dependent factor with @xmath74 for @xmath75 and @xmath76 for @xmath77 , one finds immediately the temperature dependence of the correction to the conductivity as @xmath78 , @xmath79 and @xmath80 in @xmath81 . in one and three dimensions , the correct prefactor is only found by integrating eq.([eq19 ] ) with the result @xmath82 where we subtracted temperature independent terms . here we only considered the contribution to @xmath83 from the exchange diagram shown in fig.[fig1 ] . a correction to the conductivity with the same temperature dependence as in eqs.([eq22 ] ) is also found from the direct ( hartree ) diagrams , as shown in the literature@xcite . these will then change the total amplitude of the quantum correction to the conductivity due to the interplay of disorder and interaction . we will not consider this effect here and below . for completeness , we recall also the results for the wl correction to the conductivity in a weak static field . the cooperon is given by @xmath84 here a phenomenological dephasing time @xmath85 has been introduced in order to cut off the infrared singularity . the microscopic origin of such a dephasing time will be discussed later . by inserting eq.([cooperone ] ) in eq.([wl ] ) , one arrives at the well known expressions @xmath86 the quantum corrections to the conductivity may be suppressed due to various mechanisms . we start by studying the effect of a homogeneous external electric field . the effect of an electromagnetic field on wl has been described in the literature@xcite . here we consider in addition the effect of such a field on eei . we will first consider the non - linear response to a strong dc electric field , then we will discuss the response to a dc field in the presence of a microwave field . from eqs.([wl ] ) and ( [ eei ] ) both the linear and the non - linear response due to quantum interference can be determined . note that heating effects are neglected in the present discussion . in a static electric field , the vector potential is given by @xmath87 . the cooperon is then found as @xmath88 \nonumber\\ & = & { 1\over 2 \tau } \sum_q \exp \left [ - dq^2 \eta \right ] , \end{aligned}\ ] ] i.e. the electric field drops out . as a consequence , weak localization is not affected by a static electric field@xcite . there are , however , non - linear terms in the interaction correction to the current . the diffuson with the time arguments as needed for the calculation of the current is given by @xmath89 in contrast to weak localization , the interaction correction is not related to a simple sum over the momentum entering the two - particle propagator , since the factor @xmath90 has to be taken into account . as a consequence , by shifting the @xmath91 integration by @xmath92 in eq.([eq24 ] ) it does not simplify the calculation of the current , and especially there are non - linear electric field effects . one can easily estimate beyond which electric field the non - linear effects become sizeable . equation ( [ eq24 ] ) can be expanded in the electric field under the condition @xmath93 . on the other hand , as far as the interaction correction to the current is concerned , only times @xmath94 and momenta @xmath95 contribute effectively . the electric field can thus be considered small when @xmath96 which is equivalent to the condition @xmath97 . non - linear effects are expected to dominate for @xmath98 , i.e. when the voltage drop over a thermal length exceeds the temperature . for the sake of definiteness , let us consider eq.([eei ] ) in two dimensions . performing the q - integration , the current reads @xmath99 this does not lead to a simple linear dependence of the current on the electric field and thus non - linear field effects are definitely present . in the high temperature limit @xmath100 , or weak electric field limit , it is the temperature dependent factor that cuts off the integral at large time @xmath47 , whereas the @xmath101 dependent factor varies only weakly and is approximately one . including the first non - linear correction , the current is determined as @xmath102 where @xmath103 is the approximate numerical factor from the @xmath47-integration . in the limit of large electric field we approximate the @xmath101 dependent factor as @xmath104 and then we split the @xmath47-integration in eq.([eei ] ) as @xmath105 the second integral gives only a small contribution to eq.([eq28 ] ) provided that @xmath106 , and the value of the current at large fields is found to be @xmath107 considering now the current as a function of temperature , one will observe the @xmath79 behavior for @xmath108 while the current will saturate to eq.([eq25 ] ) for @xmath109 . for illustration , we plot in fig.[fig3a ] the non - linear conductivity as a function of temperature obtained by numerically integrating eq.([cur2d ] ) . the saturation of the conductivity at low temperature is apparent . in addition fig.[fig3a ] also shows the conductivity in the presence of a slowly ( in time ) oscillating field @xmath110 which is superimposed to the current driving field @xmath111 . this situation will be examined more closely below . in fig.[fig3a ] the conductivity is determined from the average of the current over one period of the oscillating field divided by the static field . the strength of the oscillating field is @xmath112 . at high temperature , the current is not affected by the oscillating field . since the oscillating field is much stronger than the dc field , the current saturates at much higher temperature than in absence of the ac field . in more than one dimension one has to distinguish the two cases @xmath113 and @xmath114 . as it is apparent from the figure , a parallel ac field saturates the current at higher temperature . if the frequency of the ac field is not too large ( @xmath115 ) , the effect can be understood from eq.([eq25 ] ) by considering the current in a strong electric field @xmath116 and expanding in the static field @xmath117 .\ ] ] upon time averaging the contribution , in the first term , not proportional to the dc field , vanishes . the difference between parallel and perpendicular ac and dc field stems from the last term and leads to @xmath118 in agreement with the numerical result in fig.[fig3a ] in the low temperature region . we find analogous results in one and three dimensions . figure [ fig3b ] depicts the current as a function of temperature obtained by numerically integrating eq.([eei ] ) in one dimension . again we studied the current in the presence of a static field only , and the time averaged current in the presence of both a static and an oscillating field with @xmath112 . as in two dimensions , one observes a saturation of the current in the low temperature region . the temperature scale , where the saturation occurs is roughly the same in one and two dimensions . we consider now specifically the response in the presence of time - dependent electric fields . the diffuson is then given by @xmath119 we start with fields which are only slowly time dependent and discuss high frequency fields later . as we mentioned before , only small time delays , i.e. @xmath94 , are relevant for the calculation of the interaction correction to the current . therefore , when the vector potential varies only slowly on this time scale , it may be expanded in the equation above , @xmath120 and one finds that the current instantly follows the electric field in this case . the conductivity in the presence of a slowly oscillating field in figs.[fig3a ] and [ fig3b ] was determined using this approximation . in the case of a fast time dependence of the vector potential , the approximation ( [ eq27 ] ) is not sufficient and the full time dependence has to be taken into account . for @xmath121 the relevant combination is @xmath122 in the high frequency limit the sine - functions are of order one , so that compared to ( [ eq27 ] ) the vector potential is reduced by a factor @xmath123 . the linear response to a high frequency electric field , for instance , becomes temperature independent for @xmath124 , since the main contribution to @xmath125 , eq.([eei ] ) , comes from @xmath126 and processes with @xmath127 are practically cut off . the conductivity is then found as @xmath128 , @xmath129 , @xmath130 in three , two and one dimension , as it is discussed in the literature @xcite . we now concentrate on the effect of a high frequency radiation on the response to a static field . in fig.[fig3c ] we plot the static conductivity obtained by integrating eq.([eei ] ) in the presence of both a static and a high frequency field in two dimensions . the static and ac field are parallel . the temperature is @xmath131 with @xmath132 . we find that the ac field reduces the quantum correction to the conductivity . the effect is strongest for low frequencies , @xmath133 , whereas in the limit of high frequencies the effect becomes weaker . in the relevant case of weak field and low frequencies , the correction to the static conductivity can be determined from ( [ eq26 ] ) with the result that @xmath134 if the ac and dc field are parallel and one third of this for perpendicular ac and dc fields . concerning the high frequency limit we argued after ( [ eq33 ] ) that the strength of the electric field in a high frequency is reduced by a factor @xmath135 which corresponds to @xmath136 if @xmath137 . we have verified this estimate numerically and analytically and found that for week field the correction to the conductivity changes into @xmath138 in the high frequency limit . to conclude this section , for the sake of completeness and comparison , we recall the results for the weak localization correction in a microwave field@xcite . notice that in the cooperon we need the sum of the vector potential at times @xmath139 , @xmath140 instead of the difference that enters the diffuson . we average the cooperon over one period of the ac field and sum over momentum . after some algebra one arrives at @xmath141 i_0\left [ { 4t_0 ^ 3\over \omega^3 } b\left ( { \omega \eta \over 2 } \right ) \right ] \\ b(x)&= & x\left ( 1+{\sin(2x)\over 2x } - 2 { \sin^2x \over x^2 } \right ) , \end{aligned}\ ] ] where @xmath142 and @xmath143 is the imaginary argument bessel function . the external microwave field does not affect the cooperon for small @xmath47 , but does so for large @xmath47 . the estimate of the typical time scale @xmath144 , at which the external field modifies the @xmath47-integration in the equation above , is found from the condition that the argument of the exponential in the second line of eq.([eq35 ] ) is of order one , and one gets @xmath145^{1/5 } & { \rm for } \,\ , d ( e e_{\rm ac})^2 \gg \omega^3 .\end{array } \right.\ ] ] strong non - linear effects of the microwave field associated with wl will be seen when @xmath146 is smaller than the dephasing time @xmath147 . for a fixed strength of the microwave @xmath148 , the time scale @xmath146 becomes large when @xmath149 , @xmath150 , which agrees with the observation that a static field does not lead to nonlinear effects in wl . in this section we calculate the `` mass '' in the cooperon and diffuson propagator . this section does not contain new material , since we will reproduce the results of refs.@xcite . however we think that this section contains useful information in order to understand the different approaches to the calculation of the dephasing time . in the presence of time reversal invariance , the particle - particle propagator may be obtained from the particle - hole propagator by reversing the direction of one of the electron lines . as a consequence , the mass term that develops in the particle - particle propagator also enters the particle - hole propagator . a mass term in the particle - hole channel could seem to imply , at first sight , a violation of the particle conservation . however , this is not the case because such a mass term eventually disappears in the physical response function , as it has been shown by the detailed diagrammatic analysis of castellani et al.@xcite . we now review briefly the derivation of how such a mass term arises . for our convenience , we adopt the notation of ref.@xcite , and restrict the discussion to the two - dimensional case . the two - particle propagator is denoted by @xmath151 and may represent a cooperon or diffuson . according to the standard diagrammatic technique , a mass term for a propagator may be obtained by considering the appropriate self - energy . the self - energy diagrams for the two - particle propagator are those labeled ( a)-(d ) in fig.[fig4 ] . by denoting with @xmath152 the contribution of these diagrams , one can use the dyson equation to re - sum their infinite series and obtain the two - particle propagator @xmath153 as , @xmath154 successively , by expanding the self - energy in powers of @xmath155 and @xmath156 , one may identify the quantum corrections to the diffusion constant , the density of states , and the frequency . this has been done explicitly in ref.@xcite . the mass term , which will be called @xmath157 , is found from the self energy at zero momentum and frequency . the only part of the graphs in fig.[fig4 ] , which is not manifestly proportional to @xmath158 or @xmath155 , is from graph ( d ) and is given by @xmath159 here , @xmath160 , @xmath161 , and @xmath162 are matsubara frequencies with @xmath163 and @xmath164 . @xmath165 is the external frequency , which after analytical continuation @xmath166 , is sent to zero . in this limit , the number of addenda in the matsubara sum of eq.([eq36 ] ) vanishes . however , the functions in the sum contain a branch cut after the analytic continuation and eq.([eq36 ] ) yields a finite result in the @xmath167 limit . after the analytic continuation with @xmath168 and @xmath169 , by taking the limit @xmath170 , ( i.e. , particles at the fermi surface ) , and observing that the sum of the fermi and bose function @xmath171/2 = 1/\sinh(\omega / t)$ ] is an odd function of @xmath172 , one gets @xmath173 considering the screened coulomb interaction as in eq.([eq15 ] ) , @xmath174 , the integral above is well behaved in the ultraviolet limit both in frequency and momentum . furthermore , due to the presence of the hyperbolic sine , the main contribution to the inelastic scattering time comes from frequencies @xmath133 , physically corresponding to the fact that there are no available excitations with larger energy . as usual , at a given temperature , only real excitations ( @xmath133 ) generate lifetimes , while virtual excitations ( @xmath175 ) are immaterial . on the other hand , the integral ( [ 6 ] ) is infrared divergent in one and two dimensions . to cure this divergence we insert the inelastic time self - consistently in the propagator @xmath176 , and also calculate the screened coulomb interaction more accurately . in two dimensions one finds @xmath177 with @xmath178 , the inverse screening length . since the screening length is usually much shorter than the mean free path , eq.([eq40 ] ) and eq.([eq15 ] ) coincide in almost all the relevant parameter range , but not in the extreme low momentum region , where there is no singularity in @xmath179 according to eq.([eq40 ] ) . by exploiting the fact that the main contribution to the frequency integral in ( [ 6 ] ) comes from the region @xmath180 , we may write in two dimensions @xmath181 @xmath182 } .\ ] ] the frequency dependence of the integrand is controlled by the two lorentzian factors which have typical scales @xmath157 and @xmath183 . depending whether @xmath184 or @xmath185 , one can safely neglect the frequency dependence of the first or second lorentzian . performing then the @xmath172 integration , we find @xmath186\ ] ] @xmath187 where @xmath188 is given by @xmath189 . for @xmath190 , the dominant contribution to the inelastic scattering rate comes from the momentum region @xmath191 , and we find @xmath192\ ] ] this is the fa result for the inelastic scattering rate we mentioned in the introduction . the aak result for the dephasing time in the weak localization can be obtained within the same calculation , but with a different cut - off procedure of the low @xmath91 singularity . one may argue that the coulomb interaction can not contribute to dephasing on distances which are larger than the phase coherence length itself , and therefore use @xmath193 as a low momentum cut - off in order to determine the phase coherence length . one then finds @xmath194 there remains the task to justify on a more formal level , the origin of the above cut - off procedure . fukuyama suggested@xcite to calculate @xmath195 at finite @xmath196 , so that the singularities in the propagator @xmath197 and the interaction @xmath198 in the calculation of the scattering rate appear at different momenta , and their product is less divergent . however this procedure does not cure the @xmath199 divergence of the interaction and thus can not justify the difference between the fa and aak results . in the next section we will demonstrate how the low momentum cut - off arises in case of weak localization within the path - integral approach of aak@xcite . we will also see that in the case of the interaction effect , no such cancellation occurs , and the dephasing rates are different in the two cases . in the previous section we calculated the inelastic scattering rate due to the coulomb interaction . we showed that the `` self - energy '' @xmath200 , which is given by the sum of the diagrams shown in fig.[fig4 ] , is responsible of various renormalization effects associated to the coulomb interaction . in particular , inelastic scattering arises from only one type of these diagrams in the low energy region @xmath201 and within the matsubara formalism is related to the existence of a branch cut contribution . in the present section we address the determination of the dephasing time , the time scale over which the quantum correction to the conductivity decays . in order to do this calculation one needs to include also vertex corrections , i.e diagrams of the type shown in fig.[fig4]e . the analysis is strongly simplified if we restrict ourselves to frequencies @xmath202 from the beginning . in this limit one neglects processes related to the diagrams a , b , c and a part of d in fig.[fig4 ] . in the case of the inelastic scattering time , as shown in the previous section , this is immaterial since it only takes contributions from @xmath203 . hence , by restricting the energy transfers to frequencies lower than the temperature some interaction effects are lost , but those relevant for inelastic scattering and dephasing are taken into account . it is useful at this point to switch again to the real - time formalism . in the limit of low frequencies , the keldysh component of the electron - electron interaction given by @xmath204 becomes dominant due to the @xmath205 and the retarded and advanced parts can be neglected . the keldysh component of the interaction has the same structure in keldysh space as an external field , and thus the interaction can be treated formally as a fluctuating external field . at the end the fluctuation dissipation theorem is used to relate the field fluctuations to the internal electromagnetic field at thermal equilibrium . with this in mind we start with the calculation of the dephasing rate for wl within a path - integral approach , following the literature@xcite . the method allows us to include the vertex corrections in an elegant way . then it will be straightforward to calculate the dephasing rate for eei . the particle - particle propagator obeys a diffusion equation whose solution can be written in terms of a path integral @xmath206 with @xmath207 the vector potential @xmath208 is defined as the sum of the vector potentials seen by the two particles entering the propagator . to make contact with the diagrammatic approach , the potential at times @xmath209 corresponds to the retarded ( advanced ) electron lines in the diagram . the vector potential describes the effect of the electromagnetic field generated by the other electrons . in the case of a fluctuating field with only quadratic correlations , one can average over the field fluctuations , leading to @xmath210 it is useful to define the average over diffusive paths as @xmath211 \rangle_{d } = \int { \cal d } { \bf r}_t [ \cdots ] \exp(-s_0 ) / \int { \cal d } { \bf r}_t \exp(-s_0 ) , \ ] ] which allows us to write the cooperon in the presence of the fluctuation field as the product of the unperturbed cooperon times a perturbation which has to be averaged over diffusive paths : @xmath212 following ref.@xcite we then make the approximation @xmath213 in the diagrammatic language this approximation is equivalent to selecting a certain subset of graphs . as realized in ref.@xcite one selects diagrams where interaction lines do not intersect and thus correspond to graphs ( d ) and ( e ) in fig.[fig4 ] . in some cases @xmath214 is linear in @xmath215 and then the dephasing time may be easily read off , @xmath216 . in general we determine the dephasing time from the condition @xmath217 for @xmath218 . we may then write @xmath219 \label{14 } , \end{aligned}\ ] ] where @xmath220 is the electromagnetic field correlator in reciprocal space . the cosine term in @xmath221 refers to correlation of vector potentials seen by the same electron at different times , while the term in @xmath222 represents the correlation of the vector potentials of two electrons . in the diagrammatic language of the previous section these latter terms have not been considered . the electromagnetic field fluctuations may be decomposed in the longitudinal and transverse parts . here we are interested in the effect of coulomb interaction and we then consider the longitudinal part only . using the fluctuation dissipation theorem one finds the electric field fluctuations for small frequencies as ( @xmath223 ) @xmath224 inserting the retarded interaction from eq.([eq15 ] ) we arrive at @xmath225 note that the fluctuation dissipation theorem has to be used in the classical limit , @xmath226 . as it is clear from the discussion in the beginning of this section it would be inconsistent to take the quantum limit @xmath227 since this would immediately introduce contributions from @xmath175 . for example in equation ( [ 6 ] ) for the inelastic scattering time this would amount to keep the @xmath228 while neglecting the @xmath229 . various claims about the saturation of the dephasing time appear to be related to this mistake @xcite . next we will integrate eq.([14 ] ) by parts by using the relations @xmath230 @xmath231 & \nonumber\\ & = -{{\omega^2}\over 4 } \left[\cos ( \omega { { t_1+t_2}\over 2 } ) -\cos ( \omega { { t_1-t_2}\over 2})\right ] & \label{17 } .\end{aligned}\ ] ] the integration over the frequency @xmath172 may be extended to infinity when considering the asymptotic behavior @xmath232 ( which involves times @xmath22 and @xmath23 which are much longer than @xmath233 ) and one finds @xmath234-functions @xmath235 as a result of inserting eqs.([15]-[18 ] ) in eq . ( [ 14 ] ) one obtains @xmath236 eq.([19 ] ) is the first central result of this section . it shows that the small @xmath91 singularity due to long range coulomb interaction is cut off . there is therefore no need to introduce the refined form eq.([eq40 ] ) for the coulomb interaction . this happens because the oscillating term , originating from correlation between different particles , cancels with the unity factor stemming from correlations on the same particle . in the diagrammatic language this is the cancellation between self - energy and vertex corrections in the small @xmath91 limit . to proceed further , one has to average the oscillating factor over the diffusive paths . however to obtain the leading behavior is enough to proceed as follows . the oscillating factor is different from zero only when @xmath237 . then one may substitute the expression in brackets with unity with the condition that @xmath238 or @xmath239 . we note that the integral over @xmath91 is divergent in the ultraviolet limit . this is because we took the integral over @xmath172 from minus infinity to plus infinity . we may correct this error remembering that for typical frequencies @xmath240 and then use this condition as an ultraviolet cut - off in the @xmath91 integral of eq.([19 ] ) . then @xmath241 which yields the aak result in eq.([eq41 ] ) . the dephasing time in the particle - hole channel can be obtained by an analogous calculation . the particle - hole propagator is given by @xmath242 with @xmath243 this differs from the expression for the cooperon by a factor two in @xmath244 , and what is most important , the external field couples differently in @xmath245 . averaging over the fluctuating vector potential and going through the same steps as before for the particle - particle propagator we find @xmath246 \nonumber\\ & & \times \left ( 2 \delta(t_1-t_2 ) - \delta(t_1-t_2-\eta ) - \delta(t_1-t_2 + \eta ) \right ) \label{eq63 } .\end{aligned}\ ] ] we set the integration limits as needed in the calculation of @xmath247 , eq.([eei ] ) , and considered the limit @xmath248 . details of the explicit evaluation of the dephasing time in the particle - hole channel are reported in appendix a. the first delta function in ( [ eq63 ] ) corresponds to correlations of the field for the same particle ( or hole ) , whereas the two other delta functions correspond to correlations between a particle and a hole . from the integration limits it follows that these do not contribute to @xmath249 . the low momentum singularity is only cut - off if we calculate the field fluctuations more accurately , @xmath250 according to the expression for the screened coulomb interaction in our calculation of @xmath157 in eq.([eq40 ] ) . we find that the dephasing rate @xmath251 in the particle - hole channel equals the inelastic rate @xmath157 , and is thus definitely not the same as the dephasing rate in the particle - particle channel . our result that the inelastic rate is equal to the dephasing rate in the particle - hole channel is non - generic and is due to the dephasing mechanism considered here . the crucial approximation involved is that the interaction is local in time . for an interaction with retardation effects the inelastic scattering rate and the dephasing rate will not be the same . with this regard we want to mention that the processes considered here are not the only intrinsic processes leading to inelastic scattering and dephasing . castellani et al @xcite calculated the dephasing rate adding to the diagrams shown in fig.[fig4 ] the corresponding hartree diagrams . the latter correspond to exchange of spin - fluctuations . in two dimensions , a contribution to @xmath251 which is linear in @xmath8 has been found , which corrects the result given here by a numerical factor . beyond electron - electron scattering also electron - phonon scattering is a further source of inelastic scattering and dephasing . often the latter is dominant at high temperature with @xmath252 . in general any type of low lying excitation with @xmath133 that couples to the conduction electrons causes inelastic scattering and dephasing . in the zero temperature limit , however , the number of such excitations goes to zero and so does the dephasing rate . we have studied non - linear effects and dephasing of the quantum corrections to the conductivity . we worked within a time domain representation . in this representation the weak localization correction to the current at time @xmath69 is related to processes where a particle and a hole start diffusing at the same time @xmath253 around a closed loop but in opposite directions . we succeeded in deriving a new expression for the interaction correction to the current which allowed us to calculate the latter in an arbitrary time dependent electromagnetic field . in contrast to weak localization , the size of the interaction correction is not related to diffusion around a closed loop . in this case the correction to the conductivity can be written in terms of the amplitude describing the propagation of a particle and an hole with the particle moving from @xmath254 at @xmath255 to @xmath256 at @xmath253 , and the hole propagating along the same path from @xmath257 at @xmath253 to @xmath256 at @xmath69 . due to these different processes contributing to wl and eei we find substantially different non - linear effects . in a static electric field , no non - linear effects were found for wl whereas in case of eei non - linear effects arise , when the voltage drop over a thermal length is of the order of @xmath258 . note that we did not take into account heating of the sample . experimental evidence for non - linear effects on eei which are not related to heating has been found several years ago by bergmann et al@xcite in gold films . qualitative arguments why this effects should exist were given in that work , but a quantitative theory was not worked out . recently non - linear effects have been observed in two dimensional systems@xcite near the metal - insulator transition . the physical origin of the metal - insulator transition and the metallic phase in these systems is not clear , but one candidate is quantum interference in the interaction triplet channel@xcite . in this case one would expect to observe in the metallic phase the non - linear effects which we have described in this paper . the sign of the effect would depend on the relative importance of the singlet and triplet contributions to the eei corrections@xcite . however , from the available experimental data deep in the metallic phase we are not able to separate non - linear effects due to heating from those due to interference , so at present we can not provide a quantitative comparison of theory and experiment . besides the effect of a static electric field we examined the quantum corrections in the presence of a microwave field . we confirmed earlier calculations on the effect of a microwave field on wl and worked out the effects on eei . a microwave field of arbitrarily low frequency can destroy eei efficiently , if the voltage drop on a thermal length is of order of the temperature , similar to the non - linear effects in a static field . a high frequency microwave field ( @xmath259 ) is less effective in destroying eei , i.e. eei increases when increasing the microwave frequency . this behavior has been observed experimentally by liu and giordano@xcite in highly disordered gold films . in those experiments the change of the dc conductivity with the amplitude of the microwave field was measured , @xmath260 , and the factor @xmath261 was found to be frequency dependent . in ref.@xcite a frequency dependence of the interaction correction to the conductivity was suggested as a possible origin of the experimental results . whereas experimental results are qualitatively in agreement with our theory ( see fig.[fig3c ] ) the experimentally observed effect is much larger ( by a factor of about @xmath262 ) than predicted theoretically . vitkalov et al@xcite studied the conductivity of si - mos transistors in the presence of a microwave field . for electron densities above @xmath263 the experimental results agreed with the theoretical considerations on weak localization in the presence of a microwave field . at lower density , @xmath264 , they found non - heating effects which were incompatible with weak localization . from the parameters given in@xcite we conclude that non - linear effects related to electron - electron interaction become important at low density as we demonstrate here below . in a weak microwave field of low frequency @xmath172 , the weak localization correction is quadratic in the electric field and given by@xcite @xmath265 . the interaction correction on the other hand is @xmath266 . inserting the parameters given in ref.@xcite at density @xmath267 , @xmath268k , @xmath269s which corresponds to @xmath270k , and @xmath271k , it is seen that wl and eei are comparable in size . note the strong dependence of @xmath272 on the dephasing time , so varying @xmath85 by two varies the @xmath273 by almost two orders of magnitude . eei dominates if the dephasing time becomes shorter , as is the case for lower electron density . in that regime non - linear effects are only weakly magnetic field dependent ( due to zeeman splitting in the spin - triplet channel which we did not consider in this paper ) , and proportional to @xmath274 , so by reducing the temperature from 4.2k to 1.7k the non - linearities are expected to increase by more than one order of magnitude . both statements agree with the experimental observations . finally , we considered dephasing due to intrinsic electric field fluctuations . for wl we confirmed the aak result , leading to @xmath275 in two dimensions . in case of eei we found , as it was suggested earlier by castellani et al@xcite , that there is dephasing also in the particle - hole channel . however in contrast to this earlier suggestion we found different dephasing rates in the particle - hole and particle - particle channels , both of which however vanish as @xmath276 . mohanty et al @xcite observed recently a saturation of the weak localization dephasing time in gold wires . it was shown in the experiments that the saturation was not due to heating or magnetic impurities . the dephasing rate saturated below @xmath277k to values of the order @xmath27810mk . we found that the interaction correction is affected by dephasing , so when weak localization saturates , also the interaction correction is expected to saturate at still lower temperature . in the samples with the highest saturation values of @xmath279 a saturation of the interaction correction to the conductivity was indeed observed@xcite below @xmath280mk . there remains the task to determine the mechanism causing saturation . clearly it can not be equilibrium electric field fluctuations . also experimental evidences are against the idea of such an intrinsic decoherence mechanism@xcite . altshuler et al@xcite suggested an external ( non - equilibrium ) microwave field as the origin of the saturation of the weak localization dephasing time . the weak localization dephasing time , @xmath281 , and the saturation temperature of eei , @xmath282 , are functions of the amplitude and frequency of the microwave field . it was pointed out by altshuler et al that a microwave field is most efficient in destroying weak localization , when the frequency is close its `` optimum '' value i.e. , the inverse dephasing time , @xmath283 , which leads to @xmath284^{1/3}$ ] . in this case we expect saturation of eei at @xmath285 . the experimental saturation temperature however is larger and roughly found as @xmath286 . this could be explained with a microwave frequency which is not `` optimum '' . by decreasing the frequency , the wl dephasing time becomes larger according to @xmath287^{1/5}$ ] , whereas the saturation temperature of eei does not depend on frequency . from @xmath288 we estimate the frequency which is consistent with the experimentally observed saturation of wl and eei as @xmath289 . on the other hand saturation of the resistance could also occur due to heating . in ref.@xcite strong heating is assumed to set in when the voltage drop over the length of the wire is of order of the temperature . from our considerations we found a saturation of the resistance when the voltage drop over a thermal length is of the order of the temperature . in the experiments the thermal length was much shorter than the system size , so the saturation of the resistance is most probably due to heating if the conditions of ref.@xcite are satisfied . this however should be checked experimentally . the observed saturation of dephasing an resistance could also be caused by some intrinsic processes , other than the equilibrium electric field fluctuations . recently it has been suggested that two - level - systems may lead to a temperature independent weak localization dephasing rate@xcite . to assess the agreement with the experiment one has to calculate the dephasing rate in the particle - hole channel using for example the formalism we developed in this article . this however is beyond the scope of the present paper . this work was supported by murst under contract no . 9702265437 ( r.r . ) and the tmr program by the european union ( p.s . ) . c.c . and r.r . acknowledge useful conversation with m. sarachik and s. vitkalov . we thank m.leadbeater for his help in the preparation of the final version of the paper . in this appendix we report the detailed calculation of the correction to the particle - hole propagator caused by electric field fluctuations . we start from the path integral representation of the diffuson , after averaging over the fluctuating field and after the approximation ( [ eq51 ] ) : @xmath290 where @xmath244 has been defined in section v. the relevant term for dephasing is @xmath249 and is given by @xmath291 \left [ 1 -\cos(\omega\eta ) \right ] .\end{aligned}\ ] ] the @xmath292 term is due to the vertex corrections . the expression we gave in eq.([eq63 ] ) is recovered when neglecting the frequency dependence of the field fluctuations and considering large times @xmath47 only , such that the @xmath172-integration leads to delta functions in time . here we do not rely on these approximations . first we average ( [ eqa2 ] ) over diffusive paths , @xmath293\rangle_d \approx \exp ( - dq^2|t_1-t_2 | ) $ ] and switch to `` center of mass '' and `` relative '' variables in the time integrations , @xmath294 and @xmath295 , leading to @xmath296 .\end{aligned}\ ] ] note that the time integration in this expression covers only half of the region of the original integral . we cured this by multiplication by two . then it is straightforward to integrate over @xmath297 and @xmath298 , @xmath299 @xmath300 \right\ } .\ ] ] for large momenta ( @xmath301 ) the second term is negligible , so ( [ eqa4 ] ) is proportional to the inverse diffusion propagator . the second term becomes important in the low momentum region , and cancels the low-@xmath91 low-@xmath172 singularity of the propagator . explicitely we find @xmath302 & { \rm for } \,\ , dq^2 > 1/\eta \\ \eta [ dq^2+(1-\cos\omega\eta)/\eta]/\omega^2 & { \rm for } \,\ , dq^2 < 1/\eta , \omega \eta > 1 \\ \eta^2/2 & { \rm for } \,\ , dq^2 < 1/\eta , \omega \eta < 1 \end{array } \right . calculating @xmath303 for large times @xmath232 we find @xmath304 from the region of the momentum integration with @xmath305 and @xmath306 . setting @xmath307 at @xmath308 reproduces the fa result for the inelastic scattering time , see eq.([eq43 ] ) . in the relevant region of the momentum integration ( @xmath309 ) the electric field fluctuations are nearly local in time . as a consequence vertex corrections are not important in this region . the opposite limit , @xmath310 , which is relevant for @xmath311 , may be more interesting since only small times @xmath47 contribute to the interaction correction to the current . in this limit the vertex corrections are important , and the dephasing time is strongly reduced below the inelastic time , since @xmath312 , compare ( [ eqa2 ] ) . however also in this region it is essential to take the refined form ( [ eq40 ] ) of the coulomb interaction in order to determine the electric field fluctuations . the vertex corrections do not cancel the low-@xmath91 singularity of @xmath313 . going explicitly through the algebra again , we find @xmath314 the relevant region in momentum space is @xmath315 , @xmath316 . from ( [ eqa7 ] ) the dephasing time in the limit where @xmath311 is determined as @xmath317 b.l . altshuler , a.g . aronov , d.e . khmelniskii , and a.g . larkin , in _ quantum theory of solids _ , edited by i.m . lifshitz ( mir publishers , moscow , 1982 ) ; b.l . altshuler and a.g . aronov , in _ electron - electron interactions in disordered systems _ , edited by m. pollak and a.l . efros ( north - holland , amsterdam , 1985 ) , p. 1 . p.a . lee and t.v . ramakrishnan , rev . 57 * , 287 ( 1985 ) ; d. belitz and t.r . kirkpatrick , rev . phys . * 66 * , 261 ( 1994 ) . d.s . golubev and a.d . zaikin , phys . rev * 81 * , 1074 ( 1998 ) . d. cohen , and y. imry , preprint , cond - mat/9807038 ; k.a . erikson and p. hedegard , preprint , cond - mat/9810297 ; m. vavilov and v. ambegaokar , preprint , cond - mat/9902127 . altshuler , m.e . gershenson , and a.l . aleiner , physica e * 3 * , 58 ( 1998 ) ; a.l . aleiner , b.l . altshuler , and m.e . gershenson , preprint , cond - mat/9808053 . a. schmid , z. phys . * 271 * , 251 ( 1974 ) . e. abrahams , p.w . anderson , p.a . lee , t.v . ramakrishnan , phys . b * 24 * , 6783 ( 1981 ) . altshuler , a.g . aronov , and d.e . khmelnitsky , j. phys . c : solid state phys . * 15 * , 7367 ( 1982 ) . h. fukuyama and e. abrahams , phys . b * 27 * , 5976 ( 1983 ) . h. fukuyama , j. phys . . jpn . * 53 * , 3299 ( 1984 ) . w. eiler , j. low temp . phys . * 56 * , 481 ( 1984 ) . a.g . aronov , physica b * 126 * , 314 ( 1984 ) . e. abrahams in _ localization and metal - insulator transition _ , edited by h. fritsche and d. adler ( plenum , new york , 1985 ) , p.433 . c. castellani , c. di castro , g. kotliar , and p.a . lee , phys . lett . * 56 * , 1179 ( 1986 ) . a. stern , y. aharonov , and y. imry , phys . a * 41 * , 3436 ( 1990 ) . u. sivan , y. imry , and a.g . aronov , europhys . lett . * 28 * , 115 ( 1994 ) ; ya.m . blanter , phys . b * 54 * , 12807 ( 1996 ) ; b.l . altshuler , y. gefen , a. kamenev , and l.s . levitov , phys . * 78 * , 2803 ( 1997 ) . lin and n. giordano , phys . b * 35 * , 1027 ( 1987 ) . pooke , n. paquin , m. pepper , and a. gundlach , j. phys . condens . matter * 1 * , 3289 ( 1989 ) . t. hiramoto et al , appl . . lett . * 54 * , 2103 ( 1989 ) . mueller , r. stasch , and g. bergmann , solid state commun . * 91 * , 255 ( 1994 ) . p. mohanty , e.m.q . jariwala , and r.a . webb , phys . * 78 * , 3366 ( 1997 ) ; p. mohanty , and r.a . webb , phys . rev . b*55 * , 13452 ( 1997 ) ; p. mohanty , e.m.q . jariwala , and r.a . webb , fortschr . 46 * , 779 ( 1998 ) . we notice that the inclusion of the vertex diagrams is also required by ward identities , which implement the particle conservation and gauge invariance . within the quasiclassical point of view inclusion of the vertex diagrams is equivalent to the statement that the phase _ difference _ of two paths determines dephasing . g. bergmann , wei wei , yao zou , and r.m . mueller , phys . b * 41 * , 7386 ( 1990 ) . kravchenko , d. simonian , m.p . sarachik , w. meson , and g.e . fourneaux , phys . lett . * 77 * , 4938 ( 1996 ) . j. yoon , c.c . li , d. shahar , d.c . tsui , and m. shayegan , phys . lett . * 82 * , 1744 ( 1999 ) . j. liu and n. giordano , phys . b * 43 * , 1385 ( 1991 ) . vitkalov , g.m . gusev , z.d . kvon , g.i . leviev , and v.i . falko , sov . phys . jetp * 67 * , 1080 ( 1988 ) . j. rammer and h. smith , rev . * 58 * , 323 ( 1986 ) . altshuler , a.g . aronov , and p.a . lee , phys . lett . * 44 * , 1288 ( 1980 ) ; b.l . altshuler , d. khmelnitzkii , a.i . larkin , and p.a . lee , phys . b * 22 * , 5141 ( 1980 ) . c. castellani , c. di castro , p.a . lee , and m. ma , phys . b * 30 * , 527 ( 1984 ) . c. castellani , c. di castro , and p.a . lee , phys . b * 57 * , r9381 ( 1998 ) . in this paper we have not considered the quantum interferences in the interaction triplet channel . however we can obtain analogous results in the triplet channel as those discussed in this paper for the singlet channel . khavin , m.e . gershenson , and a.l . bogdanov , phys . 81 * , 1066 ( 1998 ) . y. imry , h. fukuyama and p. schwab , preprint , cond - mat/9903017 . a. zawadowski , j. von delft , and d.c . ralph , preprint , cond - mat/9902176 .

the theory to reconstruct the atomistic - level chain diffusion from the accelerated dynamics that is measured in mesoscale simulations of the coarse - grained system , is applied here to the dynamics of _ cis_-1,4-polybutadiene melts where each chain is described as a soft interacting colloidal particle . the rescaling formalism accounts for the corrections in the dynamics due to the change in entropy and the change in friction that are a consequence of the coarse - graining procedure . by including these two corrections the dynamics is rescaled to reproduce the realistic dynamics of the system described at the atomistic level . the rescaled diffusion coefficient obtained from mesoscale simulations of coarse - grained _ cis_-1,4-polybutadiene melts shows good agreement with data from united atom simulations performed by tsolou et al . the derived monomer friction coefficient is used as an input to the theory for cooperative dynamics that describes the internal dynamics of a polymer moving in a transient regions of slow cooperative motion in a liquid of macromolecules . theoretically predicted time correlation functions show good agreement with simulations in the whole range of length and time scales in which data are available . the theory provides , from data of mesoscale simulations of soft spheres , the correct atomistic - level dynamics , having as solo input static quantities .

1301.0166

computer simulations of macromolecular liquids described at the atomistic level are extremely useful because they bridge information between the microscopic molecular structure of the polymeric system , at the given thermodynamic conditions , and its macroscopic properties , such as viscosity , diffusion , and dynamical - mechanical response , which are observed experimentally.@xcite it is unfortunate that , despite the progress already occurred in the computational hardware and software , the study of the dynamics of polymeric liquids by atomistic computer simulations is still limited by the impossibility of simulating polymer dynamics in the wide range of time and length scales relevant for these systems . it is known that long simulation trajectories deteriorate with time , following a law defined by the lyapunov exponent characteristic of the system , and the resulting long - time dynamics is affected by errors.@xcite atomistic simulations of polymeric liquids are limited either in the length and number of macromolecules , or in the maximum timescale that can be reached . the long timescale is a regime of great interest for these systems , because of the relevance of dynamical - mechanical properties of macromolecules in that regime for their industrial applications . because the limitation concerns only the region of long - time , large - scale properties , it is possible to shift the focus of a simulation to this region of interest by the reduction of the simulated degrees of freedom through the averaging of local scale properties , i.e. adopting a so - called coarse - graining " procedure . while simulations of coarse - grained systems afford to represent larger length- and longer time - scales than atomistic simulations , they also are marred by the shortcoming that the measured dynamics is unphysically fast . this is the advantage that allows one to perform simulations in the long - time regime , but it is also a problem because the dynamics in the simulation becomes too fast and measured diffusion coefficients are too large . depending on the level of coarse - graining the dynamics can become orders of magnitude faster than the atomistic dynamics.@xcite being aware that the dynamics has to be rescaled to recover the correct timescale,@xcite one resorts to the most conventional method , which is to build a calibration curve calculated by superimposing the long - time dynamics of the coarse - grained and atomistic simulations.@xcite the hope is for this calibration curve to be transferable to other systems , or to be identical , and so applicable , for thermodynamic conditions close to , but different than the phase point for which the curve was optimized in the first place . otherwise , if an atomistic simulation has to be ran every time we need to rescale a coarse - grained simulation there is no advantage in running simulations of coarse - grained ( cg ) systems . the rescaling factor of the conventional calibration curve is , however , purely a numerical correction . there is no physical motivation for one to assume that this correction has to be identical for any system or thermodynamic condition different from the ones in which it was optimized . given that the effective potentials between coarse - grained units are free energies , and as such they are parameter dependent , it is very unlikely that numerically optimized corrections will be identical for different systems . however , if the calibration curve is applied for conditions of temperature , density , and degree of polymerization , very close in the phase diagram to the ones in which the calibration curve was originally optimized , there is hope that the error is very contained . alternatively , in the case that very limited levels of coarse - graining are applied , e.g. combination of a few atoms to obtain a new coarse - grained unit , the error incurred in using the calibration curve in different conditions , can be small enough to be within the numerical error of the calculation . this is in fact the reason for the observed good agreement in the rescaled dynamics of systems with contained level of coarse - graining.@xcite the goal of this paper is orthogonal to the more conventional calculation of an optimized calibration curve for systems with limited level of coarse - graining . in fact our goal is to investigate the physical motivation that leads to the accelerated dynamics in first place and to provide , through our analysis , a theoretical approach to calculate , from first principle equations , the correction factor that needs to be applied to the fast dynamics measured in coarse - grained simulations to recover the slower and more realistic dynamics of the atomistic description . we study the most extreme level of molecular coarse - graining , where one polymer chain is represented as a soft sphere , because i ) the description is analytical allowing the formal study of the rescaling problem , ii ) it provides a solid test of the reconstruction procedure because possible errors would be maximized , iii ) this level of coarse - graining is associated with the largest computational gain , so it is important . recently we have proposed an original procedure to rescale the dynamics measured in simulations of coarse - grained polymer melts.@xcite the procedure we proposed does not require to run atomistic simulations to calibrate the dynamics of the coarse - grained simulation . once one parameter , specifically the diameter of the hard - core monomer potential , is fixed , the rescaling factors are fully determined . the advantage of this method is clear , given that the coarse - grained simulation can reach larger - scale longer - time properties than the atomistic simulation . because no atomistic simulations are needed , the measured dynamics can be directly rescaled to obtain the dynamics of the real system we wanted to study . the method has been formally derived , and then applied to systems of polyethylene ( pe ) melts . the choice of pe as a test system was motivated by the wealth of experimental and simulation data available in the literature . we have shown that the proposed theory is able to predict diffusion coefficients in agreement with both atomistic simulations and experiments for systems in both the unentangled and entangled regimes , for a range of temperatures and densities . for entangled system the approach of dynamical reconstruction includes an extra loop in the calculation of the friction coefficient , which accounts for the fact that in real systems described at the atomistic level , the dynamics is slowed down by the presence of entanglements . we hypothesized for the entanglements to relax in time following the same dynamical mechanism of single chain interdiffusion . in this paper entangled chain dynamics is not investigated as most of the samples in the atomistic simulations are either unentangled or in the region of crossover from the unentangled to the entangled regimes . one of the questions that still needed to be answered was if our method was able to generate high quality predictions also for macromolecular liquids with a more complex monomeric structure than the simple polyethylene . a first study of this issue is presented in this manuscript which investigates the agreement between theoretically predicted diffusion coefficients for _ cis_-1,4-polybutadiene ( pb ) chains and data from molecular dynamics simulations performed by tsolou , mavrantzas and coworkers.@xcite the approach is completely general and can be applied to polybutadiene chains with different tacticity . for these systems the difference in local chain semiflexibility , which corresponds to different overall chain dimension , i.e. a different radius - of - gyration ( @xmath0 ) and statistical segment length ( @xmath1 ) , ultimately leads to a different diffusion coefficients for polybutadienes with different tacticity . in this paper we limit our calculations to _ cis_-1,4-polybutadiene samples because those are the samples investigated by atomistic simulations and provide a effective test of our procedure for dynamical reconstruction . our rescaling method considers two contributions emerging from the consequence of applying the coarse - graining process and derives the theoretical corrections that need to be applied to the dynamics of the coarse - grained description to recover the modalities of the original atomistic system . to calculate the analytical correction we adopt for the atomistic descriptions a simple bead - and - spring description of the chain where each pb polymer of @xmath2 monomeric units is represented as a collection of @xmath2 beads connected by semi flexible springs of length @xmath3 , and the overall chain dimension is defined by the radius - of - gyration of the molecule , @xmath0 . this model is a rouse approach modified to include chain semiflexibility and has been shown to represent well the diffusion of polymer melts in the unentangled regime.@xcite from the comparison of this model and the soft sphere representation of the chain , it is possible to evaluate the two contributions that are responsible for the accelerated dynamics , and which have to be included to correct the measured dynamics and bring it to the realistic values of the atomistic ( here bead - and - spring ) description . the first correction emerges from the consideration that due to the process of averaging the intramolecular degrees of freedom , the coarse - grained system experiences a change in its entropy , because a number of local states are neglected . while the atomistic system devotes time to explore local configurations , the coarse - grained system does nt . to recover the correct dynamics those states need to be included _ a posteriori _ in the form of an entropic contribution to the free energy that rescales the time measured in the cg simulations . the second correction comes from the change of the shape " of the molecule once it is coarse - grained . macromolecules described at the atomistic level become represented by chains of effective atoms , or as bead - and - springs , chains of soft blobs , or even each molecule as a soft sphere . in all these multiple forms the surface of each molecule available to the surrounding molecules , i.e. the solvent " , is different . the hydrodynamic radius , @xmath4 , of the coarse - grained unit in each description is different and so is its friction coefficient , @xmath5 , defined by stokes law as @xmath6 , where @xmath7 is the viscosity of the medium . by solving the memory function for the friction coefficient in each description we are able to calculate the correction factor that has to be applied to each friction coefficient to recover the description at a different level of coarse - graining . in this paper our rescaling method is first briefly summarized and then applied to study the dynamics of cis-1,4-pb and compared with simulation data by tsolou et.al.@xcite who also investigated the effect of temperature and pressure on the simulations.@xcite the global dynamics of the polymer is represented by the diffusion coefficient and the decay of the self - correlation for the end - to - end molecular vector . these dynamical quantities are reconstructed using the method by lyubimov at al.,@xcite from the information contained in the trajectories of the mesocale simulations of the coarse - grained polymer melts and show good agreement with the data from united atom simulations . for the internal dynamics the theory for cooperative motion@xcite is used to predict monomer dynamics and the dynamical structure factors , which compare well with the simulations . if atomistic simulations are not available , the theory of dynamical reconstruction is able to predict the correct dynamics from the rescaling of the mesoscale simulations having as an input the polymer radius - of - gyration , which can be calculated from simple structural models for polymers . the accuracy of adopting a freely rotating chain model for the calculation of the polymer radius of gyration is discussed in the last section . a brief summary concludes the paper . in this section we briefly present an overview of the main steps in our procedure . a complete and detailed presentation of the same has been already published@xcite and will not be repeated here . we start from the consideration that the procedure of coarse - graining corresponds to eliminating some internal degrees of freedom , by combining groups of atoms into one effective cg unit . specifically , our rescaling theory has been developed for an extended level of coarse - graining where intramolecular coordinates are fully averaged out and the polymer is represented as an isotropic sphere centered on the center - of - mass ( com ) of the macromolecule . there are some advantages in the choice of this description . first , the cg representation is fully isotropic , and even if it is known that the shape of a polymer is closer to an ellipsoid than to a sphere,@xcite the total correlation function of the polymeric liquid , i.e. the structure of the polymeric liquid , statistically averaged over all the possible configurations is well reproduced by this model ; second the formalism is analytical , which allows us to derive formally many of the physical quantities of interest for both the static and the dynamic properties of the coarse - grained system as a function of the atomistic description ; third all the consequences of the coarse - graining procedure are enhanced with this extreme level of coarse - graining and are easier to study in such a description . if the procedure adopted were not precise , the error would be maximized because of the high level of coarse - gaining ; the fact that the theory , which is predictive , is found to be in quantitative agreement with both simulated and experimental data is encouraging.@xcite one disadvantage of having such a coarse - grained representation of the molecule is that no information is collected from the mesoscale simulation on the internal dynamics of the polymer . in this way , the soft sphere representation only provides information on the diffusion coefficient of the molecule . however , we show in this paper that from the knowledge of the diffusion coefficient and the related monomer friction coefficient for unentangled chains it is possible to recover correctly the internal dynamics of the chain by applying our theory for the cooperative dynamics of a group of interacting chains , which calculates the single chain dynamics for a macromolecule whose dynamics is coupled by the presence of intermolecular forces with the other interpenetrating macromolecules in the liquid . in every cg model , the averaging of the intramolecular degrees of freedom leads to a speeding up of the dynamics . when a polymer is represented as a soft sphere the dynamics in the mesoscopic simulation is orders of magnitude faster than in the atomistic description . one of the reasons for this acceleration is the fact that while a macromolecule needs to explore a large number of internal chain configurations for each position of the center - of - mass , the soft sphere instead is free to move in the three - dimensional space undergoing brownian motion.@xcite to take into account the missing contribution of the time necessary to explore the internal degrees of freedom , the time measured in the mesoscale simulation of the soft sphere is multiplied by a rescaling factor that is calculated as the ratio of the energy due to the internal degrees of freedom in the two representations , i.e an atomistic and a soft sphere representations . for the atomistic representation we adopt an analytical bead - and - spring model , as previously described.@xcite this model has been shown to provide a quantitative description of the dynamics for polymeric liquids and for proteins in theta solvent.@xcite the first correction term is calculated from the definition of the internal energy for a simple bead - and - spring model and for the soft sphere cg model . the mesoscale ( ms ) molecular dynamic ( md ) simulations are performed using a potential expressed in dimensionless units , which is a common procedure . the time from the simulations has to be properly normalized so that the dimensionless simulation time from ms simulations , @xmath8 , once dimensionalized and rescaled reads as @xmath9 where @xmath10 is the molecular mass of one chain and @xmath0 is the radius - of - gyration of the molecule , which is an input quantity in our approach . the internal degree - of - freedom averaged out in the coarse - grained description are accounted for by introducing the @xmath11 factor where @xmath2 is the number of beads . the second rescaling is due to the change in the shape of the molecule represented in the two levels of coarse - graining , and so the effective friction in the two representations . in the long time the com mean - square displacement is related to the molecular diffusion coefficient as @xmath12 where @xmath13 is the diffusion coefficient of the system , which has to be calculated . if we use as a starting point the diffusion coefficient measured in the mesoscale simulation @xmath14 , the mean square displacement in the rescaled formalism is give by @xmath15 where the problem reduces to the calculation of the friction coefficient in the monomer and soft - sphere representations . the friction coefficients are calculated from the solution of the memory function in the two representations ( bead - and - spring and soft - sphere ) . the formalism presented so far , adopts a rouse - like description of the single chain diffusion for the atomistic level representation of the system . at the coarse - grained level the chain is represented as a soft - sphere . eq.[eq : rescaling ] shows how the diffusion coefficient measured in a mesoscale simulation of a liquid of soft - spheres , @xmath16 , needs to be rescaled to give the mean - square displacement of the real chain , @xmath17 , as a function of the rescaled time , @xmath18 . the contributions that are still unknown are the friction coefficients in the atomistic and coarse - grained descriptions , which can be solved analytically starting from the definition of the memory function.@xcite for a bead - and - spring model the monomer ( bead ) friction is given by : @xmath19 where the memory function is defined as@xcite @xmath20 where @xmath21 , @xmath22 is the monomer density and @xmath23 , @xmath24 are the unit vectors characterizing the direction of the forces acting on monomers @xmath25 and @xmath26 , respectively . @xmath27 is the projected dynamic structure factor , which includes both intra- and inter - molecular contributions , i.e. incoherent and coherent scattering , and is approximated by its non - projected form as @xmath28 , which is an acceptable approximation when the langevin equation is expressed as a function of the slow variables.@xcite in eq.([eq : mfmon ] ) @xmath29 is the monomer pair distribution function of the molecular liquid , and @xmath30 is the force due to the effective potential between two monomers , obtained from the solution of the ornstein - zernike ( oz ) equation by applying the percus yevick closure . the monomer potential , which in the atomistic - level simulation is a lennard - jones potential , is approximated here by an effective hard - sphere with an effective diameter , @xmath31 , to mimic the properties of the l - j potential . all the physical quantities that appear in the equation are known and the analytical expression for the monomer friction is @xcite @xmath32 \nonumber \\ & & + \rho \pi n h_0 \frac{1}{3\sqrt{2}(r_g^2 - 2\xi_\rho^2)^2 } \biggl [ 12\sqrt{2}\xi_\rho^7 + 12 d^4 r_g^3 \left(1 - 2\left(\frac{\xi_\rho}{r_g } \right)^2\right ) \nonumber \\ & & + 4\sqrt{2 } d^3 r_g^4 \left ( 5 - 14 \left(\frac{\xi_\rho}{r_g}\right)^2 + 2 \left(\frac{\xi_\rho}{r_g}\right)^4 \right ) + 3 d^2 r_g^5 \biggl ( 5- 14 \left(\frac{\xi_\rho}{r_g}\right)^2 - 4\sqrt{2 } \left(\frac{\xi_\rho}{r_g } \right)^5 \biggr ) \nonumber \\ & & - 12\sqrt{2 } e^{-\frac{2d}{\xi_\rho } } \xi_\rho^7 \left(1+\frac{d}{\xi_\rho } \right)^2 \biggr ] + \rho^2 \pi h_0 ^ 2 \frac{1}{12(r_g^2 - 2\xi_\rho^2)^3 } \bigg [ 40 d^3 r_g^6 + 15\sqrt{2 } d^2 r_g^7 \nonumber \\ & & - 24\sqrt{2}d^4 r_g^3 \xi_\rho^2 - 144 d^3 r_g^4 \xi_\rho^2 + 6 \sqrt{2 } d^4 r_g^5 \left(2 - 9 \left(\frac{\xi_\rho}{d}\right)^2 \right ) \nonumber \\ & & + 12 r_g^2 \xi_\rho^7 \left(4 \left(\frac{d}{\xi_\rho}\right)^3 - 7 \left(\frac{d}{\xi_\rho}\right)^2 + 9 \right ) - 8\xi_\rho^9 \left(4 \left(\frac{d}{\xi_\rho}\right)^3 - 9 \left(\frac{d}{\xi_\rho } \right)^2 + 15 \right ) \nonumber\\ & & - e^{-\frac{2d}{\xi_\rho } } 12 \xi_\rho^4 ( d + \xi_\rho ) \left ( r_g^2 ( d+3\xi_\rho)(2d+3\xi_\rho)-2\xi_\rho^2(2d^2 + 5\xi_\rho d + 5 \xi_\rho^2 ) \right ) \biggl ] \biggl ) \ . \end{aligned}\ ] ] this equation contains quantities that are well defined once the system of interest is selected . for example @xmath33 is the pair distribution function at contact , @xmath34 is the density fluctuation correlation length with @xmath22 , @xmath35 and @xmath2 already defined in the text , and @xmath36 is defined as @xmath37 . the only physical quantity that needs to be determined is the hard - sphere diameter , @xmath31 . this is defined once the lennard - jones potential is mapped onto an effective repulsive hard - sphere system as described later in this section of the paper . in the soft colloidal representation , the friction coefficient is analytically calculated from the definition of the memory function as @xmath38 where @xmath39 is the chain density . all the other physical quantities that appear in the integral , for example the pair distribution function , @xmath29 , and its related force , @xmath30 , are now defined in relation to the description of the polymeric liquid as a liquid of soft spheres , which are point particles interacting through a soft repulsive potential . each of these quantities is analytical and calculated from the solution of the oz equation with the hypernetted chain closure approximation.@xcite the regime of interest in our calculations is the diffusive limit where , in reciprocal space , the wave vector @xmath40 . here the dynamic structure factor can be approximated as @xmath41 with the total correlation function of the soft - sphere representation is given in the limit of long chains ( @xmath42 ) by the approximated expression@xcite @xmath43 the resulting equation for soft particle friction is @xcite @xmath44 \ , , \label{eq : zeta_soft}\end{aligned}\ ] ] where the value of each physical parameter ( temperature and density ) and molecular parameters ( degree of polymerization and radius - of - gyration ) that enter this equation is defined once the system to simulate is selected . these parameters are identical to the ones that are input to the equation of the monomer friction , eq.([eq : zeta_mon ] ) , given that the two equations are representations for different levels of coarse - graining of the same system . when the two equations for the monomer and the soft colloidal friction coefficients are introduced in eq.([eq : rescaling ] ) the diffusion coefficient for a single chain in a liquid is recovered as @xmath45 where we have assumed that the diffusion coefficients that appear in eq.[eq : zeta_mon ] and eq.[eq : zeta_soft ] are identical , as the long - time relaxation of the dynamic structure factor in both cg descriptions is guided by molecular diffusion . to summarize , as the first step the thermodynamic and molecular structure , i.e. the radius of gyration or equivalently the persistence length , of a polymer melt has to be defined . then , ms md simulations are performed for a liquid of point particles interacting through a soft pair potential , @xmath46 , as described in several of our previous papers , where @xmath47-c_{soft}(r ) \ , \label{eq : v(r)}\ ] ] with @xmath48^{-1 } \ .\ ] ] the mean - square - displacement of the soft spheres is measured from the ms md trajectories as a function of time ( dimensionless ) , and the resulting diffusion coefficient is rescaled following eq.[eq : diff ] to obtain the reconstructed diffusion coefficient . for the solution of the rescaling factors in eq.[eq : diff ] the only parameter that has to be defined is the value of the effective hard sphere diameter , @xmath31 , in eq.[eq : zeta_mon ] . in our formalism the elementary interaction between monomers belonging to different chains , which in the ua md is a lennard - jones potential , is approximated by a hard - sphere interaction with an effective diameter , @xmath31 , to allow for the analytical solution of the monomer memory function . under fixed thermodynamic conditions , the value of @xmath31 should depend only on the local monomer structure , and be independent of the degree of polymerization , because the monomer interaction potential is a local property of the chain . therefore in our model , once the value of @xmath31 is chosen , this value is kept fixed for all samples with different molecular weights . note that for convenience we use the term monomer to identify the @xmath49 group , where @xmath50 or 3 . to evaluate the value of @xmath31 we analyze the dimensionless quantity @xmath51 , which is identical to @xmath52 if the chain obeys strictly rouse dynamics in the long - time limit . to fulfill this requirement the chain has to have @xmath2 much smaller than the entanglement degree - of - polymerization @xmath53 and larger than the value of @xmath54 for which chains start to obey gaussian statistics , i.e. the central limit theorem applies . ) . a ) _ cis_-1,4-polybutadiene samples with n= 32 , 56 , 128 , 320 ( solid , dashed , dot - dashed , and dotted lines correspondingly ) and parameters as reported in table [ tb : ua_rg ] and ref . @xcite . b ) polyethylene samples with n=30 , 44 , 96 , 270 ( solid , dashed , dot - dashed , and dotted lines correspondingly ) and parameters as reported in refs . @xcite and @xcite . horizontal lines represent @xmath55 values , following the diffusion coefficient for unentangled chains , @xmath56 . ] figure [ fg : zmon_of_d ] a ) displays the dimensionless friction coefficient , @xmath57 , from eq.([eq : zeta_mon ] ) as a function of the hard sphere diameter , @xmath31 , for pb samples of increasing degree of polymerization , @xmath2 . the parameter values ( @xmath58 , @xmath59 , @xmath2 , @xmath0 ) are chosen to match those of ua md simulations in ref . @xcite as reported in table [ tb : ua_rg ] . horizontal lines represent the @xmath55 value for each sample , which is the rouse value of the physical quantity @xmath56 in the diffusive regime , and reproduces the scaling behavior of the diffusion for unentangled , short - chains . for comparison figure [ fg : zmon_of_d ] b ) displays the analogous plot for the pe samples , which is reproduced from refs.@xcite and @xcite . the change of @xmath31 with @xmath2 observed in figure [ fg : zmon_of_d ] is a consequence of two different effects : for short chains , end effects are important , while for long chains the crossover to entangled dynamics starts to be important . in this way , it could be argued that in our method the actual fitting parameter is not @xmath31 , but the length of the chosen sample for which @xmath31 is determined . clearly when entangled chains are forced to behave according to the rouse expression the hard sphere diameter needs to be artificially decreased in order to compensate the overlooked increase in the monomer friction , which is associated with entanglement effects growing with increasing @xmath2 . whereas the range of lengths in the unentangled region can be rather wide , it is always possible to choose the chain length knowing experimentally measured or theoretically estimated value of @xmath53,@xcite and by optimizing our choice of @xmath31 by testing the predictions of the theory against data from simulations of short chains . however comparison with simulations is not necessary , as the only needed information is the value of @xmath53 and the value of @xmath0 for a sample of short , unentangled , chains ( @xmath60).@xcite in our previous study of pe melts we fixed the hard sphere diameter following the same procedure described here , and we selected a sample with @xmath61 to calculate @xmath31 . the obtained value of @xmath62 ( see figure [ fg : zmon_of_d]b ) , was larger than the carbon - carbon bond length @xmath63 , but smaller than the lennard - jones parameter @xmath64 typically used in ua md simulations , which seems to be a reasonable choice of the parameter . compared to pe , pb chains are more flexible and so they are less entangled . the estimated entanglement degree - of - polymerization , @xmath53 , is calculated from the generalized formula for the plateau value in the shear relaxation modulus,@xcite which gives @xmath65 backbone carbon atoms for pb , compared to @xmath66 for pe . notice that the value of @xmath53 from shear measurements is different from the value obtained from the analysis of scattering experiments which gives , for example , for pe samples @xmath67 . it is important to notice that even if there is some freedom in choosing the reference sample , the deviations of @xmath31 from the chosen value is not large . however , small deviations of the value of @xmath31 can lead to different values of the diffusion coefficient . figure [ fg : dofn ] displays a comparison of the predicted diffusion coefficient as a function of @xmath2 when different values are chosen for the hard - sphere parameter @xmath31 . it is possible to see that for values of @xmath31 obtained by enforcing rouse dynamics for three different chain lengths that obey diffusive rouse dynamics in the long - time regime , the predicted diffusion coefficients are all in reasonable agreement with the values of the ua md simulation . , calculated by enforcing rouse diffusive behavior for n=56 ( d=1.4672 ) , 64 ( d=1.4051 ) , or 80 ( d=1.374 ) ( circles , diamond , triangles correspondingly ) leads to diffusion coefficients in good agreement with the ua md simulation data ( filled squares ) . ] notwithstanding the fact that for pb there is a wide range of degrees - of - polymerization , @xmath68 , for which the hard - sphere diameter , @xmath31 , could be optimized , the best agreement between theory and simulations is obtained for values of the parameter @xmath31 optimized for short chains , @xmath69 . this can be explained by considering that in pb samples the crossover region from unentangled to entangled dynamics is extended ( see for example figure [ fg : dofn_pb ] , where deviations from the @xmath70 scaling occur already for @xmath71 ) . samples with @xmath2 approaching @xmath72 are not following the unentangled scaling behavior , which is necessary condition for the optimization of @xmath31 using rouse - like diffusion . considering the broader range of unentangeled pb chains , we select for the calculation of the hard sphere diameter the pb sample with @xmath73 . the obtained value of @xmath74 ( which is smaller than the value for pe of @xmath75 given the flexibility of pb ) is consistent with the fact that in pb chains half of the carbon bonds are shorter double bond and @xmath76 . the solution of eq.([eq : rescaling ] ) gives the diffusion coefficient for the center - of - mass of a polymer described at the atomistic level , having as an input the diffusion coefficient calculated from the mesoscale simulation of the cg polymer liquid . the purpose of this procedure is to have ultimately mesoscale simulations that , once properly rescaled , can provide directly the diffusion coefficient , without need of running atomistic simulations . to test our procedure we first run the mesoscale simulations , then we rescale the diffusion coefficient , and finally we compare the predicted values of @xmath77 against atomistic simulations and/or experiments . thermodynamic and molecular parameters entering our equations have to be consistent with the parameters of the system against which we compare our approach . we study the dynamics of _ cis_-1,4-polybutadiene ( pb ) melts of increasing degree of polymerization and compare our results with the data from the simulations performed by tsolou et al.@xcite the input parameters to our theory are displayed in table [ tb : ua_rg ] , including the degree - of - polymerization , @xmath2 , and the monomer density @xmath22 . table [ tb : ua_rg ] also includes the mean - square - radius - of - gyration calculated from the united atom simulations @xmath78 . from this radius - of - gyration the value of the semiflexibility parameter , @xmath79 , is derived , as the value that corresponds to chains with the desired overall dimension , i.e. @xmath0 . in this way the parameter @xmath79 is not an independent parameter but is determined by the @xmath0 value . the parameter @xmath79 does not enter the equation , but is used in the calculations reported in the last section of this paper . it is important to notice that the values for @xmath0 could be taken from experimental data , and that , in principle , no atomistic simulations of the system under study are necessary for our approach . the ua md simulations were performed in the @xmath80 ensemble , at the constant pressure , @xmath81 , and temperature , @xmath59 , reported in the table.@xcite note that the values of @xmath0 for systems pb160 , pb200 , pb240 were corrected from 257@xmath82 to 275@xmath82 , from 304@xmath82 to 340@xmath82 and from 401@xmath82 to 410@xmath82 correspondingly , after consultation with the authors of ref.@xcite . the correction was motivated by the inconsistency observed with the data reported in reference @xcite when the calculation of @xmath0 is performed using a standard model , such as the freely - rotating - chain , described later in in this paper . lccc & @xmath22[sites / @xmath83 & @xmath78[@xmath82 ] & g + pb 32 & 0.0352375 & 45@xmath855 & 0.6237 + pb 48 & 0.0363767 & 70@xmath855 & 0.6243 + pb 56 & 0.0367159 & 85@xmath857 & 0.6339 + pb 64 & 0.0369744 & 95@xmath8510 & 0.6246 + pb 80 & 0.0373425 & 125@xmath8510 & 0.6374 + pb 96 & 0.0375921 & 152@xmath8516 & 0.6394 + pb 112 & 0.0377723 & 184@xmath8515 & 0.6490 + pb 128 & 0.0379087 & 215@xmath8518 & 0.6544 + pb 140 & 0.0379910 & 234@xmath8518 & 0.6523 + pb 160 & 0.0381012 & 275@xmath8525 & 0.6595 + pb 200 & 0.0382567 & 340@xmath8520 & 0.6551 + pb 240 & 0.0383610 & 410@xmath8530 & 0.6557 + pb 320 & 0.0384923 & 576@xmath8530 & 0.6695 + pb 400 & 0.0385714 & 678@xmath8530 & 0.6519 + + [ tb : ua_rg ] in the mesoscale simulations of a polymer melt , each chain is represented as a point particle interacting through a soft - core potential derived from the solution of the ornstein zernike equation applying the hyper netted chain closure . ms md simulations were implemented in the microcanonical @xmath86 ensemble on a cubic box with periodic boundary conditions . we used reduced units such that all the units of length were scaled by @xmath0 ( @xmath87 ) and energies were scaled by @xmath88 . more details of our simulation procedure have been reported in previous papers.@xcite table [ tb : dofn_ua ] reports the diffusion coefficient directly calculated from the ms md in the soft sphere representation , @xmath89 , and the diffusion coefficient calculated using the dynamical reconstruction procedure , @xmath77 . finally , for comparison , the table shows the diffusion coefficient measured in the united atom simulation , @xmath90 . the diffusion coefficient measured in the mesoscale simulations is several orders of magnitude faster than in the atomistic simulations . however , once it is rescaled the diffusion coefficient becomes very similar to the one obtained directly from the atomistic simulation . it is important to notice that , once the parameter @xmath31 , which is characteristic of the polymer considered , is determined the diffusion coefficient is calculated without any input from the dynamics of the atomistic simulations , so the procedure is predictive . the samples here are relatively short and these calculations are performed for unentangled and slightly entangled systems . for strongly entangled systems , we proposed in a previous paper a perturbative version of our approach that accounts for the fact that both the tagged chain and the surrounding chains , that provide the entanglements , relax following the same dynamics.@xcite in this manuscript the pb samples are in the unentangled and slightly entangled regimes , which are well represented by the theory without the one - loop perturbation . lccc & @xmath91[@xmath82/ns ] & @xmath77[@xmath82/ns ] & @xmath90[@xmath82/ns ] + 32 & 3875 & 184.5 & 152.2 + 48 & 3400 & 68.1 & 69.4 + 56 & 3425 & 51.9 & 46.4 + 64 & 3275 & 37.3 & 40.1 + 80 & 3174 & 25.1 & 24.9 + 96 & 3114 & 18.0 & 23.2 + 112 & 3235 & 15.1 & 16.7 + 128 & 3283 & 12.6 & 15.5 + 140 & 3221 & 10.5 & 12.0 + 160 & 2963 & 9.1 & 11.0 + 200 & 3245 & 6.0 & 6.6 + 240 & 3012 & 4.5 & 5.4 + 320 & 3386 & 3.2 & 2.6 + 400 & 3016 & 1.8 & 1.6 + + [ tb : dofn_ua ] . diffusion coefficients reconstructed from ms md by applying our procedure ( open symbol ) are compared against ua md data ( filled symbol ) from reference @xcite . in analogy with the figure from the source , three scaling regimes in terms of power law dependence of @xmath92 are shown as dashed ( @xmath93 ) , solid ( @xmath94 ) , and dot - dot - dashed ( @xmath95 ) lines . ] on figure [ fg : dofn_pb ] the diffusion coefficient is presented as a function of degree of polymerization @xmath2 . filled symbols represent ua md simulations from ref . @xcite and open symbols represent ms md simulations rescaled as in eq.([eq : rescaling ] ) . in analogy with the figure in ref.@xcite from which the ua md data are taken , we report three scaling regimes in terms of power law dependence of @xmath96.for @xmath97 , there is a faster than rouse regime with @xmath77 with @xmath98 , which is attributed to the free - volume effects due to chain ends which is significant only for very short chains @xcite . in the intermediate regime for @xmath99 a rouse - like behavior can be observed where the scaling exponent @xmath26 is close to 1 . for longer chains with @xmath100 the value of @xmath101 is typical of the crossover to reptation dynamics . . ] the data in this plot are not closely following the reported scaling exponents because as the degree of polymerization increases the density of the system changes ( fig.[fg : dens_of_n ] ) . the different regimes are even less prominent in the ms md simulations , where we observe a smoother transition between regimes than in ua md data . in our calculations features like free volume effects due to chain ends ( causing faster than rouse decay of the diffusion coefficient for short chains ) enter only indirectly as a consequence of the interplay of the input parameters , i.e. @xmath22 , @xmath0 , @xmath2 . the increasing of the density with @xmath2 enters our analytical equation both directly , through the liquid density , and indirectly through the effective radius of gyration of the polymer , given that the compactness of the chain changes with the density ( see also section [ frc ] of this paper ) . the _ cis_-1,4-pb chains are more flexible in comparison to pe chains , therefore we should expect larger entanglement length @xmath53 ( for pe from scattering @xmath102 ) . figure [ fg : dofn_pb ] shows that crossover from rouse to reptation regime starts at @xmath103 . the largest system shown on figure [ fg : dofn_pb ] is for @xmath104 which is still a very weakly entangled system ( 1 or 2 entanglements per chain ) . overall the agreement between simulated and reconstructed diffusion coefficients is quite good . the coarse - grained model adopted in this paper simplifies the macromolecular structure by reducing the description of the molecule to an effective site that is centered on the center - of - mass of the molecule . for this reason , the only dynamics that can be predicted by this cg description , through the rescaling procedure , is the long - time diffusive behavior of the center - of - mass of the molecule . however from the knowledge of the diffusion coefficient the monomer friction coefficient can be obtained , and the latter can be used as an input to the langevin equation that describes the internal dynamics of the polymer chain at any length scale of interest . specifically we follow our approach for the cooperative dynamics of a group of macromolecules , or cooperative dynamics generalized langevin equation ( cdgle).@xcite conventional theories of polymer dynamics , such as the rouse theory for unentangled polymer dynamics and the reptation " model for entangled dynamics , predict diffusive motion at short time following the ballistic regime ; this is however in disagreement with simulations and experiments , which show a sub - diffusive regime before the com starts to follow brownian motion . both theories are mean - field approaches of single chain dynamics in an uniform bath.@xcite our approach focuses on the dynamics of a group of polymer chains in a melt and relates the anomalous subdiffusive behavior to the presence of cooperative motion of a group of polymer chains in a dynamically heterogeneous liquid , as observed in simulations data of polymer melts and experiments.@xcite theoretical predictions of the com and monomer mean - square displacements are shown to be in quantitative agreement with computer simulations of unentangled@xcite and slightly entangled polymer melts,@xcite and with scattering experiments of neutron spin echo.@xcite the physical picture underlying the theory builds on the fact that in polymer melts the dynamics appears heterogeneous with the motion of a tagged chain correlated to the dynamics of a group of @xmath105 chains comprised inside the range of the intermolecular potential . the latter has a range of the order of the correlation hole , i.e. of the radius - of - gyration of the polymer , as it emerges from the solution of the oz equation . the number of chains comprised in a volume of the order of @xmath106 is given by @xmath107 , where @xmath1 , the statistical segment length , and the other quantities have been defined earlier on . the number of interpenetrating chains , @xmath105 , increases with increasing density , degree of polymerization , and the stiffness of the polymer . in the cooperative dynamic approach the dynamics is described by a set of coupled equations of motion ( eom ) . each equation is expressed in the space coordinates of the monomer @xmath25 , belonging to molecule @xmath108 and positioned at @xmath109 , and contains a balance of different forces acting on the monomer . these include the viscous force , @xmath110 , the intramolecular force @xmath111,which contains the structural matrix @xmath112 , the time - dependent intermolecular mean - field force @xmath113 $ ] , and the random interactions with the surrounding liquid , given by the random force @xmath114 . @xmath115\nonumber\\ & & + \mathbf{f}_a^{(i)}(t ) \ .\end{aligned}\ ] ] here @xmath116 is the matrix of intramolecular connectivity , which reduces to the rouse matrix when infinitely long and completely flexible macromolecules are considered , and is defined as @xmath117 here @xmath118 is the equilibrium averaged bond correlation matrix @xmath119 and @xmath120 is a structural matrix , with all the elements equal zero except @xmath121for @xmath122 , @xmath123 with @xmath124 , and @xmath125 with @xmath124 . the @xmath118 matrix is a function of the local semiflexibility parameter @xmath126 , which is related to the persistence length of the polymer . for our samples the values of @xmath79 are reported in table [ tb : ua_rg ] and are calculated from the molecule radius of gyration , which can be obtained from experiments or from atomistic simulations.@xcite through eq.([equ ] ) the eom includes a complete microscopic description of the structure and local flexibility of the specific molecule under investigation , containing all the relevant parameters that define the intramolecular mean - force potential . equations of motion for different beads belonging to the same chain or to two chains undergoing slow cooperative dynamics are coupled by the presence of intra- and inter - molecular interaction potentials . the intrinsic , chemical dependent semiflexibility of the macromolecule enters explicitly through the description of the @xmath127 matrix . the intermolecular potential is time dependent , as it is a function of the relative position coordinates of the centers - of - mass of a pair of molecules . as the two molecules move with respect to each other the intermolecular force changes . from the initial ensemble of @xmath105 dynamically correlated chains , molecules diffuse in time and finally move outside the range of the potential , @xmath0 , in a timescale of @xmath128 . the set of coupled equations is solved after transformation into normal modes of motion and numerically using a self - consistent procedure that calculate the effective potential at any given distance . once @xmath105 and @xmath129 are defined , the first optimized against the simulation data and the second from our rescaling procedure , the set of equations has no adjustable parameters . by applying consideration of symmetry the set of coupled eoms in normal coordinates can be simplified and reduced to two sets of @xmath2 uncoupled equations in the relative and collective normal - mode coordinates . we assume that the eoms for internal modes ( @xmath130 ) do not contain intermolecular forces . this approximation is justified on the basis that polymer local dynamics is affected in a similar way by inter- and intra - molecular excluded volume interactions , which in polymer liquids tend to compensate each other . intermolecular forces , which enter the dynamics through the eom for the first ( @xmath131 ) normal mode , still perturb the dynamics on the local scale through the linear combination of the normal modes . in the following , we present an overview of some of the quantities analyzed in the original simulations@xcite and their comparison with the predictions of the cooperative dynamics theory having as an input parameter the reconstructed friction coefficient calculated from the soft - sphere simulations and the rescaling procedure as described in the first part of this paper . figure [ fg : petepb112 ] presents the end - to - end vector , time decorrelation function for the @xmath132 sample of pb . the cooperative dynamic theory with the reconstructed monomer friction coefficient is directly compared against ua md data . this function represents the rotational relaxation of the chain and can not be measured from the soft colloid simulation directly . the @xmath132 sample has chains shorter than the entanglement length and the cooperative dynamics theory reproduces the atomistic simulation data rather well . and the monomer friction coefficient is reconstructed from ms md simulations , using the procedure described in this paper . ] figure [ fg : commsd ] shows the center of mass mean square displacement ( com msd ) for three samples of _ cis_-1,4-polybutadiene melts with chains of increasing degree of polymerization , @xmath133 and @xmath134 . the long time diffusive dynamics is calculated from the diffusion coefficients obtained from the rescaling procedure . for time shorter than the longest relaxation time , @xmath135 , the mean - square displacement of the com shows a subdiffusive behavior that can not be reproduced by the rouse approach , i.e. the conventional theory of unentangled chain dynamics . in the cooperative dynamics theory the subdiffusive dynamics is a consequence of the cooperative motion of the interpenetrating chains inside the correlation hole region , coupled by the effective intermolecular potential between the coms of the macromolecules . the theory , with the monomer friction from rescaled ms md simulations as an input , is compared against the ua md simulation data . the diffusion coefficient for pb240 is underestimated and for pb320 is slightly overestimated by the theory in comparison with the simulations , which is also can be seen in table [ tb : dofn_ua ] and figure [ fg : dofn_pb ] . these deviations are within the error related to @xmath0 . dot - dash lines in the inset of figure [ fg : commsd ] represent results calculated with the upper and the lower values of @xmath0 taken from the data of the atomistic simulations , as reported in table [ tb : ua_rg ] for the pb240 sample . the number of correlated chains is optimized to reproduce the subdiffusive behavior giving @xmath136 for pb240 , pb320 and pb400 respectively , with a trend that qualitatively agree with the predicted scaling behavior . the error in the radius - of - gyration does not allow for a more precise calculation of the parameter @xmath105 . as reported in table [ tb : ua_rg ] . ] starting from the theory of cooperative dynamics , the single chain dynamic structure factor can be calculated as @xmath137 \right > \nonumber\\ & \approx & \frac{1}{n } \sum_{i , j}^n exp \left [ -\frac{\mathbf{q}^2}{6 } \left < \left(\mathbf{r}_i(t)-\mathbf{r}_j(0 ) \right)^2 \right > \right ] \ . \label{eq : skt}\end{aligned}\ ] ] eq.[eq : skt ] describes the dynamics of a `` tagged '' polymer chain , which moves in a group of @xmath105 interpenetrating chains . the intermolecular interaction between chains affects the dynamics resulting in the subdiffusive motion of the single - chain center - of - mass , as depicted in figure [ fg : commsd ] , and in the subdiffusive decay of the low @xmath138 dynamics of @xmath139 . the average time - dependent distance between two monomers in the tagged chain depends on the presence of the @xmath140 other chains dynamically correlated , and it is expressed as a function of relative and collective mode coordinates . once @xmath141 and @xmath105 are determined , the theory predicts the full decay of the dynamic structure factor , at the different experimental wave vectors . figures [ fg : sofk_pb96 ] and [ fg : sofk_pb400 ] display the normalized dynamic structure factor obtained from the theory and compared with ua md simulations for @xmath142 @xmath143 . for unentangled system pb96 shown on figure [ fg : sofk_pb96 ] the agreement is excellent for all values of scattering vector @xmath138 and time range . the rouse model ( not shown in this figure , @xcite ) fails to describe these data with satisfactory agreement , while the cooperative dynamics theory is found to be in quantitative agreement for samples at different molecular weights , over the whole time range , and for different wave vectors . . the data from ua md simulations ( symbols ) are compared against the cooperative dynamics theory ( solid lines ) where the number of correlated chains is set to @xmath144 and the monomer friction coefficient is reconstructed from ms md simulations , using the procedure described in this paper . ] on figure [ fg : sofk_pb400 ] the normalized dynamic structure factor is shown for the weakly entangled system pb400 . the agreement for the smallest scattering vector which represents global dynamics is excellent at all times , while for intermediate and large values of @xmath138 the agreement between theory and simulations is slightly less accurate . the theoretical decay , which is slightly faster than simulations at long times , suffers from the lack of properly describing the entanglement effects . a more recent version of the cooperative dynamics theory includes chain uncrossability , i.e. entanglements , and shows the correct slowing down of the relaxation in the long - time regime for entangled samples.@xcite . the data from ua md simulations ( symbols ) are compared against the theory for cooperative dynamics ( solid lines ) where the number of correlated chains is set to @xmath145 and the monomer friction coefficient is reconstructed from the ms md simulations , using the procedure described in this paper . ] figure [ fg : monmsd ] displays the monomer mean square displacement for _ cis_-1,4-polybutadiene sample with @xmath132 . theoretical prediction is compared against ua md data from ref . the agreement in both subdiffusive and diffusive regions is very good . the theory accounts for the local semiflexibility and chain connectivity , together with the cooperative motion of the chains due to the presence of intermolecular interactions . the purpose of the rescaling procedure is to predict diffusion coefficients from the mesocale simulations properly rescaled when the atomistic simulations or the experiments are not available . one key input quantity to the procedure is the molecular radius of gyration . in the previous sections we adopted the values of @xmath0 determined from the atomistic simulations . when there are no data available for the radius of gyration , neither from atomistic simulations nor from experiments , it is convenient to use some simple model , like the freely rotating chain model to provide @xmath0 , as it captures the semiflexibility of the polymer chain . in this section we present a critical discussion of adopting a freely rotating chain model to describe the pb chain , using a constant flexibility parameter @xmath79 . for the _ cis_-1,4-polybutadiene chain with @xmath2 carbon atoms along its backbone there are @xmath146 double bonds with @xmath76 , @xmath147 single bonds with @xmath63 and @xmath148 single bonds with @xmath149 . the average bond length over all @xmath150 bonds is calculated as @xmath151/(n-1)\ , , \ ] ] which in the @xmath152 limit gives the value of @xmath153 . the radius of gyration for a freely rotating chain ( frc ) model with semiflexibility parameter @xmath79 is given as @xmath154 figure [ fg : g_param_of_rhom ] shows the density dependence of the semiflexibility parameter @xmath79 in the ua md simulations . calculated from eq.([eq : frc ] ) using values of the radius - of - gyration @xmath0 measured in ua md and reported in table [ tb : g_parampb ] . ] lcccc & @xmath155[@xmath82 ] & @xmath156 & @xmath78[@xmath82 ] & @xmath157 + pb 32 & 270@xmath8510 & 0.6237 & 45@xmath855 & 0.6237 + pb 48 & 440@xmath8510 & 0.6391 & 70@xmath855 & 0.6243 + pb 56 & 530@xmath8510 & 0.6458 & 85@xmath857 & 0.6339 + pb 64 & 600@xmath8515 & 0.6406 & 95@xmath8510 & 0.6246 + pb 80 & 790@xmath8530 & 0.6531 & 125@xmath8510 & 0.6374 + pb 96 & 950@xmath8520 & 0.6516 & 152@xmath8516 & 0.6394 + pb 112 & 1150@xmath8520 & 0.6609 & 184@xmath8515 & 0.6490 + pb 128 & 1335@xmath8530 & 0.6643 & 215@xmath8518 & 0.6544 + pb 140 & 1430@xmath8540 & 0.6576 & 234@xmath8518 & 0.6523 + pb 160 & 1640@xmath8560 & 0.6578 & 275@xmath8525 & 0.6595 + pb 200 & 2100@xmath8580 & 0.6634 & 340@xmath8520 & 0.6551 + pb 240 & 2480@xmath85100 & 0.6581 & 410@xmath8530 & 0.6557 + pb 320 & 3100@xmath85200 & 0.6382 & 576@xmath8530 & 0.6695 + pb 400 & 4138@xmath85100 & 0.6568 & 678@xmath8530 & 0.6519 + + [ tb : g_parampb ] fitting simultaneously all samples to @xmath158 data gives an average value of the semiflexibility parameter for all densities , @xmath159 . figure [ fg : rgyrpb_of_n ] displays the radius of gyration squared over degree of polymerization @xmath160 as a function of @xmath2 from united atom simulations and from the freely - rotating - chain model . the curvature at small @xmath2 is due to chain - end effects . samples with small @xmath2 present the largest deviation between the two sets of data , while the constant flexibility hypothesis works best for the large @xmath2 samples . the deviation is due to the fact that the united atom simulations are performed in the npt ensemble where increasing @xmath2 corresponds to an increase of the liquid density , as shown in figure [ fg : dens_of_n ] . the hypothesis of a constant semiflexibility parameter works well for the long chains for which coarse - graining methods are most useful . as a function of @xmath2 . ua md data ( circles ) , with statistical error , are compared against data calculated with the freely rotating chain model using an averaged semiflexibility parameter , @xmath159 ( solid line ) . ] with this in mind , we report in figure [ fg:4 ] the diffusion coefficient calculated from the mesoscale simulations , properly rescaled , where now the input radius - of - gyration is taken from the freely rotating chain model . as it is expected the best agreement between ua simulations and theoretical predicted diffusion is for the samples with larger @xmath2 . the disagreement at small @xmath2 is related to the fact that the density is changing with increasing @xmath2 in that region of the plot , and the radius of gyration should be properly corrected for this difference . in conclusion , if an exact value of the radius of gyration as a function of thermodynamic conditions and degree of polymerization is known , the described procedure should be able to provide an accurate estimate of the dynamics starting from the mesoscale simulations of the coarse - grained system , which are computationally very efficient . the figure also reports a number of samples for which neither experiments nor simulations have been performed ( @xmath161 , @xmath162 , @xmath163 , and @xmath164 ) . the most relevant conclusion of this part of our study is the fact that it is possible to make predictions for data that are not reported in the ua md simulations , obtaining information for new systems for which neither simulations nor experiments are available . in a couple of recent papers we have presented an original approach to rescale the dynamics measured in mesoscale simulations of coarse - grained systems and recover the realistic dynamics as measured in atomistic simulations performed in identical thermodynamic conditions . the relevance of this method is in its ability of predicting diffusion coefficient , with good precision , directly from the simulation of a coarse - grained system , without the need of performing an atomistic simulation . simply put , the method does not need a calibration curve . while the previous papers focused on the coarse - grained dynamics of polyethylene samples , this is the first paper in which the method is applied to a polymer with a different monomeric structure . in this paper we use an extreme " level of coarse - graining as the whole chain is represented as a soft colloidal particle . the choice of this representation is motivated by two reasons . the first is that a possible error in the procedure would be clearly made evident in this level of coarse - graining , and we believe that the procedure would produce an even better agreement if the macromolecule is coarse - grained at a lower level , e.g. as a collection of soft particles , or blobs.@xcite the second is that for this representation all the physical quantities that enter our approach are analytically solved and the rescaling procedure is formally derived . the paper presents a comparison between theoretical predictions of coarse - grained simulations properly rescaled following our first - principles procedure , and data from united - atom simulations for _ cis_-1,4-polybutadiene melts . the simulations were performed by tsolou et al . in a @xmath80 ensemble.@xcite the rescaled diffusion coefficient shows good agreement with the diffusion measured in atomistic simulations , which supports the validity of the proposed procedure . as the coarse - graining model represents the whole polymeric chain as a soft colloidal particle , the only dynamical information that can be collected from the rescaled mesoscale simulation is the diffusion coefficient of the center - of - mass of the polymer . from this , however , the friction coefficient of the monomer is derived . this paper shows how the monomer friction coefficient is used as an input to traditional langevin equations for the dynamics of polymeric liquids . specifically we use here the cooperative dynamics model for unentangled chains , given that the samples in the ua simulations are in the regime of unentangled and slightly entangled dynamics . from the theory of cooperative dynamics , which represents the cooperative motion of a group of interacting polymers in a dynamically heterogeneous liquid , all time correlation functions of interest can be calculated . specifically we present here the comparison with the quantities reported in the original paper by tsolou et al.@xcite and show that the theory , with the rescaled friction coefficient , is able to reproduce quantitatively the dynamics observed in the end - to - end time decorrelation function , and in the dynamic structure factor for unentangled and slightly entangled chains , and from the global to the local intramolecular dynamics . for entangled chains the local dynamics , represented by the high @xmath138 regime in the dynamic structure factor and in the monomer mean - square - displacement , shows a slight departure from the theory as the cooperative dynamics approach adopted here does not yet include the effect of entanglements.@xcite we acknowledge support from the national science foundation through grant dmr-0804145 . computational resources were provided by trestles through the xsede project supported by nsf . we are grateful to v. g. mavrantzas for the informative discussion of his paper . 99 k. binder ed . _ monte carlo and molecular dynamic simulations in polymer science _ ( oxford university press , new york , 1995 ) . d. frenkel , and b. smit , _ understanding molecular simulation _ ( academic , new york , 2000 ) . t. sauer , c. grebogi , and j. a. yorke phys . lett . * 79 * , 59 ( 1997 ) . m. p. allen , and d. j. tildesley _ computer simulation of liquids _ ( oxford science publications , oxford , 1992 ) . i. y. lyubimov , j. mccarty , a. clark , and m. g. guenza , j. chem . phys . * 132 * , 224903 ( 2010 ) . i. y. lyubimov , m.g . guenza phys . e * 84 * , 031801 ( 2011 ) . p. ilg , h. c. ttinger , and m. krger phys . e * 79 * , 011802 ( 2009 ) . p. ilg , and m. krger j. rheol . * 55 * , 69 ( 2011 ) . izvekov and voth , j. chem . 125 , 151101 ( 2006 ) . j. t. padding , w. j. briels j. phys . * 23 * 233201 ( 2011 ) . s. o. nielsen , c. f. lopez , g. srinivas , m. l. klein j. phys . condens . mat . * 16 * , r481 ( 2004 ) . y. li , s. tang , b. c. abberton , m. krger , c. burkhart , b. jiang , g. j. papakonstantopoulos , m. poldneff , w. k. liu polymer * 53 * , 5935 ( 2012 ) . j. t. padding and w. j. briels j. chem . phys . * 117 * , 925 ( 2002 ) . v. a. harmandaris and k. kremer soft matter * 5 * , 3920 ( 2009 ) . g. tsolou , v. g. mavrantzas , d. theodorou macromol . * 38 * , 1478 ( 2005 ) . g. tsolou , v.a . harmandaris , v.g . mavrantzas macromol . theory simul . * 15 * , 381 ( 2006 ) . stephanou , c. baig , g. tsolou , v.g . mavrantzas , and m. kroger j.chem.phys * 132 * , 124904 ( 2010 ) . doi , m. ; edwards , s. f. _ the theory of polymer dynamics _ ( clarendon press , oxford , 1986 ) . m. g. guenza j. phys : cond . matt . * 20 * , 033101 ( 2008 ) . marina guenza , phys . lett . * 88 * , 025901 ( 2002 ) . m. zamponi , a. wischnewski , m. monkenbusch , l. willner , d. richter , p. falus , b. farago and m. g. guenza j. phys . chem . * 112 * , 16220 ( 2008 ) . e. caballero - manrique , j. k. brey , w. a. deutschman , f. w. dahlquist , and m. g. guenza , bioph . j. * 93 * , 4128 ( 2007 ) . y. rosenfeld phys . a * 15 * , 2545 ( 1977 ) . j. a. armstrong , c. chakravarty , p. ballone j. chem . phys . * 136 * , 124503 ( 2012 ) . r. zwanzig _ nonequilibrium statistical mechanics _ ( oxford university press , new york , 2001 ) . k. s. schweizer , j. chem . phys . * 91 * , 5802 ( 1989 ) . g. yatsenko , e. j. sambriski , m. a. nemirovskaya , and m. guenza , phys . . lett . * 93 * , 257803 ( 2004 ) . fetters , d.j . lohse , c.a . garcia - franco , p. brant , and d. richter macromol . * 35 * , 10096 ( 2002 ) . j. mccarty , i. y. lyubimov , m. g. guenza macromol . * 43 * , 3964 ( 2010 ) . j. mccarty , i. y. lyubimov , and m. g. guenza j. phys . b * 113 * , 11876 ( 2009 ) . f. bueche , physical properties of polymers ( interscience , new york , 1962 ) . harmandaris , m. doxastakis , v.g . mavrantzas , d.n . theodorou j.chem . phys . * 116 * , 436 ( 2002 ) . e. von meerwall , s. beckman , j. jang , and w.l . mattice j.chem . * 108 * , 4299 ( 1998 ) . m. g. guenza , j. chem . phys . * 110*,7574 ( 1999 ) . b. frick , g. dasseh , a. cailliaux , and c. alba - simionesco chem . phys . * 292 * , 311 ( 2003 ) . a. arbe , j. colmenero , b. farago , m. monkenbusch , u. buchenau , and d. richter chem . phys . * 292 * , 295 ( 2003 ) . j. colmenero , f. alvarez , and a. arbe , phys . e * 65 * , 041804 ( 2002 ) . g. tsolou , v. a. harmandaris , and v. mavrantzas j. non - newtonian fluid mech . * 152 * , 184 ( 2008 ) . marina guenza , macromol . * 35 * , 2714 ( 2002 ) . p. debnath , and m. g. guenza , phil . mag . * 88 * , 33 ( 2008 ) . m. g. guenza ( in preparation ) . a. j. clark , and m. g. guenza , j. chem . * 132 * , 044902 ( 2010 ) . a. j. clark , j. mccarty , i. y. lyubimov , m. g. guenza phys . lett . * 109 * , 168301 ( 2012 ) .

taking into account the contributions of the quark - quark , quark - gluon , and gluon - gluon condensate corrections , the strong form factors and coupling constants of , , , , , and vertices are investigated within the three - point qcd sum rules method , without and with the symmetry .

1509.01715

in high energy physics , investigation of meson interactions depends on information about the proper functional form of the strong form factors . among all vertices , the charmed meson ones , which play an important role in understanding the final state interactions , are much more significant . therefore , researchers have concentrated on computing the strong form factors and coupling constants connected to these vertices . until now , the vertices involving charmed mesons such as @xmath7@xcite , @xmath8 @xcite , @xmath9@xcite , @xmath10@xcite , @xmath11 @xcite , @xmath12 @xcite , @xmath13 , @xmath14 , @xmath15 , @xmath16 @xcite , @xmath17 , @xmath18 , @xmath19 @xcite , @xmath20 @xcite , @xmath21 , @xmath22 @xcite , @xmath23 @xcite , @xmath24 , @xmath25 @xcite , and @xmath26 @xcite have been studied within the framework of the qcd sum rules . the effective lagrangians for the interaction vertices @xmath27 , @xmath28 , and @xmath29 are @xcite : @xmath30,\end{aligned}\ ] ] where @xmath31 , @xmath32 , and @xmath33 are the strong form factors . from these lagrangians , the elements related to the @xmath27 , @xmath34 and @xmath29 vertices can be derived in terms of the strong form factors as : @xmath35\nonumber\\&\times&\varepsilon_{\alpha } ( p)\varepsilon'_{\mu}(p')\varepsilon''_{\nu}(q),\end{aligned}\ ] ] where @xmath36 . in this work , we decide to calculate the strong form factors and coupling constants associated with the @xmath37 , @xmath38 , @xmath39 , @xmath40 , @xmath41 , and @xmath42 vertices in the frame work of the three - point qcd sum rules ( 3psr ) . the plan of the present paper is as follows : in section ii , the strong form factor calculation of the @xmath38 vertex is derived in the frame work of the 3psr ; computing the quark - quark , quark - gluon and gluon - gluon condensate contributions in the borel transform scheme . in section iii , using necessary changes in the expression obtained for the @xmath43 , the strong form factors @xmath44 , @xmath45 , @xmath46 , @xmath47 , and @xmath48 are presented . the next section depicts our numerical analysis of the strong form factors as well as the coupling constants , without and with the @xmath6 symmetry . to compute the strong form factor of the @xmath27 vertex via the 3psr , we start with the correlation function . when @xmath49 meson is off - shell , the correlation function is as follows . @xmath50 for off - shell charmed meson , the correlation function is : @xmath51 where @xmath52 , and @xmath53 are interpolating currents with the same quantum numbers of @xmath54 , and @xmath49 mesons , respectively . also @xmath55 is time ordering product , @xmath56 and @xmath57 are the four momentum of the initial and final mesons , respectively as depicted in fig . [ f1 ] . the correlation functions in eqs . ( [ eq21 ] ) and ( [ eq22 ] ) are complex functions of which the real part comprises the computations of the theoretical part or qcd and imaginary part comprises the computations of the physical or phenomenological . in the qcd representation , the correlation function is evaluated in quark - gluon language like quark - quark , gluon - gluon condensate , etc using the wilson operator product expansion ( ope ) . in the phenomenological part , the representation is in terms of hadronic degrees of freedom which is responsible for the introduction of the form factors , decay constants and masses . the qcd part of the correlation functions can be calculated by expanding it in terms of the ope , in the deep euclidean region , as : @xmath58 where @xmath59 are the wilson coefficients , i is the unit operator , @xmath60 is the local fermion field operator and @xmath61 is the gluon strength tensor . the wilson coefficient @xmath62 is contribution of the perturbative part of the qcd and the other coefficients are contribution of the non - perturbative part . the diagrams corresponding to the perturbative ( bare loop ) , are depicted in fig . [ f1 ] . ( 90,20 ) ( 0,-10 ) to compute the portion of the perturbative part using the double dispersion relation for coefficient of the lorentz structure @xmath63 appearing in eq . ( [ eq12 ] ) , we obtain : @xmath64 where @xmath65 is spectral density . performing the fourier transformation and using the cutkosky rules , i.e. , @xmath66 , the spectral densities are calculated for the @xmath67 structure related to the @xmath27 vertex . @xmath68 for the off - shell @xmath49 ( fig . @xmath69 ( a ) ) : @xmath70.\end{aligned}\ ] ] @xmath68 for the off - shell @xmath71 ( fig . @xmath69 ( b ) ) : @xmath72.\end{aligned}\ ] ] the explicit expressions of the coefficients in the spectral densities are given in appendix - a . to compute the contribution of the non - perturbative part of the correlation function for the off - shell @xmath49 meson , six diagrams of dimension @xmath73 are considered . these diagrams named gluon - gluon condensate , shown in fig . [ f2 ] . in this case the gluon - gluon diagrams are more important than the other terms in the ope , since the heavy @xmath74 quark is a spectator @xcite . ( 100,20 ) ( 0,-25 ) when @xmath71 is off - shell , the quark - quark and quark - gluon diagrams of dimension @xmath75 and @xmath76 are more important than the gluon - gluon condensate , since the light @xmath77 quark is a spectator @xcite . [ f3 ] shows these diagrams related to the quark - quark and quark - gluon condensate . ( 100,00 ) ( 0,-20 ) after some straightforward calculations and applying the double borel transformations with respect to the @xmath78 and @xmath79 , where @xmath80 and @xmath81 are borel parameters , the following results are obtained for the non - perturbative contributions corresponding to fig . [ f2 ] and fig . [ f3 ] , respectively : @xmath82 where the explicit expressions for @xmath83 and @xmath84 are given in appendix - b . it should be noted that to obtain the gluon - gluon condensate contributions , we will follow the same procedure as stated in @xcite . in order to calculate the phenomenological part of the correlation functions in eqs . ( [ eq21 ] ) and ( [ eq22 ] ) , three complete sets of intermediate states with the same quantum number should be inserted in these equations . performing the fourier transformation , for the phenomenological parts , we have : @xmath85 the matrix elements @xmath86 , and @xmath87 are defined as : @xmath88 where @xmath89 , @xmath90 , @xmath91 , and @xmath92 are the masses and decay constants of mesons @xmath49 and @xmath93 , respectively . @xmath94 is the polarization vector of the vector meson @xmath49 . inserting eqs . ( [ eq12 ] ) and ( [ eq29 ] ) in eq . ( [ eq28 ] ) and after some calculations , we obtain @xmath95 and @xmath96 in terms of strong form factors @xmath97 and @xmath98 as : @xmath99 the strong form factors are calculated by equating two representations of the correlation function and applying the borel transformations with respect to the @xmath100 and @xmath79 on the phenomenological as well as the perturbative and nonperturbative parts of the correlation function in order to suppress the contributions of the higher states and continuum . the equations for the strong form factors are obtained as follows : @xmath101 where @xmath102 and @xmath103 are the continuum thresholds in @xmath49 and @xmath104 mesons , respectively . @xmath105 and @xmath106 are the lower limits of the integrals over @xmath77 as @xmath107 also @xmath108 and @xmath109 are defined as : @xmath110 following the previous steps in section ii , phrases similar to eq . ( [ eq212 ] ) can be obtained for the strong form factors of the @xmath37 , @xmath111 , @xmath40 , @xmath112 , and @xmath42 vertices via the 3psr . for this purpose , the appropriate terms of @xmath113 , the spectral density , and quark - gluon condensate should be replaced in eq . ( [ eq212 ] ) . proper expressions for @xmath113 , the spectral density , and quark - gluon condensate , related to the strong form factors @xmath44 , @xmath114 , @xmath46 , @xmath47 , and @xmath115 , have been listed in the tables [ t331]-[t332 ] and appendix - b , respectively . .expressions for the coefficient @xmath113 related to the strong form factors @xmath44 , @xmath114 , @xmath46 , @xmath47 , and @xmath115 . [ cols="^,^",options="header " , ] in summary , taking into account the contributions of the quark - quark , quark - gluon and gluon - gluon condensate corrections , the strong form factors @xmath116 , @xmath117 , @xmath118 , @xmath119 , @xmath120 , and @xmath121 were estimated within the 3psr without and with the @xmath6 symmetry . for instance , the dependence of the strong form factors @xmath117 on the transferred momentum square @xmath122 were plotted . also the coupling constants of these vertices were evaluated . partial support of the isfahan university of technology research council is appreciated . in this appendix , the explicit expressions of the coefficients in the spectral densities are given as : @xmath123,\nonumber \\ c_2 & = & \frac{1}{\lambda(s , s',q^2)}[2 s \delta -u \delta '' ] , \nonumber \\ a & = & -\frac{1}{2\lambda(s , s',q^2)}[4ss'm_s^2-s\delta^2-s'\delta''^2-m_s^2 u^2+u\delta\delta''],\nonumber \\ e_1 & = & \frac{1}{2\lambda^2(s , s',q^2 ) } [ 8 ss'^2 m_s^2\delta '' -2 s'm_s^2 u^2 \delta''-4 s s ' m_s^2 u \delta + m_s^2 u^3 \delta -2 s'^2 \delta''^3 \nonumber\\ & + & 3 s ' u \delta \delta''^2 -2s s ' \delta^2 \delta '' -u^2 \delta^2 \delta '' + s u \delta^3],\nonumber\\ e_2 & = & \frac{1}{2\lambda^2(s , s',q^2 ) } [ 8 s^2 s ' m_s^2 \delta-2 s m_s^2 u^2 \delta''- 4 s s ' m_s^2 u \delta '' + m_s^2 u ^3\delta''-2 s^2 \delta ^3 \nonumber\\ & + & 3 s u \delta^2 \delta''-2 s s ' \delta \delta''^2 -u^2 \delta \delta''^2 + s ' u \delta''^3 ] , \end{aligned}\ ] ] also @xmath124 , and @xmath125 . in this appendix , the explicit expressions of the coefficients of the quark and gluon condensate contributions of the strong form factors in the borel transform scheme for all the vertices are presented . @xmath126 @xmath127 @xmath128 @xmath129 @xmath130 @xmath131 @xmath132}(3,2,2)m_{c}^{4}-\hat{i}_{1}(3,2,1)m_{c}^{4}+6\hat{i}_{0}(4,1,1)m_{c}^{3}m_{s } \\ & & -\hat{i}_{1}(3,1,2)m_{c}^{3}m_{s}+2\hat{i}_{2}(2,2,2)m_{c}^{3}m_{s}+\hat{i}_{2}(3,1,2)m_{c}^{3}m_{s}+2\hat{i}_0^{[1,0]}(3,2,2)m_{c}^{3}m_{s } \\ & & + 2\hat{i}_{0}(3,2,1)m_{c}^{3}m_{s}-2\hat{i}_{1}(2,2,2)m_{c}^{3}m_{s}+8\hat{i}_{0}(3,1,2)m_{c}^{2}m_{s}^{2}-\hat{i}_1^{[1,0]}(3,2,2)m_{c}^{2}m_{s}^{2 } \\ & & + 6\hat{i}_{1}(1,1,4)m_{c}^{2}m_{s}^{2}+3\hat{i}_{0}(4,1,1)m_{c}^{2}m_{s}^{2}-6\hat{i}_{2}(1,1,4)m_{c}^{2}m_{s}^{2}-6\hat{i}_{0}(1,1,4)m_{c}^{2}m_{s}^{2 } \\ & & -\hat{i}_2^{[1,0]}(3,2,2)m_{c}^{2}m_{s}^{2}-\hat{i}_2^{[1,0]}(3,2,2)m_{c}m_{s}^{3}+2\hat{i}_{2}(3,1,2)m_{c}m_{s}^{3}-\hat{i}_1^{[1,0]}(3,2,2)m_{c}m_{s}^{3 } \\ & & -12\hat{i}_{0}(1,1,4)m_{c}m_{s}^{3}-2\hat{i}_{1}(3,1,2)m_{c}m_{s}^{3}-2\hat{i}_0^{[1,0]}(3,2,2)m_{c}m_{s}^{3}+8\hat{i}_{0}(3,1,2)m_{c}m_{s}^{3 } \\ & & -6\hat{i}_{0}(1,1,4)m_{s}^{4}+3\hat{i}_{0}(3,1,2)m_{s}^{4}+3\hat{i}_{0}(2,2,1)m_{c}^{2}-3\hat{i}_1^{[0,1]}(3,1,2)m_{c}^{2 } \\ & & -2\hat{i}_{1}(1,2,2)m_{c}^{2}-3\hat{i}_0^{[0,1]}(4,1,1)m_{c}^{2}+\hat{i}_0^{[0,1]}(3,2,1)m_{c}^{2}-3\hat{i}_2^{[0,1]}(3,1,2)m_{c}^{2 } \\ & & -\hat{i}_0^{[0,2]}(3,2,2)m_{c}^{2}+5\hat{i}_1^{[1,1]}(3,2,2)m_{c}^{2}+5\hat{i}_2^{[1,1]}(3,2,2)m_{c}^{2}+2\hat{i}_{2}(1,2,2)m_{c}^{2 } \\ & & + 2\hat{i}_{1}(1,2,2)m_{c}m_{s}-3\hat{i}_2^{[1,0]}(3,2,1)m_{c}m_{s}-2\hat{i}_{2}(1,2,2)m_{c}m_{s}-\hat{i}_{2}(2,2,1)m_{c}m_{s } \\ & & + \hat{i}_1^{[1,0]}(3,2,2)m_{c}m_{s}+2\hat{i}_0^{[1,1]}(3,2,2)m_{c}m_{s}+6\hat{i}_{0}(3,1,1)m_{c}m_{s}+\hat{i}_{1}(2,2,1)m_{c}m_{s } \\ & & -2\hat{i}_1^{[0,1]}(2,1,3)m_{c}m_{s}-2\hat{i}_2^{[0,1]}(2,1,3)m_{c}m_{s}+\hat{i}_2^{[1,1]}(3,2,2)m_{c}m_{s}-3\hat{i}_1^{[1,0]}(3,2,1)m_{c}m_{s } \\ & & -\hat{i}_1^{[1,0]}(2,2,2)m_{s}^{2}-3\hat{i}_0^{[1,0]}(3,2,1)m_{s}^{2}+6\hat{i}_0^{[0,1]}(1,1,4)m_{s}^{2}+4\hat{i}_{0}(3,1,1)m_{s}^{2 } \\ & & -5\hat{i}_{0}(2,1,2)m_{s}^{2}-\hat{i}_2^{[1,0]}(2,2,2)m_{s}^{2}-2\hat{i}_0^{[1,0]}(2,2,2)m_{s}^{2}-\hat{i}_{2}(2,1,2)m_{s}^{2}+\hat{i}_{1}(2,1,2)m_{s}^{2 } \\ & & -6\hat{i}_0^{[0,1]}(3,1,2)m_{s}^{2}-\hat{i}_0^{[1,0]}(1,2,2)+2\hat{i}_0^{[1,1]}(2,2,2)-2\hat{i}_0^{[0,1]}(2,2,1 ) \\ & & -2\hat{i}_{2}(2,1,1)+3\hat{i}_{0}(2,1,1)-\hat{i}_0^{[0,1]}(1,2,2)-6\hat{i}_0^{[0,1]}(1,3,1)+2\hat{i}_{1}(2,1,1 ) \\ & & + 2\hat{i}_{1}(1,2,1)+2\hat{i}_{1}(1,1,2)-\hat{i}_{0}(1,2,1)-\hat{i}_0^{[1,0]}(2,2,1)-2\hat{i}_{2}(1,2,1 ) \\ & & + \hat{i}_0^{[0,2]}(2,2,2)-2\hat{i}_{2}(1,1,2)-3\hat{i}_0^{[0,1]}(2,1,2),\end{aligned}\ ] ] @xmath133}(3,2,2)m_c^4 + 3\hat{i}_1(2,2,2)m_c^4 - 3\hat{i}_2(2,2,2)m_c^4+\hat{i}_1(3,1,2)m_c^4 \\ & & -3\hat{i}_0(3,2,1)m_c^4 + 2\hat{i}_2(2,2,2)m_c^3m_s+2\hat{i}_0(3,2,1)m_c^3m_s+2\hat{i}_0^{[1,0]}(3,2,2)m_c^3m_s \\ & & -2\hat{i}_1(2,2,2)m_c^3m_s+\hat{i}_2^{[1,0]}(3,2,2)m_c^3m_s+6\hat{i}_0(4,1,1)m_c^3m_s+4\hat{i}_0(2,2,2)m_c^3m_s \\ & & + \hat{i}_2(3,2,1)m_c^3m_s-\hat{i}_1(3,2,1)m_c^3m_s+\hat{i}_0(3,1,2)m_c^3m_s+\hat{i}_1^{[1,0]}(3,2,2)m_c^3m_s \\ & & -2\hat{i}_0(2,2,2)m_c^2m_s^2 + 2\hat{i}_1(3,1,2)m_c^2m_s^2 - 3\hat{i}_0(4,1,1)m_c^2m_s^2 - 2\hat{i}_2(3,1,2)m_c^2m_s^2 \\ & & -2\hat{i}_0^{[0,1]}(3,2,2)m_c^2m_s^2 + 5\hat{i}_0(3,1,2)m_cm_s^3 + 2\hat{i}_2(2,1,3)m_cm_s^3 + 2\hat{i}_2(3,1,2)m_cm_s^3 \\ & & -2\hat{i}_1(2,1,3)m_cm_s^3 - 2\hat{i}_1(3,1,2)m_cm_s^3-\hat{i}_0(2,2,2)m_s^4+\hat{i}_0^{[1,0]}(3,2,2)m_s^4 \\ & & -5\hat{i}_1^{[1,1]}(3,2,2)m_c^2 + 2\hat{i}_0^{[0,1]}(3,1,2)m_c^2 - 5\hat{i}_2^{[1,1]}(3,2,2)m_c^2 + 3\hat{i}_2^{[1,0]}(3,2,1)m_c^2 \\ & & + 2\hat{i}_0^{[0,1]}(2,2,2)m_c^2 + 3\hat{i}_1(1,3,1)m_c^2 - 2\hat{i}_0(1,2,2)m_c^2 + 3\hat{i}_1^{[1,0]}(3,2,1)m_c^2 \\ & & -3\hat{i}_2(1,3,1)m_c^2 - 12\hat{i}_0(1,1,3)m_cm_s+\hat{i}_2^{[1,1]}(3,2,2)m_cm_s-3\hat{i}_1^{[1,0]}(3,2,1)m_cm_s \\ & & -9\hat{i}_0(2,1,2)m_cm_s+3\hat{i}_1(1,3,1)m_cm_s-2\hat{i}_1^{[0,1]}(2,1,3)m_cm_s-\hat{i}_2(2,2,1)m_cm_s \\ & & -3\hat{i}_2^{[1,0]}(3,2,1)m_cm_s+\hat{i}_1(2,2,1)m_cm_s+4\hat{i}_1(2,1,2)m_cm_s-2\hat{i}_2^{[0,1]}(2,1,3)m_cm_s \\ & & -4\hat{i}_2(2,1,2)m_cm_s-3\hat{i}_2(1,3,1)m_cm_s+\hat{i}_1^{[1,1]}(3,2,2)m_cm_s-2\hat{i}_0(3,1,1)m_s^2 \\ & & + \hat{i}_2(2,1,2)m_s^2-\hat{i}_1(2,1,2)m_s^2 + 2\hat{i}_0^{[1,0]}(2,2,2)m_s^2 + 3\hat{i}_0(1,1,3)m_s^2 \\ & & + \hat{i}_2^{[1,0]}(2,2,2)m_s^2 - 2\hat{i}_0(2,2,1)m_s^2+\hat{i}_1^{[1,0]}(2,2,2)m_s^2 + 2\hat{i}_0^{[0,1]}(2,2,2)m_s^2 \\ & & + 3\hat{i}_0(2,1,2)m_s^2 - 2\hat{i}_0^{[1,1]}(2,2,2)-2\hat{i}_0^{[0,2]}(3,1,2)-2\hat{i}_1(1,2,1)+2\hat{i}_0^{[0,1]}(2,2,1 ) \\ & & + \hat{i}_0^{[0,1]}(1,2,2)+2\hat{i}_2(1,2,1)-3\hat{i}_0^{[1,1]}(3,2,1)-3\hat{i}_0^{[0,1]}(1,1,3)+\hat{i}_0(2,1,1 ) \\ & & + 2\hat{i}_2(1,1,2)+\hat{i}_0^{[0,1]}(2,1,2)-2\hat{i}_1(1,1,2),\end{aligned}\ ] ] @xmath134}(3,2,2)m_{c}^{4 } \\ & & + 3\hat{i}_{0}(4,1,1)m_{c}^{4}-\hat{i}_2^{[1,0]}(3,2,2)m_{c}^{4}+2\hat{i}_{6}(3,2,2)m_{c}^{4}+2\hat{i}_{2}(2,1,3)m_{c}^{3}m_{s } \\ & & + 2\hat{i}_{0}(2,1,3)m_{c}^{3}m_{s}-2\hat{i}_{1}(2,2,2)m_{c}^{3}m_{s}-2\hat{i}_{1}(2,1,3)m_{c}^{3}m_{s}-\hat{i}_{1}(3,2,1)m_{c}^{3}m_{s } \\ & & + \hat{i}_{2}(3,2,1)m_{c}^{3}m_{s}-3\hat{i}_{0}(4,1,1)m_{c}^{3}m_{s}+2\hat{i}_{2}(2,2,2)m_{c}^{3}m_{s}+\hat{i}_{0}(2,2,2)m_{c}^{2}m_{s}^{2 } \\ & & -2\hat{i}_{6}(3,2,2)m_{c}^{2}m_{s}^{2}+\hat{i}_0^{[0,1]}(3,2,2)m_{c}^{2}m_{s}^{2}+\hat{i}_2^{[1,0]}(3,2,2)m_{c}^{2}m_{s}^{2}-\hat{i}_1^{[1,0]}(3,2,2)m_{c}^{2}m_{s}^{2 } \\ & & + 2\hat{i}_{2}(3,1,2)m_{c}m_{s}^{3}-\hat{i}_2^{[1,0]}(3,2,2)m_{c}m_{s}^{3}-2\hat{i}_{1}(3,1,2)m_{c}m_{s}^{3}+6\hat{i}_{0}(1,1,4)m_{c}m_{s}^{3 } \\ & & + \hat{i}_1^{[1,0]}(3,2,2)m_{c}m_{s}^{3}+\hat{i}_{0}(3,1,2)m_{s}^{4}+2\hat{i}_6^{[1,0]}(3,2,2)m_{c}^{2}+\hat{i}_{0}(2,1,2)m_{c}^{2 } \\ & & + 8\hat{i}_{8}(3,2,1)m_{c}^{2}+4\hat{i}_8^{[0,1]}(3,2,2)m_{c}^{2}+2\hat{i}_6^{[0,1]}(3,2,2)m_{c}^{2}-4\hat{i}_{6}(3,2,1)m_{c}^{2 } \\ & & -4\hat{i}_7^{[0,1]}(3,2,2)m_{c}^{2}+3\hat{i}_1^{[1,0]}(4,1,1)m_{c}^{2}+6\hat{i}_{6}(4,1,1)m_{c}^{2}+6\hat{i}_{6}(3,1,2)m_{c}^{2 } \\ & & -8\hat{i}_{7}(3,2,1)m_{c}^{2}-12\hat{i}_{7}(4,1,1)m_{c}^{2}+2\hat{i}_{0}(2,2,1)m_{c}^{2}+12\hat{i}_{8}(4,1,1)m_{c}^{2 } \\ & & -3\hat{i}_2^{[1,0]}(4,1,1)m_{c}^{2}+8\hat{i}_{6}(2,1,3)m_{c}m_{s}+4\hat{i}_{6}(2,2,2)m_{c}m_{s}-\hat{i}_0^{[1,1]}(3,2,2)m_{c}m_{s } \\ & & -3\hat{i}_{1}(1,3,1)m_{c}m_{s}-2\hat{i}_1^{[0,1]}(3,1,2)m_{c}m_{s}+3\hat{i}_1^{[1,0]}(3,2,1)m_{c}m_{s}-2\hat{i}_{6}(3,2,1)m_{c}m_{s } \\ & & + 3\hat{i}_{2}(1,3,1)m_{c}m_{s}+2\hat{i}_2^{[0,1]}(3,1,2)m_{c}m_{s}-\hat{i}_{0}(2,1,2)m_{c}m_{s}-2\hat{i}_{6}(3,1,2)m_{c}m_{s } \\ & & -3\hat{i}_2^{[1,0]}(3,2,1)m_{c}m_{s}-2\hat{i}_6^{[1,0]}(3,2,2)m_{s}^{2}-12\hat{i}_{6}(1,1,4)m_{s}^{2}-\hat{i}_{0}(2,1,2)m_{s}^{2 } \\ & & -3\hat{i}_{0}(1,1,3)m_{s}^{2}-2\hat{i}_0^{[0,1]}(3,1,2)m_{s}^{2}+3\hat{i}_0^{[1,0]}(1,1,4)m_{s}^{2}+\hat{i}_0^{[1,1]}(3,2,2)m_{s}^{2 } \\ & & + 6\hat{i}_2^{[1,0]}(1,1,4)m_{s}^{2}+\hat{i}_{0}(3,1,1)m_{s}^{2}+24\hat{i}_{7}(1,1,4)m_{s}^{2}-24\hat{i}_{8}(1,1,4)m_{s}^{2 } \\ & & + 4\hat{i}_{6}(2,2,2)m_{s}^{2}-6\hat{i}_1^{[1,0]}(1,1,4)m_{s}^{2}-2\hat{i}_0^{[1,0]}(2,2,2)m_{s}^{2}+3\hat{i}_0^{[0,1]}(1,1,4)m_{s}^{2 } \\ & & + 2\hat{i}_{6}(3,1,2)m_{s}^{2}+3\hat{i}_1^{[0,1]}(2,1,2)-2\hat{i}_{2}(2,1,1)+3\hat{i}_{6}(1,3,1 ) \\ & & -3\hat{i}_2^{[0,1]}(2,1,2)-4s_{1},1(3,2,2)+4\hat{i}_7^{[1,0]}(3,2,1)-8\hat{i}_{8}(2,2,1 ) \\ & & -2\hat{i}_6^{[0,1]}(2,2,2)-2\hat{i}_6^{[0,1]}(3,1,2)+8\hat{i}_{7}(2,2,1)-4\hat{i}_8^{[1,0]}(3,2,1 ) \\ & & -4\hat{i}_6^{[1,0]}(3,2,1)+2\hat{i}_6^{[1,1]}(3,2,2)-12\hat{i}_{8}(3,1,1)+3\hat{i}_1^{[1,0]}(3,1,2 ) \\ & & + 2\hat{i}_{1}(2,1,1)-2\hat{i}_2^{[1,0]}(1,2,2)+2\hat{i}_1^{[1,0]}(1,2,2)+4\hat{i}_7^{[0,1]}(2,2,2 ) \\ & & -4\hat{i}_8^{[0,1]}(2,2,2)+4\hat{i}_{6}(2,2,1)-\hat{i}_0^{[0,1]}(3,1,1)-3\hat{i}_2^{[1,1]}(3,1,2 ) \\ & & + \hat{i}_2^{[2,0]}(2,2,2)-\hat{i}_1^{[2,0]}(2,2,2)+12\hat{i}_{7}(3,1,1)-8\hat{i}_{6}(2,1,2 ) \\ & & + 4\hat{i}_8^{[1,1]}(3,2,2)+\hat{i}_0^{[2,0]}(3,2,1)-2\hat{i}_0^{[1,0]}(1,2,2)-\hat{i}_0^{[0,1]}(2,1,2 ) \\ & & -3\hat{i}_{0}(2,1,1)-4\hat{i}_{6}(3,1,1)-2\hat{i}_6^{[1,0]}(2,2,2)+4\hat{i}_7^{[1,0]}(2,2,2 ) \\ & & -4\hat{i}_8^{[1,0]}(2,2,2)+2\hat{i}_{0}(1,1,2)-\hat{i}_0^{[1,0]}(3,1,1)-6\hat{i}_{7}(1,3,1)\\ & & + 6\hat{i}_{8}(1,3,1),\end{aligned}\ ] ] @xmath135}(3,2,2)m_{c}^{3}m_{s}-\hat{i}_{2}(3,2,1)m_{c}^{3}m_{s } \\ & & -\hat{i}_{0}(3,1,2)m_{c}^{3}m_{s}-2\hat{i}_{6}(3,2,2)m_{c}^{2}m_{s}^{2}-\hat{i}_1^{[1,0]}(3,2,2)m_{c}^{2}m_{s}^{2}-6\hat{i}_{0}(1,1,4)m_{c}^{2}m_{s}^{2 } \\ & & + \hat{i}_2^{[1,0]}(3,2,2)m_{c}^{2}m_{s}^{2}+6\hat{i}_{2}(1,1,4)m_{c}m_{s}^{3}-6\hat{i}_{1}(1,1,4)m_{c}m_{s}^{3}+\hat{i}_{0}(3,1,2)m_{s}^{4 } \\ & & + \hat{i}_{1}(2,1,2)m_{c}^{2}-\hat{i}_{1}(3,1,1)m_{c}^{2}+2\hat{i}_{0}(2,2,1)m_{c}^{2}+\hat{i}_1^{[2,0]}(3,2,2)m_{c}^{2 } \\ & & + 8\hat{i}_{8}(3,2,1)m_{c}^{2}+4\hat{i}_8^{[0,1]}(3,2,2)m_{c}^{2}+2\hat{i}_6^{[0,1]}(3,2,2)m_{c}^{2}-12\hat{i}_{7}(4,1,1)m_{c}^{2 } \\ & & -4\hat{i}_7^{[0,1]}(3,2,2)m_{c}^{2}+4\hat{i}_8^{[1,0]}(3,2,2)m_{c}^{2}+\hat{i}_{2}(3,1,1)m_{c}^{2}+2\hat{i}_8^{[1,0]}(3,2,2)m_{c}^{2 } \\ & & -\hat{i}_2^{[2,0]}(3,2,2)m_{c}^{2}-4\hat{i}_{6}(3,2,1)m_{c}^{2}-8\hat{i}_{7}(3,2,1)m_{c}^{2}-\hat{i}_0^{[0,1]}(2,2,2)m_{c}^{2 } \\ & & + \hat{i}_{0}(2,1,2)m_{c}^{2}+6\hat{i}_{6}(4,1,1)m_{c}^{2}-\hat{i}_{2}(2,1,2)m_{c}^{2}-4\hat{i}_7^{[1,0]}(3,2,2)m_{c}^{2 } \\ & & + 6\hat{i}_{6}(3,1,2)m_{c}^{2}+12\hat{i}_{8}(4,1,1)m_{c}^{2}-4\hat{i}_{1}(1,1,3)m_{c}m_{s}+4\hat{i}_{2}(1,1,3)m_{c}m_{s } \\ & & -4\hat{i}_{6}(2,2,2)m_{c}m_{s}-3\hat{i}_{2}(1,3,1)m_{c}m_{s}+2\hat{i}_{0}(1,2,2)m_{c}m_{s}+2\hat{i}_{6}(3,1,2)m_{c}m_{s } \\ & & + \hat{i}_{0}(2,1,2)m_{c}m_{s}-2\hat{i}_2^{[0,1]}(3,1,2)m_{c}m_{s}-8\hat{i}_{6}(2,1,3)m_{c}m_{s}+2\hat{i}_0^{[1,0]}(2,1,3)m_{c}m_{s } \\ & & + 2\hat{i}_{6}(3,2,1)m_{c}m_{s}+\hat{i}_{2}(2,1,2)m_{c}m_{s}+2\hat{i}_1^{[0,1]}(3,1,2)m_{c}m_{s}-\hat{i}_{1}(2,1,2)m_{c}m_{s } \\ & & + 3\hat{i}_{1}(1,3,1)m_{c}m_{s}+3\hat{i}_{0}(2,2,1)m_{c}m_{s}-3\hat{i}_{0}(1,1,3)m_{s}^{2}+4\hat{i}_{6}(2,2,2)m_{s}^{2 } \\ & & -6\hat{i}_1^{[1,0]}(1,1,4)m_{s}^{2}+3\hat{i}_0^{[0,1]}(1,1,4)m_{s}^{2}-12\hat{i}_{6}(1,1,4)m_{s}^{2}-2\hat{i}_0^{[0,1]}(3,1,2)m_{s}^{2 } \\ & & -\hat{i}_{0}(2,1,2)m_{s}^{2}+6\hat{i}_2^{[1,0]}(1,1,4)m_{s}^{2}+2\hat{i}_{6}(3,1,2)m_{s}^{2}-2\hat{i}_6^{[1,0]}(3,2,2)m_{s}^{2 } \\ & & -2\hat{i}_2^{[1,0]}(1,2,2)+\hat{i}_2^{[1,0]}(3,1,1)-4\hat{i}_6^{[1,0]}(3,2,1)-\hat{i}_0^{[1,0]}(3,1,1 ) \\ & & -8\hat{i}_{8}(2,2,1)-\hat{i}_0^{[0,1]}(2,1,2)-12\hat{i}_{8}(3,1,1)-2\hat{i}_6^{[1,0]}(2,2,2)+12\hat{i}_{7}(3,1,1 ) \\ & & -8\hat{i}_{8}(2,1,2)-8\hat{i}_{6}(2,1,2)-\hat{i}_1^{[1,0]}(3,1,1)+8\hat{i}_{7}(2,1,2)+4\hat{i}_7^{[1,0]}(3,2,1 ) \\ & & -4\hat{i}_8^{[1,0]}(3,2,1)+4\hat{i}_7^{[0,1]}(3,1,2)-4\hat{i}_8^{[0,1]}(3,1,2)+2\hat{i}_1^{[0,1]}(1,2,2 ) \\ & & + 2\hat{i}_{1}(2,1,1)-4\hat{i}_7^{[1,1]}(3,2,2)+4\hat{i}_8^{[1,1]}(3,2,2)+3\hat{i}_{6}(1,3,1)-2\hat{i}_6^{[0,1]}(3,1,2 ) \\ & & -2\hat{i}_6^{[0,1]}(2,2,2)-4\hat{i}_{6}(3,1,1)+\hat{i}_0^{[0,2]}(3,1,2)-\hat{i}_0^{[0,1]}(3,1,1 ) \\ & & -3\hat{i}_{0}(2,1,1)+4\hat{i}_{6}(2,2,1)+2\hat{i}_6^{[1,1]}(3,2,2)-2\hat{i}_{2}(2,1,1 ) \\ & & + 8\hat{i}_{7}(2,2,1)-2\hat{i}_0^{[0,1]}(2,2,1),\end{aligned}\ ] ] @xmath136}(3,2,2)m_{c}^{3}+6\hat{i}_{1}(4,1,1)m_{c}^{3}+2\hat{i}_2^{[1,0]}(3,2,2)m_{c}^{3 } \\ & & + 6\hat{i}_{0}(4,1,1)m_{c}^{3}+4\hat{i}_{0}(2,2,2)m_{c}^{3}+2\hat{i}_{0}(3,2,1)m_{c}^{3}+2\hat{i}_0^{[1,0]}(3,2,2)m_{c}^{3 } \\ & & + 2\hat{i}_1^{[0,1]}(3,2,2)m_{c}^{3}+2\hat{i}_2^{[0,1]}(3,2,2)m_{c}^{3}-2\hat{i}_{1}(3,1,2)m_{c}^{3}+2\hat{i}_{1}(3,2,1)m_{c}^{3 } \\ & & + 2\hat{i}_1^{[1,0]}(3,2,2)m_{c}^{3}+4\hat{i}_{1}(2,2,2)m_{c}^{3}+6\hat{i}_{2}(4,1,1)m_{c}^{3}-4\hat{i}_{2}(2,2,2)m_{c}^{2}m_{s } \\ & & -2\hat{i}_2^{[1,0]}(3,2,2)m_{c}^{2}m_{s}+8\hat{i}_{0}(2,1,3)m_{c}^{2}m_{s}+2\hat{i}_{1}(3,1,2)m_{c}^{2}m_{s}-2\hat{i}_{0}(3,1,2)m_{c}^{2}m_{s } \\ & & + 4\hat{i}_{2}(2,1,3)m_{c}^{2}m_{s}-2\hat{i}_{2}(3,1,2)m_{c}^{2}m_{s}-6\hat{i}_{2}(4,1,1)m_{c}^{2}m_{s}-2\hat{i}_2^{[0,1]}(3,2,2)m_{c}^{2}m_{s } \\ & & + 4\hat{i}_{1}(3,1,2)m_{c}m_{s}^{2}+4\hat{i}_{0}(3,1,2)m_{c}m_{s}^{2}-2\hat{i}_1^{[1,0]}(3,2,2)m_{c}m_{s}^{2}+12\hat{i}_{1}(1,1,4)m_{c}m_{s}^{2 } \\ & & -2\hat{i}_2^{[1,0]}(3,2,2)m_{c}m_{s}^{2}+12\hat{i}_{2}(1,1,4)m_{c}m_{s}^{2}-2\hat{i}_0^{[1,0]}(3,2,2)m_{c}m_{s}^{2}+6\hat{i}_{2}(3,1,2)m_{c}m_{s}^{2 } \\ & & + 12\hat{i}_{0}(1,1,4)m_{c}m_{s}^{2}-4\hat{i}_{2}(3,1,2)m_{s}^{3}+2\hat{i}_2^{[1,0]}(3,2,2)m_{s}^{3}-12\hat{i}_{2}(1,1,4)m_{s}^{3 } \\ & & -4\hat{i}_{2}(2,1,3)m_{s}^{3}-2\hat{i}_{2}(2,2,2)m_{s}^{3}+2\hat{i}_{2}(3,1,1)m_{c}-4\hat{i}_2^{[1,0]}(3,2,1)m_{c } \\ & & -4\hat{i}_1^{[0,1]}(3,1,2)m_{c}+4\hat{i}_{0}(1,2,2)m_{c}-2\hat{i}_1^{[1,0]}(3,1,2)m_{c}+4\hat{i}_{2}(1,2,2)m_{c } \\ & & + 8\hat{i}_{0}(2,1,2)m_{c}+8\hat{i}_{2}(2,1,2)m_{c}+2\hat{i}_{0}(3,1,1)m_{c}+4\hat{i}_{1}(1,2,2)m_{c } \\ & & -6\hat{i}_1^{[1,0]}(3,2,1)m_{c}+2\hat{i}_{0}(2,2,1)m_{c}-4\hat{i}_0^{[0,1]}(3,1,2)m_{c}-4\hat{i}_2^{[0,1]}(3,1,2)m_{c } \\ & & + 2\hat{i}_2^{[1,1]}(3,2,2)m_{c}+2\hat{i}_1^{[1,1]}(3,2,2)m_{c}+2\hat{i}_{1}(2,2,1)m_{c}+6\hat{i}_{1}(2,1,2)m_{c } \\ & & + 2\hat{i}_0^{[1,1]}(3,2,2)m_{c}-6\hat{i}_0^{[1,0]}(3,2,1)m_{c}-2\hat{i}_{1}(3,1,1)m_{c}+8\hat{i}_1^{[1,0]}(2,1,3)m_{s } \\ & & + 4\hat{i}_2^{[1,0]}(3,2,1)m_{s}+12\hat{i}_{1}(1,1,3)m_{s}-4\hat{i}_{2}(1,2,2)m_{s}+2\hat{i}_2^{[0,1]}(2,2,2)m_{s } \\ & & -2\hat{i}_{2}(2,2,1)m_{s}+4\hat{i}_2^{[0,1]}(3,1,2)m_{s}-2\hat{i}_2^{[1,1]}(3,2,2)m_{s}+4\hat{i}_{1}(2,1,2)m_{s } \\ & & + 4\hat{i}_{2}(1,1,3)m_{s}+4\hat{i}_{0}(2,1,2)m_{s}-10\hat{i}_{2}(2,1,2)m_{s}+2\hat{i}_2^{[1,0]}(2,2,2)m_{s } \\ & & + 4\hat{i}_2^{[0,1]}(2,1,3)m_{s}+20\hat{i}_{0}(1,1,3)m_{s},\end{aligned}\ ] ] @xmath137}(3,2,2)m_{c}^{3}-2\hat{i}_{2}(3,2,1)m_{c}^{3}+2\hat{i}_2^{[0,1]}(3,2,2)m_{c}^{3}+2\hat{i}_0^{[0,1]}(3,2,2)m_{c}^{3 } \\ & & + 4\hat{i}_{1}(2,2,2)m_{c}^{3}+2\hat{i}_0^{[1,0]}(3,2,2)m_{c}^{3}+4\hat{i}_{0}(2,2,2)m_{c}^{3}+6\hat{i}_{2}(4,1,1)m_{c}^{3 } \\ & & + 2\hat{i}_2^{[1,0]}(3,2,2)m_{c}^{3}+4\hat{i}_{2}(2,2,2)m_{c}^{3}+2\hat{i}_{2}(3,1,2)m_{c}^{3}+2\hat{i}_{0}(3,1,2)m_{c}^{3 } \\ & & + 6\hat{i}_{1}(4,1,1)m_{c}^{3}+2(\hat{i}_0^{[0,1]}(3,2,2)m_{c}^{3}+6\hat{i}_{0}(4,1,1)m_{c}^{3}+2\hat{i}_2^{[0,1]}(3,2,2)m_{c}^{2}m_{s } \\ & & + 6\hat{i}_{2}(4,1,1)m_{c}^{2}m_{s}-2\hat{i}_{1}(2,2,2)m_{c}^{2}m_{s}+2\hat{i}_{0}(3,1,2)m_{c}^{2}m_{s}-2\hat{i}_{1}(3,1,2)m_{c}^{2}m_{s } \\ & & + 8\hat{i}_{0}(2,1,3)m_{c}^{2}m_{s}+2\hat{i}_{2}(3,1,2)m_{c}^{2}m_{s}+8\hat{i}_{1}(2,1,3)m_{c}^{2}m_{s}+2\hat{i}_{2}(2,2,2)m_{c}^{2}m_{s } \\ & & + 2\hat{i}_2^{[1,0]}(3,2,2)m_{c}^{2}m_{s}+4\hat{i}_{2}(2,1,3)m_{c}^{2}m_{s}-2\hat{i}_{0}(2,2,2)m_{c}^{2}m_{s}-2\hat{i}_1^{[1,0]}(3,2,2)m_{c}m_{s}^{2 } \\ & & + 12\hat{i}_{1}(1,1,4)m_{c}m_{s}^{2}+6\hat{i}_{0}(3,1,2)m_{c}m_{s}^{2}+4\hat{i}_{1}(3,1,2)m_{c}m_{s}^{2}+12\hat{i}_{2}(1,1,4)m_{c}m_{s}^{2 } \\ & & + 2\hat{i}_{2}(3,2,1)m_{c}m_{s}^{2}-2\hat{i}_0^{[1,0]}(3,2,2)m_{c}m_{s}^{2}-2\hat{i}_2^{[1,0]}(3,2,2)m_{c}m_{s}^{2}+12\hat{i}_{0}(1,1,4)m_{c}m_{s}^{2 } \\ & & + 4\hat{i}_{2}(3,1,2)m_{c}m_{s}^{2}-2\hat{i}_2^{[1,0]}(3,2,2)m_{s}^{3}+4\hat{i}_{2}(3,1,2)m_{s}^{3}+12\hat{i}_{2}(1,1,4)m_{s}^{3 } \\ & & + 4\hat{i}_{2}(2,2,2)m_{s}^{3}-4\hat{i}_{2}(2,1,3)m_{s}^{3}-6\hat{i}_2^{[0,1]}(3,1,2)m_{c}-2\hat{i}_{2}(2,2,1)m_{c } \\ & & + 4\hat{i}_{0}(1,2,2)m_{c}-4\hat{i}_1^{[1,0]}(3,2,1)m_{c}+4\hat{i}_{2}(1,2,2)m_{c}+10\hat{i}_{0}(2,1,2)m_{c } \\ & & + 2\hat{i}_1^{[1,1]}(3,2,2)m_{c}+4\hat{i}_{1}(1,2,2)m_{c}+8\hat{i}_{1}(2,1,2)m_{c}+10\hat{i}_{2}(2,1,2)m_{c } \\ & & -4\hat{i}_2^{[1,0]}(3,2,1)m_{c}-4\hat{i}_1^{[0,1]}(3,1,2)m_{c}-6\hat{i}_0^{[0,1]}(3,1,2)m_{c}+2\hat{i}_{0}(3,1,1)m_{c } \\ & & + 2\hat{i}_0^{[1,1]}(3,2,2)m_{c}-2\hat{i}_2^{[0,1]}(3,2,1)m_{c}-4\hat{i}_0^{[1,0]}(3,2,1)m_{c}+2\hat{i}_2^{[1,1]}(3,2,2)m_{c } \\ & & + 2\hat{i}_{2}(1,2,2)m_{s}-4\hat{i}_2^{[1,0]}(3,2,1)m_{s}-2\hat{i}_{1}(1,2,2)m_{s}+2\hat{i}_0^{[1,0]}(2,2,2)m_{s } \\ & & + 6\hat{i}_{1}(1,3,1)m_{s}+4\hat{i}_{2}(1,1,3)m_{s}-2\hat{i}_2^{[1,0]}(2,2,2)m_{s}-4\hat{i}_{1}(1,1,3)m_{s } \\ & & -2\hat{i}_{1}(2,1,2)m_{s}-4\hat{i}_2^{[0,1]}(2,2,2)m_{s}-6\hat{i}_{0}(2,1,2)m_{s}-2\hat{i}_{0}(1,2,2)m_{s } \\ & & + 4\hat{i}_2^{[0,1]}(2,1,3)m_{s}-4\hat{i}_{0}(1,1,3)m_{s}+2\hat{i}_1^{[1,0]}(2,2,2)m_{s}-4\hat{i}_2^{[0,1]}(3,1,2)m_{s } \\ & & + 2\hat{i}_2^{[1,1]}(3,2,2)m_{s}+2\hat{i}_{2}(2,1,2)m_{s}+4\hat{i}_{2}(2,2,1)m_{s},\end{aligned}\ ] ] where @xmath138 } ( a , b , c)&= & [ m_1 ^ 2]^{\alpha}[m_2 ^ 2]^{\beta}\frac{d^\alpha}{d(m_1 ^ 2)^{\alpha } } \frac{d^\beta}{d(m_2 ^ 2)^{\beta}}[m_1 ^ 2]^{\alpha}[m_2 ^ 2]^{\beta}\hat i_{\mu } ( a , b , c ) , \nonumber \\ \hat{i}_k(a , b , c ) \!\!\ ! & = & \!\!\ ! i \frac{(-1)^{a+b+c+1}}{16 \pi^2\,\gamma(a ) \gamma(b ) \gamma(c ) } ( m_1 ^ 2)^{1-a - b+k } ( m_2 ^ 2)^{4-a - c - k } \ , { u}_0(a+b+c-5,1-c - b ) , \nonumber \\ \hat{i}_m(a , b , c ) \!\!\ ! & = & \!\!\ ! i \frac{(-1)^{a+b+c+1}}{16 \pi^2\,\gamma(a ) \gamma(b ) \gamma(c ) } ( m_1 ^ 2)^{-a - b-1+m } ( m_2 ^ 2)^{7-a - c - m } \ , { u}_0(a+b+c-5,1-c - b ) , \nonumber\\ \hat{i}_6(a , b , c ) \!\!\ ! & = & \!\!\ ! i \frac{(-1)^{a+b+c+1}}{32 \pi^2\,\gamma(a ) \gamma(b ) \gamma(c ) } ( m_1 ^ 2)^{3-a - b } ( m_2 ^ 2)^{3-a - c } \ , { u}_0(a+b+c-6,2-c - b ) , \nonumber\\ \hat{i}_n(a , b , c ) \!\!\ ! & = & \!\!\ ! \pi^2\,\gamma(a ) \gamma(b ) \gamma(c ) } ( m_1 ^ 2)^{-4-a - b+n } ( m_2 ^ 2)^{11-a - c - n } \ , { u}_0(a+b+c-7,2-c - b),\end{aligned}\ ] ] where @xmath139 , @xmath140 and @xmath141 . we can define the function @xmath142 as : @xmath143,\end{aligned}\ ] ] where @xmath144

a special asymptotic solution of the painlev-2 equation with small parameter is studied . this solution has a critical point corresponding to a bifurcation phenomenon . when the constructed solution varies slowly and when the solution oscillates very fast . we investigate the transitional layer in detail and obtain a smooth asymptotic solution , using a sequence of scaling and matching procedures . [ kiselev - firstpage ]

math-ph0101037

in this work a special asymptotic solution for the equation painlev-2 @xmath3 is constructed as @xmath4 . the behaviour of wanted solution differs in different intervals of the parameter @xmath5 . the qualitative behaviour of numerical solution is indicated in the figure ( at @xmath6 ) . calculations for another small values of @xmath7 give pictures like this . let us explain these numeric results using asymptotic theory for small @xmath7 . in the area i the special solution is approximated by an asymptotic solution in which the leading term is a least root of a cubic equation @xmath8 . corrections are algebraic functions of @xmath5 . such algebraic asymptotics becomes invalid near the point @xmath0 , where two of the roots of the cubic equation coalesce . in the left neighborhood of the point @xmath0 this asymptotics goes into an asymptotics of a special solution of the painlev-1 equation with respect to new scaling variable @xmath9 . this is the area ii in the figure . this special solution of the painlev-1 equation has poles at @xmath10 , @xmath11 . in the neighborhoods of these poles one more scaling is done . new variable is @xmath12 ( area iii ) . in this area the asymptotics is defined by separatrix solution of a nonlinear autonomous equation . this combined asymptotic structure becomes invalid as @xmath13 , because the poles of the solution painlev-1 equations close to each other . as @xmath13 a fast oscillating asymptotics is valid . it is the area iv in the figure . the qualitative behaviour of solutions of second - order ordinary differential equations with respect to an additional parameter was explained , for example , in the book @xcite . in @xcite the various types of bifurcations for equilibrium positions of conservative second - order ordinary differential equations were described also . the equation ( [ p2 ] ) is nonautonomous , however one can separate the dependence on slow @xmath5 and fast variables ( @xmath14 ) in the asymptotic solution of this equation . we may speak about the phase plane and phase trajectory of this equation with respect to the fast variable @xmath15 . the equilibrium positions on the phase plane depend on @xmath5 . when @xmath1 there are three equilibrium positions . at the critical point @xmath0 the bifurcation saddle - center " occurs . it means that one of stable and unstable equilibrium positions coalesce . when @xmath2 only one equilibrium position exists . such bifurcation leads to instability @xcite . the bifurcations of slowly varying equilibrium positions of second order equation with the algebraic nonlinearity and slowly varying parameters were considered in @xcite only in a preliminary fashion . when this text was written , new work @xcite became known . in that work , a change of an energy and a phase jump has been studied for a solution in a very narrow layer near a saddle - center bifurcation point in general case . however results of the work @xcite are inapplicable to the painlev-2 equation , because this equation has degenerate behaviour with respect to equations considered in @xcite . amongst another works , in which the asymptotic solutions of the nonlinear equations with varying coefficients were studied we should note the work @xcite . in that work a changing of an adiabatic invariant was studied in a problem when a solution passed through separatrix in the nondegenerate case . one more work where the passage through the separatrix in nondegenerate case was studied is @xcite . more complicated bifurcation is the pitchfork in the equation painlev-2 . asymptotics with respect to a small parameter of the solutions for the equation painlev-2 with zero in the right hand side of the equation ( [ p2 ] ) was investigated in works @xcite . in this case a solution in an interior layer near a bifurcation value of the parameter @xmath0 is determined by the equation painlev-2 , but already without a small parameter ; and the problem , generally speaking , does not become simpler . the asymptotics of the solutions for the painlev equations with a leading term as an elliptic function with the modulated parameters were studied , for example , in works @xcite@xcite . we must mention some of the works about the scaling limits or double asymptotics for the painlev-2 equation . the general approach to the scaling limits of the painlev equations based on the bcklund transformations was studied in @xcite . the scaling limit passage from the equation painlev-2 to painlev-1 was studied on the level of classical solutions in the work @xcite . in work @xcite different approaches to the double asymptotics were developed . the qualitative analysis for the relation of the algebraic and fast oscillating asymptotic solutions of the equation ( [ p2 ] ) was done in the work @xcite also . however , the asymptotic solutions constructed by this way are non - uniform with respect to two variable @xmath5 and @xmath7 . the major difference of the presented work in comparsion with the others cited is constructing uniform asymptotic solution with respect to two parameters @xmath5 and @xmath7 as @xmath16 into an interval of @xmath5 , where the main term of asymptotics ( elliptic function ) is degenerated . this uniform asymptotic solution is valid on a segment containing the saddle - center bifurcation point . in this work we specify a different types of asymptotic approximations of the being studied solution , their valid intervals and orders of neglected terms which result when these asymptotics are substituted into a being solved equation . the contents of the various sections are as follows . section 2 states the main problem . the results are formulated in section 3 . an asymptotic expansion to be valid before the bifurcation point @xmath0 is written in section 4 . section 5 is devoted to inner asymptotic expansions which are matched with each other and with asymptotics from section 4 . a fast oscillating asymptotic expansion of kuzmak - type , a degeneration of the oscillations and the matching of the fast oscillations with inner expansions from section 5 are described in section 6 . open problems are discussed in section 7 . let s consider a cubic equation : @xmath17 which is obtained as a rejection of the term with the small parameter in the equation ( [ p2 ] ) . there exist the point @xmath0 and the value @xmath18 such , that if @xmath19 , then @xmath18 is double root of the equation ( [ kub ] ) . the values @xmath18 and @xmath0 are easy to obtain by solving of the equations : @xmath20 there are @xmath21 , @xmath22 . the discriminant of the equation ( [ kub ] ) has the form : @xmath23 the discriminant @xmath24 when @xmath1 and hence the cubic equation ( [ kub ] ) has three real roots @xmath25 . if @xmath2 then @xmath26 and the cubic equation ( [ kub ] ) has one real root and two complex conjugate roots . at @xmath19 the roots @xmath27 and @xmath28 coalesce @xmath29 . when @xmath1 it is possible to construct a real formal solution of the equation ( [ p2 ] ) : @xmath30 by taking any of roots @xmath31 as a leading term @xmath32 . these three formal solutions are slowly varying equilibrium positions for the equation ( [ p2 ] ) . consider an equation @xmath33 which describe small perturbations of the leading term @xmath31 of the asymptotic expansions ( [ form ] ) . if @xmath34 , then @xmath35 , so that the equation ( [ l1 ] ) has one exponentially growing solution and one exponentially decreasing solution . hence corresponding asymptotic expansion is unstable with respect to small perturbation of the leading term . otherwise if @xmath36 then the coefficients in the equation ( [ l1 ] ) are positive and the formal solution ( [ form ] ) with @xmath37 as the leading term of the asymptotics is stable . when @xmath19 the roots @xmath27 and @xmath28 coalesce : @xmath29 and when @xmath2 , there exists only one equilibrium @xmath38 . therefore @xmath0 is the bifurcation point for the asymptotic solution ( [ form ] ) in case the leading term @xmath39 . our proposal is to construct a smooth asymptotic solution of the equation ( [ p2 ] ) with the leading term @xmath27 when @xmath1 on a segment @xmath40 $ ] , @xmath41 . here we describe a smooth asymptotic solution constructed in this work . following by v p maslov @xcite we will use the words asymptotic solution with respect to @xmath42 " , namely , a function is said to be an asymptotic solution of @xmath43 of the equation ( [ p2 ] ) if after its substitution into this equation the latter is satisfied up to the terms of the order @xmath44 . when @xmath45 @xmath46 and @xmath47 the asymptotic solution with respect to @xmath48 has the form @xmath49 the last term of the formal asymptotic solution as @xmath50 can be written as : @xmath51 when @xmath52 the asymptotic solution is defined by two different types of the formal asymptotic expansions . the first one has the form @xmath53 here the variable @xmath54 is defined by the formula @xmath9 , the function @xmath55 is defined as the solution of the equation painlev-1 : @xmath56 with the pure algebraic asymptotic behavior as @xmath57 : @xmath58 the formula ( [ exp20 ] ) is asymptotic solution with respect to @xmath59 as @xmath60 . the function @xmath55 has poles of second order at some points @xmath61 , @xmath62 ( see , e.g. @xcite ) : @xmath63 near the poles the last term of the asymptotic solution can be written as @xmath64 the expansion ( [ exp20 ] ) is suitable at @xmath65 . the formula ( [ exp20 ] ) is asymptotic solution with respect to @xmath66 . as @xmath13 the main term of the asymptotics ( [ exp20 ] ) is @xmath67 here @xmath68 the phase shift @xmath69 is defined in section 5.4.1 by a formula ( [ shift1 ] ) . the parameter of the weierstrass elliptic function @xmath70 . the second parameter @xmath71 is defined by a solution of an equation ( see , e.g. @xcite ) : @xmath72 where @xmath73 is any circle on an algebraic curve @xmath74 . the last term of the asymptotics ( [ exp20 ] ) has the form : @xmath75 the formula ( [ exp20 ] ) is asymptotic solution with respect to @xmath76 @xmath77 . the expansion ( [ exp20 ] ) is suitable as @xmath78 and @xmath79 . the second one , which is valid in the neighborhoods @xmath80 of poles @xmath61 of the function @xmath55 , reads as @xmath81 where @xmath82 . the phase shift @xmath83 is defined by a formula ( [ shift2 ] ) . the function @xmath84 is defined by the formula : @xmath85 the last term of the formal asymptotics ( [ exp30 ] ) at @xmath86 can be written as @xmath87 the formula ( [ exp30 ] ) is asymptotic solution with respect to @xmath88 . when @xmath89 and @xmath90 the asymptotic solution with respect to @xmath91 has the fast oscillating behavior : @xmath92 here the last term of the asymptotics ( [ exp40 ] ) at @xmath93 can be written as @xmath94 the leading term of the asymptotic solution satisfies the cauchy problem : @xmath95 here @xmath96 . the function @xmath97 is defined by the equation @xmath98 where @xmath99 and @xmath100 ( @xmath101 ) are two real roots of the equation @xmath102 , other roots of this equation are complex . the phase function @xmath103 is the solution of the cauchy problem : @xmath104 where @xmath105 is the constant defined by the formula @xmath106 where @xmath107 is the unique solution of the equation @xmath108^{5/2}}y^{5/2}=0,\ ] ] and @xmath109}}.\ ] ] the phase shift @xmath110 is defined by an equation ( see @xcite ) : @xmath111 * remark 1 . * in this work the constant @xmath112 remains out side of our analysis . its value may be defined by using the monodromy - preserve method for the painlev-2 equation @xcite . * remark 2 . * the domains of validity of the asymptotic solution ( [ exp10 ] ) and the asymptotic solution ( [ exp20 ] ) intersect , so that these expansions match . the solution of the painlev-1 equation which defines the asymptotics ( [ exp20 ] ) has infinite sequence of the poles @xmath61 , @xmath113 near all of these poles we match the asymptotic solutions ( [ exp20 ] ) and ( [ exp30 ] ) . as the number of the pole @xmath114 the domain of validity of this complicated combine asymptotics ( [ exp20 ] ) and ( [ exp30 ] ) intersects with the domain of validity for the fast oscillating asymptotic solution ( [ exp40 ] ) . its allows to match the sandwiched asymptotics with the fast oscillating asymptotic solution . the algebraic asymptotic solution ( [ exp10 ] ) of the equation ( [ p2 ] ) is constructed here . this asymptotic solution is suitable when @xmath1 and asymptotic behavior of this solution is investigated as @xmath115 . we construct the asymptotic solution of the equation ( [ p2 ] ) as : @xmath116 let s formulate the result of this section . the asymptotic solution ( [ exp10 ] ) with respect to @xmath117 , where @xmath39 is least of the solutions of the equation ( [ kub ] ) , is suitable when @xmath118 and @xmath119 where @xmath41 . let s obtain the coefficients of the asymptotics ( [ exp1 ] ) . substituting the ansatz ( [ exp1 ] ) into the equation ( [ p2 ] ) and equating coefficients at identical powers of @xmath7 we find the sequence of the formulas for @xmath120 , @xmath121 . @xmath122 @xmath123 the cubic equation for @xmath32 when @xmath124 has three real roots @xmath25 . as the leading term of asymptotic expansion ( [ exp1 ] ) we choose @xmath27 . the second derivative of @xmath32 has the form : @xmath125 this allows to obtain the formula for @xmath126 . it is easy to get the expressions for the following terms of the asymptotic solution ( [ exp1 ] ) . in an explicit form they are not adduced here , however , it is important to note , that the power of the denominator @xmath127 in the coefficients of the asymptotics grows with each next step . the @xmath128-th term of the asymptotic expansion as @xmath129 has the form @xmath130 let s write the asymptotic behavior of the asymptotic expansion ( [ exp1 ] ) as @xmath131 . for this purpose we shall calculate the asymptotics of the expression @xmath127 : @xmath132 using this formula and @xmath32 , @xmath126 we obtain : @xmath133\ ! \vspace{3mm}\\ \ds \phantom{u(t,\varepsilon)= } + o\left(\varepsilon^4(t_*-t)^{-9/2}\right)+o\left((t_*-t)^{3/2}\right ) . \ea\ ] ] the domain of validity for this expansion as @xmath134 is determined from the relation @xmath135 . it follows from the formula ( [ un ] ) , that the expansion ( [ exp1 ] ) is suitable when @xmath136 . evaluate the residual which is obtained when one substitutes the asymptotic solution ( [ exp10 ] ) into the equation ( [ p2 ] ) @xmath137 using the asymptotic behaviour of the @xmath138 , @xmath139 as @xmath140 one can obtain @xmath141 in this section the asymptotic expansions of solution of ( [ p2 ] ) which are suitable in the small neighborhood of a point @xmath0 are constructed . by following terminology of the matching method @xcite , they are called the inner asymptotic expansions " . it follows from the consideration of the validity of the outer expansion , made in the previous section , that it is natural to make the following scaling of variables : @xmath142 as a result we write the equation ( [ p2 ] ) as @xmath143 in the limit as @xmath4 we obtain the equation painlev-1 . this asymptotic reduction is known as one of the scaling limits for the painlev-2 equation @xcite ( see also @xcite ) . a solution of this equation has the asymptotic expansion as @xmath144 : @xmath145 the asymptotic solution of the equation ( [ p1 ] ) we build as : @xmath146 where the function @xmath55 is the solution of the painlev-1 equation . here it is shown , that the asymptotic solution ( [ exp2 ] ) is suitable in the neighborhood of infinity ( when @xmath147 ) and in the neighborhood of the poles for the function @xmath55 : @xmath148 . the coefficients of the asymptotics are calculated from the matching condition for the asymptotic expansion ( [ exp1 ] ) as @xmath131 and the expansion ( [ exp2 ] ) as @xmath144 . in particular , @xmath55 has the algebraic asymptotics : @xmath150 in the book @xcite it is shown , that there exists the solution of the painlev-1 equation with the asymptotics ( [ asp11 ] ) . the data of a monodromy for the solution of the painlev-1 equation with the asymptotics ( [ asp11 ] ) are calculated in the work @xcite . the first correction in the asymptotics ( [ exp2 ] ) satisfies the equation @xmath151 the asymptotics of the solution for this equation as @xmath144 has the form @xmath152 asymptotics of the higher corrections is constructed by ordinary way . the @xmath128-th correction as @xmath144 has an order : @xmath153 the requirement of validity for the asymptotics is @xmath154 . it reduces to the condition @xmath155 . the residual of the asymptotic solution ( [ exp20 ] ) has the form : @xmath156 using the asymptotic behaviour of @xmath157 , @xmath158 as @xmath155 one can obtain @xmath159 the function @xmath55 has the poles when @xmath160 . let s denote these poles by @xmath61 . in the neighborhood of the pole @xmath161 the function @xmath55 is defined by the converging power series ( see e.g. @xcite ) @xmath162 the constants @xmath61 and @xmath163 are the parameters of this solution . in the review @xcite it is noted , that the problem on the connection between the asymptotics of this solution at infinity and the constants @xmath61 and @xmath163 is not investigated yet . the points of the poles @xmath61 and appropriate constants @xmath163 can be obtained with the help of the numerical calculation using the given asymptotics at infinity ( [ asp11 ] ) . the asymptotics of @xmath164 as @xmath161 may be written as a sum of a certain solution of a nonhomogeneous linearized painlev-1 equation @xmath165 @xmath166 and two solutions of a homogeneous linearized equation @xmath167 , @xmath168 : @xmath169 @xmath170 thus : @xmath171 here @xmath172 and @xmath173 are constants . higher corrections have the same form : @xmath174 where @xmath175 by using the asymptotics ( [ asp12 ] ) and ( [ aslp1 ] ) it is easy to see , that the asymptotic expansion ( [ exp2 ] ) is suitable at @xmath177 the residual of the asymptotic solution as @xmath178 when @xmath65 is @xmath179 for the construction of the uniform asymptotics in the neighborhood of the pole of the function @xmath180 it is necessary to make one more scaling of the independent variable and the function ( see @xcite ) : @xmath181 for function @xmath182 we obtain the equation : @xmath183 the solution of this equation has following asymptotic expansion as @xmath184 : @xmath185 this long asymptotic formula shows , that the constant @xmath186 appears only in the correction of an order @xmath187 . if we want to construct the first correction of the asymptotics for the first inner expansion after the pole @xmath61 , we must construct the correction in order @xmath187 for the second inner expansion . it is convenient to include a time shift depended on @xmath7 into the main term , and construct the asymptotic expansion depended on a new time variable : @xmath188 where @xmath189 . we search the asymptotic expansion for the solution of this equation as a segment of an asymptotic series @xmath190 in this case the equation for the @xmath191 looks like : @xmath192 it is shown here , that the asymptotic solution ( [ exp30 ] ) is the formal asymptotic solution of the equation ( [ de3 ] ) with respect to @xmath193 when @xmath194 . the solution of the equation for the leading term of the asymptotics ( [ exp3 ] ) is defined by the asymptotics as @xmath178 of the asymptotic expansion ( [ exp2 ] ) , which is outer with respect to ( [ exp3 ] ) . this solution has the form @xmath195 the constants @xmath196 are defined by asymptotics of the function @xmath197 . using the formula ( [ w - left - as ] ) we obtain : @xmath198 the corrections in the expansion ( [ exp3 ] ) satisfy the linearized equations @xmath199 @xmath200 @xmath201 @xmath202 the expression for @xmath203 can be used to obtain two linearly independent solutions of the homogeneous equation for the corrections : @xmath204 @xmath205 \frac{1}{\left(1 + 4u_*^2\theta_k^2\right)^2}.\ ] ] by using these solutions of the homogeneous equation it is easy to get the solutions of the nonhomogeneous equations for the corrections . the asymptotics of the corrections as @xmath206 has the form : @xmath207 @xmath208 @xmath209 @xmath210 it is important to note that the leading term of the asymptotics of @xmath211 as @xmath212 is the same . this term defines the constants @xmath163 and hence the solution of the painlev-1 equation before and after the pole . we have the same value of the constant @xmath163 as @xmath213 and @xmath214 and then we have the same asymptotic solution of painlev-1 equation before and after the pole @xmath61 . an opposite result takes place for the coefficient @xmath215 in order @xmath216 as @xmath214 . this coefficient is changed . it is equal @xmath217 as @xmath213 and @xmath218 as @xmath214 . using the asymptotics for the corrections and the leading term , we obtain , that the expansion ( [ exp3 ] ) is suitable when @xmath219 . on the other hand , the expansion ( [ exp2 ] ) is suitable at @xmath220 . hence , the domains of the applicability for the expansions ( [ exp2 ] ) and ( [ exp3 ] ) are intersected at realization of the condition @xmath221 . if we take into account also the requirement of fitness of the asymptotic expansion ( [ exp3 ] ) , then get the restriction @xmath222 . from this inequality it follows , that in this section the formal asymptotic expansions to be suitable when @xmath223 are constructed . we calculate the residual of second internal asymptotic expansion using the asymptotic behaviour of @xmath224 , @xmath225 : @xmath226 when @xmath227 . using the asymptotic expansion of the second inner expansion as @xmath206 and the first inner expansion as @xmath228 we find that the first inner expansion after the pole has the form ( [ exp2 ] ) where @xmath229 here @xmath230 the shift @xmath231 may be calculated from the asymptotics of the second inner expansion as @xmath206 , for @xmath232 we have obtained : @xmath233 thus a behaviour of the asymptotic solution in the internal layer is combined by the first and the second inner asymptotic expansions . in the above sections we demonstrate the asymptotic behaviour of the asymptotic solution at @xmath144 and near the poles of the solution for the painlev-1 equation . below we study the asymptotic behaviour of the solution as @xmath13 . the regular asymptotic expansion on @xmath7 is constructed in the previous section concerning the first inner asymptotic expansion . this asymptotics is not valid as large @xmath54 . to use the first inner asymptotic expansion as @xmath13 we must modulate the parameters of the solution of the painlev-1 equation and cancel secular terms in the first and the second corrections in the asymptotic expansion ( [ exp2 ] ) using singular perturbation theory . instead of modulating the parameters of the solution of the painlev-1 equation it is more convenient to study the modulation equation for parameters of the main term of the asymptotics for the solution of the painlev-1 equation as @xmath13 this study will be developed in this subsection . the elliptic asymptotics of the solution for the painlev-1 equation as @xmath13 was obtained by p boutroux @xcite . here we are interested in a connection formula for the solution of the painlev-1 equation . namely we have the asymptotic behaviour of the solution as @xmath144 and we need the asymptotic behaviour of the same solution as @xmath13 . the painlev-1 equation is integrable by the monodromy - preserving method @xcite . if the monodromy data are known , then the solution of the painlev-1 equation is uniquely defined . in the correspondence with @xcite , in our case the monodromy data are constants @xmath235 and @xmath236 and these constants are equal to zero . the asymptotics of the function @xmath55 outside of the poles has the form ( see e.g. @xcite ) @xmath237 where @xmath238 is some constant , the function @xmath239 is determined by the weierstrass elliptic function @xmath240 the phase function @xmath241 . it is important to note , that in the formula ( [ asp1 ] ) the shift of the phase function @xmath242 is equal to zero . a parameter is @xmath70 and a parameter @xmath71 is defined as a solution of the equation : @xmath243 where @xmath244 is any circle on an algebraic curve : @xmath245 . we are interested in the asymptotic solution of the perturbed painlev-1 equation ( [ p1 ] ) . therefore the asymptotics of the works @xcite for unperturbed painlev-1 equation are the main term of our asymptotics on @xmath7 . substitute into ( [ p1 ] ) : @xmath246 where @xmath247 . as a result we obtain an equation : @xmath248 let us to construct the asymptotics of the perturbed equation as a segment of the asymptotic series : @xmath249 where @xmath250 @xmath69 is modulated phase shift and @xmath251 is once more slow variable . substitute the formula ( [ w - as ] ) into the equation ( [ w - p ] ) . then let s equate the coefficients with identical powers of @xmath7 . as a result we obtain the sequence of the equations : @xmath252 , \label{w - e - t}\ ] ] @xmath253 , \ea \label{lame1-t}\ ] ] @xmath254 here @xmath255 are linear operators . solutions of these equations may be represented as asymptotic expansions as @xmath256 . we will assume that the corrections of these expansions are small as @xmath256 . therefore we will neglect the correction terms of this asymptotics and consider only the main terms of these asymptotics with respect to @xmath242 . the equations for the main terms @xmath257 , @xmath258 , @xmath259 have the form : @xmath260 @xmath261 @xmath262 the requirement of validity for the first inner asymptotics at @xmath13 is @xmath263 , where @xmath264 we find the modulated equation for the @xmath69 such that the first inner expansion is valid as large @xmath54 . the solution of the equation for the main term is the function @xmath265 which is defined by ( [ ro0 ] ) . the equations for the corrections is the lame equations with external force . to write the solutions of these equations we will use two linear independent solutions of the lame equation . denote one of these solutions by @xmath266 we denote the second solution as @xmath267 . the solutions @xmath268 and @xmath267 are such that a wronskian : @xmath269 the second solution is aperiodic : @xmath270 the solution of the equation ( [ lame1 ] ) can be written as : @xmath271 here @xmath272 , where @xmath273 is pole of the function @xmath274 and @xmath275 is a real period of the @xmath276 . the @xmath277 and @xmath278 are constants . the first correction @xmath279 is bounded as @xmath280 if @xmath281 in this formula the integral must be regularized . namely : @xmath282.\ ] ] using a perturbation theory for the second order equation developed in works @xcite@xcite , we can show , that the solution for the @xmath283 has an order @xmath284 , @xmath285 , @xmath286 , if the function @xmath69 is a solution of the differential equation : @xmath287 this equation and the equation for the @xmath278 allow to write the cauchy problem for the @xmath69 in the form : @xmath288 where @xmath289 the constants @xmath277 and @xmath278 are defined by matching conditions of the asymptotics ( [ p1-as - elliptic ] ) , ( [ w - as ] ) and ( [ exp2 ] ) as @xmath290 : @xmath291 @xmath292 as a result we obtain the formula ( [ exp201 ] ) for the main term of the asymptotics ( [ w - as ] ) . let us denote by @xmath293 then the condition of the validity @xmath294 , @xmath295 fulfills as @xmath296 or as the same : @xmath297 near the pole @xmath298 we obtain the asymptotics : @xmath299 hence the first inner asymptotic expansion is suitable near the poles @xmath61 as @xmath300 the residual of the first inner expansion is @xmath301 using the results of the above section outside of the poles of @xmath180 one can obtain : @xmath302 in the second inner expansion the asymptotics as @xmath13 corresponds to the asymptotics at @xmath304 . it is easy to see , that in this case the first correction grows . this growth limits the value @xmath61 , at which the asymptotics ( [ exp3 ] ) is correct : @xmath222 . one can use the formula ( [ res - int2 ] ) to obtain the residual of the second inner asymptotics as @xmath304 . in this section we apply formulas obtained in @xcite@xcite , @xcite to the fast oscillating formal asymptotic solution of the painlev-2 equation . these formulas are usable when @xmath2 . the fast oscillating asymptotics is constructed as @xmath305 as the argument @xmath306 we use expression @xmath307 where @xmath103 and @xmath110 are unknown functions . the equations for the leading term and the corrections of the asymptotics ( [ exp4 ] ) look like : @xmath308 @xmath309 @xmath310 integrating once with respect to @xmath306 the equation for @xmath311 we obtain : @xmath312 where @xmath97 is the constant of integration " . we study the equation when the right - hand side has two real roots @xmath313 . we define the initial data as @xmath314 solution of the equation ( [ eq - u ] ) is an elliptic function . the left hand side of the formula ( [ eq - u ] ) is positive for real functions @xmath315 and @xmath311 . the term of highest order in the right hand side is @xmath316 . therefore the real solution of the equation ( [ eq - u ] ) has no poles when @xmath317 and @xmath318 . to construct an uniform asymptotics ( [ exp4 ] ) we choose the unknown functions @xmath103 , @xmath110 and the constant of integration " @xmath97 by a special way . they must satisfy to anti - resonant conditions for equations ( [ eq-2u])([eq - u2 ] ) . the first condition is the boundedness of the right hand side of ( [ eq-1u ] ) as @xmath319 . it is satisfied if a period of the oscillates of the function @xmath320 on @xmath306 is a constant ( see for example @xcite ) : @xmath321 the next condition is a boundedness of the first correction @xmath322 when @xmath319 . it gives equation defining a main term of an action @xmath323 ( see , @xcite ) : @xmath324 ^ 2dt_1= \mbox{const}.\ ] ] using the explicit expression for the derivative with respect to @xmath306 we present this formula in a some other form : @xmath325 where @xmath99 and @xmath100 are the solutions of the equation @xmath326 . the equation ( [ eq - s ] ) means that the action is a constant . at the other hand - side the equation ( [ eq - s ] ) defines the energy @xmath97 as a function with respect to slow time @xmath5 . at last the necessary condition of boundedness of the second correction @xmath327 with respect to @xmath306 is the equation ( see , @xcite ) : @xmath328 the equations ( [ eq - e])([eq - phi ] ) define the parameters of the main term of the asymptotics as functions with respect to @xmath5 . to find these functions one should define corresponding constants . notice that to define the value of the action @xmath323 we construct the asymptotic solution of the equation ( [ p2 ] ) when @xmath2 . thus the polynomial of the fourth power on @xmath311 in the right hand side of the equation ( [ eq - u ] ) can have no more than two various real roots @xmath99 and @xmath100 . hence this polynomial can be submitted as : @xmath329 the degeneration of the elliptic integral at @xmath19 corresponds to the case @xmath330 and @xmath331 . for this case it is easy to calculate the constant in the right hand side of the equation ( [ eq - s ] ) , which is equal to @xmath332 ( i.e. @xmath333 ) and the value of the parameter @xmath334 . in this subsections we calculate the asymptotic behaviour of the phase functions @xmath103 as @xmath93 . the oscillating solution is degenerated as @xmath93 . let s construct the asymptotics of this solution in the neighborhood of the degeneration point . for this purpose we calculate the asymptotics of the phase function @xmath103 and the function @xmath97 . let s write the equation ( [ eq - s ] ) as : @xmath335}\ , dx=\pi , \label{eq - e1}\ ] ] where @xmath336 , @xmath337 , @xmath338 , @xmath128 are real functions when @xmath339 . these functions satisfy the vita equations : @xmath340 the equation ( [ eq - e ] ) and three equation from ( [ vieta ] ) define the dependency @xmath341 , @xmath337 , @xmath338 , @xmath342 on the parameter @xmath5 . the last equation in ( [ vieta ] ) defines the function @xmath97 . let s make changes of variables : @xmath343 , @xmath344 , @xmath345 after simple transformations of the equations ( [ vieta ] ) we obtain : @xmath346 + \left[2m^2_1 - 2n^2+\eta\right]2m_1=0 , \vspace{1mm}\\ \ds m_*^2\left(12m_1 ^ 2 - 4n^2+\eta\right)+2m_*m_1 \left(6m_1 ^ 2 - 2n^2+\eta\right ) \vspace{1mm}\\ \ds \qquad \qquad + \left(3m^2_1-n^2+\eta\right)\left(m_1 ^ 2+n^2\right)=-g_1 . \label{v2 } \ea\ ] ] construct the solution of this system as @xmath93 as : @xmath347 let s substitute these expressions in ( [ v2 ] ) , equate the coefficients at the identical powers of @xmath348 . as a result we obtain : @xmath349 to define the constants @xmath350 and @xmath351 it is necessary to construct the asymptotics as @xmath352 of the left hand side of the equation ( [ eq - e1 ] ) . the asymptotics of the outside the integral coefficient in the equation ( [ eq - e1 ] ) has the form @xmath353 . \label{a - b}\ ] ] the integral in the equation ( [ eq - e1 ] ) is presented as @xmath354 where @xmath355 the value of the constant @xmath356 will be defined from an asymptotics below . the asymptotics of an integral @xmath357 as @xmath358 has the form @xmath359 where @xmath360^{5/2}}y^{5/2}.\ ] ] first three terms in this formula are calculated by standard way . let s show as we can obtain the function @xmath361 . for this purpose the following trick ( @xcite ) is applicable . let s calculate third derivative with respect to @xmath362 of the function @xmath357 : @xmath363^{5/2}}.\ ] ] on the right hand side we replace @xmath364 by @xmath365 and we present the integral as @xmath366^{5/2}}+o(1 ) . \label{de1}\ ] ] solving the ordinary differential equation ( [ de1 ] ) in the neighborhood of @xmath367 , we get : @xmath368 where @xmath369^{5/2}}y^{5/2}.\ ] ] after that it is easy to obtain the asymptotics ( [ asi ] ) . to define the value of @xmath356 we substitute the asymptotics ( [ a - b ] ) and ( [ asi ] ) in ( [ eq - e1 ] ) and equate to zero the coefficients at identical powers of @xmath348 . in the result we get at @xmath370 the equation @xmath371 this is the transcendental equation for the definition of the parameter @xmath356 . the numerical solution gives @xmath372 . using the formula ( [ k1 ] ) , we get : @xmath373 to construct the asymptotics of @xmath103 as @xmath93 we use the equation connecting the period of fast oscillations with its phase ( @xcite ) : @xmath374}}. \label{eq - s2}\ ] ] present the integral in the right hand side as @xmath375}}.\ ] ] after the same replacements , as at the construction of the asymptotics @xmath376 , as @xmath358 we get : @xmath377 } } + o(1).\ ] ] we substitute this expression into the equation ( [ eq - s2 ] ) , use the asymptotics @xmath362 and @xmath378 as @xmath379 and in the result we get : @xmath380 where @xmath381}}.\ ] ] the period of the oscillations for the function @xmath320 with respect to the variable @xmath306 in the krylov bogolubov s method is an arbitrary constant . let s choose it such , that @xmath382 : @xmath383 in the result the phase of the oscillations as @xmath131 has a form @xmath384 where @xmath385 is some constant . its value will be defined below at the matching of the asymptotics ( [ exp4 ] ) and inner asymptotics ( [ exp2 ] ) , ( [ exp3 ] ) as @xmath93 . in this subsection we establish the domain of validity of the fast oscillating asymptotics and compute the residual of this asymptotic solution . the validity of the asymptotics is defined by the formula @xmath386 . let us check this requirement . for this we must obtain the order of the first correction as @xmath93 . evaluate the order of the right hand side of the equation for the first correction : @xmath387 from the equation for @xmath311 one can evaluate second term in @xmath388 as @xmath93 : @xmath389 one must reduce formula for the derivative of @xmath311 with respect to @xmath5 to evaluate of the first term in the formula for @xmath388 . the function @xmath311 is the inverse function with respect to the elliptic integral @xmath390 both limits of the integration are functions with respect to @xmath5 , it is not convenient for us . make the substitution @xmath391 . then we obtain @xmath392 now we differentiate this formula with respect to @xmath5 , as a result we obtain the formula for the @xmath393 $ ] : @xmath394 = \frac{1}{(\alpha-\beta)s ' } \sqrt{\left(\alpha-\overset{0}{u}\right)\left(\overset{0}{u}-\beta\right ) \left(\left(\overset{0}{u}-m\right)^2+n^2\right ) } \vspace{3mm}\\ \ds \phantom{\p_t \left[\frac{\overset{0}{u}-\beta}{\alpha-\beta}\right ] = } \times \left [ -\p_t\left(\frac{s'}{(\alpha-\beta)}\right ) \int_0^{\frac{\overset{0}{u}-\beta}{\alpha-\beta } } \frac{dz}{\sqrt{z(1-z)\left((z-\gamma)^2+\delta^2\right)}}\right . \vspace{3mm}\\ \ds \phantom{\p_t \left[\frac{\overset{0}{u}-\beta}{\alpha-\beta}\right ] = } \left . + \frac{s'}{(\alpha-\beta ) } \int_0^{\frac{\overset{0}{u}-\beta}{\alpha-\beta } } \frac{dz\,\delta\delta'-(z-\gamma)\gamma ' } { \sqrt{z(1-z)}\left((z-\gamma)^2+\delta^2\right)^{3/2}}\right ] , \ea\ ] ] where @xmath395 . in the same way we can obtain the formula : @xmath396 = \frac{1}{(\alpha-\beta)s ' } \sqrt{\left(\alpha-\overset{0}{u}\right)\left(\overset{0}{u}-\beta\right ) \left(\left(\overset{0}{u}-m\right)^2+n^2\right ) } \vspace{3mm}\\ \ds \phantom{\p_t \left[\frac{\overset{0}{u}-\beta}{\alpha-\beta}\right ] = } \times \left [ -\p_t\left(\frac{s'}{(\alpha-\beta)}\right ) \int_{-1}^{\frac{\overset{0}{u}-\alpha}{\alpha-\beta } } \frac{dz}{\sqrt{z(1-z)\left((z-\gamma)^2+\delta^2\right)}}\right . \vspace{3mm}\\ \ds \phantom{\p_t \left[\frac{\overset{0}{u}-\beta}{\alpha-\beta}\right ] = } \left.+ \frac{s'}{(\alpha-\beta ) } \int_0^{\frac{\overset{0}{u}-\alpha}{\alpha-\beta } } \frac{dz\,\delta\delta'-(z-\gamma)\gamma ' } { \sqrt{z(1-z)}\left((z-\gamma)^2+\delta^2\right)^{3/2}}\right ] , \ea\ ] ] where @xmath397 . these formulas will be useful when we will reduce the formula for the second derivative of @xmath311 with respect to @xmath5 . the first derivative of @xmath311 with respect to @xmath5 has the form : @xmath398 \vspace{3mm}\\ \ds \phantom{\p_t \overset{0}{u}= } + \frac{1}{(\alpha-\beta)s ' } \sqrt{\left(\alpha-\overset{0}{u}\right)\left(\overset{0}{u}-\beta\right ) \left(\left(\overset{0}{u}-m\right)^2+n^2\right ) } \vspace{3mm}\\ \ds \phantom{\p_t \overset{0}{u}= } \times \left [ -\p_t\left(\frac{s'}{(\alpha-\beta)}\right ) \int_0^{\frac{\overset{0}{u}-\beta}{\alpha-\beta } } \frac{dz}{\sqrt{z(1-z)\left((z-\gamma)^2+\delta^2\right)}}\right . \vspace{3mm}\\ \ds \phantom{\p_t \overset{0}{u}= } \left.+\frac{s'}{(\alpha-\beta ) } \int_0^{\frac{\overset{0}{u}-\beta}{\alpha-\beta } } \frac{dz\,(z-\gamma)\gamma'+\delta\delta ' } { \sqrt{z(1-z)}\left((z-\gamma)^2+\delta^2\right)^{3/2}}\right ] . \ea\ ] ] now we can evaluate the second derivative @xmath399 . @xmath400.\ ] ] using the formula for @xmath401 one can obtain as @xmath93 : @xmath402 this formula allows to evaluate the right hand side in the equation ( [ eq-1u ] ) as @xmath93 : @xmath403 the first correction is periodical function with respect to @xmath306 . one can derive the solution of the equation for the first correction using two linear independent solution of the equation @xmath404 here our goal is to write these solutions in the terms of @xmath311 because then we evaluate the order of derivatives of the first correction of asymptotic solution ( [ exp4 ] ) using the formula for @xmath405 . the first one is @xmath406 here the sign before the root is @xmath407 , when @xmath408 and vice versa . the second solution of the homogeneous linearized equation for the first correction is @xmath409 integral in this formula must be regularized because integrand has second order poles at points @xmath410 and @xmath411 . one of the possibile way of the regularization is done in @xcite . here we will follow @xcite . near the singular points @xmath412 @xmath413 one must represent the integral as @xmath414 the parameter @xmath415 may be for example @xmath416 , where @xmath105 is the period of oscillations of the function @xmath311 with respect to @xmath306 . change the second derivative of @xmath417 in the last integrand as @xmath418 and the first derivative of @xmath417 as @xmath419 as a result we obtain formula for regularization of the integral : @xmath420 using the functions @xmath417 and @xmath421 one can solve the equation ( [ eq-1u ] ) for the first correction of the asymptotic solution ( [ exp4 ] ) and obtain the solution the terms of @xmath311 . one can see the first correction @xmath422 has the order of the right hand side of the equation for the first correction multiplyed on @xmath315 . it means that @xmath423 as @xmath93 . this formula allows to obtain the restriction for the validity of the formal asymptotic solution ( [ exp4 ] ) : @xmath424 evaluate the residual of the asymptotic solution ( [ exp4 ] ) . for this we must evaluate the function @xmath425 for all @xmath426 $ ] the order of @xmath427 . but the order of @xmath428 grows as @xmath93 , because the derivatives with respect to @xmath5 have singularity at @xmath19 . in the right hand side of this formula there is only functions on the @xmath311 , then we can differentiate the right hand side for deriving of the second derivation of @xmath311 with respect to @xmath5 . the formula for @xmath429 will be very large if we will write it in these definitions but now one can evaluate the order of @xmath429 as @xmath93 using the formulas @xmath430 as a result one obtain : @xmath431 using the same formulas one can evaluate the order of the @xmath432 as @xmath93 : @xmath433 the matching of this asymptotics with the inner asymptotics ( [ exp2 ] ) and ( [ exp3 ] ) is carried out . from the matching condition for the phase function we obtain the initial condition @xmath434 . now we turn to the evaluation of the asymptotics for the function @xmath311 as @xmath93 . the function @xmath311 may be written in the implicit form : @xmath435 we remind the parameter @xmath306 is equal to @xmath436 and the additional term @xmath385 is undefined in the function @xmath103 . this term we define in this subsection . the formula ( [ imu-0 ] ) allows us to obtain the main term of asymptotics for the @xmath311 at @xmath93 . denote @xmath437 and @xmath438 using asymptotics of @xmath97 and @xmath99 as @xmath93 one can get @xmath439 this formula allows to write the asymptotic expansion of @xmath311 as @xmath93 in the form : @xmath440 the main term of the asymptotics is defined by formula : @xmath441 this asymptotics is applicable as @xmath442 . the function @xmath320 is periodic with respect to @xmath306 . it means the asymptotic ( [ exp5 ] ) is applicable on some segments of the interval @xmath443 . the argument of the function @xmath444 in the neighborhood of some point @xmath445 as @xmath446 is : @xmath447 where @xmath448 . it is easy to see the argument of the function @xmath444 may be represented as @xmath449 where @xmath450 is some constants depending on @xmath385 and number @xmath356 . one can see the main term of the asymptotics ( [ exp5 ] ) coincides up to shift @xmath450 with the main term of the second inner asymptotic expansion which is the function @xmath84 . it is easy to see for full definition of the function @xmath320 one must find the phase shift @xmath385 . we defined the additional constant @xmath385 by matching the functions @xmath320 and the first inner asymptotic expansion . the formula ( [ exp5 ] ) is suitable when @xmath451 . when @xmath452 is small , we consider other asymptotic formula for the function @xmath320 : @xmath453 substitute this formula to the second - order equation for the function @xmath320 ( [ eq-2u ] ) . expand the function @xmath454 with respect to the small parameter @xmath455 . in a result the equation for the main term of the asymptotic expansion is @xmath456 this equation coincides with the equation for the asymptotics of the first correction of the first inner asymptotic expansion . the boundary conditions for the function @xmath457 is obtained from the condition of the matching ( [ exp6 ] ) with the asymptotics of the expansion ( [ exp5 ] ) as @xmath86 . the additional constant @xmath385 in the formula ( [ phase ] ) is finally defined at the matching of the asymptotic expansions ( [ exp5 ] ) , ( [ exp6 ] ) with asymptotics of the inner asymptotic expansions . this get : @xmath458 . in this work the bifurcation of the slowly varying equilibrium of the painlev-2 equation was studied by matching method on the formal asymptotic approach . however it is necessary to note two important problems which remind out of side of our analysis . * the phase shift of the oscillating asymptotic solution is undefined in our approach . its definition demands much more thin calculations for the corrections of the asymptotic formulas . * a problem of a justification of the remainder term for the constructed asymptotic solution remains open as well . i am grateful to a n belogrudov , s g glebov , l a kalyakin , v yu novokshenov and b i suleimanov for stimulating discussions and also v e adler for the help in the realization of the numerical calculating . also i would like to thank r haberman kindly sending his ( collaborate with d c diminnie ) work @xcite . i am grateful to a v kitaev for many valuable comments , which resulted in substantial improvement of the original version of this work . andronov a a and chaikin c e , theory of oscillations , princenton university press , princenton , nj , 1949 . arnold v i , afraimovich v s , ilyashenko yu s and shilnikov l p , theory of bifurcations , in results of the science and engineering . modern problems of mathematics . fundamental directions , nauka , moscow , 1986 , v.5 , 9218 . haberman r , slowly varying jump and transmision phenomena associated with algebraic bifurcation problems , _ siam j. appl . math . _ , 1979 , v.37 , n 1 , 69106 . diminnie d c and haberman r , slow passage through a saddle - center bifurcation , _ j. nonlin . science _ , 2000 , v.10 , 197221 . neishtadt a i , about changes of adiabatic invariant at passage through seperatrix , _ plasma physics _ , 1986 , v.12 , n 8 , 9921001 . bourland f j and haberman r , separatrix crossing : time potentials with dissipation , _ siam j. appl . _ , 1990 , v.50 , n 6 , 17161744 . haberman r , nonlinear transition layers the second painlev transcendent , _ stud . _ , 1977 , v.57 , 247270 . karasyev m v and pereskokov a v , one - dimensional equations of a self - consistent field with cubic nonlinearity in a semiclassical approximation , _ mathematical notes _ , 1992 , v.52 , n 2 , 6682 . kapaev a a , an asymptotics of solutions of the equation the painlev of the first type , _ differentsialnye uravneniya _ , 1988 , v.24 , n 10 , 16841659 . kitaev a v , isomonodromic technique and elliptic asymptotics for the first painlev transcendent , _ algebra i analiz _ , 1993 , v.5 , 117151 ( in russian ) , english translation in _ petersburg math . journ . _ , 1994 , v.5 , 117151 . kitaev a v , elliptic asymptotics first and second painlev - transcendents , _ usp . nauk _ , 1994 , v.49 , n 1 , 77140 . boutroux p , rechercles sur les transcedantes de m. painlev et letude asymptotique des equations differentielles du second orde , _ ann . ecol . norm . super . _ , 1913 , v.30 , 255 - 376 ; 1914 , v.31 , 99159 . ioshi n and kruskal m d , on asymptotic approach to the connection problem for the first and the second painlev equations , _ phys . a _ , 1988 , v.130 , 129137 . novokshenov v yu , ansatz boutroux for the second equation the painlev equation in complex area , _ izv . an ussr , ser . matem . _ , 1990 , v.54 , 12291251 . moore g , geometry of the string equation , _ comm . phys . _ , 1990 , v.133 , n 3 , 261304 . novikov s p , quantonization of finite - zone potentials and nonlinear quasiclassics , originating in nonperturbative strings theory , _ funk . i ego pril . _ , 1990 , v.24 , n 4 , 4353 . krichever i m , on heisenberg relations for the ordinary linear differetial operators , preprint ihes , bur - sur - yvette , 1990 . kawai t and takei j , wkb analysis of painlev trnascendents with a large parameter . i , _ adv . _ , 1996 , v.86 , 315376 . kapaev a a , painlev - transcendents as nonlinear special functions , diss . of doct . of sci . , spb , 1997 . vereschagin v l , global asymptotics for fourth painlev transcenden , _ matem . sbor . _ , 1997 , v.188 , n 12 , 1132 . kitaev a v , turning points of linear systems and double asymptotics of the painlev transcendents , in painlev transcendents their asymptotics and physical applications , editors d levi and p winternitz , nato asi series , b 278 , plenum , new york , 1982 , 8196 ; _ zap . lomi _ , 1991 , v.187 , 5374 ( in russian ) . kapaev a a and kitaev a v , limit transition @xmath459 , _ zap . lomi _ , 1991 , v.187 , 7587 ( in russian ) , english translation in _ j. math . , 1995 , v.73 , n 4 , 460467 . joshi n , the second painlev equation in the large parameter limit i : local asymptotic analysis , _ stud . appl . math . _ , 1999 , v.102 , 345373 . kiselev o m and suleimanov b i , the solution of the painlev equations as special functions of catastrophes , dafined by rejection in these equations of terms with derivative , http://xxx.lanl.gov/solv-int/9902004 . maslov v p and omelyanov g a , asymptotic soliton - like solutions with small dispersion , _ uspekhi matem . nauk _ , 1981 , v.36 , n 3 , 63126 . gromak v i and lukashevich n a , analytical properties of solutions of the painlev equations , minsk , 1990 . bourland f j and haberman r , the modulated phase shift for strongly nonlinear , slowly varying and weakly damped oscillators , _ siam j. appl . _ , 1988 , v.48 , n 3 , 737748 . flaschka h and newell a c , monodromy and spectrum preserving deformations , _ comm . phys . _ , 1980 , v.76 , 65116 . ilin a m , matching of asymptotic expansions of solutions of boundary value problem , ams , 1992 . ince e l , the ordinary differential equations , dover publications viii , ny , 1944 . kuzmak g e , asymptotic solutions of nonlinear differential second - order equations with variable coefficients , _ appl . mech . _ , 1959 , v.23 , n 3 , 515526 . luke j c , a perturbation method for nonlinear dispersive wave problems , _ proc . royal soc . , ser . a _ , 1966 , v.292 , 403412 . whitham g b , a general approach to linear and nonlinear dispersive waves using a lagrangian , _ j. fluid mech . _ , 1965 , v.27 , 273283 ; two timing variational principles and waves , _ j. fluid mech . _ , 1970 , v.44 , 373395 . dobrokhotov s yu and maslov v p , finite - zone almost - periodic solutions in wkb approximations , _ itogi nauki , ser . _ , 1980 , v.15 , 394 ( _ j . soviet . math . _ , 1980 , v.15 , 14331487 ) . fedoryuk m v , asymptotics . integrals and serieses , nauka , moscow , 1987 . fedoryuk m v , method wkb for a nonlinear second - order equation , _ j. vych . i matem fiz . _ , 1986 , v.26 , n 2 , 198210 .

we undertake an investigation into the spatial structure of the m31 satellite system utilizing the distance distributions presented in a previous publication . these distances make use of the unique combination of depth and spatial coverage of the pan - andromeda archaeological survey ( pandas ) to provide a large , homogeneous sample consisting of of m31 s satellites , as well as m31 itself . we find that the satellite distribution , when viewed as a whole , is no more planar than one would expect from a random distribution of equal size . a disk consisting of 15 of the satellites is however found to be highly significant , and strikingly thin , with a root - mean - square thickness of just kpc . this disk is oriented approximately edge on with respect to the milky way and almost perpendicular to the milky way disk . it is also roughly orthogonal to the disk like structure regularly reported for the milky way satellite system and in close alignment with m31 s giant stellar stream . a similar analysis of the asymmetry of the m31 satellite distribution finds that it is also significantly larger than one would expect from a random distribution . in particular , it is remarkable that of the satellites most likely lie on the milky way side of the galaxy , with the asymmetry being most pronounced within the satellite subset forming the aforementioned disk . this lopsidedness is all the more intriguing in light of the apparent orthogonality observed between the satellite disk structures of the milky way and m31 .

1301.7131

the possibility that irregular distributions of satellite galaxies may be a common feature of large galaxy halos was originally bolstered by several studies of the anisotropic distribution of our own galaxy s satellites . @xcite found that the magellanic stream along with sculptor and the draco - ursa minor stream and their associated dwarf spheroidal galaxies all appear to lie in the orbital plane of the magellanic clouds . in @xcite , all the then known dwarf spheroidal companions of the milky way are identified as lying in one of two streams . @xcite examined the likelihood of producing the observed disk - like distribution of milky way satellites from a spherical or oblate dark matter halo . from comparisons with theoretical isotropic satellite distributions produced from such a halo , they find that the chance of producing the observed distribution from the dark - matter sub - halos of cold - dark - matter ( cdm ) cosmology is less than 0.5 % . they examine various combinations of the inner most satellites and find a best - fit plane that is almost perpendicular to the plane of the milky way with a root - mean - square height ranging from only about @xmath3 to @xmath4 kpc . @xcite , whilst finding a similar plane to @xcite for the milky way satellites , disagree with their assumption that such a plane is unlikely to arise from a conventional cdm dark matter halo . they argue that that the most luminous satellites can not be taken for granted as forming randomly from the isotropic sub - halo distribution but instead , lie preferentially at smaller distances from the halo centre and co - planar with the major axis of the host halo . coupled with the finding that galaxies preferentially align themselves with their major - axis highly - inclined or even perpendicular to that of the surrounding matter ( e.g. @xcite ; @xcite ) , this would provide a good explanation for the observed orientation of the best fit plane . the arguments of @xcite are contended by @xcite however , and it should be noted that in contrast to @xcite , @xcite found that the distribution of the most massive subhaloes is not as flattened as the distribution of the milky way s satellites . more recently , @xcite , using the six halo models in the aquarius simulations @xcite , find that all six halos produce a significant population of sub - halos with quasi - planar orbits aligned with the main halo spin . this , they argue , is a natural explanation for the observed satellite distribution of the milky way . @xcite argue against this however . with the calculation of the angular momenta of 8 milky way satellites @xcite revealing a strong alignment between 6 of the orbital poles , @xcite examine the likelihood of randomly drawing 6 sub - halos from each of the 6 aquarius simulations ( among other halo simulations ) , and finding a similar degree of alignment . more precisely , they draw @xmath5 sets of 8 satellites from each of the 6 simulations , and select the 6 with the highest degree of alignment between their orbits , thus emulating the findings of @xcite . they then look at the degree of clumping of the orbital poles @xmath6 as well as the angular distance of the average of the orbital pole inclinations from the model equator @xmath7 and find that the actual degree of planarity observed for the six satellites identified by @xcite ( @xmath8 and @xmath9 ) are equalled or exceeded in the random draws in less than @xmath10 of cases when @xmath6 is considered and less than @xmath11 of cases for @xmath7 . @xcite also find that the degree of planarity observed for the milky way satellites is uncommon in all six of the aquarius halos ( see fig . 7 of that study ) . in addition to the revelation that the milky way s satellites appear to inhabit highly - inclined great planes , they also appear to corroborate the finding of @xcite , namely that the companions of spiral galaxies preferentially congregate at high galactic latitudes ( the holmberg effect ) , as observed in his study of 174 galaxy groups . it is not clear why this should be the case , or even if it truly is the case , although if the apparent adherence of satellite systems to polar great planes is typical of galaxies in general , then the holmberg effect seems to be an extension of this . @xcite proposed that dynamical friction may be responsible for the observed polar great planes , with those orbits spending the most time in close proximity to the galactic disk , experiencing the fastest decay , while those that take the most direct route through the disk environs , namely the polar orbits , experiencing the slowest orbital decay . it should be noted however that dynamical friction , whilst producing more polar orbits , would not produce planes of satellites . nor would it be effective on the young globular clusters which are shown to be co - planar with the `` vast plane of satellites '' identified around the milky way by @xcite . indeed , @xcite show that dynamical friction did not play a role in the formation of the milky way satellite orbits . besides the conjecture that satellite great planes trace the major - axis of the dark - matter halo in which the parent galaxy resides , there are other proposed mechanisms for their creation . one hypothesis is that these planes trace the orbits of ancient galaxies that have been cannibalized by the host galaxy . @xcite have investigated this hypothesis by looking for planes among groups of satellite galaxies and globular clusters in the milky way s outer halo and find various members to be co - planar with either the magellanic or sagittarius streams . the findings of @xcite are also consistent with such a hypothesis . indeed , it is this hypothesis which is most strongly supported by @xcite , wherein the @xmath6 and @xmath7 of satellites drawn from various tidal models equal or exceed @xmath12 and @xmath13 in over @xmath14 of draws in some cases . a similar hypothesis , which in some regards links the galaxy - cannibalization and dark - matter hypotheses , proposes that the observed planes result from the orientation of the large - scale filamentary structure of galaxy clusters ( e.g. @xcite ) , an orientation traced out by those minor galaxies which fall into the halo of a major galaxy . @xcite argue however that extra - galactic associations of dwarf galaxies are too extended to account for the high degree of planarity observed for the milky way satellites . this argument is supported by the findings of @xcite based on the aquarius simulations . the great obstacle to a conclusive resolution of these issues is the lack of systems for which reliable spatial ( and kinematic ) data exists . while some such data does exist for large galaxy clusters such as virgo and coma , accurate 3d distributions of galaxies within their halo have for a long time been known only for our own galaxy s halo , ascertainable due to our central position within it . it has only been in recent times that a second system has opened up to us - that of our counterpart in the local group , m31 . whilst various databases of photometry and other data have been available for m31 and some of its brighter companions for over a decade , it is the pan - andromeda archaeological survey ( pandas - @xcite ) - a deep photometric , 2-colour survey providing a uniform coverage of the m31 halo out to approximately 150 kpc - that has provided a new level of detail for this system . it is from this survey that we obtained our distances to m31 and 27 of its companions , following the method developed in @xcite ( henceforth cli11 ) and further adapted for this purpose in @xcite ( henceforth cil12 ) . the distances themselves and their associated uncertainty distributions are presented in cil12 and it is these distributions that are utilized for all analysis contained in this paper . with regard to previous studies of the anisotropy in the m31 satellite distribution , two investigations warrant consideration at this point . @xcite , making use of wide field camera ( wfc ) photometry from the isaac newton telescope ( int ) in what was essentially the forerunner to the pandas survey , focus on ghostly streams " of satellite galaxies following a similar approach as @xcite used for the milky way . in addition , they characterize the large degree of asymmetry in the satellite distribution , a feature also noted in cil12 , and examine the radial distribution of the satellites , noting a ( statistically insignificant ) larger average distance from m31 than that observed between the milky way and its satellites . they find a large number of candidate satellite streams , with some favoring the dwarf spheroidal members . utilize distance measurements from a variety of sources and focus particularly on planes of satellites and , whilst they do not find a particularly significant best fit plane when their whole satellite sample is considered , it is rather interesting that they find a 99.7 % statistical significance to their best fit plane when the then - known dwarf spheroidal galaxies dominate their sample . furthermore , this plane is near - polar - as has been observed for the milky way , although they find little support for the holmberg effect . @xcite utilize a particularly robust method in their search for high - significance planar fits to subsets of galaxies by considering every possible combination of a given number of satellites from their sample . in the current study we employ a similar approach , but with the great advantage of having a considerably extended sample of galaxies in our sample , with all distances derived by the same method and from the same data as described in cli11 and cil12 . ( distances are sampled from the distributions without the density prior applied - e.g. cil12 , table 1 , column 2 - whilst the best fit distances are drawn from cil12 , table 2 , column 4 . ) as a result , we are able to give full consideration to the effects of selection bias on the observed satellite distribution . this then presents an excellent opportunity to greatly improve our knowledge of the three - dimensional structure of the m31 satellite distribution , with important implications regarding the recent evolution of the system . a breakdown of the structure of the paper is as follows . in section [ s_method ] , we outline our method for plane fitting ( [ ss_method_pf ] ) and locating significant planes of satellites as well as the orientation , magnitude and significance of the asymmetry of the distribution . a method for generating random realizations of satellites subject to the same selection biases as the real data is also discussed in this section ( [ ss_method_rr ] ) as is the selection bias itself ( [ ss_method_sdb ] ) . [ s_results ] then presents the results of applying these methods , first to the sample as a whole , and then to subsets of galaxies . specifically , [ ss_bfp_all_sats ] presents a study of planarity within the satellite system when all satellites contribute to the determination of the best fit plane ; [ ss_asy_all_sats ] examines the asymmetry in a similar way ; [ sat_subsets ] examines the orientations of planes of smaller subsets of satellites within the distribution ; and [ great_plane ] concludes this section with a determination of the significance of a ` great plane ' of satellites emerging from the preceding sections . sections [ s_discussion ] and [ s_conclusions ] then follow with discussion and conclusions . note that this paper was written in conjunction with a shorter contribution ( @xcite ; hereafter ilc13 ) which announced some of the key discoveries resulting from the analysis we present here . in particular , the process of identifying the member satellites of the ` great plane ' discussed in ilc13 is described here in more detail . in this analysis however , we concern ourselves with the _ spatial _ structure of the satellite system only and so the reader should refer to ilc13 for the interesting insight provided by the addition of the velocity information . in order to find planes of satellites within the m31 satellite system , our first concern is to convert the satellite distances as presented in cil12 into three - dimensional positions . to do this , we begin with an m31-centered , cartesian coordinate system oriented such that the x and y axes lie in the m31 tangent plane with the z - axis pointed toward the earth . specifically , the x - axis corresponds to @xmath15 which is the projection of m31 s declination onto the tangent plane . the y - axis then corresponds to @xmath16 - the projection of m31 s right ascension onto the tangent plane . the z - axis then points along the earth - to - m31 vector , with magnitude increasing with distance from earth . this orientation can be seen in fig . 10(c ) of cil12 . thus : @xmath17 where @xmath18 and @xmath19 are the distances from earth to m31 and from earth to the satellite respectively , @xmath20 is the angular separation on the sky between m31 and the satellite , and @xmath21 and @xmath22 are the real - angle equivalents of the tangent plane projection angles @xmath23 and @xmath24 respectively . next , we rotate this reference frame to the conventional m31 reference frame such that the positive z - axis points toward m31 s north galactic pole(i.e . @xmath25 ) and the @xmath26 meridian passes through the earth . so as to be consistent with the earlier work of @xcite , we have adopted the same values for m31 s position angle ( @xmath27 ) and inclination ( @xmath28 - @xcite ) . each object is hence rotated by @xmath27 about the z - axis to counter the effect of its position angle , and then @xmath28 about the x - axis to account for m31 s inclination . a final rotation of @xmath29 about the z - axis is then necessary to bring @xmath26 into alignment with the direction of earth ( which hence lies at @xmath30 ) . the resulting spherical coordinates for each object in the sample are plotted onto an aitoff - hammer projection in fig . [ sat_aitoff ] . this same figure also shows the uncertainties in position associated with each object , generated via sampling of the respective distance posterior probability distributions ( ppds ) of each object and subsequent conversion of each drawn distance into a three - dimensional position . with the satellites positions determined in cartesian coordinates , it is straight forward to determine the minimum distance of each satellite from a given plane as follows : @xmath31 where @xmath32 is the distance of a satellite at a point @xmath33 from a plane whose normal vector is @xmath34 and is of unit length . for simplicity , we invoke the reasonable requirement that all planes must pass through the center of m31 and so in our case , @xmath35 and the plane normal vector points out from the center of m31 . hence , in order to find the best - fit or maximum significance plane to a set of satellites , we need simply minimize @xmath32 for the satellites to be fitted . this can be done via a variety of means , some of which are compared in the following section , but perhaps the most robust and the predominant method employed in this study , is that of minimizing the root - mean - square ( rms ) of the distances to the fitted satellites . in order to measure the asymmetry of the satellite distribution about a given plane , we need only count the number of satellites on one side of the plane . to do this , we can simply remove the absolute value signs from equation [ e_plane_dist ] , so that the side of the plane on which a satellite lies can be determined by whether @xmath32 is positive or negative . the plane of maximum asymmetry is then taken to be that which divides the sample such that the difference in satellite counts for opposite sides of the plane is greatest . whether we wish to determine the best fit plane through a sample of satellites or the plane of maximum asymmetry , we require a system by which a large number of planes can be tested on the sample so that the goodness of fit ( or asymmetry ) can be calculated for each . to do this , we define each tested plane by its normal vector or pole @xmath34 so that eq . [ e_plane_dist ] can be applied directly . we then rotate this pole to different orientations around the sky in such a way as to ` scan ' the whole sphere evenly and at a suitably high resolution . in practice , we need to be able to apply this routine many thousands of times for a large number of samples and so a fast computational time is of the essence . to this end , for a given sample , our algorithm determines the desired plane following a two step procedure . firstly , a low resolution scan of the sphere is made to determine the approximate direction on the sky of the pole to the best - fit plane . only half the sphere actually needs to be scanned since poles lying on the opposite hemisphere correspond to the identical planes flipped upside down . the low resolution scan tests 2233 different poles across the hemisphere . a near - uniform coverage is achieved by decreasing the number of planes tested in proportion to the cosine of the latitude of the planes pole . this prohibits what would otherwise be an increased coverage at the higher latitudes of the coordinate system . with the pole to the best - fit plane determined in low - resolution , a high resolution search is then made around the identified coordinates at 10 times the resolution . in this way a pole can effectively be found at any of approximately 250,000 evenly spread locations on the hemisphere . whilst we are now equipped to identify best - fit planes to our sample and subsamples thereof , it is necessary to have some means of determining the significance of these planes in an absolute sense . the most intuitive way to do this is to perform the same analysis on a randomly generated sample of equal size . in particular , when we are concerned with all possible combinations of a particular number of satellites that can be produced from the whole sample , we are often dealing with a very large number of subsamples and so it is inevitable that some of these subsets of satellites will exhibit a very high degree of planarity . identical analysis must therefore be performed on random distributions , to see if there are similar numbers of subsets with equal degrees of planarity . for this reason , considerable care was taken to design an algorithm capable of providing a unique random realization of the desired number of satellites whenever it is called . the algorithm makes use of the distance ppd for each satellite , and also takes into account the irregular window function ( i.e. useable portion ) of the pandas survey . each time a satellite is to be added to the random realization , one of the 27 actual satellites is chosen at random and a distance is drawn from its associated ppd . this distance ( @xmath19 ) is then converted into a three dimensional position @xmath33 following equation set [ e_cart_coord ] and this satellite - to - m31 separation vector is then spun around to a new , random location in the m31 sky . note that for each random realization , a new value of @xmath18 is similarly drawn from the m31 distance ppd . once again , care must be taken in this step to ensure that the whole sphere is given equal weight , otherwise there is a higher likelihood for the artificial satellites to be positioned at high latitude . again , this is remedied by weighting the likelihood by the cosine of the latitude . with the new , random location for the satellite chosen , it is then projected back onto the sky as it would appear from earth and a check is made to ensure that it does indeed lie within the boundaries of the pandas survey area , and outside of the central ellipse ( @xmath36 major axis , @xmath37 minor axis - see fig . 10 ( c ) of cil12 ) where the disk of m31 inhibits reliable measurements . if the satellite does not meet these requirements , it is rejected and the satellite drawing process is repeated until a suitable position is generated . by repeating this process until the desired number of satellites are produced , a new , random comparison sample is generated which gives full account to the constraints on the actual data . in order for the random satellite realizations to mimic the actual data most closely , it is necessary that each artificial satellite is represented not by just one point , but rather a string of points reflecting the uncertainty in the earth - to - object distance . hence once acceptable positions for each satellite are drawn as described above , the distance distributions for each object are sampled and projected to their equivalent positions along the line of sight about the initially placed point . for sections [ ss_bfp_all_sats ] , [ ss_asy_all_sats ] and [ great_plane ] each artificial satellite s distance distribution is represented by @xmath38 points such that each plane - fitting measurement is made for @xmath38 possible positions of the object and then the average value of the measurements is taken . the only exception to this number is where the maximum - likelihood approach is used in [ ss_bfp_all_sats ] . due to the inclusion of a second fitting parameter in this case , only @xmath39 samples are taken for each satellite . for [ sat_subsets ] , as we are not concerned with comparisons of plane significance between the real sample and the random realizations , it is sufficient to use a single drawn position for each artificial satellite . by employing a similar method to that described above , it is also possible to explore the effect of the pandas survey area boundaries on the satellite detection bias as viewed from the center of m31 . it is intuitive that more satellites are likely to be detected along the line of sight to earth , since even satellites at a large distance from m31 will still appear within the survey boundaries if they lie along this line . we can visualize this effect by generating a large number of randomly distributed satellites and plotting them on the m31 sky after first rejecting those satellites that would appear outside the survey area ` mask ' if viewed from earth . to do this , one million satellites were drawn from a spherically symmetric halo potential with density falling off as a function of the square of the distance from the halo center . satellites were hence drawn at distances between @xmath40 and @xmath41 kpc from m31 with equal probability . the satellites were then projected onto the m31 tangent plane and those satellites lying outside the survey area or inside the m31 disk obstruction area were excised from the density map . the resulting anisotropy of the satellites on the m31 sky is presented in fig . [ survey_mask_effects ] . as can be seen from the figure , the probability of detection is indeed higher along a great circle oriented edge - on with respect to the direction of earth , and perpendicular to the m31 disk ( @xmath42 ) . this great circle has its pole/ anti - pole at @xmath43 and hence we would expect a predisposition toward finding planes of satellites with a pole in this vicinity . we would also expect , though to a lesser extent , to find an excess of satellite planes oriented edge - on with respect to earth at any inclination . such planes would have poles lying anywhere on the great circle whose normal is directed toward earth . the drop in the satellite density at @xmath30 and @xmath44 is a consequence of the hinderance to detection caused by the m31 disk . due to the increased volume of space covered by the survey at greater distances from earth , unhindered satellite detection is possible over a larger range of angles on the far side of m31 in comparison to the earth - ward side . in order to find the best - fit plane to the satellite system as a whole , the procedure of [ ss_method_pf ] is applied to the whole sample of 27 satellites presented in cil12 . the rms thickness of the sample is used here , as in subsequent sections , as the statistic of planarity ; we find it to be a robust measure and it has the convenient property of being computationally inexpensive . since we are dealing with only one sample in this case , two other measures are also used for comparison . the first calculates the sum of the absolute values of the distances of each of the satellites from the tested plane . the second is essentially a maximum likelihood approach and replaces the plane of zero - thickness with a ` gaussian plane ' such that a satellite s position within the gaussian determines the plane s goodness - of - fit to that satellite . this second approach requires that different gaussian widths @xmath45 be tested for each plane orientation in order to find the width that best matches the satellite distribution . values between 5kpc and 150 kpc were tested at 5 kpc intervals for each tested plane orientation . hence an additional characteristic of the satellite distribution is obtained , but at the expense of a considerably longer computation time . for each of the three measures of goodness - of - fit described above , the first step is to find the best - fit plane to the satellite positions with their positions determined from their best - fit distances . when either the rms or maximum likelihood approach is used , the same best - fit plane is found as @xmath46 with pole at @xmath47 . this plane is plotted as a great circle on the m31 sky in fig . [ bfp_all_sats ] with the poles of the plane indicated . when the absolute distance sum is used instead , the pole is found farther from the plane of the galaxy , at @xmath48 . nevertheless , the polar - plane described by @xcite is supported by either measurement , and is reminiscent of the satellite streams identified in the milky way satellite system . in light of the detection biases imposed by the pandas survey area as illustrated in fig . [ survey_mask_effects ] , the result in this case must clearly be treated with suitable caution however . like @xcite , we find little evidence for the holmberg effect , with only 3 best - fit satellite positions falling within @xmath49 of the m31 galactic poles , and only 6 of the @xmath50 error trails from fig . [ sat_aitoff ] pass beyond @xmath51 . to determine the uncertainty in the plane s goodness - of - fit , we need to repeat the procedure for a large number of realizations of the satellite sample , with the best - fit satellite distances replaced with a distance drawn at random from their respective satellite distance ppds . a density map of the best - fit plane poles identified from 200,000 such realizations is presented in fig . [ rms_distribution ] . this figure was generated using the distribution rms as the goodness - of - fit statistic , and contains @xmath52 of all poles within a @xmath53 radius of the best - fit pole stated above . when the sum of absolute distances is used in place of the rms , this fraction falls to @xmath54 , or to @xmath55 when the maximum likelihood approach is used . it should be noted that the distribution of poles lies in close proximity to the pole of maximum detection bias at @xmath56 , again suggesting that the detection bias is having a strong influence on the polar orientation of the best - fit plane . in order to determine whether the goodness - of - fit of the best - fit plane is really physically significant , similar analysis should be performed on a large number of random realizations of satellites , to see how often distributions of satellites arise with a comparable degree of planarity . figure [ bfp_ppds ] presents probability distributions of the plane significance for possible realizations of the real satellite sample along with average values from random realizations of the satellites ( as per [ ss_method_rr ] ) , obtained using the three measures of goodness - of - fit stated above . @xmath57{figures / rms_err_samp_ppd2.ps } \put(10,63){\small a ) } \end{overpic } \begin{overpic}[width = 0.26\textwidth , angle=-90]{figures / rms_sig_ppd2.ps } \put(10,63){\small b ) } \end{overpic } \\ \begin{overpic}[width = 0.26\textwidth , angle=-90]{figures / abval_err_samp_ppd2.ps } \put(10,63){\small c ) } \end{overpic } \begin{overpic}[width = 0.26\textwidth , angle=-90]{figures / abval_sig_ppd2.ps } \put(10,63){\small d ) } \end{overpic } \\ \begin{overpic}[width = 0.26\textwidth , angle=-90]{figures / ml_err_samp_ppd2.ps } \put(10,63){\small e ) } \end{overpic } \begin{overpic}[width = 0.26\textwidth , angle=-90]{figures / ml_sig_ppd2.ps } \put(10,63){\small f ) } \end{overpic } \end{array}$ ] it is immediately clear from fig . [ bfp_ppds ] that regardless of the choice of the measure of goodness - of - fit , the range of values obtainable from possible realizations of the real satellite positions are similar to the most likely values to be expected from completely random realizations of the satellites . hence , whilst a prominent plane of satellites comprising roughly half of the sample is suggested in fig . [ sat_aitoff ] , it would seem that the sample as a whole is no more planar than would be expected from a strictly random distribution . again , this is in keeping with the findings of @xcite , and detracts from any physical significance that should be attributed to the plane s polar orientation . further to this finding , the overall width of the ` plane ' is again in keeping with that expected from a purely random satellite distribution . from fitting the gaussian plane to the best - fit satellite positions , a @xmath50 width of @xmath58 kpc is found to produce the best fit to the data . when the 200,000 ppd - sampled realizations were tested , a @xmath50 of @xmath58 kpc was found preferential in @xmath59 of cases , with a @xmath50 of @xmath60 kpc being preferred in @xmath61 of cases . values of @xmath62 kpc make up the remaining @xmath63 almost entirely . the average value for the actual satellite distribution was thus determined as @xmath64 kpc . this value is similar to the most likely width identified from the 10,000 random realizations , as can be seen in fig . [ bfp_sigma ] . to determine the plane of maximum asymmetry and its significance , we employ an identical approach as in the preceding section , but with the goodness - of - fit statistic replaced with a count of the number of satellites on each side of the plane as per [ ss_method_pf ] . as was suggested by the three - dimensional satellite distribution generated in cil12 , the asymmetry about the m31 tangent plane is close to a maximum , with @xmath65 satellites on the near - side of the plane but only @xmath66 on the other when the best - fit satellite positions are assumed . the highest asymmetry plane possible from this same distribution has @xmath67 satellites on one side and @xmath68 on the other , with the equation of the plane identified by the algorithm as @xmath69 . the anti - pole of this plane lies @xmath70 away from the milky way at @xmath71 . this plane is plotted as a great circle on the m31 sky in fig . [ bfp_asy ] . when @xmath72 realizations of the satellite sample are generated using the satellite s respective distance probability distributions , the most likely asymmetry of the sample is actually found to be greater than this , with @xmath73 satellites on one side and only @xmath74 on the other . such a scenario is more than twice as likely as the @xmath75 scenario . in one realization , a plane was identified which could divide the sample such that all 27 satellites lay in a single hemisphere , while an asymmetry of @xmath76 was found possible for @xmath77 @xmath78 of the realizations . the distribution of maximum - asymmetry poles on the sky , as determined from realizations of possible satellite positions , is illustrated in fig . [ asy_distribution ] , whilst fig . [ asy_ppds ] @xmath79 plots the probability distribution for the greatest number of satellites that can be found in one hemisphere for a given realization of the observed satellite sample . the average value of this distribution is @xmath80 ( shown as a dashed line in fig . [ asy_ppds ] @xmath81 ) , a value which is equalled or exceeded for @xmath82 out of the @xmath83 random realizations represented in fig . [ asy_ppds ] @xmath81 . a maximum asymmetry ratio of @xmath75 , as was observed for the best - fit satellite distribution plotted in fig . [ bfp_asy ] , is more common however , falling inside the @xmath50 credibility interval . what is particularly striking about the satellite distribution however , is the orientation of the asymmetry , with the majority of satellites lying on the near - side of the m31 tangent plane . from fig . [ asy_ppds ] ( c ) , it is clear that the effect of the distance uncertainties lying along the line of sight is to create quite a broad distribution in the level of asymmetry about the tangent plane , though the average is markedly high at @xmath84 . to investigate the likelihood of this scenario arising from a random satellite distribution , we measure the average number of satellites on either side of the m31 tangent plane for each of @xmath83 random realizations as per [ ss_method_rr ] . the results are illustrated in fig . [ asy_ppds ] @xmath85 . the observed profile is more - or - less as expected , with a maximum probability close to the minimum possible asymmetry at @xmath86 and then a rapid fall off toward higher asymmetries . it is therefore clear that the distance uncertainties lying along the line of sight have no significant bearing on the orientation of the asymmetry . yet the observed degree of asymmetry about the m31 tangent plane is equalled or exceeded in only @xmath87 of the @xmath83 random satellite realizations and hence is very significant . the possibility that this asymmetry may be a consequence of data incompleteness is currently being examined more closely ( see @xcite ) , although it seems very unlikely . the high degree of asymmetry is still observed even when only the brightest satellites are considered . furthermore , the data incompleteness appears to be dominated by the boundaries of the pandas survey area and obstructed regions which are already taken into account by our analysis . indeed , one would expect more satellites to be observed on the far side of the m31 tangent plane on account of the increased volume of space covered by the survey at greater distances , an effect clearly visible in fig . [ survey_mask_effects ] . @xmath57{figures / err_samp_ppd2.ps } \put(10,60){\small a ) } \end{overpic } \begin{overpic}[width = 0.26\textwidth , angle=-90]{figures / asymm_ppd2.ps } \put(10,60){\small b ) } \end{overpic } \\ \begin{overpic}[width = 0.26\textwidth , angle=-90]{figures / err_samp_fp_ppd2.ps } \put(10,60){\small c ) } \end{overpic } \begin{overpic}[width = 0.26\textwidth , angle=-90]{figures / asymm_fp_ppd2.ps } \put(10,60){\small d ) } \end{overpic } \end{array}$ ] it is perhaps not surprising that the satellite system of m31 , when treated as a whole , is no more planar than one would expect from a random sample of comparable size . indeed , a similar result was noted for the m31 system by @xcite . the existence of outliers in our satellite sample was already clear from fig . [ sat_aitoff ] and furthermore , if multiple planes of differing orientation are present as has been suggested for both the milky way s satellite system ( e.g. @xcite ; @xcite ) and the m31 system @xcite , then the goodness of fit of the best - fit plane to the entire distribution is of little consequence . for this reason , we now concentrate our analysis on subsets or _ combinations _ of satellites . specifically , we perform a pole - count analysis by determining the pole of the best - fit plane to every possible satellite combination of a particular size that can be drawn from the entire sample . a pole - count analysis is an excellent way of mapping the degree of prominence of various planes that exist within the distribution as a whole , whatever their orientation may be . the choice of combination size is not trivial however . the number of combinations @xmath88 of a particular number of satellites @xmath89 that can be drawn from the entire sample of @xmath90 satellites can be determined as follows : @xmath91 for reasons that shall be discussed shortly , we will effectively be working with a sample of @xmath92 satellite positions . it is clear from this equation however that with @xmath92 satellites forming the entire sample , the total number of combinations that can be drawn may be very large , depending on the number of satellites forming the combinations . for instance , if @xmath93 and @xmath94 , there are over @xmath95 million possible combinations that can be drawn . additionally , if we are to properly account for the uncertainties in the satellite positions , it will be necessary to sample from the distance distributions of each satellite a large number of times for every combination . given that we must test every possible plane orientation ( as per [ ss_method_pf ] ) for every rendition of every combination , the computation times can become impracticable . it is therefore necessary to limit our combination sizes as much as possible . we note however , that the final pole - plot distribution showing the poles of the best - fit planes to each combination , is not so dependent on the combination size as might at first be thought . with all the planes tested as per [ ss_method_pf ] having to pass through the center of m31 , the minimum number of satellites that can not be fitted exactly is @xmath96 . this is therefore the smallest combination size we consider . there are @xmath97 combinations of @xmath96 satellites that can be drawn from the full sample of @xmath92 satellites . if we increase the combination size considerably to @xmath98 satellites , there are @xmath99 satellite combinations that can be drawn . due to an excessive number of combinations beyond this point , this is the largest combination size we consider . but it is critical to note that even if we produce our pole - plot map from combinations of only @xmath96 satellites we _ do not _ exclusively find planes consisting of @xmath96 satellites . if a plane of @xmath98 satellites exists for instance , then by eq . [ combinations ] , such a plane will produce @xmath100 poles at the same location on the pole plot , where a plane consisting of only @xmath96 satellites would contribute only one pole . conversely if we take combinations of @xmath98 satellites , despite the larger number of possible combinations in total , we become less sensitive to planes made up of less than @xmath98 satellites . so in a sense , the combination size we choose depends on the satellite planes we wish to be most sensitive to . in practice , we have found that the smaller combination sizes of @xmath96 and @xmath74 satellites are particularly useful for identifying the lowest rms planes congregating around the band of satellites visible in fig . [ sat_aitoff ] . the larger combination sizes of @xmath101 , @xmath68 and @xmath98 satellites gradually shift toward finding planes closer to the best - fit plane to the entire satellite sample illustrated in fig . [ bfp_all_sats ] . noting these points , we proceed as follows . first , the number of satellites per combination @xmath89 is chosen ( @xmath102 ) and then for each combination , distances are drawn for each of the satellites from their respective posterior distance distributions as provided in cil12 . to give a satisfactory representation of the form of the distributions , each combination is sampled @xmath39 times . as such , each satellite combination contributes not @xmath103 pole to the pole density map for the chosen combination size but @xmath39 , with the spread of poles relating the possible orientations of the best - fit plane to the combination , given the error in the individual satellite positions . the contribution of each pole to the density map is also weighted by dividing it by the rms of the best - fit plane it represents . thus each pole does not contribute @xmath103 count , but rather some fraction , depending on how good a fit the plane it represents is to the satellites in the combination . this fraction is also further divided by @xmath39 , since it represents only @xmath63 of the samples for the combination , as just discussed . as stated above , it should also be noted that we effectively limit the total number of satellites in our sample to @xmath92 for all analysis in this subsection . this is to account for the bound group of satellites consisting of ngc147 , ngc185 and and xxx ( henceforth the ngc147 group - see @xcite ) . since we suspect that these satellites orbit m31 as a group and since they all lie along the apparent plane identified in fig . [ sat_aitoff ] , it is preferable to treat the group as a single object when we are not concerned with measurements of the significance of particular planes . to do this , we take the luminosity weighted centre as an approximation for the center of mass , and treat this determined position as though it were the location of a single satellite . to calculate the luminosity weighted center , we can ignore the contribution from and xxx since it is negligible compared with the contributions of the two dwarf ellipticals . from the third reference catalogue of bright galaxies @xcite , ngc185 is @xmath104 magnitudes brighter than ngc147 in the v - band . each time the ngc147 group is chosen as one of the ` satellites ' for a combination , distances to each of ngc147 and ngc185 are sampled from their respective distributions and the luminosity weighted center of the group is determined . as for any other combination , this position , along with all other satellites in the combination , is sampled @xmath39 times . the results of applying the above procedure to all combinations of @xmath96 , @xmath74 , @xmath101 , @xmath68 and @xmath98 satellites that can be drawn from the total sample is presented in fig . [ actual_pole_plots ] . the left - hand column shows the fit to the most planar combination determined from the best - fit positions whilst the right - hand column shows the corresponding pole density plots for all combinations of that particular number of satellites , based on @xmath39 samples of each combination as per the discussion above . it is noteworthy that the best - fit planes to the most planar combinations are almost identical in every case , except for that of the @xmath96 satellite combinations , where the rms values are so small for so many combinations as to make this result not particularly important . it should also be noted that these best - fit planes trace out the same approximate great circle as the prominent plane indicated in fig . [ sat_aitoff ] , a result that shall be investigated a little later in [ great_plane ] . it is particularly interesting that the pole shared by each of these planes , located at @xmath105 corresponds to a pole count maximum in each of the pole plots . this indicates that many of the satellite combinations are aligned along this plane , hence further suggesting that the plane applies to more satellites than the combination sizes tested here . the other , lower latitude principle maximum in the pole plots is that corresponding approximately to the best fit to _ all _ the satellites and hence it grows more prominent in the plots made from larger combination sizes as discussed earlier . besides the pole count maxima that are strongly indicative of a highly planar subset of satellites , the other principle feature of the pole plots in fig . [ actual_pole_plots ] is the great circle along which the pole count density is highest . this great circle is very prominent but great caution must be exercised in attributing any significance to it . it is centered on the milky way indicating that the constituent poles result from a majority of satellites lying along the earth to m31 line of site . but this reflects the anisotropy predicted from fig . [ survey_mask_effects ] , the result of the bias incurred by the finite area of the pandas survey . hence it would seem that the progenitor of this prominent great circle is not physical but rather the result of selection effects . to further investigate the significance of the patterns observed in the pole plots , @xmath38 random realizations of @xmath92 satellites were generated as per [ ss_method_rr ] , and a similar pole count analysis performed on each of them . specifically , the pole density distribution resulting from the best fit planes to all combinations of @xmath101 satellites was generated for each of them . the resulting pole plots for @xmath66 of the @xmath38 random realizations ( chosen at random ) are presented in fig . [ random_reals ] along with an enlarged version of the equivalent plot from fig . [ actual_pole_plots ] generated from the real distribution . a bias toward a similar high - density great circle is indeed observed in these plots , but the plot generated from the actual data features a conspicuously narrower great circle , and a much more constrained distribution in general . this appears to be primarily the result of the large fraction of satellites that lie along the prominent plane that is repeatedly identified and plotted in the left - hand column of fig . [ actual_pole_plots ] . it should also be noted that this plane , whilst being oriented perfectly edge - on with respect to the earth , contains a significant fraction of satellites lying well outside the region of the m31 sky where the detection bias is large , and hence it is unlikely that its prominence is due to our observational constraints . @xmath106{figures / sat_combo_3_sats.ps } \put(-5,45){\small 3 satellites } \end{overpic } \begin{overpic}[width = 0.52\textwidth]{figures / pole_density_3_sats.eps } \end{overpic } \\ \begin{overpic}[width = 0.25\textwidth , angle=-90]{figures / sat_combo_4_sats.ps } \put(-5,45){\small 4 satellites } \end{overpic } \begin{overpic}[width = 0.52\textwidth]{figures / pole_density_4_sats.eps } \end{overpic } \\ \begin{overpic}[width = 0.25\textwidth , angle=-90]{figures / sat_combo_5_sats.ps } \put(-5,45){\small 5 satellites } \end{overpic } \begin{overpic}[width = 0.52\textwidth]{figures / pole_density_5_sats.eps } \end{overpic } \\ \begin{overpic}[width = 0.25\textwidth , angle=-90]{figures / sat_combo_6_sats.ps } \put(-5,45){\small 6 satellites } \end{overpic } \begin{overpic}[width = 0.52\textwidth]{figures / pole_density_6_sats.eps } \end{overpic } \\ \begin{overpic}[width = 0.25\textwidth , angle=-90]{figures / sat_combo_7_sats.ps } \put(-5,45){\small 7 satellites } \end{overpic } \begin{overpic}[width = 0.52\textwidth]{figures / pole_density_7_sats.eps } \end{overpic } \end{array}$ ] @xmath107{figures / pole_density_5_sats.eps } \end{overpic } \\ \begin{overpic}[width = 0.50\textwidth]{figures / pole_density_5_sats_rr1.eps } \end{overpic } \begin{overpic}[width = 0.50\textwidth]{figures / pole_density_5_sats_rr2.eps } \end{overpic } \\ \begin{overpic}[width = 0.50\textwidth]{figures / pole_density_5_sats_rr3.eps } \end{overpic } \begin{overpic}[width = 0.50\textwidth]{figures / pole_density_5_sats_rr4.eps } \end{overpic } \\ \begin{overpic}[width = 0.50\textwidth]{figures / pole_density_5_sats_rr5.eps } \end{overpic } \begin{overpic}[width = 0.50\textwidth]{figures / pole_density_5_sats_rr6.eps } \end{overpic } \\ \begin{overpic}[width = 0.50\textwidth]{figures / pole_density_5_sats_rr7.eps } \end{overpic } \begin{overpic}[width = 0.50\textwidth]{figures / pole_density_5_sats_rr8.eps } \end{overpic } \end{array}$ ] figure [ pole_radial_densities ] provides for a comparison between the concentration of poles around the principle maximum in the pole distributions of the actual satellite distribution and the average of the @xmath38 random satellite distributions . from line ( a ) in fig . [ pole_radial_densities ] we see that @xmath108 of all combinations of the actual satellite positions are fitted by a best - fit plane with pole within @xmath109 of the principal maximum ( located at @xmath110 ) . this is in stark contrast to the @xmath111 that lie within @xmath109 of the principal maximum for the average random realization of satellite positions ( fig . [ pole_radial_densities ] line ( b ) ) . furthermore , we find that only @xmath112 of the @xmath38 random realizations exhibited the degree of concentration of poles within @xmath109 of the principal maximum that was observed for the actual satellite distribution . hence it would seem that a large percentage of satellite combinations are fitted by best - fit planes that all have strikingly similar orientations when compared with what one could expect from a random distribution of satellites . again , this points toward a significant plane of satellites that includes a large fraction of the whole satellite sample . in order to obtain a better understanding of the satellites that this plane consists of , it is of particular interest to explore the number of times each satellite is included in a combination that is best fit by a plane with pole in close proximity to the principal maximum in the pole distribution for the entire sample . once again , we use the pole distribution for all combinations of @xmath101 satellites , and we count the number of times each satellite contributes to a pole within @xmath113 of the principal maximum at @xmath110 . the counts are divided by @xmath39 to account for the @xmath39 samples that are taken of each combination . the result can be seen in fig . [ sat_cont_prin_max ] . from this figure , it can be seen that the main contributors to the principal maximum in pole counts are those same satellites identified as forming a prominent plane in fig [ sat_aitoff ] , namely andromedas i , xi , xii , xiii , xiv , xvi , xvii , xxv , xxvi , xxvii and the ngc147 group , along with andromeda iii and andromeda ix . hence the conclusion of our analysis thus far must be that there is indeed a significant plane in the satellite distribution of m31 and that it broadly consists of the aforesaid satellites . we therefore investigate the numerical significance of the best - fit plane to these satellites in [ great_plane ] . as yet there is still more to be gleaned from a study of the pole density distribution however . from fig . [ sat_cont_prin_max ] we have been able to determine the principle contributing satellites to the principal maximum in the pole density distribution , but what of the remaining satellites ? do the positions of these satellites follow any particular trend ? the best way to determine this is to construct pole density plots of the two halves of the complete sample , namely the major contributors to the principal maximum and the minor contributors . the resulting pole plots are presented in fig . [ sat_cont_pole_plots ] . @xmath114{figures / main_cont.eps } \end{overpic } \ \begin{overpic}[width = 0.50\textwidth]{figures / minor_cont.eps } \end{overpic } \end{array}$ ] the left - hand plot of fig . [ sat_cont_pole_plots ] shows the pole density distribution generated from the major contributing satellites to the principal maximum at @xmath110 . this half - sample includes andromedas i , iii , ix , xi , xii , xiii , xiv , xvi , xvii , xxv , xxvi , xxvii and the ngc147 group . as expected , this plot reflects the existence of the aforementioned plane with all combination poles lying in the vicinity of the principal maximum . the right - hand plot , with poles generated from the remaining 12 satellites , namely andromedas ii , v , x , xv , xviii , xix , xx , xxi , xxii , xxiii , xxiv and m33 , paints a very different picture however . there is a much greater spread in the distribution of poles , with the great circle induced by the survey area bias once again conspicuous . also prominent in this plot are @xmath115 density maxima with their corresponding mirror images in the opposite hemisphere . the maximum lying midway between andromedas xix and xx lies very close to the pole of maximum detection bias at @xmath116 and so it is not unexpected , now that the prominent plane of satellites is effectively removed from the distribution . the elongated maximum passing through @xmath117 is more interesting however , and suggests the possibility of a second plane , roughly orthogonal to the major plane represented in the left - hand plot , though much less conspicuous . the planes represented by this maximum pass close to the error trails on the m31 sky of andromedas ii , iii , xix , xx , xxiii and xxiv . this maximum is faintly discernible in the pole distribution for combinations of @xmath68 satellites presented in fig . [ actual_pole_plots ] but is no more pronounced than anywhere else along the high - density great circle in any of the other pole plots . on account of this , it would appear that this plane is likely no more significant than one would expect to find from a random satellite distribution subject to the same detection biases , such as those illustrated in fig . [ random_reals ] . throughout the investigation undertaken thus far , all evidence has repeatedly pointed toward a conspicuously planar sub - set of satellites consisting of roughly half the total sample of satellites . andromedas i , xi , xii , xiii , xiv , xvi , xvii , xxv , xxvi , xxvii and xxx as well as the dwarf ellipticals ngc147 and ngc185 all appeared to lie along a plane in fig . [ sat_aitoff ] . the reality of this co - planarity was verified in [ sat_subsets ] and in particular fig . [ sat_cont_prin_max ] , which also suggested that andromeda iii and andromeda ix should be considered as plane members . hence it is of great interest to ascertain whether this ` great plane ' is in fact significant . to do this , it is necessary to determine how likely such a plane is to arise from a random satellite distribution subject to the same selection biases . the plane itself and the satellites of which it is constituted are illustrated in fig . [ best_15_aitoff ] . the plane shown is that calculated from the best - fit satellite positions and has equation of the form : @xmath118 with pole at @xmath119 . note that for this section , we re - instate ngc147 , ngc185 and andromeda xxx as separate objects since we are again concerned with measurements of the significance of the planarity of the distribution . our ` great plane ' thus consists of @xmath120 satellites out of the entire sample of @xmath0 . using the method of [ ss_method_rr ] , we again generate @xmath83 independent random realizations of @xmath0 satellites and seek the most planar combination of @xmath120 satellites from each . for each random realization , we sample @xmath38 possible positions for each satellite as in previous sections and take the average value for the rms of the best fit plane through the most planar combination . since there are more than @xmath121 million ways that @xmath120 satellites can be drawn from @xmath0 , and since we are not concerned with the orientation of each fitted plane as we have been in all previous sections , we depart from the plane fitting method of [ ss_method_pf ] for this section and instead proceed as follows . for each sample of satellite positions from each realization , @xmath83 randomized planes are generated and the @xmath120 closest satellites of the @xmath0 to the plane are stored in each case and the rms recorded . the lowest rms achieved is hence taken to be that for the most planar combination of 15 satellites in the sample . these minimum rms values from each of the @xmath38 samples of the particular random realization are then averaged to provide the best representation for the realization , given the distance uncertainties . [ best_15_ppds ] provides probability distributions in the rms for the observed ` great plane ' @xmath79 together with those for the average rms for the most planar combination from each random realization @xmath81 . the average rms for the observed plane is plotted in ( b ) for comparison . @xmath122{figures / err_samp_best_15.ps } \put(10,65){\small a ) } \end{overpic } \\ \begin{overpic}[width = 0.35\textwidth , angle=-90]{figures / best_15_sats_sig.ps } \put(10,65){\small b ) } \end{overpic } \end{array}$ ] as can be seen from fig . [ best_15_ppds ] , the rms for the observed plane is very low compared to what one could reasonably expect from a chance alignment . indeed , the average rms of @xmath123 kpc for the observed plane is found to be equalled or exceeded in only @xmath124 out of the @xmath83 random realizations . the chances of obtaining such a planar group of 15 satellites from a sample of @xmath0 at random is thus estimated as @xmath125 . hence we can conclude from this test that the observed plane is very unlikely to be a chance alignment , but rather the result of some underlying physical mechanism . note that an independent but equivalent investigation is presented in ilc13 where such an alignment is found to occur in only @xmath126 of instances . this is due to the larger central obstruction adopted in that analysis ( 19.6 vs. 7.9 sq . deg . ) which rejects more satellites in close proximity to the plane pivot point ( m31 ) where small plane distances are most likely . throughout the analysis conducted in [ s_results ] , the presence of a prominent plane of satellites has been a consistent feature . this is not the first time that a significant plane of satellites has been identified from among the denizens of the m31 halo however . as early as 1995 , @xcite had noted that the majority of the then known satellites lay conspicuously in a plane oriented edge - on with respect to the milky way . @xcite identified a highly significant plane lying within @xmath53 to @xmath127 of being polar . furthermore , they identify a subset of @xmath128 satellites from this plane lying within a thin disk with an rms of @xmath129 kpc . @xcite and later @xcite similarly identify a disk of satellites , this time not so markedly polar , with pole ( in our coordinate system ) at @xmath130 . they find this disk to have an rms height of @xmath131 kpc . this disk is clearly the same structure that we identify here , being tilted by only @xmath132 with respect to our ` great plane . ' our plane is found to have a much smaller rms of just @xmath1 kpc however , despite including a comparable number of satellites . it is particularly noteworthy however , that their satellite sample is significantly different to that used here , with their disk including m32 , ngc205 , ic10 , lgs3 and ic1613 - all of which lie outside the portion of the pandas survey region used in this study ( see fig . 10 ( c ) of cil12 ) . indeed , it is clear from fig . 4 of @xcite that the galaxies m32 , ic10 , lgs3 and ic1613 all lie along the same great circle as our ` great plane ' in fig . [ best_15_aitoff ] , as do their entire error trails . their conformity along with andromeda i to a thin disk is noted in the said paper as one of @xmath66 possible ` streams of satellites , ' thus providing another early detection of the plane identified by this study . both @xcite and @xcite also draw attention to the linear distribution of many of the plane - member satellites on the sky , a consequence of the edge - on orientation of the plane as indicated by the present study . the plane of @xcite does however include a significant number of satellites that , whilst included in our sample , we exclude due to their looser association with our plane . this then accounts for the much smaller rms height observed in our study . unlike previous studies of the m31 satellite system , we have a significant advantage in this study on account of the greatly improved sample of satellites available to us . our sample is not only more numerous , but the positions are all determined via the same method applied to the same data as per cli11 and cil12 . we are thus afforded unprecedented knowledge of the satellite detection biases , as well as the uncertainties in the object positions and have factored this knowledge into the analysis . an understanding of this bias is of particular importance when it comes to ascertaining the significance of any substructure identified , since a physically homogeneous satellite distribution will inevitably appear anisotropic after ` folding in ' the selection function and it is important that we do not attribute physical significance to this anisotropy . even after taking these effects into account however , there can be little doubt that the plane described in [ great_plane ] is a real physical object . the component satellites extend well into the regions of low detection bias in fig . [ survey_mask_effects ] and the analysis of the last section makes it clear that such a thin disk of satellites has very little chance of arising within a random satellite distribution of the same size , even when subject to the same observation biases . furthermore , it should be noted that the study of the plane s significance in [ great_plane ] is likely to be conservative , given that if the satellites m32 , ic10 , lgs3 , ic1613 and ngc205 were to be included in the analysis , the significance of our observed plane would likely grow still further . what is also particularly interesting is that subsequent research has shown @xmath133 of the @xmath120 objects to be co - rotating . this result is discussed in more detail in ilc13 . what then could be the progenitor of this ` great plane ' ? the polar orientation one might expect to arise had the satellites formed within the dark matter halo or had the dynamical friction proposed by @xcite had sufficient time to take effect is not observed . similarly , the findings of @xcite seemingly preclude the possibility that the structure might be the result of the accretion of an external galactic association . furthermore , there is apparently no marked distinction in the metallicities of the disk members compared with the non - disk members as one might expect from this scenario , though it is possible that some of the non - disk satellites may share the same origin as the disk members . there remains however the possibility that the satellites trace out the tidal debris of a galaxy merger . this is a particularly interesting possibility , especially since the plane , when projected onto the m31 tangent plane , is in close alignment with the giant stellar stream . indeed , @xcite show that the giant stellar stream could feasibly be the product of a major merger event that began around @xmath128 gyr ago , sustained by the returning stars from a tidal tail oriented similarly to our ` great plane . ' if a link is to be established , the observed asymmetry of the system must also be compatible with the tidal scenario however . it is of particular interest that , of the @xmath133 co - rotating satellites in the plane , all but one lie on the near side of the m31 tangent plane . indeed , if we removed all of the plane member - satellites from the system , the remaining satellite distribution would no longer be significantly asymmetric . with almost all of the satellites currently on the near side of m31 , it would seem on first consideration that the progenitor event could not have occurred substantially more than a typical orbital time ago or else the satellites would have had sufficient time to disperse . this suggests the event responsible must have occurred within the last @xmath101 gyr . other studies however have supported the proposition that a group of tidal dwarf galaxies could survive for extended periods whilst retaining the asymmetry inherent from the time of their formation . for instance , @xcite find three moderately old tidal dwarf candidates with this asymmetry preserved . it must also be remembered that the precise orbits of the satellites are undetermined and so it is not clear how close they have come to m31 in the past . as discussed by @xcite , there is also evidence that tidal material can survive in a bridge between the interacting galaxies for an extended period . nevertheless , it must be cautioned that there is no established precedent for tidal dwarfs with the longevity implied by the stellar populations of the m31 satellites , and the fact that they continue to adhere to such a thin structure is even more perplexing under such a scenario . there is also another striking characteristic of the observed plane . as one will note from examination of fig . [ best_15_aitoff ] ( and indeed the left - hand column of plots in fig . [ actual_pole_plots ] ) , it is oriented perfectly edge - on with respect to the milky way . whilst there is a noted bias toward detection of satellites positioned along planes oriented in this way , it must be remembered that this bias arises primarily due to the propensity for detecting satellites close to the line of sight passing through m31 . many of the satellites observed to lie on our plane are located a good distance from this line of sight however and well into the low - bias portions of the m31 sky . in any case , the random realizations of [ great_plane ] suffer from the same biases and yet show unequivocally that the observed plane is very unlikely to arise by chance . hence if we are to accept these results , we must also accept the plane s orientation . further to this strikingly edge - on orientation , it is also noteworthy that the plane is approximately perpendicular to the milky way disk . this fact can be easily seen if the constituent satellites are traced out in galactic coordinates ( i.e. all lie on approximately the same galactic longitude ) . this of course raises the question - how does the orientation of the milky way s polar plane of satellites compare with this plane ? noting that the average pole of the ` vast polar structure ' described by @xcite points roughly in the direction of m31 , the two planes are approximately orthogonal . these precise alignments are discussed in more detail in ilc13 , but suffice to say here that this alignment is particularly interesting and suggests that the milky way and m31 halos should not necessarily be viewed as fully isolated structures . it is entirely conceivable that our current ignorance as to the coupling between such structures may be to blame for our inability to pin down the precise mechanism by which such planes arise . finally , in consequence of these findings , most particularly with respect to the highly significant , very thin disk of satellites that has been identified , it is clear that if @xmath134 is to remain the standard model of cosmology , the occurrence of such structures has to be explicable within it . the possible deeper implications of the satellite anisotropies are discussed in @xcite , @xcite , @xcite and @xcite wherein other alternative cosmologies are also highlighted . it is clear that whilst the satellites of m31 when taken as a whole are no more planar than one can expect from a random distribution , a subset consisting of roughly half the sample _ is _ remarkably planar . the presence of this thin disk of satellites has been conspicuous throughout the analysis contained in this paper . the degree of asymmetry determined from the satellite distribution is also found to be relatively high . of particular note , the orientation of the asymmetry is very significant , being aligned very strongly in the direction of the milky way . when this fact is combined with the apparent orthogonality observed between the milky way and m31 satellite distributions and the milky way disk , it appears that the two halos may in fact be coupled . regardless , the great plane of satellites identified in this study , and its clear degree of significance , is not directly expected from @xmath134 cosmology . this finding provides strong evidence that thin disks of satellites do indeed exist in galaxy halos , and whether or not the standard model can account for such structures remains to be seen . a. r. c. would like to thank macquarie university for their financial support through the macquarie university research excellence scholarship ( mqres ) and both the university of sydney and universit de strasbourg for the use of computational and other facilities . g. f. l. thanks the australian research council for support through his future fellowship ( ft100100268 ) and discovery project ( dp110100678 ) . r. a. i. and d. v. g. gratefully acknowledge support from the agence nationale de la recherche though the grant pommme ( anr 09-blan-0228 ) . based on observations obtained with megaprime / megacam , a joint project of cfht and cea / dapnia , at the canada - france - hawaii telescope ( cfht ) which is operated by the national research council ( nrc ) of canada , the institut national des science de lunivers of the centre national de la recherche scientifique ( cnrs ) of france , and the university of hawaii .

we consider the non - linear sigma model for large as an effective theory for low - energy qcd at finite temperature , in the chiral limit . at this formulation provides a good description of scattering data in the scalar channel and generates dynamically the pole , the pole position lying within experimental determinations . previous results with this model are updated using newer analysis of pion scattering data . we calculate the pion scattering amplitude at finite and show that it satisfies exactly thermal unitarity , which had been assumed but not formally proven in previous works . we discuss the main differences with the result and we show that one can define a proper renormalization scheme with counterterms such that the renormalized amplitude can be chosen to depend only on a few parameters . next , we analyze the behaviour of the pole at finite , which is consistent with chiral symmetry restoration when the scalar susceptibility is saturated by the state , in a second - order transition scenario and in accordance with lattice and theoretical analysis .

1511.00031

the study of hadronic properties at finite temperature @xmath2 is one of the theoretical ingredients needed to understand the behaviour of matter created in relativistic heavy ion collision experiments , such as those in rhic and lhc ( alice ) . in particular , the qcd transition involving chiral symmetry restoration and deconfinement plays a crucial role , as it is clear from the many recent advances of lattice groups in the study of the phase diagram and other thermodynamical properties @xcite . for vanishing baryon chemical potential , the qcd transition is a crossover for 2 + 1 flavours with physical quark masses , the transition temperature being about @xmath5 150 -160 mev . in the chiral limit ( vanishing light quark masses for fixed strange mass ) it is believed to become a second - order phase transition belonging to the universality class of the @xmath6 model @xcite . lattice simulations also support this fact . actually , in @xcite it is shown that the lattice results are compatible with a @xmath7-like restoration pattern in the chiral limit and for physical masses , by studying the scaling of different thermodynamical observables near @xmath8 . the expected reduction in the transition temperature from the physical mass case to the chiral limit one based on those analysis is about 15 - 20@xmath9 , although subject to many lattice uncertainties @xcite . from the theoretical side , it is important to provide solid analysis of this chiral restoration pattern based on effective theories , given the limitations of perturbative qcd at those temperature scales . a simple model realization was historically the linear sigma model ( lsm ) based on @xmath10 spontaneous symmetry breaking @xcite , where the @xmath11-component of the @xmath6 field acquires a thermal vacuum expectation value and mass , both of them vanish at the transition in the chiral limit , and @xmath12 mesons degenerate as chiral partners . however , such a simple description is nowadays in conflict with observations : on the one hand , the @xmath13 broad resonance produced in pion - pion scattering and listed in the pdg @xcite is not compatible with a particle - like state ( see @xcite for a recent review ) . on the other hand , to reproduce consistently pion data , the lsm requires working in a strong coupling regime , invalidating the perturbative description . nevertheless , it is clear that the @xmath14 state must play an important role in chiral restoration , since it shares the quantum numbers of the qcd vacuum . chiral symmetry restoration has also been studied within qcd inspired models like the nambu - jona - lasinio or gross - neveu ones @xcite . a systematic and model - independent framework that takes into account the relevant light meson degrees of freedom and their interactions is chiral perturbation theory ( chpt ) @xcite . the effective chpt lagrangian is constructed as a derivative and mass expansion @xmath15 , where @xmath16 denotes generically a meson energy scale compared to the chiral scale @xmath17 1 gev . the lowest order lagrangian @xmath18 is the non - linear sigma model ( nlsm ) . the use of energy expansions in chiral effective theories is also justified at finite temperature to describe heavy ion physics . pions are actually the most copiously produced particles after a heavy ion collision and most of their properties from hadronization to thermal freeze - out can be reasonably described within the temperature range where these theories are applicable . thus , the chiral restoring behaviour in terms of the quark condensate is qualitatively obtained within chpt @xcite . moreover , the introduction of realistic pion interactions by demanding unitarity through the inverse amplitude method ( iam ) @xcite extended at finite @xmath2 @xcite improves chpt , providing a more accurate description of several effects of interest in a heavy - ion environment , such as thermal resonances , transport coefficients and electromagnetic corrections @xcite . this approach also provides a novel understanding of the role of the @xmath13 broad resonant state in chiral symmetry restoration , without having to deal with the typical lsm drawbacks . thus , the unitarized @xmath19 scattering amplitude within chpt at finite temperature develops a @xmath20 thermal pole at @xmath21 ^ 2 $ ] , which for @xmath3 corresponds to the pdg state , and whose trajectory in the complex plane as @xmath2 varies shows some interesting features : the sudden drop of @xmath22 towards the two - pion threshold can be interpreted in terms of chiral symmetry restoration , as opposed for instance to the @xmath23 @xmath24-channel where the mass drop is much softer . in addition , it has been recently shown @xcite that the scalar susceptibility saturated with this @xmath11-like state , with squared mass @xmath25 , develops a maximum near @xmath26 compatible with lattice data , unlike the pure chpt prediction which is monotonically increasing . moreover , chiral partners in the scalar - pseudoscalar sector are understood through degeneration of correlators and susceptibilities @xcite , something which is also directly seen in lattice data @xcite . the role of the @xmath4 state for chiral restoration could become more complicated if its possible tetraquark component is also considered at finite temperature @xcite . a crucial step in the unitarized approach is the condition of exact thermal unitarity for the partial waves , with a thermal space factor modified by the bose - einstein distribution function . this condition holds perturbatively in chpt @xcite and the unitarized amplitude is constructed by requiring thermal unitarity to all orders , based on the physical collision processes occurring in the thermal bath @xcite . however , it is important to emphasize that thermal unitarity for the full amplitude was not formally proven in those works ; in fact , that will be one of the relevant issues discussed in the present work . although the approaches based on effective theories in terms of the lightest mesons provide a good description of the physics involved , especially in what concerns the effect of the lightest resonances ( as discussed above ) , a more accurate treatment near @xmath26 would require including heavier degrees of freedom . that is for instance the framework of the hadron resonance gas , which describes the system through the statistical ensemble of all free states thermally available , and where corrections due to interactions and lattice masses can be also accounted for @xcite . effective chiral models including explicitly vector and axial - vector resonances have also been successfully used to depict several hadron thermal properties relevant for observables such as the dilepton and photon spectra and @xmath27 mixing / degeneration at the chiral transition @xcite . in this work , we will consider an alternative approach to the thermal pion scattering amplitude , namely the limit of large number of nambu - goldstone bosons @xmath1 , or in other words , large number of light flavours with no strangeness , as treated before at @xmath3 in @xcite . previous large-@xmath1 analysis at @xmath28 can be found in @xcite . within this framework , the lowest order chiral effective lagrangian for low - energy qcd will be the @xmath0 nlsm , whose corresponding symmetry breaking pattern is @xmath29 . as we have just commented , the latter is believed to take place in chiral symmetry restoration for @xmath30 , since @xmath6 and @xmath31 are respectively isomorphic to the isospin groups @xmath32 and @xmath33 . this technique has the advantage of allowing for a partial resummation of the scattering amplitude preserving many physical properties such as unitarity and the dynamical generation of the @xmath4 pole , which will help us to shed more light on the chiral restoring issues discussed before . we will work in the chiral limit , since it simplifies considerably the analysis , besides enhancing chiral - restoring effects , as explained above . at this point , it is important to remark that massless pions remain massless at finite temperature @xcite , unlike many other instances in thermal field theory where elementary massless excitations acquire mass in the thermal bath , like high-@xmath2 fermions @xcite , gauge fields @xcite including large-@xmath34 analysis @xcite or when electromagnetic corrections are switched on @xcite . the study of the large-@xmath1 approach in low - energy qcd implies a simplification of the pion dynamics @xcite without changing essential features such as analyticity , unitarity and the low - energy behaviour for pion scattering . this is fully accomplished when working in the functional formalism of the theory @xcite so that the lagrangian is built as @xmath35 covariant and @xmath36 invariant ( in the chiral limit ) . furthermore , as we will see in detail here , this approach will allow to describe consistently the @xmath4 state through its pole in the second riemann sheet , where the parameters of the model are fitted to pion - pion scattering phase shift data . thanks to the fact that the model is exactly unitary , we will be granted to go beyond the standard perturbative chpt description for the scattering @xcite , as a complementary description of unitarization methods such as the iam . an additional observation that makes this approach suitable for studying chiral restoring effects is that in order to reproduce correctly the @xmath37 pole ( linked to chiral restoration as mentioned before ) , the dominant contributions to @xmath19 scattering are the loop diagrams from the leading order chiral lagrangian , rather than the particular form of higher order terms needed to renormalize the amplitude @xcite . thus , the large-@xmath1 limit framework provides a resummation of the dominant loop contributions needed to maintain exact unitarity , so that the scalar pole can be correctly described . with the above motivations kept in mind , we will analyze in this work elastic pion - pion scattering at finite temperature within the large-@xmath1 @xmath0 model in the chiral limit . we will show that at @xmath3 one gets reasonable values for the @xmath37 pole from a fit to experimental data of a two - parameter partial wave . the extension to @xmath28 includes a formal discussion of the renormalizability of the model , which as expected can be carried out in terms of @xmath3 counterterms , although with important subtleties to be taken into account . the important feature of exact unitarity is demonstrated , including thermal corrections , something that allows us to define the second - sheet pole . having fixed the @xmath3 pole position , its @xmath2 dependence is obtained and it is shown that the results are compatible with a second - order chiral restoring phase transition , consistently with previous determinations and lattice data . the paper is organized as follows : in section [ sec : scattering ] we introduce our large @xmath1 approach within the framework of the massless nlsm and work out the diagrammatic expansion for pion scattering , both at zero and at finite temperature ; section [ sec : ren ] is devoted to explain the renormalization procedure , for which technical details are relegated to appendix [ app : ren ] . in section [ sec : pw ] we perform the analysis of the @xmath20 partial wave , providing a fit to @xmath3 data and showing that the large-@xmath1 amplitude satisfies exactly unitarity at zero and finite temperature . the latter grants us to define the riemann second - sheet pole corresponding to the @xmath4 state , which we study in detail in section [ sec : pole ] , paying special attention to its thermal evolution and the connection with chiral symmetry restoration . our conclusions are presented in section [ sec : conc ] . in a theory with spontaneous symmetry breaking @xmath29 , the coset space where the nambu - goldstone bosons ( ngb ) are defined is the @xmath1-dimensional sphere @xmath38 . in such a theory , the most general @xmath36-invariant and @xmath39 covariant lagrangian in the chiral limit can be obtained as a derivative expansion of the ngb modes , whose lowest order expression is given by the nlsm @xcite : @xmath40 where @xmath41 is a metric in the @xmath39 manifold which is parametrized in the ngb @xmath42 coordinates as @xmath43 with @xmath44 . as explained in @xcite , the advantage of choosing a @xmath39 covariant formalism is that we can easily construct @xmath36 invariant lagrangians of higher order by properly contracting indices with the @xmath45 metric . in addition , this formalism ensures the independence of the green functions on ngb field reparametrizations , i.e. , when changing coordinates in @xmath46 . the covariance of the quantum model is guaranteed as long as we work in the dimensional regularization ( dr ) scheme with @xmath47 , since the metric factor appearing in the ngb quantum measure @xmath48 , with @xmath49 the metric determinant , amounts to add to the lagrangian a term proportional to @xmath50 @xcite which vanishes in dr @xcite . although there is no need to invoke the lsm in the above construction , one can understand the nlsm as the kinetic part of the lsm when the vacuum constraint @xmath51 is imposed , where @xmath52 is the vacuum expectation value acquired by the @xmath11 field at tree level . in that way , it is easy to understand the @xmath1-scaling of the @xmath53 constant , where @xmath54 is the pion decay constant at this order for the usual @xmath30 case @xcite . the lagrangian provides the standard kinetic term for the ngb fields and along with it , an infinite set of self - interaction terms with an arbitrary even number of ngb . these interactions are obtained when the metric ( [ metric ] ) is expanded and written as a function of two field derivatives and powers of the pion field . hence , to obtain the relevant feynman rules for pion scattering , we can write the lagrangian ( [ nlsm ] ) as follows : @xmath55 . \label{lagrangian1}\ ] ] the first term in the latter expansion gives the standard kinetic lagrangian and the rest are the @xmath56-vertices with @xmath57 , represented in fig . [ treel ] . the feynman rule in momentum space for each vertex is @xmath58 with @xmath59 and @xmath60 the isospin indices of all possible choices of four different lines in the diagram and where the dots indicate the rest of pair contractions , i.e. , all products @xmath61 with @xmath62 the isospin indices of pairs different from @xmath63 and @xmath64 . with these rules , we proceed to calculate the @xmath65 scattering that , as customary , is parametrized as : @xmath66 here , @xmath67 are the mandelstam variables @xmath68 , @xmath69 , @xmath70 and use of isospin and crossing symmetry has been made to parametrize the amplitude . we will denote @xmath71 . the dominant contribution to @xmath72 in the large-@xmath1 limit comes from the diagrams showed in fig . [ 4pioneff ] . thus , when considering diagrams of arbitrary loop order with just the four pion vertex in fig.[treel ] , the dominant contribution of isospin flux is that where the pairs of lines are chosen as @xmath73 for an internal vertex , i.e. , with no attachments to any external line ( @xmath74 and @xmath75 are the four - momenta of the loops attached to that vertex ) and @xmath76 for the external ones . in that way , an additional factor of @xmath1 is generated for every pair of vertices between a given loop , coming from a contraction @xmath77 , where @xmath78 ( @xmath79 ) is one of the two free indices in the first ( second ) vertex . the result is a net factor @xmath80 for every vertex and an additional @xmath1 for every loop , so that the resulting amplitude is @xmath81 . other loop contributions are subdominant according to this counting . for those vertices in fig.[treel ] with more than four pions , the only way to compensate the additional @xmath82 factors of the @xmath56-pion vertex is to close @xmath83 of them in tadpole - like contributions , giving rise to a @xmath84 factor , so that this contribution would count the same @xmath85 as the four - pion vertex . those tadpole insertions correspond to fields sharing the same isospin index with no derivatives , since @xmath86 , with @xmath87 the free pion propagator . at @xmath3 , they actually vanish , since the tadpole contribution @xmath88 in the chiral limit . that will be not the case at @xmath89 , as we discuss in section [ sec : diagft ] and will become one of the main novelties of the present calculation . finally , note also that in the chiral limit , the pion propagator is not corrected by loop effects and hence needs no renormalization @xcite . for instance , a tadpole correction to the self - energy would require contracting two pion lines with the same isospin index to produce a @xmath1 factor , but that gives @xmath86 , and other contributions are non - dominant with respect to the tree level propagator , which is @xmath90 . from the latter considerations , and after including the proper combinatoric factors , the @xmath72 function in ( [ generalt ] ) depends only on @xmath91 to leading order in @xmath85 , which is actually one of the main simplifications of this approach , and is given by @xmath92= \frac{s}{nf^{2}}\sum_{k=0}^\infty \left[\frac{s j(s)}{2f^2}\right]^k= \frac{s}{nf^{2}[1-s\,j(s)/2f^{2 } ] } , \label{zerotamp}\ ] ] where @xmath93 is the usual logarithmically - divergent loop integral that in the dr scheme reads @xcite @xmath94 and @xmath95 contains the divergent part : @xmath96+{{\cal o}}(\epsilon ) . \label{jepsilon}\ ] ] here @xmath97 , @xmath98 is euler s constant and @xmath99 is the renormalization scale . note that we follow the convention in @xcite to define the pole contribution @xmath100 but we include the @xmath101 contribution in the divergent part , unlike in @xcite , in order to compare easily with previous large-@xmath1 works @xcite . we recall that for @xmath102 and @xmath103 ( i.e. , above the two - pion threshold which in the chiral limit is at @xmath104 ) we can easily obtain the imaginary parts of the loop integral as the usual unitarity cut contribution ( see section [ sec : unit ] ) : @xmath105 while @xmath106 for @xmath107 . in section [ sec : ren ] we will discuss the renormalization procedure implemented to absorb the divergent part ( [ jepsilon ] ) , but before that , let us explain the main distinctive features of the @xmath28 calculation . we will work within the imaginary - time formalism of thermal field theory @xcite so that the thermal scattering amplitude is understood as the analytic continuation to continuous energies of the corresponding four - point green function after performing the loop matsubara sums and applying the lsz standard reduction formula for @xmath3 asymptotic states @xcite . comparing with the analysis performed in the previous section , the first observation is that the absence of renormalization for the pion propagator remains the same at @xmath28 , so pions do not acquire effective thermal masses , following the same diagrammatic argument as before . however , there is an important difference with the @xmath3 case and is that the tadpole contribution is now different from zero in the chiral limit , namely @xcite : @xmath108 with the matsubara frequencies @xmath109 , @xmath110 . this means that from now on , the diagrams coming from closing pairs of extra pion lines in the vertices with 6 or more legs in fig.[treel ] have to be considered . to accomplish this in an efficient way , we construct the effective thermal tadpole vertex given in fig.[effvertx ] , and we rebuild the scattering amplitude to all perturbative orders with the associated feynman rule of the thermal vertex , something that we show schematically in fig.[finitetamp ] . in addition , we have to take into account that the loop integral is also @xmath2-dependent , so that the thermal amplitude to leading order in @xmath85 is given by @xmath111 which depends now separately on the space and time components of @xmath16 due to the loss of lorentz covariance in the thermal bath . the vertex function reads @xmath112 and the finite-@xmath2 loop integral @xmath113 is the analytic continuation of the external matsubara frequency @xmath114 of @xmath115 where @xmath116 in ( [ js ] ) . explicit expressions for the above @xmath117 integral for arbitrary three - momentum @xmath118 can be found for instance in @xcite . its uv divergent part is the same as for @xmath3 , since bose - einstein factors regulate exponentially the uv behaviour , so that we will write in general @xmath119 with @xmath95 in and @xmath120 finite and whose scale dependence is contained only in the @xmath3 part , namely @xmath121 . in this work we will be interested only in calculations in the center of mass frame ( corresponding to @xmath122 ) , where partial waves are defined ( see section [ sec : pw ] ) and moreover , we have @xcite @xmath123 where @xmath124 and @xmath125 is the usual bose - einstein distribution function @xmath126 note that @xmath127 is uv finite ( @xmath128 ) thanks to the @xmath129 term ; besides , we can easily separate the real and imaginary parts of @xmath127 for @xmath102 and @xmath130 by isolating the pole contribution at @xmath131 in the integrand in ( [ jthermal2 ] ) as @xmath132 while for @xmath107 there is no pole in the integrand in ( [ jthermal2 ] ) . finally , we recall that for @xmath133 , @xmath127 is ir finite ( @xmath134 ) while for @xmath135 , it diverges as @xmath136 so that @xmath137 ( as it appears in the thermal amplitude ( [ thermalampbare ] ) ) remains finite ( and vanishing ) in that limit . we will first review the main points regarding the renormalization of the model in the @xmath3 case , as discussed in @xcite . the scattering amplitude can be renormalized by choosing an appropriate ( infinite ) set of counterterm lagrangians of higher orders in derivatives and summing their contribution to the amplitude to all orders . the philosophy behind this approach is to include only those lagrangians needed to obtain a renormalized amplitude to leading order in @xmath85 , although from the symmetry arguments explained above , many other operator structures are possible . consistently , we can consider formally the couplings , or low - energy constants ( lec ) of those additional lagrangians to be suppressed in the @xmath85 counting . in the conventional chpt approach @xcite all possible terms are included to a given order in the derivative / momentum expansion and then the lec are fixed with experimental data , although the predictions are limited to low energies . the energy applicability range can be enlarged when additional conditions such as unitarization are implemented , and then the lec can take in general different values from the perturbative ones . here we are considering a partial resummation of the series for the amplitude , namely the leading @xmath85 contribution , which in the end can be given in a finite form that depends only on a few parameters , to be fixed to experimental data . nevertheless , there will be some important subtleties to be taken into account in the @xmath28 case , as we will explain below , in order to ensure a renormalized amplitude with a @xmath3 renormalization scheme . we will discuss the main results in this section , while additional details are given in appendix [ app : ren ] . let us consider , for instance , the possible counterterms lagrangians to fourth - order in derivatives . it is clear that one of them satisfying the symmetry constraints would be just proportional to @xmath138 : @xmath139 ^ 2=\frac{g_1}{2nf^4}\left[\left(\partial_\mu{\bm{\pi}}^a\partial^\mu{\bm{\pi}}_a\right)^2+{{\cal o}}\left(\frac{{\bm{\pi}}^6}{n}\right)\right ] , \label{l4}\ ] ] where the normalization proportional to the bare coupling @xmath140 has been conveniently chosen . at this order there is another term allowed , i.e. @xmath141\left[g_{cd}({\bm{\pi}})\partial^{\mu}{\bm{\pi}}^{c}\partial^{\nu}{\bm{\pi}}^{d}\right]$ ] . the lec multiplying the two allowed terms are the counterpart of the @xmath142 constants of chpt @xcite . the main result at @xmath3 is that the scattering amplitude can be rendered finite by a set of infinite counterterm lagrangians which to @xmath143 have the form @xcite @xmath144 . \label{lk}\ ] ] this reduces to for @xmath145 . it is indeed always possible to find at each order an adequate contraction with the @xmath45 metric that gives rise to the terms ( [ lk ] ) , for instance @xmath146 . consistently , the lec multiplying other possible terms such as @xmath147 can be considered formally suppressed in the @xmath85 counting . the dominant contributions of the new terms to the amplitude in the @xmath85 counting arise from all possible @xmath148 insertions in diagrams of the form showed in fig.[fig : insertions ] . it is actually not difficult to see that with the covariant structures discussed above , the @xmath149 terms and higher in give subdominant contributions , which also holds at @xmath28 . as it is explained in appendix [ app : ren ] , each lagrangian ( [ lk ] ) insertion produces a @xmath150 power in the vertex at @xmath3 . thus , summing up all the possible @xmath148 insertions in the dominant loop diagrams in fig.[fig : insertions ] is equivalent to the following redefinition of the four - pion vertex @xcite : @xmath151 with @xmath152 , which gives for the @xmath3 amplitude : @xmath153 or equivalently , @xmath154 , \label{invamp0}\ ] ] written in a more suitable way to implement its renormalization , as we discuss below . now , we can renormalize the bare divergent ( and scale independent ) lec @xmath155 correctly to absorb order by order the loop divergences coming from the @xmath93 function . thus , we denote @xmath156 for @xmath157 the renormalized ( and scale dependent ) couplings that are renormalized in terms of the @xmath158 with @xmath159 ( see details in appendix [ app : ren ] ) . equivalently , we define the renormalized function @xmath160 which replaced in gives rise to the renormalized amplitude : @xmath161 where the subscript @xmath162 is merely added to emphasize that the amplitude is finite . recall also that the renormalized amplitude is independent of the scale @xmath99 , since it was so from the original expression , the scale dependence of @xmath163 being cancelled by that of the renormalized @xmath156 . the function @xmath164 can also be written as a formal series in powers of @xmath91 by expanding both sides of eq . so that @xmath165 taking @xmath166 would give the order - by - order renormalization of the @xmath155 , presented explicitly in appendix [ app : ren ] up to @xmath167 ( specifically and ) . at @xmath89 , from general grounds , we should be able to renormalize the amplitude with @xmath3 counterterms @xcite . however , we notice that only with the renormalization of the four - pion vertex in and is not enough to renormalize the thermal amplitude in ( even though the divergent part of the @xmath117 integral is the same as at @xmath3 ) unless powers of @xmath168 were attached to the counterterm lagrangians , something that would violate the above mentioned @xmath89 renormalization principle . actually , things become more complicated , since the feynman rules arising from terms like are not as simple as the @xmath3 ones , which in particular means that a given diagram mixes different @xmath169 powers . in appendix [ app : ren ] we present a detailed analysis of the renormalization scheme that has to be applied to the @xmath28 case . the main conclusion is that the thermal amplitude can be rendered finite with a @xmath3 renormalization where not only the four - point vertex is involved , but also all @xmath170-pion vertices of the nlsm lagrangian for arbitrary integer @xmath171 as follows : @xmath172 the above renormalization , which can be understood either in terms of an addition of effective diagrams renormalizing the vertices or as a formal renormalization of the metric function at the lagrangian level ( see appendix [ app : ren ] ) , amounts to the redefinition of the effective thermal vertex in fig.[effvertx ] as given in when the corresponding tadpole diagrams are summed up . in fact , one can also arrive to the same four - pion effective vertex renormalization in starting by redefining the thermal effective vertex with an unknown bare function of @xmath91 and @xmath2 that absorbs the divergent part of the @xmath117 integral in the total amplitude . thus , we finally have for the thermal amplitude : @xmath173}{nf^2}\frac{1}{1-\frac{s g_{0}(s)f[g_{0}(s ) i_\beta]}{2f^2}j(p;t)}.\ ] ] after using exactly the same renormalization method given in , we obtain a finite renormalized thermal amplitude given by @xmath174}{nf^2}\frac{1}{1-\frac{s g_{r}(s;\mu)f[g_{r}(s;\mu ) i_\beta]}{2f^2}j_{fin}(p;t;\mu ) } , \label{renampt}\ ] ] where @xmath175 is the finite part of the thermal @xmath117 function , defined in . recall that our final finite thermal amplitude is also independent of the @xmath99 scale , as in the @xmath3 case , since the dependence in @xmath175 cancels out that of @xmath164 encoded by the finite renormalization constants @xmath176 through . we point out also that the renormalization scheme discussed here is the same as for @xmath3 and is of course consistent with the previous analysis of the scattering amplitude in the large-@xmath1 nlsm in @xcite , but at @xmath3 it is enough to consider the four - point vertex renormalization because the @xmath170 vertices with @xmath157 simply do not show up in the scattering amplitude at leading @xmath85 order . the renormalized thermal amplitude is one of our main results . as it happened in the @xmath3 case @xcite the infinite couplings @xmath177 parametrize different choices of effective theories sharing basic properties such as renormalizability , the lowest order energy expansion ( low - energy theorems ) and unitarity ( see section [ sec : pw ] ) . indeed , at @xmath3 one can for instance choose the @xmath177 to recover the amplitude of the lsm with explicit exchange of a scalar particle . an alternative approach , which is the one we will follow here , is to fix the scale and the renormalization conditions such as only a finite number of the @xmath177 are nonzero . in its simplest version , we can choose @xmath178 . it is not difficult to see that this condition is compatible with the renormalization conditions of the @xmath179 function , namely , and its corresponding renormalization group evolution @xcite and leaves the thermal amplitude as dependent only on two free parameters , @xmath99 and @xmath180 : @xmath181}{nf^2}\frac{1}{1-\frac{s f [ i_\beta]}{2f^2}j_{fin}(p;t;\mu)}. \label{renamptfinal}\ ] ] we will show in section [ sec : pwfit ] that it is possible with this approach to fit scattering data fairly well , considering that this is a chiral - limit approach ( the finite pion mass case has been analyzed with the same method in @xcite ) . that will be enough for our present purposes , since our main goal is to show that with a @xmath3 amplitude which complies with the above physical requirements , we can obtain a thermal behaviour compatible with different theoretical and lattice expectations regarding chiral symmetry restoration , as we discuss in section [ sec : pole ] . it is also important to stress that following the guide principle that the @xmath3 renormalization should be enough to render a finite amplitude ( proven perturbatively up to @xmath167 in appendix [ app : ren ] ) , the insertion of counterterms with bare renormalization constants @xmath182 , the subsequent absorption of the divergent part of the loop integrals to define the @xmath183 and taking @xmath184 , would have been equivalent to take the thermal amplitude with @xmath117 replaced by @xmath175 . what we have derived here is an explicit construction of such a renormalization scheme . to end this section and before proceeding with the analysis of partial waves and thermal poles , we provide a first result related to the pion decay constant at finite temperature . taking the low - energy limit of the thermal amplitude from its general renormalized form gives simply : @xmath185 which we can compare with the low - energy expression of the scattering amplitude given by weinberg s theorem @xcite to define a @xmath2-dependent pion decay constant , namely , @xmath186 so that : @xmath187 ( n=3 ) . \label{fpit}\ ] ] where we have used that @xmath188 . the result coincides with the known chpt result to @xmath189 @xcite and with the leading @xmath1 contribution studied in @xcite , which are additional consistency checks of our present analysis . a more careful analysis of @xmath190 beyond @xmath191 would require to analyze the residue of the axial - axial correlator @xcite . as discussed above , we will fix the undetermined constants in the scattering amplitude from experimental information . for that purpose and for the subsequent analysis of the @xmath4 pole and chiral restoration , we will consider partial waves with well - defined values of total isospin @xmath192 and angular momentum @xmath117 , which at @xmath3 are defined in the center of mass ( com ) frame @xmath122 : @xmath193 where @xmath194 are legendre polynomials , @xmath195 is the scattering angle and @xmath196 is a particular combination involving @xmath72 ( defined in ) which gives the scattering amplitude at given isospin @xmath192 , with @xmath197 , @xmath198 in the com frame . the large-@xmath1 analysis is specially adequate for the @xmath20 channel ( see below ) , which on the other hand is the one we are interested in this work . in that case , we have : @xmath199 , \label{t0def}\ ] ] with @xmath200 given for @xmath3 in and where we have made use of the fact that @xmath72 depends only on @xmath91 to leading order in @xmath85 and consistently , we only take the leading order also for the partial wave combination . the final result is that the @xmath20 partial wave is independent of @xmath1 ( to leading order ) . at @xmath3 we have then : @xmath201 recall that within the large-@xmath1 framework , the other possible isospin channels for pion scattering , namely @xmath202 and @xmath203 , are subdominant , since they are proportional to @xmath85 . this analysis is therefore particularly suited for @xmath204 @xcite which for the thermal case is the most relevant one regarding chiral symmetry restoration from the point of view of thermodynamic quantities such as the scalar susceptibility , as explained above . as discussed in section [ sec : ren ] , we will work within the minimal approach for which @xmath205 for @xmath157 so that we end up with the @xmath3 partial wave in with @xmath206 . defining the phase shift as customary for elastic channels @xmath207 , we can use this result to try to fit experimental phase shift data . there are several comments that are pertinent at this point : first , the choice of data sets is delicate because there have been several experiments over the years with results sometimes incompatible among them for this channel . in addition , we have to take into account that we are working in the chiral limit , so that we expect our amplitude to describe more naturally data sufficiently away from threshold , i.e. , typically for large @xmath208 . on the other hand , there is also a natural upper limit of applicability for @xmath209 , namely below the next resonance mass in this channel , which is the @xmath210 . this would need the inclusion of the strange sector . thus , we have chosen as data for the fit the sets given by @xcite in the @xmath209 range 450 - 800 mev . in addition , we consider also the parametrization of the scattering amplitude described in @xcite in the same energy region . that parametrization provides a precise description of @xmath19 scattering from a combined analysis based on dispersion relations and provides an accurate prediction of the @xmath4 and @xmath210 pole parameters . we do not include in the fit the recent ( and also more precise ) data of the na48 experiment @xcite which are very low - energy data below @xmath211 400 mev . those low - energy data are very well described by the parametrization @xcite . in the fit to the parametrization @xcite , we select points with a 5 mev energy interval and take into account the small uncertainties given in that paper . another important point is that in principle the values obtained for @xmath180 in our fits should not be far from the physical value of the pion decay constant @xmath212 . in the chiral limit , @xmath213 87 mev @xcite , so that @xmath180 would be around @xmath214 51 mev . however , once again it is important to stress that we are forcing our chiral limit amplitude to fit data for massive pions and it is then not surprising that we need a higher value for @xmath180 since mass corrections increase the @xmath190 value , so that we are encoding in @xmath180 a great part of the uncertainty we have in our chiral limit analysis . in any case , it is not the purpose of this work to provide a very precise fit to experimental data , as in other unitarized or dispersive approaches @xcite which can even be compared to lattice results by suitable mass extrapolations @xcite . after all , this is just a two - parameter fit in the chiral limit . we just need some reasonable reference values for the parameters such as we generate dynamically a pole in the second riemann sheet with consistent values for the pole position , so that the thermal behaviour corresponds to a physically realistic situation . in fig.[phaseshift ] we show the @xmath20 phase shift as a function of the com energy obtained with our large-@xmath1 amplitude with our best fits to grayer data and to pelez et al parametrization . the corresponding fit parameters are given in table [ table : fittable ] . channel phase shift for different fits , as explained in main text . pelez parametrization is given in @xcite , na48 low - energy data in @xcite and grayer data in @xcite . ] .parameters for the grayer and pelez data fits and their respective coefficients of determination . [ cols="^,^,^,^",options="header " , ] another chiral - restoration property we can examine is the scaling law for the scalar susceptibility defined through , i.e. , calculate the critical exponent @xmath98 determined as @xmath215 for @xmath216 . the results for the best fits are showed in table [ tab : scaling ] . we can compare this analysis , on the one hand , with the exact result for the nonlinear @xmath7 model for @xmath217 in four dimensions , @xmath218 @xcite . on the other hand , the critical exponent of the @xmath6 three - dimensional heisenberg model , which lattice qcd results resemble within uncertainties @xcite , is @xmath219 0.54 for @xmath220 in the chiral limit @xcite . our results and the iam one lie close to those values , providing then a consistency check of our approach to define the scalar susceptibility from the @xmath4 pole . we have studied pion scattering in the large-@xmath1 @xmath0 model at finite temperature in the chiral limit and its consequences regarding the @xmath4 pole and chiral symmetry restoration . our analysis gives rise to interesting theoretical and phenomenological results , consistent with previous analysis and lattice data . after calculating the relevant feynman diagrams , which include an effective thermal vertex from tadpole resummation , an important part of our work has been devoted to show that it is possible to find a renormalization scheme rendering the thermal amplitude finite with a @xmath3 renormalization of the corresponding vertices this is a nontrivial extension of the @xmath3 renormalization of the scattering amplitude , since the breaking of lorentz covariance in the thermal bath induces crossed terms between tadpole - like and @xmath221 loop functions . in the low - energy expansion of the model , up to @xmath167 , we have checked explicitly this renormalization scheme , providing a diagrammatic and lagrangian interpretation . another relevant result is that the large-@xmath1 thermal amplitude satisfies exactly the thermal unitarity relation , imposed in previous works as a physical condition for the exact amplitude . its low - energy properties are also preserved , being consistent for instance with the thermal dependence of the pion decay constant . by a suitable choice of the low - energy constants , similarly to the @xmath3 case , compatible with the scale evolution of the renormalized couplings , we end up with a phenomenological unitary amplitude depending only on two parameters , @xmath180 and @xmath99 . by fitting those parameters to experimental data in the @xmath20 channel , which is more reliable for data not very close to threshold in the elastic region , we reproduce the pole position of the @xmath4 in the second riemann sheet fairly consistently with pdg values and recent determinations . the chiral limit character of our approach implies a larger value for @xmath180 than phenomenologically expected , but it allows to obtain pole position parameters @xmath222 closer to the physical case . the fits to data are actually very good in the chosen region , precisely the most relevant energy range concerning this resonant state . once the @xmath3 pole has been fixed to physical values , we have studied its evolution with temperature . the @xmath4 pole remains a wide state for all the temperature range of interest , the real and imaginary parts @xmath223 and @xmath224 behaving similarly to the iam analysis , showing the signature of chiral restoration . in order to explore this further , we define a scalar susceptibility @xmath225 saturated by the inverse of @xmath226 , corresponding to the real part of the scalar state self - energy at zero four - momentum , which diverges at a given @xmath8 with a power law , as it corresponds to a continuous second - order phase transition in the chiral limit . the values obtained for @xmath8 , as well as the critical exponent of @xmath227 , are consistent with those obtained with other analytical approaches , such as the iam , and with lattice analysis , being compatible with a @xmath6 scaling . the combination of the large-@xmath1 framework with the phenomenological features of the @xmath4 pole allows to improve the predictions of previous approaches based on the partition function . thus , we obtain a very reasonable description of the chiral restoration transition within this approach , given the different uncertainties involved , such as possible @xmath85 corrections near the physical @xmath30 case or the absence of heavier degrees of freedom , which should play an important role near the transition and improve our simple pion gas scenario . work partially supported by the spanish research contracts fpa2011 - 27853-c02 - 02 , fpa2014 - 53375-c2 - 2-p . we also acknowledge the support of the eu fp7 hadronphysics3 project , the spanish hadron excelence network ( fis2014 - 57026-redt ) and the ucm - santander project gr3/14 910309 . santiago corts thanks prof . jos rolando roldn and the high energy physics group of universidad de los andes and colciencias for financial support . we are also grateful to jos ramn pelez and jacobo ruiz de elvira for useful comments and for providing us with their parametrization values for the phase shift . a lagrangian of the form ( [ lk ] ) gives rise to the feynman rule @xmath228\delta_{ab}\delta_{cd}$ ] where @xmath229 are the four - momenta of the four legs and @xmath230 their isospin indices . we will consider insertions of these counterterms in diagrams of the form depicted in fig.[fig : insertions ] , which will be the dominant ones in @xmath85 . for those insertions , we will have to deal then with integrals of the type : first , consider the @xmath3 case . since @xmath239 in dr @xcite , the only remaining terms after expanding the numerator in ( [ jn ] ) are the @xmath240 one and the contributions @xmath241 and @xmath242 . the latter vanish also in dr since @xmath243 ( @xmath244 even ) is formally proportional to @xmath245 , with @xmath246 ( using the standard parametrization for @xmath247 ) . therefore , in that case we have simply : thus , at @xmath3 , any @xmath148 insertion is proportional to @xmath150 , regardless of the vertex being internal of external . as stated in the main text , this allows to renormalize the amplitude at every order . as an example , let us show here the diagrams contributing up to @xmath167 at @xmath3 , which are those showed in fig.[fig : t0diags3 ] and where we have indicated the order of every diagram . summing up these contributions , the amplitude is finite with the following renormalization of @xmath140 , @xmath249 : @xmath250 ^ 2+g^r_1(\mu)j_\epsilon(\mu)=g_2+\frac{1}{4}\left[j_\epsilon(\mu)\right]^2+g_1j_\epsilon(\mu ) . \label{reng2}\end{aligned}\ ] ] for the renormalized amplitude at @xmath3 . we plot the different topological configurations contributing , so that diagram ( e ) is multiplied by two , corresponding to the possible vertex insertions of @xmath140 . ] when the @xmath3 amplitude is written in terms of the renormalized constants , it adopts the form ( [ renormamp1 ] ) , where to this order @xmath251 . as it was assured above , this is equivalent to renormalize the amplitude by the functional renormalization of the four - pion vertex given in eqns . and . at @xmath28 , there are additional complications that need to be analyzed in detail . first of all , the simple relation for the integrals @xmath252 in does no longer hold . namely , for @xmath253 we get directly from : where we have used that in dr @xmath258 for @xmath259 ( although not for any real @xmath260 as it happened in the @xmath3 case ) since @xmath261 for @xmath262 . note also that in , the two contributions @xmath263 cancel among them , and also independently by parity . however , this does not happen for @xmath264 , meaning that the @xmath252 s are not simply linear combinations of @xmath265 and @xmath255 as in the previous cases . for instance , for @xmath264 the following integral contributes : with @xmath267 and @xmath268 . at @xmath3 , one has @xmath269 in dr , but at @xmath28 the timelike and spacelike contributions decouple , from the loss of lorentz covariance in the thermal bath , and they are in general nonzero . besides , where we have made use of the standard feynman parametrization as well as poisson s summation formula @xmath271 . the function @xmath272 is defined in @xcite . on the other hand , using again the dr properties , @xmath273 . note that @xmath274 so that these pure thermal contributions would not give rise to new type of divergences , i.e. different from those coming from the standard loop integral in . therefore , the feynman rules at @xmath28 for @xmath148 insertions change with respect to the @xmath3 ones . namely , a @xmath148 insertion in the generic diagram of fig.[fig : insertions ] produces an integral of the type for the internal loop momenta @xmath275 and then is not equivalent to a simple @xmath276 power as for @xmath3 . one of the consequences of the above results is that when considering all the diagrams contributing to a given @xmath276 order , the @xmath91 and @xmath277 powers mix , so that a larger number of diagrams has to be considered . in fig.[fig : finitetdiags3 ] we have displayed all the diagrams that would give @xmath167 contributions , all of them including @xmath140 and @xmath249 insertions according to the results and . the vertex with no @xmath278 insertions is the effective thermal vertex in fig.[effvertx ] . attached to each diagram , we have indicated the different powers of @xmath279 that it gives rise to . for the renormalized amplitude at @xmath89 . we plot the different topological configurations contributing , so that diagrams ( e ) , ( h ) , ( j ) , ( k ) , ( l ) , ( p ) are multiplied by two , corresponding to the possible vertex insertions . ] we have calculated all diagrams in fig.[fig : finitetdiags3 ] with the feynman rules discussed above . the result is that the amplitude to that order remains finite with the same @xmath3 renormalizations of @xmath140 and @xmath249 given in - , which is a nontrivial consistency check . furthermore , the analysis of the result reveals some interesting features that will shed light on the renormalization scheme to be followed in the general case . first of all , we show the @xmath280 calculation . in addition to diagrams ( a ) , ( b ) , ( c ) of fig.[fig : finitetdiags3 ] , which are the counterparts of ( a ) , ( b ) , ( c ) in fig.[fig : t0diags3 ] with the thermal vertex , we have to consider as well diagram ( e ) in fig.[fig : finitetdiags3 ] , since it includes a @xmath281 contribution , as well as diagram ( g ) , whose @xmath282 part has to be taken into account . altogether , we obtain for the amplitude at that order : @xmath283\right\}+{{\cal o}}(s^3)\nonumber \\ & = \frac{s}{nf^2}f(i_\beta)\left\{1+\frac{s}{2f^2}f(i_\beta)\left[2g_1^r(\mu)+j_{fin}(p;t;\mu)\right]\right\}+{{\cal o}}(s^3 ) , \label{ampts2}\end{aligned}\ ] ] where we have separated the loop integral into its divergent and finite parts according to and and we have used exactly the same renormalization of the @xmath140 constant as for @xmath3 , namely . the thermal amplitude in is explicitly finite and scale independent to this order . moreover , note that we can write the thermal amplitude to that order in a form similar to the renormalized @xmath3 case in as follows : @xmath287}-\frac{s j(p;t)}{2f^2}\right]=\frac{nf^2}{s}\left[\frac{1}{g_0(s)}-\frac{i_\beta}{f^2}-\frac{s j(p;t)}{2f^2}\right ] \nonumber\\ = & \frac{nf^2}{s}\left[\frac{1}{g_r(s;\mu)}-\frac{i_\beta}{f^2}-\frac{s j_{fin}(p;t;\mu)}{2f^2}\right]\nonumber\\ \rightarrow & a_r(p;t)=\frac{s g_r(s;\mu)f[g_r(s;\mu ) i_\beta]}{nf^2}\frac{1}{1-\frac{s g_r(s;\mu)f[g_r(s;\mu ) i_\beta]}{2f^2}j_{fin}(p;t;\mu)}+{{\cal o}}(s^3).\end{aligned}\ ] ] @xmath290\right . \nonumber\\+&\left.\frac{s^2}{f^4}f^2(i_\beta)\left[g_1j(p;t)+\frac{1}{4}j^2(p;t)+g_2\left(1-\frac{i_\beta}{f^2}\right)+g_1 ^ 2\frac{i_\beta}{f^2}\right]\right\}+{{\cal o}}(s^4)\nonumber \\= & \frac{s}{nf^2}f(i_\beta)\left\{1+\frac{s}{2f^2}f(i_\beta)\left[2g_1^r(\mu)+j_{fin}(p;t;\mu)\right]\right.\nonumber\\ + & \left.\frac{s^2}{f^4}f^2(i_\beta)\left[g^r_1 ( \mu)j_{fin}(p;t;\mu)+\frac{1}{4}j_{fin}^2(p;t;\mu)+g_2^r(\mu)\left(1-\frac{i_\beta}{f^2}\right)+\left[g_1^r(\mu)\right]^2\frac{i_\beta}{f^2}\right]\right\}+{{\cal o}}(s^4 ) , \label{ampts3}\end{aligned}\ ] ] @xmath291}{nf^2}\frac{1}{1-\frac{s g_0(s)f[g_0(s ) i_\beta]}{2f^2}j(p;t)}+{{\cal o}}(s^4 ) \nonumber\\\rightarrow & a_r(p;t)=\frac{s g_r(s;\mu)f[g_r(s;\mu ) i_\beta]}{nf^2}\frac{1}{1-\frac{s g_r(s;\mu)f[g_r(s;\mu ) i_\beta]}{2f^2}j_{fin}(p;t;\mu)}+{{\cal o}}(s^4),\end{aligned}\ ] ] what is interesting for our purposes is that the renormalization given in can actually be achieved by a @xmath3 renormalization of each of the @xmath56-pion vertices in the original lagrangian by assigning them a @xmath295 factor in momentum space , as displayed in table [ feynruletable ] , thus generalizing the @xmath296 vertex renormalization in . cccc vertex & lagrangian & bare rule & renormalized rule + 4 pions & @xmath297 & @xmath298 & @xmath299 + + 6 pions & @xmath300 & @xmath301 & @xmath302 + + @xmath170 pions & @xmath303 & @xmath304 & @xmath305 + it is actually possible to trace the origin of this renormalization scheme in terms of the contributing diagrams . consider for instance diagrams with just one @xmath140 insertion . at @xmath280 , one has to sum the contributions to the amplitude from diagrams ( c ) , ( e ) and ( g ) in fig.[fig : finitetdiags3 ] , namely , with @xmath307 . now , the infinite contributions in the sum in amount to four - point diagrams with @xmath308 tadpole contractions ( @xmath309 ) i.e. , like the diagrams in fig.[effvertx ] with a @xmath310 vertex . but we can interpret each of those diagrams as the contribution to the scattering amplitude of a multiplicative renormalization of the @xmath311-pion vertex of the lagrangian given by @xmath312+{{\cal o}}(s^3)=\frac{s}{(nf^2)^k } g_0^k(s)+{{\cal o}}(s^3)$ ] , for which closing @xmath313 lines gives @xmath314 to leading order in @xmath1 . this is precisely the rule given in table [ feynruletable ] . we have checked that we find the same rule by analyzing the remaining @xmath182 insertions in the graphs in fig.[fig : finitetdiags3 ] . specifically , the @xmath167 contribution with one @xmath140 insertion ( diagrams ( e ) , ( g ) , ( h ) and ( k ) ) give the renormalization rule in table [ feynruletable ] for diagrams with the one - loop @xmath117 function , the @xmath167 with one @xmath249 insertion ( diagrams ( j ) and ( n ) ) give the linear part in @xmath249 of the @xmath315 contribution which we have just analyzed above to @xmath280 , while the @xmath167 with two @xmath140 insertions ( diagrams ( i ) , ( l ) , ( m ) , ( o ) , ( p ) , ( q ) ) reproduce precisely the @xmath316 part of those @xmath317 terms . therefore , in this way we are able to reinterpret all @xmath318 insertions in fig.[fig : finitetdiags3 ] giving @xmath279 mixed powers with @xmath319 and @xmath320 , in terms of the @xmath2-independent renormalization scheme in table [ feynruletable ] and , from the contributing diagrams without mixed terms . for higher insertions , we would need higher order diagrams with respect to those of fig . [ fig : finitetdiags3 ] , i.e. , up to @xmath321 . the crucial conclusion is that following the above renormalization scheme also in the general nonperturbative case , namely starting from the full amplitude , yields a finite scattering amplitude with a @xmath3 renormalization , as discussed in the main text . @xmath322 \notag \\ & = -\frac{1}{4}\left[1+\frac{1}{2}\sum_{k=0}^{\infty}{\frac{1}{k+1}\left(\frac{{\bm{\pi}}^2}{nf^{2}}\right)^{k+1}}\right]\square{\bm{\pi}}^2=-\frac{1}{4}\left[1-\frac{1}{2}\ln\left(1-\frac{{\bm{\pi}}^2}{nf^{2}}\right)\right]\square{\bm{\pi}}^2 , \label{laglogform}\end{aligned}\ ] ] @xmath323-({\bm{\pi}}^2)^{j+1}\square{\bm{\pi}}^2\right\ } ; \\ \partial_{\mu}{\bm{\pi}}_{a}\partial^{\mu}{\bm{\pi}}^{a}&=\partial_{\mu}({\bm{\pi}}_{a}\partial^{\mu}{\bm{\pi}}^{a})-\frac{1}{2}\square{\bm{\pi}}^2,\end{aligned}\ ] ] from the previous expression , we see that the feynman rules for the renormalized lagrangian listed in table [ feynruletable ] , as long as @xmath19 scattering in the large-@xmath1 limit is concerned , are equivalent to replace : therefore , the renormalization scheme we are analyzing here is very natural in the sense that every @xmath325 power in the expansion is renormalized with the same power of the @xmath326 function , acting as showed in . since those powers come from the same metric covariant function , as indicated in , this is consistent with the introduction of the counterterm lagrangians in a covariant fashion as we have discussed in section [ sec : ren ] . m. i. buchoff , m. cheng , n. h. christ , h .- ding , c. jung , f. karsch , z. lin and r. d. mawhinney _ et al . _ , phys . d * 89 * , 054514 ( 2014 ) . t. bhattacharya , m. i. buchoff , n. h. christ , h .- ding , r. gupta , c. jung , f. karsch and z. lin _ et al . _ , phys . lett . * 113 * , 082001 ( 2014 ) . k. a. olive _ et al . _ [ particle data group collaboration ] , chin . c * 38 * , 090001 ( 2014 ) . j. r. pelaez , arxiv:1510.00653 [ hep - ph ] . v. bernard , u. g. meissner and i. zahed , phys . d * 36 * , 819 ( 1987 ) . j. p. blaizot , r. mendez - galain and n. wschebor , annals phys . * 307 * , 209 ( 2003 ) . a. gomez nicola , f. j. llanes - estrada and j. r. pelaez , phys . b * 550 * , 55 ( 2002 ) . a. dobado , a. gomez nicola , f. j. llanes - estrada and j. r. pelaez , phys . c * 66 * , 055201 ( 2002 ) . d. fernandez - fraile and a. gomez nicola , eur . j. c * 62 * , 37 ( 2009 ) . j. a. oller , e. oset and j. r. pelaez , phys . lett . * 80 * , 3452 ( 1998 ) . j. r. pelaez and g. rios , phys . lett . * 97 * , 242002 ( 2006 ) . j.zinn-justin , quantum field theory and critical phenomena " , oxford university press 2002 .

starbursts are episodes of intense star - formation that occur in the central regions of galaxies , and dominate the integrated emission from the galaxy . they are a significant component of the present- day universe , being the site of 25% of the high - mass star- formation . they offer unique ` laboratories ' for testing our ideas about star - formation , the evolution of high - mass stars , and the physics of the interstellar medium . they serve as local analogs of the processes that were important in the origin and early evolution of galaxies and in the heating and chemical enrichment of the inter - galactic medium . in this contribution i review starbursts from this broad cosmogonical perspective , stressing several key lessons we have learned from starbursts : 1 ) violent , transient events play a significant role in the origin and evolution of galaxies . 2 ) galaxies do not evolve as ` island universes ' : starbursts are triggered by galaxy interactions and produce outflows of hot chemically - enriched gas that ` pollute ' the inter- galactic medium . 3 ) dust dramatically affects of view of high - mass star - formation in starbursts and ( probably ) in high - redshift galaxies . throughout this review i emphasize the importance of space - based observations in understanding starbursts .

astro-ph9708263

starbursts are brief episodes of intense star - formation that occur in the central - most 10@xmath1 to 10@xmath2 pc - scale regions of galaxies , and dominate the overall luminosity of the galaxy . their star - formation rates are so high that the existing gas supply can sustain the starburst for only a small fraction of the age of the universe ( in agreement with detailed models of the observed properties of starbursts , which imply typical burst ages of - order 10@xmath3 years or less ) . comprehensive reviews of the starburst phenomenon may be found in the volumes edited by leitherer et al ( 1991 ) and franco ( 1997 ) . starbursts are powered by high - mass stars , which i will define to have masses greater than 8 m@xmath4 . these stars are very luminous ( l @xmath5 few thousand l@xmath4 ) , and so ( when present in significant numbers ) they dominate the energy output of a population of stars . they are also short - lived ( a few million to a few - tens - of - million years ) and therefore trace relatively recent star - formation . high - mass stars dominate the heating of the interstellar medium , the gas out of which new stars form . they are also the ` thermonuclear furnaces ' that forge most of the elements heavier than he , and then disperse these heavy elements ( hereafter ` metals ' ) back into the interstellar medium when the dying high- mass stars explode as supernovae . in the local universe , high - mass stars are found in both normal galaxies like our own milky way and in starburst galaxies . in normal galaxies they are distributed in the spiral arms throughout the @xmath0 30-kpc - scale disk of the galaxy . in such galaxy disks , the existing supply of gas can sustain the current rate of star- formation for many gigayears , consistent with a relatively slow evolution of the star - formation rate over much of the history of the universe . as already stated above , the high - mass stars formed in starburst galaxies are concentrated into a small region in the galaxy center ( @xmath0 100 times smaller than the galaxy - as - a - whole ) and are the consequence of a transient event with a duration @xmath6 10@xmath3 years ( i.e. @xmath6 1% the present age of the universe ) . high - mass stars are typically very hot ( t @xmath5 25,000 k ) , and thus the luminous output of a starburst should have its peak in the vacuum - ultraviolet range ( @xmath7 @xmath0 912 to 3000 ) . on the other hand , starbursts are rich in interstellar gas and dust . the effective absorption cross - section of these dust grains is a strong inverse function of wavelength ( cf . calzetti 1997 ) . the dust grains therefore absorb much of the space - ultraviolet starlight and are thereby heated to temperatures of - order 10 to 100 k. the grains then cool by emitting far - infrared radiation . thus , samples of low- redshift starburst galaxies have come mostly from surveys that select galaxies with unusually bright ultraviolet or far - infrared emission ( cf . huchra 1977 ; soifer et al 1989 ) . first of all , starbursts are a very significant component of our present - day universe , and as - such deserve to be understood in their own right . the most complete and well - characterized sample of starbursts is that drawn from the far - infrared iras survey ( e.g. soifer et al 1989 ) . the rather strong inverse correlation between gas - depletion timescales and luminosity means that we may consider that starbursts dominate the set of galaxies lying above the ` knee ' in the far - infrared galaxy luminosity function ( l@xmath8 few @xmath9 l@xmath4 ) . by this reckoning , starbursts provide about 10% of the bolometric emissivity of the local universe . thus , starbursts are an energetically - significant phenomenon . it is also instructive to estimate the rate of high - mass star- formation in starbursts compared to that in the disks of normal galaxies in the local universe . high - mass stars are so hot and luminous that they are the dominant source of photons energetic enough to photoionize hydrogen in the interstellar medium . as the ionized hydrogen recombines and the captured electron cascades to the ground - state , it emits photons . thus , the luminosity of these hydrogen emission - lines is a measure of the total number of high- mass stars in a galaxy . since we know the lifetimes of these stars , this yields an average rate at which such stars must have formed in the recent past ( cf . kennicutt 1983 ; leitherer & heckman 1995 ) . of course the observed fluxes of the hydrogen emission - lines must be corrected for the effects of extinction by dust . in normal galaxies ( where the extinction due to dust is not severe ) , the optical balmer emission - lines can be used . in dusty starbursts , infrared or mm- wave lines ( which are much less affected by dust - extinction ) must be used . applying this technique to the kraan - korteweg & tammann ( 1979 ) volume - limited catalog of galaxies in the nearby universe ( d @xmath6 10 mpc ) , i find that the four most luminous circum - nuclear starbursts in the local universe ( m 82 , ngc 253 , m 83 , and ngc 4945 ) together comprise about 25% of the total high - mass star- formation in this volume , and the six most - actively - star - forming disks of normal galaxies contribute a comparable amount . in specific terms , the rate of high - mass star - formation in the few- hundred - parsec - scale starburst in m 82 actually exceeds the rate in the entire disk of the spiral galaxy m 101 ( which has a surface area approximately four orders - of - magnitude larger than the m 82 starburst ) ! thus , both in terms of total energy production and rate of high - mass star - formation , starbursts are indeed highly significant components of the present universe . while starbursts are therefore fascinating and significant objects , they are even more important when placed in the broader context of contemporary stellar and extragalactic astrophysics . they are ideal laboratories in which to study : 1 ) the formation and evolution of massive stars 2 ) the physics of the interstellar medium under extreme conditions 3 ) the processes involved in the formation and early evolution of galaxies 4 ) the processes likely responsible for chemically - enriching and heating the intergalactic medium . the cosmological relevance of starbursts has been dramatically underscored by one of the most spectacular discoveries in years : the existence of a population of high - redshift ( z @xmath5 2 ) field galaxies ( cf . steidel et al 1996 ; lowenthal et al 1997 ; pettini , this volume ) . the number density of these galaxies implies that they almost certainly represent precursors of typical present - day galaxies in an early actively - star - forming phase . this discovery moves the study of the star - forming history of the universe into the arena of direct observations ( cf . madau et al 1996 ; madau , this volume ) , and gives added impetus to the quest to understand local starbursts . it is these aspects of the starburst phenomenon - the way in which nearby starburst galaxies can used to address major issues in cosmogony - that are the subject of the rest of this contribution . before discussing starbursts in more detail , it is useful to briefly summarize the major lessons in extragalactic astronomy and cosmology that we have learned from the study of starbursts in the local universe . + * i ) violent , transient events play a significant role in the origin and evolution of galaxies . * current interest in starbursts has been stimulated by ( and has helped to create ) a paradigm shift in which the importance of cataclysmic events in the evolution of galaxies has become increasingly recognized . rather than simply evolving in a steady ` clockwork ' fashion , galaxies may also evolve discontinuously through bursts of star - formation . this is reminiscent of the concept of ` punctuated equilibrium ' for the evolution of biological systems . + * ii ) galaxies do not evolve as ` island universes ' or ` closed boxes . ' * there is instead two - way causal communication between galaxies and their environment : strong starbursts are triggered by the interaction or merger of two galaxies and the subsequent starbursts produce outflows of ionizing radiation and metal - enriched matter that travel into the galactic halos and possibly beyond . further pursuing the biological analogy , this might be thought of as the ` ecology ' of galaxy evolution . + * iii ) dust in galaxies drastically affects our view of star- formation in local starbursts and ( most likely ) in galaxies in the early universe . * as noted earlier , the intrinsic spectral energy distribution of a young stellar population peaks in the vacuum - ultraviolet , just where the opacity of the interstellar dust grains that permeate the region of star - formation is maximal . this effect is especially relevant at high - redshifts , since the visible - light observations that provide the bulk of our information about these galaxies sample the vacuum - ultraviolet portion of their rest - frame spectrum . starburst galaxies require the presence of substantial amounts of gas within the central - most few hundred parsecs of the galaxy . the relatively low efficiency of stars for energy production ( e @xmath10 m@xmath11c@xmath1 ) , coupled with the severe energetic demands ( ranging from @xmath010@xmath12 ergs for a modest starburst like m 82 up to @xmath0 10@xmath13 ergs for an ` ultraluminous ' starburst like arp 220 ) mean that interstellar gas masses of at least 10@xmath3 to 10@xmath14 m@xmath4 are needed to ` fuel ' a starburst ( even assuming 100% efficiency for the conversion of gas into stars ) . indeed , mm - wave interferometric maps of the molecular gas in powerful starbursts provide direct observational evidence for such material ( cf . sanders & mirabel 1997 ) . as larson ( 1987 ) , elmegreen ( 1997 ) , and others have argued , the surface mass density of the cold interstellar matter in starburst nuclei is so high ( typically of - order 10@xmath2 m@xmath4 pc@xmath15 ) , that the growth time for gravitational instabilities that would lead to star - formation in the starburst is extremely short ( @xmath0 a million years ) . moreover , the timescales for gas depletion in a starburst via star - formation and/or supernova - driven outflows ( see section 3.2 below ) are also short compared to the minimum time it would take gravity to move material into the starburst from the large - scale disk of the galaxy ( 10@xmath16 to 10@xmath3 years vs. 10@xmath3 to 10@xmath17 years respectively ) . larson therefore argues that since the gas that fuels the starburst must be assembled at least as fast as it is consumed , and since the gas has a mass comparable to the entire mass of the interstellar medium in a normal galaxy , powerful starbursts can only occur when some process allows a substantial fraction of the interstellar medium of a galaxy to flow inward by at least an order - of - magnitude in radius at velocities that are comparable to the orbital velocities in the galaxy s disk . by way of illustration , larson emphasizes that the collapse of a self - gravitating system implies a maximum infall rate of roughly 25 ( v@xmath18/50 km s@xmath19)@xmath2 m@xmath4 per year . this can be compared to typical estimated star - formation rates of 10 to 100 m@xmath4 per year in starbursts and typical orbital velocities of roughly 150 km s@xmath19 in the ` host ' galaxy of the starburst ( e.g. lehnert & heckman 1996b ) . to summarize , the fueling of a starburst requires a mechanism that can induce non- circular motions that are both large in amplitude and involve a substantial fraction of the interstellar medium of the galaxy . fortunately , these theoretical or heurstic arguments are supported by observational evidence . as samples of increasingly luminous starbursts are examined , the strength of the evidence that they are strongly interacting or merging systems also increases ( cf . sanders & mirabel 1997 ) . such evidence is mostly morphological : 1 ) the presence of two or more nuclei within a single distorted envelope ( consistent with a nearly - complete merger of two galaxies ) . 2 ) the presence of a second galaxy with the requisite proximity and relative brightness , together with the bridges and long linear ` tails ' that are the hallmark of tidally- interacting galactic disks ( toomre & toomre 1972 ) . the physical picture that has emerged from the theoretical interpretation of inceasingly - sophisticated numerical simulations ( cf . mihos & hernquist 1994a , b ) is that during the close passage of two galaxies , tidal stresses act to strongly perturb the orbits of the stars and gas in the galaxy disk . the dissipation of kinetic energy as gas collides with gas , allows the gas to become sufficiently displaced from the stars that gravitational torques act between the stars and gas to transfer significant amounts of angular momentum from the gas to the stars . the gas can thereby fall into the center of the galaxy , where it can fuel a starburst . if the passage of the two galaxies is slow and inter - penetrating enough , dynamical friction can transfer enough kinetic energy from the stars to the dark - matter halo to allow the two galaxies to merge into a single galaxy . such mergers or strong interactions should take a few times the galaxy rotation period ( e.g. @xmath20 years ) , with the intense starburst phase being significantly shorter ( cf . mihos & hernquist 1994a , b ) . these timescales are loosely consistent with independent estimates of starburst lifetimes , and ( in any case ) are quite consistent with the transient nature of starbursts . perhaps the most spectacular aftermath of a starburst is the ` life - changing ' event associated with a galaxy merger . as first suggested by toomre ( 1977 ) , it is now clear both observationally and theoretically that the merger between two disk galaxies is at least one avenue for producing an elliptical galaxy ( cf . schweizer 1992 ; barnes 1995 ) . the presence of ` kinematically - decoupled ' cores in elliptical galaxies ( central regions in which the stars have a very different angular momentum axis from the rest of the galaxy ) may be a fossil record of an merger - induced starburst in the distant past ( e.g. franx & illingworth 1988 ) . likewise , the so- called ` e+a ' galaxies , whose spectra appear to be the sum of the stars found in normal elliptical ( e ) galaxies plus a population dominated by a stars ( stars having a lifetime of 0.1 to 1 gigayears ) , may be ` post - starbursts ' systems that are intermediate in evolutionary state between merger - driven starbursts and bona- fide elliptical galaxies ( cf . zabludof et al 1997 ) . another very intruiging consequence of starbursts is the ` super star cluster ' phenomemon . meurer et al ( 1995 ) find that typically 20% of all the vacuum - ultraviolet light ( and hence @xmath020% of the high - mass stars ) in starbursts is produced by very luminous stellar clusters ( dubbed ` super star clusters ' by oconnell et al 1994 ) . these clusters are so luminous that it is very tempting to describe them as the possible progenitors of globular clusters ( cf . whitmore & schweizer 1995 ) . while the relationship between the super star clusters and classical globular clusters is a matter of on - going debate ( cf . van den bergh 1995 ; meurer 1995 ) , work on the nearest such objects demonstrates that they indeed have sizes , velocity dispersions , and masses that are similar to those of typical globular clusters ( ho & filippenko 1996a , b ) . more generally , the census of high - mass star - formation in the local universe described in section 1.2 above has an very interesting implication . adopting the ` copernican principle ' ( that is , assuming that our local piece of the present - day universe is representative of the universe elsewhere and elsewhen ) , the fact that @xmath025% of all high - mass stars are formed in starbursts would mean one of the following : + * i ) operating over the history of the universe , starbursts make about 25% of _ all _ the stars in galaxies ( not just high - mass stars ) . * given the central location of starbursts in galaxies and the fact that the most powerful starbursts seem to be associated with the building of elliptical galaxies via mergers , the only plausible fossil record of these ancient starbursts would be the central parts of disk galaxies ( bulges and inner disks ) and elliptical galaxies . however , bulges and ellipticals are redder and older than the average disk of a spiral galaxy , so it may be difficult to make this idea hang - together quantitatively . that is , starbursts may have indeed produced @xmath025% of the stars in galaxies on - average integrated over the history of the universe , but could not be doing - so today or the central parts of galaxies would be too blue and too bright . + * ii ) starbursts ( unlike normal galactic disks ) make only high- mass stars and therefore leave little long - term residue . * by this i mean that high - mass stars are short - lived ( @xmath6 few @xmath21 years ) compared to low - mass stars ( e.g. 10@xmath14 years for the sun ) . if starbursts form only high - mass stars , then shortly after the burst is over , the starburst would fade away completely , leaving only a residue of neutron stars and black holes ( the ` chesire cat ' model ) . leitherer ( 1997 ) reviews the evidence that starbursts may ` manufacture ' only high - mass stars ( cf . rieke et al 1993 vs. satyapal et al 1997 ) . in contrast , star - formation with a normal initial mass - function extending down to well below 1 m@xmath4 produces stars ( and hence starlight and stellar mass ) that remain behind essentially forever ( @xmath5 the age of the universe ) . this is in fact how we believe the large - scale disks of normal galaxies are constructed over the course of many gyrs . the inter - galactic medium ( ` igm ' ) represents the majority of the baryonic matter in clusters of galaxies ( cf . donahue , this volume ) . comparison of the total baryonic content of the universe ( cf . turner et al 1996 ) to the baryons located inside the visible parts of galaxies implies that perhaps as much as 80 to 90% of the baryons in the universe might reside in the igm . the source for the heating and re - ionization of the igm at high - redshifts is a puzzle , since the igm is already ionized at the time the first known qsos appear at z @xmath0 5 . ionization due to a pre - qso phase of galaxy formation has been suggested as possibility ( cf . madau & shull 1996 ; miralda - escude & rees 1997 ; giroux & shapiro 1996 ) . as i will argue below , starbursts may play a vital role not only in heating the igm , but also in ` polluting ' it with the heavy elements forged by high - mass stars . the high - mass stars in starbursts may be a copious source of photons that are energetic enough to ionize hydrogen and neutral helium in the igm . put most simply , for star - formation with a normal initial - mass function that extends down to 0.1 m@xmath4 ( 8 m@xmath4 ) , each baryon that is incorporated into stars yields the emission of roughly 3000 ( 20000 ) photons that are capable of ionizing hydrogen . thus , a little bit of star - formation could go a long way in photoionizing the igm ( especially during very early times before the earliest known qsos appear on the scene ) . however , direct observations of local starburst galaxies below the lyman limit at 912 using the hopkins ultraviolet telescope show that only a small fraction of the total ionizing radiation produced by these starbursts ( as measured by the luminosity of the hydrogen recombination - lines ) escapes into the igm ( leitherer et al 1995 ; hurwitz et al 1997 ) . giallongo et al ( 1997 ) have recently set an upper limit of 20% on the average fraction of ionizing radiation due to high - mass stars escaping from all galaxies out to z @xmath0 1 based on a comparison of the total amount of star - formation between z @xmath22 0 and z @xmath22 1.3 ( cf . madau , this volume ) and the upper limits on the brightness of the cosmic background ionizing - radiation field in the present universe . thus , the role of starbursts ( and high - mass stars in general ) in photoionizing the igm is still unclear . the discovery of star- forming galaxies at z @xmath5 3 in principle allows the spectral region below the lyman - break to be probed directly with the new generation of 8 and 10-meter - class telescopes , and so the relative importance of high - mass stars and qsos in photoionizing the igm can be determined for this early epoch . over the last few years , observations have provided convincing evidence of the existence ( and even the ubiquity ) of ` superwinds ' - galactic - scale outflows of gas driven by the collective effect of multiple supernovae and stellar winds in a starburst ( cf . heckman , lehnert , & armus 1993 ; lehnert & heckman 1996a ; veilleux , cecil , & hawthorn 1996 ) . x - ray data have proved particularly crucial since they are the only direct probe of the hot gas that contains most of the mass and energy in the flow . soft x - ray emission ( hot gas ) is a generic feature of the halos of the nearest starburst galaxies ( dahlem , weaver , & heckman 1997 ) . the estimated thermal energy content of this gas represents a significant fraction of the time - integrated mechanical energy supplied by the starburst , while the soft x - ray luminosity is only a few % of the supernova heating rate . these results mean that little of the mechanical energy supplied by supernovae and stellar winds in starbursts is radiated away . thus , in principle superwinds may efficiently transport much of the mechanical energy supplied by high - mass stars into the igm , making such stars an important heating source for the igm ( see donahue s contribution to this volume ) . the temperature of the hot outflowing gas in starbursts is considerably cooler ( few to ten million degrees ) than would be expected for pure thermalized supernova+stellar wind ejecta ( 10@xmath3 k ) . given these temperatures , will this gas escape the galaxy gravitational potential , or will the gas remain bound and perhaps cool and return as a galactic fountain ? following wang ( 1995 ) , the ` escape temperature ' for hot gas in a galaxy potential with an escape velocity v@xmath23 is given by t@xmath23 @xmath0 4 @xmath24 10@xmath25 ( v@xmath23/600 km s@xmath26)@xmath1 k ( where i have roughly chosen parameters appropriate to a typical spiral galaxy like our own ) . the gas temperatures in superwinds are observed to be independent of the galaxy rotation speed , while t@xmath23 should scale as the square of the rotation speed ( _ modulo _ the extent of the dark matter halo ) . thus , it appears that the x - ray emitting gas can easily escape from dwarf galaxies undergoing starbursts ( della - ceca et al 1996 ) but possibly not from the halos of the most massive starburst galaxies ( dahlem , weaver , & heckman 1997 ; wang 1995 ) . as larson & dinerstein ( 1975 ) and many others have proposed , this selective loss of metal - enriched gas from the shallowest galaxy potential wells may be the physical mechanism that underlies the strong correlation between the metal abundance of the stellar population and the escape velocity in elliptical galaxies ( e.g. franx & illingworth 1990 ) . in this sense , superwinds should have a particularly devastating impact on dwarf galaxies ( e.g. dekel & silk 1986 ; marlowe et al 1995 ) . if indeed superwinds carry substantial amounts of metal - enriched gas out of starbursts , we should see the cumulative effect of these flows in the form of a metal - enriched igm and/or metal - enriched gaseous halos around galaxies . by now it is clear that typical mgii absorption - line systems at z @xmath6 1 seen in the spectra of distant qsos arise in the metal - enriched halos of intervening galaxies ( cf . churchill , steidel , & vogt 1996 ) . these galaxies appear to be normal systems ( not starbursts ) , so it is unlikely that ` living , breathing ' superwinds are implicated . nevertheless , it is plausible that the metals in these galaxy halos may trace ` fossil ' superwinds . that is , galactic halos may be polluted primarily by the episodic eruptions associated with powerful starbursts ( heckman 1997 ) . can we say anything about the gas _ outside _ galaxies ? the existence of an igm in clusters of galaxies whose metal content exceeds that of all the stars in all the cluster s galaxies is one of the most remarkable phenomena in extragalactic astronomy ( see donahue , this volume ) . recent asca x - ray spectra of this gas show super - solar abundance ratios for the @xmath27-process elements like o , ne , and si relative to fe ( mushotzky et al 1996 ) . this implicates ` core - collapse ' supernovae ( the end product of high - mass stars ) and - by inference - superwinds as the source of the metals in the cluster igm ( loewenstein & mushotzky 1996 ) . if most of the metals in clusters of galaxies are floating around outside galaxies , could this be true more globally in the universe ? we do nt know the answer in the present - day universe , but the situation at high - redshift is very intruiging . burles & tytler ( 1996 ) use the detection of ovi absorption - lines in the spectra of background qsos to estimate that the minimum metallicity of the entire gas - phase baryonic component of the universe at z @xmath0 1 is @xmath5 0.02 h solar for @xmath28 0.01 h@xmath29 . this is a lower limit , because it assumes all the oxygen is the form of ovi and that the detected ovi absorption - lines are optically - thin . at z @xmath5 2 , the presence of metals in the ` ly@xmath27 forest ' ( the ` cloudy ' component of the early igm ) is certainly suggestive of the dispersal of chemically - enriched material by early superwinds ( cf . et al 1995 ; tytler et al 1995 ; madau & shull 1996 ) . the ly@xmath27 forest material appears to have a metallicity of about 10@xmath29 solar , but a high ratio of si / c that is suggestive of core - collapse supernovae as the source of metals ( cowie et al 1995 ; giroux & shull 1997 ) . given the above indications of a metal - enriched igm at high - z , do we see any _ direct _ evidence for the outflow of metal - enriched gas from starburst galaxies in the early universe ? we will return briefly to this in section 4.2 below . the relatively large inferred masses of the x - ray emitting gas in local superwinds imply that we are primarily observing emission from ambient gas that has been ` mass - loaded ' in some way into the superwind ( cf . suchkov et al 1996 ; hartquist et al 1997 ; della ceca et al 1996 ) . that is , each supernova explosion on - average must heat and eject a few hundred m@xmath4 worth of gas up to of - order 10@xmath16 k. the large amount of mass - loading implies that the ratio of gas that is blown out of the starburst compared to the mass that is turned into stars ranges from of - order unity ( if a normal complement of low- mass stars is formed in the starburst ) to of - order ten ( if only high - mass stars are formed ) . to the extent that local starbursts are analogs to forming galaxies , this suggests that galaxy formation may have been an inefficient process with only a minority of the initial complement of baryons being retained and converted into stars and the majority expelled into galactic halos or the igm . observations in the vacuum - uv spectral regime ( @xmath7 @xmath0 912 to 3000 ) are crucial for both understanding local starbursts , and for relating them to galaxies at high - redshift . not only is this the energetically - dominant spectral region for the hot stars that power the starburst , this is the spectral regime where we can most clearly observe the direct spectroscopic signatures of these hot stars . moreover , the vacuum - uv contains a wealth of spectral features including the resonance transitions of most cosmically- abundant ionic species . these give uv spectroscopy a unique capability for diagnosing the ( hot ) stellar population and the physical and dynamical state of gas in starbursts . since ground - based optical observations of galaxies at high- redshifts sample the vacuum - uv portion of their rest - frame spectrum , we can not understand how galaxies evolved without documenting the vacuum - uv properties of galaxies in the present epoch . in particular , a thorough understanding of how to exploit the diagnostic power of the rest - frame uv spectral properties of local starbursts will give astronomers powerful tools with which to study star - formation and galaxy - evolution in the early universe . the vacuum - uv spectra of starbursts are characterized by strong absorption features , as seen in figure 1 . these absorption features can have three different origins : stellar winds , stellar photospheres , and interstellar gas . detailed analyses of hubble space telescope ( hst ) and hopkins ultraviolet telescope ( hut ) spectra of starbursts show that the resonance lines due to species with low - ionization potentials ( oi , cii , siii , feii , alii , etc . ) are primarily interstellar in origin . in contrast , the resonance lines due to high - ionization species ( nv , siiv , civ ) can contain significant contributions from both stellar winds and interstellar gas , with the relative importance of each varying from starburst to starburst ( cf . conti et al 1996 ; leitherer et al 1996 ; heckman & leitherer 1997 ; gonzalez - delgado et al 1997 ) . the most unambiguous detection of stellar photospheric lines is provided by excited transitions , but these lines are usually rather weak ( cf . heckman & leitherer 1997 ) . 10truept we ( heckman et al 1997 - hereafter h97 ) have just completed an analysis of the vacuum - uv spectroscopic properties of a large sample of starburst galaxies in the local universe using the data archives of the international ultraviolet explorer ( iue ) satellite . taken together with the results of previous studies of starbursts in the vacuum - uv , the principal lessons we have learned are as follows : + * first , dust has a profound effect on the emergent uv spectrum . * as noted above , the luminous output of a starburst should have its peak in the vacuum - ultraviolet range ( @xmath7 @xmath0 912 to 3000 ) . on the other hand , starbursts are rich in interstellar gas and dust , and the effective absorption cross - section of these dust grains is a strong inverse function of wavelength . this means that the effect of dust on the vacuum - uv properties of starbursts is profound . previous papers have established that various independent indicators of dust extinction in starbursts observed with iue correlate strongly with one another . calzetti et al ( 1996 ) show that the spectral slope in the vacuum - uv continuum ( as parameterized by @xmath30 , where f@xmath31 ) correlates strongly with the nebular extinction measured in the optical using the balmer decrement . meurer et al ( 1995;1997 ) show that @xmath30 correlates well with the ratio of far - ir to vacuum - uv flux : the greater the fraction of the uv that is absorbed by dust and re - radiated in the far - ir , the redder the vacuum - uv continuum . the interpretation of these correlations with @xmath30 in terms of the effects of dust are particularly plausible because the _ intrinsic _ value for @xmath30 in a starburst is a robust quantity . figures 31 and 32 in leitherer & heckman ( 1995 ) show that @xmath30 should have a value between about -2.0 and -2.6 for the range of ages and initial mass functions appropriate for starbursts ( cf . leitherer 1997 ) . our detailed understanding of the above results is incomplete , since they must involve both the geometrical distribution of the dust , stars , and gas in the starburst and the vacuum - uv extinction law for the dust . however , the available data strongly suggest that ( quite surprisingly ) much of the dust responsible for the vacuum- uv extinction is apparently distributed around the starburst in the form of a moderately inhomogeneous foreground screen or ` sheath ' surrounding the starburst ( gordon , calzetti , & witt 1997 ) . interestingly , h97 find that the amount of vacuum - uv extinction in starbursts correlates strongly with the bolometric luminosity of the starburst : only starbursts with l@xmath32 @xmath6 few @xmath33 l@xmath4 have colors expected for an unreddened starburst and have vacuum - uv luminosities that rival their far - ir luminosities . starbursts that lie at or above the ` knee ' in the local starburst luminosity function ( l@xmath32 @xmath5 few @xmath9 l@xmath4 - cf . soifer et al 1987 ) have red uv continua ( @xmath34 -1 to + 0.4 ) and are dominated by far - ir emission ( l@xmath35 @xmath0 10 to 100 l@xmath36 ) . we also find that the amount of vacuum - uv extinction in starbursts correlates well with the absolute blue magnitude and the rotation speed of the galaxy ` hosting ' the starburst : starbursts in more massive galaxies are more dust - shrouded . + * second , the metallicity of the starburst also strongly affects the uv spectrum . * apart from the effects of dust , a starburst s metallicity is the single most important parameter in determining its vacuum - uv properties . in fact , metallicity and the effects of dust are well- correlated , as shown in figure 2 . 10truept at low metallicity ( @xmath610% solar ) a significant fraction of the intrinsic vacuum - uv actually escapes the starburst ( l@xmath35/l@xmath36 @xmath0 unity ) , and the vacuum - uv colors are consistent with the intrinsic ( unreddened ) colors expected for a starburst population ( @xmath34 -2 ) . in contrast , at high metallicities ( @xmath5 solar ) 90% to 99% of the energy emerges in the far - ir ( l@xmath35/l@xmath36 = 10 to 100 ) and the vacuum - uv colors are very red ( @xmath34 0 ) . storchi - bergmann , calzetti , & kinney ( 1994 ) had previously noted the correlation between metallicity and uv color . these correlations have a straightforward interpretation : the vacuum - uv radiation escaping from starbursts suffers an increasing amount of reddening and extinction as the dust - to - gas ratio in the starburst ism increases with metallicity . this will be true provided that neither the gas column density towards the starburst , nor the fraction of interstellar metals locked into dust grains are strong inverse functions of metallicity . the properties of the vacuum - uv absorption - lines are also strongly dependent on metallicity . figure 1 shows that both the high - ionization ( e.g. civ@xmath71550 and siiv@xmath71400 ) and low - ionization ( e.g. cii@xmath71335 , oi@xmath71302 , and siii@xmath7@xmath71260,1304 ) resonance absorption - lines arei significantly stronger in starbursts with high metallicity . the metallicity - dependence of the high - ionization lines ( noted previously by storchi - bergmann , calzetti , & kinney 1994 ) is not surprising , given the likely strong contribution to these lines from stellar winds . theoretically , we expect that since stellar winds are radiatively driven , the strengths of the vacuum - uv stellar wind lines will be metallicity - dependent . this is confirmed by available hst and hut spectra of lmc and especially smc stars ( walborn et al 1995 ; puls et al 1996 ) . if post - main - sequence stars ( supergiants and wolf - rayet stars ) contribute significantly to the integrated light , the stellar wind properties enter in a second , indirect way . that is , the integrated uv spectrum heavily depends on the evolutionary history of the ob population , which in turn is critically dependent on the stellar mass - loss rates , which in turn are a function of the metal abundance ( see maeder & conti 1994 ) . figure 1 also shows a metallicity - dependence for the strengths of the uv absorption - lines that are of stellar - photospheric rather than interstellar origin ( we know they are photospheric lines because they correspond to transitions out of highly excited states ) . such lines are generally rather weak in starburst spectra and/or blended with strong interstellar features . they include ciii@xmath71175 , siiii@xmath71417 , ciii@xmath7@xmath71426,1428 , sv@xmath71502 , siiii@xmath71892 , and feiii@xmath371925,1960 . the weak but statistically - significant correlation between metallicity and the strength of the low - ionization resonance lines ( which are primarily formed in the interstellar medium of the starburst ) is also unsurprising . analyses of hst spectra ( cf . pettini & lipman 1995 ; heckman & leitherer 1997 ; sahu & blades 1997 ; gonzalez - delgado et al 1997 ) show that the strong interstellar lines are saturated ( highly optically - thick ) . in this case , the equivalent width of the absorption - line ( w ) is only weakly dependent on the ionic column density ( n@xmath38 ) : w @xmath27 b[ln(n@xmath38/b)]@xmath39 , where b is the normal doppler line - broadening parameter . over the range that h97 sample well , the starburst metallicity increases by a factor of almost 40 ( from 0.08 to 3 solar ) , while the equivalent widths of the strong interstellar lines only increase by an average factor of about two . this is consistent with the strong interstellar lines being quite optically - thick . + * third , the properties of the strong interstellar absorption- lines reflect the hydrodynamical consequences of the starburst , and do not straightforwardly probe the gravitational potential of the galaxy . * as noted above , analyses of hst and hut uv spectra of starbursts imply that the interstellar absorption - lines lines are optically- thick . their strength is therefore determined to first - order by the velocity dispersion in the starburst ( see above ) . thus , these lines offer a unique probe of the kinematics of the gas in starbursts . the enormous strengths of the starburst interstellar lines ( equivalent widths of 3 to 6 in metal - rich starbursts ) require very large velocity dispersions in the absorbing gas ( few hundred km s@xmath19 ) . are these gas motions primarily due to gravity or to the hydrodynamical ` stirring ' produced by supernovae and stellar winds ? both processes probably contribute to the observed line - broadening . h97 find only a very weak ( but still statistically significant ) correlation between the strengths ( widths ) of the interstellar absorption - lines and the rotation- speed of the host galaxy . the weakness of the correlation suggests that gravity alone is not the whole story . moreover , for any astrophysically - plausible ionic column density , the interstellar lines are typically at least twice as broad as can be explained by optically - thick gas orbiting in the galaxy potential well . the most direct evidence for a non - gravitational origin of the gas motions comes from analyses of hst and hut spectra , which show that the interstellar lines are often blueshifted by one - to - several - hundred km s@xmath19 with respect to the systemic velocity of the galaxy ( heckman & leitherer 1997 ; gonzalez - delgado et al 1997 ; sahu & blades 1997 ; lequeux et al 1995 ) . this demonstrates directly that the absorbing gas is flowing outward from the starburst , probably ` feeding ' the superwinds described in section 3.2 above ) . the results summarized in section 4.1 have a variety of interesting implications for the interpretation of the rest - frame - uv properties of galaxies at high - redshift . powerful starbursts in the present universe emit almost all their light in the far - infrared , not in the ultraviolet . thus , an ultraviolet census of the local universe would significantly underestimate the true star - formation - rate and would systematically under - represent the most powerful , most metal - rich starbursts occuring in the most massive galaxies . this may also be true at high - redshift , where the current estimates of star - formation rely almost exclusively on data pertaining to the rest - frame vacuum - uv ( see madau , this volume ) . for example , current samples might under- represent young / forming massive elliptical galaxies . using the strong correlation between the vacuum - uv color of local starbursts ( @xmath30 ) and the ratio of far - ir to vacuum - uv light emitted by local starbursts , meurer et al ( 1997 ) estimate that an average vacuum - uv - selected galaxy at high - redshift ( e.g. steidel et al 1996 ; lowenthal et al 1997 ) suffers 2 to 3 magnitudes of extinction . the ` correct ' prescription for de - extincting the high- z galaxies in order to correctly obtain the bolometric luminosity and star - formation rate is a matter of on - going debate ( see the contributions by pettini and madau to this volume ) . it is worth emphasizing that the purely empirical method of meurer et al bypasses the substantial uncertainties about the dust extinction law , and the initial mass function and age of the stellar population in the high - z galaxies ( it assumes only that the high - z galaxies behave like local starbursts ) . as shown recently by burigana et al ( 1997 ) , the existing limits on the far - ir / sub - mm cosmic background are consistent with the global star - formation rates inferred by meurer et al at z@xmath52 due to dusty starbursts unless the dust in these galaxies is quite cool ( t@xmath40 @xmath6 20 k ) compared to the dust in local starbursts ( t@xmath40 @xmath0 30 to 60 k ) . this seems very unlikely , since the bolometric surface - brightnesses of the high - redshift galaxies are similar to local starbursts ( meurer et al 1997 ) , implying that the energy density of the radiation field that heats the grains is similar in the two types of objects ( cf . figure 5 and the relevant discussion in lehnert & heckman ( 1996b ) . in any case , it seems fair to conclude that the history of star - formation in the universe at early times ( z @xmath5 1 ) will remain uncertain until the effects of dust extinction are better understood . the strong correlation shown in figure 2 between vacuum - uv color ( @xmath30 ) and metallicity in local starbursts - if applied naively to high - z galaxies - would suggest a broad range in metallicity from substantially subsolar to solar or higher and a median value of 0.3 to 0.5 solar . this is signifcantly higher than the mean metallicity in the damped ly@xmath27 systems ( the major repository of hi gas at these redshifts ) , but this may be due to selection effects : the uv - selected galaxies are the most actively star- forming regions of galaxies , while the damped ly@xmath27 systems tend to sample the outer , less - chemically - enriched parts of galaxies or perhaps proto - galactic fragments ( e.g. pettini et al 1997 ) . it would be interesting to use the correlations between absorption- line strengths and metallicity in local starbursts to ` guesstimate ' the metallicity of the high - z galaxies ( see pettini , this volume ) . one prediction based on the local starbursts ( h97 ) is that the high - z galaxies should show a strong correlation between the strength of the uv absorption - lines ( stellar and interstellar ) and @xmath30 ( the more metal - rich local starbursts are both redder and stronger - lined ) . as noted above , meurer et al ( 1997 ) argue that the uv - selected galaxies at high - redshift suffer substantial amounts of extinction . if their proposed extinction - corrections are applied , the high - z galaxies have very large bolometric luminosities ( @xmath0 10@xmath41 to 10@xmath42 l@xmath4 for h@xmath43 = 75 km s@xmath19 mpc@xmath19 and q@xmath43 = 0.1 ) . interestingly , the bolometric surface- brightnesses of the extinction - corrected high - z galaxies are very similar to the values seen in local starbursts : @xmath0 10@xmath14 to 10@xmath41 l@xmath4 kpc@xmath29 . the high - redshift galaxies appear to be ` scaled - up ' ( larger and more luminous ) versions of the local starbursts . the physics behind this ` characteristic ' surface- brightness is unclear ( cf . meurer et al 1997 ; lehnert & heckman 1996b ) . however , it is intruiging that the implied average surface- mass - density of the stars within the half - light radius ( @xmath0 10@xmath1 to 10@xmath2 m@xmath4 pc@xmath29 ) is quite similar to the values in present - day elliptical galaxies . are we witnessing the formation of elliptical and bulges ? finally , based on local starbursts - it seems likely that the gas kinematics that are measured in the high - z galaxies using the interstellar absorption - lines are telling us a great deal about the hydrodynamical consequences of high - mass star - formation on the interstellar medium , but rather little ( at least directly ) about the gravitational potential or mass of the galaxy . even the widths of the nebular emission - lines in local starbursts are not always reliable tracers of the galaxy potential well ( cf . lehnert & heckman 1996b ) . this means that it will be tricky to determine masses for the high - z galaxies without measuring real rotation- curves via spatially - resolved spectroscopy . on the brighter side , if the kinematics of the interstellar absorption - lines can be generically shown to arise in outflowing metal - enriched gas , we can then directly study high - redshift star- forming galaxies caught in the act of ` polluting ' the intra - cluster medium and inter - galactic medium with metals in the early universe ( see section 3.2 above ) . in fact , there is now rather direct observational evidence that this is the case . i will need to briefly digress to explain this evidence . as emphasized above , the interstellar absorption - lines are significantly blue - shifted with respect to the systemic velocity of the galaxy ( v@xmath44 ) in many local starbursts . in the high - redshift galaxies there is rarely a good estimator of v@xmath44 ( although the weak stellar photospheric lines listed in section 4.1 above are a promising possibility in spectra with adequate signal - to - noise ) . it is also the case in local starbursts that the true galaxy systemic velocity lies between the velocity of the uv interstellar absorption - lines and the ly@xmath27 _ emission _ line ( lequeux et al 1995 ; gonzalez - delgado et al 1997 ) . this is due to outflowing gas that both produces the blue - shifted absorption - lines and absorbs - away the blue side of the ly@xmath27 emission - line . thus , a purely - uv signature of outflowing gas is a blueshift of the interstellar absorption - lines with respect to the ly@xmath27 emission - line ( even though neither is at v@xmath44 ) . . recently , franx et al ( 1997 ) have seen just this effect in a spectrum of the highest - redshift object in the universe : a gravitationally - lensed galaxy at z = 4.92 . they find that the ly@xmath27 emission line is redshifted by about 400 km s@xmath19 relative to the siii@xmath71260 interstellar absorption - line across the entire face of the galaxy . more generally , lowenthal et al ( 1997 ) have constructed a composite uv spectrum of 12 high - z galaxies . in figure 3 i overplot this spectrum on an hst spectrum of the local starburst iras 0833 + 6517 ( gonzalez - delgado et al 1997 ) . the spectra have been aligned to force a coincidence in wavelength between the strong interstellar absorption - lines . the similarity of the two spectra is striking : a strong redshifted ly@xmath27 emission- line and a weak blue - shifted ly@xmath27 _ absorption _ line . this composite spectrum strongly suggests that the outflow of metal- enriched gas at velocities of a few hundred km s@xmath19 is a generic feature of the high - z galaxies . _ if _ the outflowing gas escapes into the igm , such flows could bring an igm with @xmath45 @xmath0 0.01 h@xmath29 up to a mean metallicity of @xmath5 10@xmath29 solar by a redshift of 2.5 ( cf . madau & shull 1996 ) . starbursts are defined as brief episodes ( @xmath6 10@xmath3 years ) of intense star - formation that occur in the central - most 0.1 to 1 kpc- scale regions of galaxies and dominate the integrated emission from the galaxy . they are a significant component of the present - day universe : they provide roughly 10% of the bolometric emissivity of the local universe and are the sites of @xmath025% of the high- mass star - formation . thus , they deserve to be understood in their own right . they also offer unique ` laboratories ' for testing our ideas about star formation , the evolution of high - mass stars , and the physics of the interstellar medium . they serve as local analogs of the processes that were important in the origin and early evolution of galaxies and in the heating and chemical enrichment of the inter- galactic medium ( igm ) . in this contribution i have reviewed starbursts from this broad cosmogonical perspective , stressing several key lessons we have learned from starbursts : + * i ) violent , transient events play a significant role in the origin and evolution of galaxies . * rather than simply evolving in a steady ` clockwork ' fashion , galaxies also evolve discontinuously through powerful bursts of star - formation , triggered by galaxy interactions . + * ii ) galaxies do not evolve as ` island universes ' or ` closed boxes . ' * powerful starbursts are triggered by galaxy interactions and mergers that can create an elliptical galaxy out of two spirals . the starbursts produce outflows of hot metal - enriched gas ( ` superwinds ' ) that pollute the inter - galactic medium . there is now direct observational evidence for a metal - enriched igm and for the outflows at low- and high - redshift that are responsible for this enrichment . study of superwinds in the local universe suggest that galaxy formation may have been an inefficient process with the majority of gas being ejected . dust dramatically affects of view of high - mass star- formation in starbursts and ( probably ) in high - redshift galaxies . * powerful starbursts in the present universe emit almost all their light in the far - infrared . an ultraviolet census of the local universe would significantly underestimate the true star - formation- rate and could systematically under - represent the most powerful , most metal - rich starbursts occuring in the most massive galaxies . applying the empirical relations followed by local starbursts to the uv - selected galaxies at high - redshift implies that the latter typically suffer 2 to 3 magnitudes of extinction , and that the star - formation rate in the universe need not be smaller at z@xmath0 3 than at z@xmath01 . the high - z galaxies would then have estimated bolometric surface - brightnesses and sizes consistent with a population of forming elliptical galaxies . + * iv ) space - based observations are crucial in understanding starbursts . * x - ray data ( rosat , asca , axaf , xmm ) are vital for studying the superwind phenomenon , since they directly probe the hot gas that comprises most of the mass and energy in the flow . ultraviolet data ( iue , hst , hut , fuse ) offer detailed information about both the hot stars that power the starburst and the dynamics of the interstellar medium . visible observations with the high - angular resolution of hst have revealed the importance of galaxy interactions and mergers at low and medium redshift and given us a tantalizing glimpse of the star - forming history of the early universe . finally , the infrared ( iras , iso , cobe , wire , sirtf , sofia , ngst ) represents the primary channel for energy loss for the dusty , metal - rich starbursts . such data not only document the presence of dusty starbursts in the local universe ( z @xmath46 1 ) , but also constrain their global importance at high - redshift . i would like to thank my collaborators over the past few years ( especially d. calzetti , m. dahlem , r. della ceca , r. gonzalez- delgado , m. lehnert , c. leitherer , g. meurer , c. robert , a. suchkov , and k. weaver ) for hard work and inspiration . special thanks to j. lowenthal for providing a composite spectrum of high- z galaxies and to m. carollo , m. dickinson , p. madau , c. martin , c. mihos , c. norman , m. pettini , c. steidel , and m. voit for stimulating discussions and constructive suggestions . this work has been partially supported by nasa grant nagw-3138 . barnes , j. 1995 , in the formation of galaxies , ed . c. munoz - tunoz , cup burigana , c. , danese , l. , de zotti , g. , franceschini , a. , mazzei , p. , & toffolati , l. 1997 , mnras , 287 , l17 burles , s. , & tytler , d. 1996 , apj , 460 , 584 calzetti , d. 1997 , in the uv universe at low and high redshift , ed . w. waller , m. fanelli , hollis , j. , & danks , a. 1997 , woodbury : aip , in press calzetti , d. kinney , a. , & storchi - bergmann , t. 1996 , apj , 458 , 132 churchill , c. , steidel , c. , & vogt , s. 1996 , apj , 471 , 164 conti , p. , leitherer , c. , & vacca , w. 1996 , apj , 461 , l87 cowie , l. , songaila , a. , kim , t. , & hu , e. 1995 , aj , 109 , 1522 dahlem , m. , weaver , k. , & heckman , t. 1997 , in preparation dekel , a. , & silk , j. 1986 , apj , 303 , 36 della ceca , r. , griffiths , r. , heckman , t. , & mackenty , j. 1996 , apj , 469 , 662 elmegreen , b. 1997 , in rev . mex . de astronom . y astrofis . serie de conferencias , vol . 6 , starburst activity in galaxies , p. 165 franco , j. 1997 , rev . . de astronom . y astrofis . serie de conferencias , vol . 6 , starburst activity in galaxies franx , m. , & illingworth , g. 1990 , apj , 359 , l41 franx , m. , & illingworth , g. 1988 , apj , 327 , l55 franx , m. , illingworth , g. , kelson , d. , van dokkum , p. , & tran , k .- v . 1997 , preprint giallongo , e. , fontana , a. , & madau , p. 1997 , mnras , in press giroux , m. , & shapiro , p. 1996 , apjs , 102 , 191 giroux , m. , & shull , s.m . 1997 , aj , 112 , 1505 gordon , k. , calzetti , d. , & witt , a. 1997 , apj , in press gonzalez - delgado , r. , leitherer , c. , heckman , t. , ferguson , h. , & lowenthal , j. 1997 , apj , in press hartquist , t. , dyson , j. , & williams , r. 1997 , apj , 482 , 182 heckman , t. 1997 in galactic halos , ed . d. zaritzky , asp , in press heckman , t. , & leitherer , c. 1997 , aj , 114 , 69 heckman , t. , lehnert , m. , & armus 1993 , in the evolution of galaxies and their environments , ed . m. shull and h. thronson , kluwer , 455 heckman , t. , robert , c. , leitherer , c. , van der rydt , f. , garnett , d. , & kinney , a. 1997 ( h97 ) , in preparation ho , l. , & filippenko , a. 1996a , apjl , 466 , l83 ho , l. & filippenko , a. 1996b , apj , 472 , 600 huchra , j. 1977 , apjs , 35 , 171 hurwitz , m. , jelinsky , p. , & van dyke dixon , w. 1997 , apj , 481 , l31 kennicutt , r. 1983 , apj , 272 , 54 kraan - korteweg , r. , & tammann , g. 1979 , astronnach , 300 , 181 larson , r. , 1987 in starbursts & galaxy evolution , ed . t. thuan , t. montmerle , & j. tran thanh van , editions frontieres , 467 larson , r.b . , & dinerstein , h.l . 1975 , pasp , 87 , 911 lehnert , m. , & heckman , t. 1996a , apj , 462 , 651 lehnert , m. , & heckman , t. 1996b , apj , 472 , 546 leitherer , c. , & heckman , t. 1995 , apjs , 96 , 9 leitherer , c. , walborn , n. , heckman , t. , & norman , c. 1991 , massive stars in starburst galaxies , cambridge university press leitherer , c. , ferguson , h. , heckman , t. , & lowenthal , j. 1995 , apj , 454 , l19 leitherer , c. , vacca , w. , conti , p. , filippenko , a. , robert , c. , & sargent , w. 1996 , apj , 465 , 717 leitherer , c. 1997 , in the stellar initial mass function , ed . g. gilmore , san francisco : asp , in press lequeux , j. , kunth , d. , mas - hesse , j. , & sargent , w. 1995 , a&a , 301 , 18 loewenstein , m. , & mushotzky , r. 1996 , apj , 466 , 695 lowenthal , j. , koo , d. , guzman , r. , gallego , j. , phillips , a. , faber , s. , vogt , n. , & illingworth , g. 1997 , apj , 481 , 673 oconnell , r. gallagher , j. , & hunter , d. 1994 , apj , 433 , 65 madau , p. , & shull , s.m . 1996 , apj , 457 , 551 madau , p. , ferguson , h. , dickinson , m. , giavalisco , m. , steidel , c. , & fruchter , a. 1996 , mnras , 283 , 1388 marlowe , a. , heckman , t. , wyse , r. , & schommer , r. 1995 , apj , 438 , 563 maeder , a. , & conti , p. 1994 , araa , 32 , 227 meurer , g. 1995 , nature , 375 , 742 meurer , g. , heckman , t. , leitherer , c. , kinney , a. , robert , c. , & garnett , d. 1995 , aj , 110 , 2665 meurer , g. , heckman , t. , leitherer , c. , lowenthal , j. , & lehnert , m. 1997 , aj , 114 , 54 mihos , j.c . , & hernquist , l. 1994a , apj , 425 , l13 mihos , j.c . , & hernquist , l. 1994b , apj , 431 , l9 miralda - escude , j. & rees , m. 1997 , apj , 478 , 57 mushotzky , r. , loewenstein , m. , arnaud , k. , tamura , t. , fukazawa , y. , matsushita , k. , kikuchi , k. , & hatsukade , i. 1996 , apj , 466 , 686 pettini , m. , & lipman , k. 1995 , a&a , 297 , 63 pettini , m. , smith , l. , king , d. , & hunstead , r. 1997 , apj , 486 , 665 puls , j. et al . 1996 , a&a , 305 , 171 rieke , g. , loken , l. , rieke , m. , & tamblyn , p. 1993 , apj , 412 , 99 sahu , m. , & blades , j.c . 1997 , apj , 484 , l125 sanders , d. , & mirabel , i.f . 1997 , araa , 34 , 749 satyapal , s. , et al . 1997 , apj , 483 , 148 schweizer , f. 1992 , in physics of nearby galaxies : nuture or nature ? t. thuan et al . , editions frontieres : gif- sur - yivette , 283 soifer , b.t . , sanders , d. , madore , b. , neugebauer , g. , lonsdale , c. , persson , s.e . , & rice , w. 1987 , apj , 320 , 238 steidel , c. , giavalisco , m. , pettini , m. , dickinson , m. , & adelberger , k. 1996 , apj , 462 , l17 storchi - bergmann , t. , calzetti , d. & kinney , a. 1994 , apj , 429 , 572 suchkov , a. berman , v. , heckman , t. , & balsara , d. 1996 , apj , 463 , 528 toomre , a. 1977 , in the evolution of galaxies and stellar populations , ed . b. tinsley & r. larson , yale university observatory , 401 toomre , a. & toomre , j. 1972 , apj , 178 , 623 turner , m. , truran , j. , schramm , d. , & copi , c. 1996 , apj , 466 , l59 tytler , d. , et al . 1995 , in qso absorption lines , ed . g. meylan , springer : berlin , 289 van den bergh , s. 1995 , nature , 374 , 215 veilleux , s. , cecil , g. , & bland - hawthorn , j. 1996 , sciam , 274 , 86 walborn , n. , lennon , d. , haser , s. , & kudritzki , r. 1995 , pasp , 107 , 104 wang , b. 1995 , apj , 444 , 590 whitmore , b. , & schweizer , f. 1995 , aj , 109 , 960 zabludoff , a. et al . 1996 , apj , 466 , 104

the adaptive module of the 4.1-m soar telescope , sam , corrects ground - layer turbulence using a uv laser guide star . it has been commissioned in 2013 and it is in regular science operation since 2014 . sam works with the ccd imager covering a field or with the speckle camera . it operates routinely and stably , delivering resolution in the band equal to the free - atmosphere seeing . this paper describes the sam system as a whole , providing essential reference for its users and technical information of interest to instrumentalists . operation of the instrument , its performance , and science projects done with sam so far are reviewed .

1608.05593

soar adaptive module ( sam ) is a facility adaptive optics ( ao ) instrument at the 4.1-m southern astrophysical research telescope , soar . it improves the natural seeing by partial compensation of turbulence near the ground . although various aspects of sam have been covered in the literature , there has been no single reference with a comprehensive description of the entire instrument and its performance . the present article fills this lacuna . it is primarily intended for sam users and may be of interest to instrument developers . on the other hand , the format of a journal article is not well suited for technical details such as full optical prescription , software algorithms , etc . ; this paper is _ not _ an instrument reference manual . from the outset , the soar telescope was designed to deliver high angular resolution over a relatively narrow field , unlike other wide - angle 4-m telescopes @xcite . it contains a fast actuated tertiary mirror ( m3 ) for compensation of atmospheric tilts . extending the compensation order using the ao technology appeared natural . however , soar is used mostly at visible wavelengths , where the standard ao compensates turbulence only over a narrow few - arcsec field and needs a bright star or a powerful laser to do so . in 2002 , when soar was completed , the ao technology enabled good turbulence correction only in the infrared ( ir ) . however , low - order ir ao systems still provided _ partial _ correction in the visible and could improve the resolution by a factor of @xmath2 , as demonstrated by the pueo instrument @xcite . such partial correction could benefit a large range of science projects done at soar at visible wavelengths . the idea of correcting only low - altitude turbulence , known as ground - layer ao ( glao ) , was first formulated by @xcite . in glao , the compensation quality is traded against uniformity of correction over a wide field . quantitative analysis of glao was done later by @xcite , see also @xcite . although several natural guide stars ( ngss ) or laser guide stars ( lgss ) located at the perimeter of the field of view are optimal for sensing ground - layer turbulence , even a single lgs at low altitude is an acceptable solution for glao . in such case , the so - called cone effect effectively reduces sensitivity to high - altitude turbulence . the concept of sam , first presented by @xcite , is based on a single lgs created by rayleigh scattering of an ultraviolet ( uv ) laser . the use of a uv laser , advocated by @xcite , has several advantages : efficient rayleigh scattering proportional to @xmath3 , easy separation of the laser light from longer science wavelengths , and no ocular hazards , making the beam safe to airplanes . the lgs is needed for a high sky coverage ; an alternative wave - front sensor ( wfs ) using several ngss would deliver a much inferior compensation . at that time , an ao system with a uv laser was being built for the 2.5-m telescope at mt . wilson @xcite , while a similar ao instrument with a green rayleigh laser was developed at the 4.2-m william hershel telescope @xcite . for technical reasons , those two projects have not produced any science results , but we learned from their experience . the glao system with a green rayleigh laser at the 6.5-m mmt telescope @xcite is not used in regular observing programs . on the other hand , the robo - ao instrument with a uv laser at the palomar 1.5-m telescope @xcite is highly productive ( e.g. * ? ? ? the sam instrument design was driven by the science goal ( improve seeing over a moderate field of view at optical wavelengths ) , availability of suitable technology ( e.g. uv lasers and fast ccd detectors ) , and technological trade - offs . in the development of sam , we tried to use commercial or otherwise proven components , whenever possible . new and/or critical elements were first prototyped and tested . we consulted with the ao team at the mmt telescope who generously shared their experience . the sam system as a whole is presented in [ sec : overview ] . the main ao module is described in [ sec : aom ] , the laser subsystem is covered in [ sec : lgs ] . [ sec : sw ] describes the sam software , and [ sec : perf ] provides information on the sam performance . science operation is covered in [ sec : sci ] . in [ sec : sum ] we discuss the place of sam among other ground - based and space facilities , its future instruments and upgrades . the choice of the deformable mirror ( dm ) has a large impact on the instrument design . an adaptive secondary mirror , like in the mmt , would be an ideal technical solution , but it was beyond our budget and , in hindsight , would have substantially delayed the project . initially we hoped to use a small electrostatic dm from oko - tech and to build a small instrument around it , but the stroke of this dm turned out to be insufficient @xcite . we selected instead the bimorph dm bim-60 made by cilas . the initial optical relay based on refractive optics was replaced by a classical scheme with a pair of off - axis parabolic mirrors ( oaps ) and a pupil diameter of 50 mm . then the ao - corrected beam has the same focal ratio and pupil location as at the uncorrected telescope focus , allowing to use with sam any instrument built for soar ( within allowable mass and space limits ) . sam acts as a seeing - improvement interface between soar and its science instrument . sam feeds corrected images to its internal ccd detector , sami ( 4k@xmath44k ccd ) or to a visitor instrument ( this port is now occupied by the speckle camera ) . figure [ fig : all ] presents a simplified block - diagram of the sam system . light from celestial objects reflected by the soar primary mirror m1 experiences three more reflections before reaching the sam adaptive - optics module ( aom ) , attached to the instrument - selector box ( isb ) at the nasmyth focal station of soar . during observations , sam rotates to compensate for parallactic angle . the optical relay inside aom includes reflection from the dm . the corrected beam is sent either to sami or to the visitor instrument , both mounted on the opposite sides of the aom . the lgs is created by the pulsed uv laser located in a box on the telescope truss and also subjected to variable gravity depending on the telescope elevation . the laser power supply and chiller are housed in a thermally insulated rack attached to the fork of the soar mount , in fixed gravity . the laser beam is transported to the laser launch telescope ( llt ) located behind the soar secondary mirror , m2 . each pulse of the laser light propagates through the atmosphere and part of it is scattered back . photons scattered from the distance of 7 km are selected by the fast range - gate shutter inside aom , synchronized with the laser pulses . the uv light is analyzed by the wfs , and the correction is sent to the dm in closed loop . other essential elements of sam shown in figure [ fig : all ] are its electronics , mostly housed in a rack near the aom , and its computers with software . the rayleigh scattering preserves polarization . the aom rotates relative to the telescope and the laser , hence its polarization - sensitive wfs can not be adjusted to match the polarization of the laser . to mitigate this effect , the laser beam is propagated circularly polarized , and the scattered circularly - polarized beam is converted back to linear polarization in the wfs . only the oblique reflection from m3 slightly affects the polarization tuning . sam is different from the majority of ao systems in several ways . first , the tip and tilt ( tt ) compensation is provided by m3 , the actuated tertiary mirror of soar , upstream from the instrument . second , tt is sensed by two small guide probes deployed in the input ( un - corrected ) focal plane . this avoids the need to split photons in wavelength between science and guide channels . the probes sample the @xmath5 square patrol field outside the @xmath0 science field . in this design the tilts introduced by the deformable mirror ( dm ) are not sensed , but the dm control is tilt - free . the uv light from the lgs is focused well behind the nominal telescope focal plane , meaning that the light path of lgs photons inside sam is very different from the starlight path . nevertheless , major non - common - path errors between the science and wfs paths were avoided . readers who are not interested in the sam design can skip the following material and go to section 6 . in this section , the aom and its subsystems are covered . the optical layout of the sam aom is shown in figure [ fig : oap ] , and the optical elements are listed in table [ tab : opt ] . the beam , focused by the telescope and reflected by the m4 mirror inside the isb , propagates to the focal plane inside aom , where guide probes ( gps ) are located , and then hits the first oap . simulated light source can be inserted in the path before oap1 for tests and dm flattening . the collimated beam after oap1 propagates towards the dm , with an incidence angle of 1416 . after the dm , the uv light is reflected by the dielectric coating of the dichroic plate , while longer wavelengths are transmitted through the dichroic towards oap2 and re - focused on the science instrument ( after the fold mirrors ) . the sami fold mirror is installed permanently . the visitor fold mirror is mounted on a translation stage ; when inserted , it intercepts the beam and directs it to the visitor port . the atmospheric dispersion corrector ( adc ) can be inserted in the collimated beam after the dichroic . it consists of two cemented zero - deviation prisms , each made of the bak2 ( angle 445 ) and caf2 ( angle 554 ) glasses @xcite . immediately behind the adc , the collimated space is available for insertion of other elements , for example a non - redundant pupil mask or a fabry - perot etalon . the uv photons from the lgs at 7-km distance are focused well behind the science focal plane , very close to the oap1 . the lgs beam on the dm is divergent . upon reflection from the dichroic , it is focused by the custom uv triplet lens , then the beam is folded by a mirror . the triplet also works well for the visible light , enabling sam operation with the ngs . however , this mode was used only during initial commissioning , and sam can not be reconfigured back to work with ngs without losing the lgs capability . the mirrors in the aom , including the dm , have a protected silver coating with dielectric layers enhancing reflectivity at the laser wavelength 355 nm . the overall transmission in the science path was measured to be 0.90 at 633 nm . during optical integration , the strehl ratio in the science channel at this wavelength reached 0.60 after alignment and with a carefully flattened dm . this means that , despite substantially different light path in the wfs , the non - common - path errors are not significant ( remember that sam optics does not need to be diffraction - limited ) . the optical relay with two oaps has a slightly curved field with a quadratic distortion . these effects , determined by the optical design , were confirmed by mapping the position and focus of the point light source . the curvature is small and it can be safely neglected , while the distortion is correctable by post - processing the images . the maximum lateral shift caused by the distortion is 0.6 mm or 18 . sam uses the bimorph or `` curvature '' dm bim-60 made by the french company cilas for the 8-m vlt telescope . the stroke of this dm is hence more than sufficient for the 4.1-m soar aperture , even considering the reduced pupil diameter of 50 mm instead of the nominal 60 mm . the curvature radius with a maximum voltage of 400v on all electrodes is measured to be 6.7 m ; a radius of 30 m would suffice to correct a 1 seeing in sam . the electrodes are arranged in radial geometry with five rings . without any voltage , the dm has a large static aberration , including defocus . about half of the dynamic range is used to `` flatten '' the dm . the `` flat '' voltages depend on ambient temperature and instrument orientation , because the thin dm is deformed by gravity as sam rotates , creating mainly defocus and trefoil . the gravity deformation is compensated by applying to all electrodes pre - calculated corrections varying as sines and cosines of the rotation angle . with those corrections , sam can work in open loop , which is useful in its normal operation when the lgs is not needed or can not be used for some reason . l cc l element & diam . next & notes + & ( mm ) & ( mm ) & + + focal plane & 60x60 & 810 & science field + oap1 & 150 & 890 & @xmath6 mm + dm & 60 & 150 & bim-60 + + dichroic & 80 & 32 & 614.7 mm to oap2 + adc prism1 & 80 & 3 & thickness 14.5 mm + adc prism2 & 80 & 550.6 & thickness 14.5 mm + oap2 & 150 & 345.5/455.5 & to visitor / sami + visitor fold & 210x150 & 478.6 & visitor instrument + sami fold & 210x150 & 368.6 & sami + science focus & 60x60 & n / a & + + dichroic & 80 & 180 & + wfs triplet & 80 & 65 & uv triplet + wfs fold & 100 & 616.3 & mirror + @xmath7 plate & 25 & 6.6 & + f - adjuster lens & 25 & 53.3 & single lens + field stop & 0.55x0.55 & 61.3 & square aperture + wfs collimator & 25 & 10.2 & single lens + polarizer-1 & 10 & 3 & foctek polarizer + pockels cell & 10 & ( 20 ) & qx1020 + polarizer-2 & 10 & 10.2 & foctek polarizer + re - imager l1 & 25 & 5.6 & + re - imager l2 & 25 & 43.3 & + s - h focal plane & n / a & 12.2 + s - h coll . lens & 25 & 14.7 + lenslet array & 2.2 & 6.5 & pupil 1.989 mm + ccd-39 & 2.2 & n / a & wfs detector + sam was first tested on the sky in 2009 with a simplified wfs , using an ngs for turbulence sensing . the uv dichroic was replaced then by a neutral beamsplitter . in the beginning of 2011 , when sam was finally commissioned in the ngs mode , the definitive lgs wfs was installed ; it is described here . both versions of the wfs used the same detector , ccd-39 , and the same lenslet array ( lla ) . the lgs wfs optics ( figure [ fig : lgs - wfs ] ) is tailored to match the commercial pockels cell qx1020 manufactured by cleveland crystals . constraints on the beam angular divergence in the cell imposed by its birefringent material , kd*p , and the clear aperture of the cell , 10 mm , called for a @xmath86-mm diameter beam inside the cell . so , after the field stop at the wfs focus the beam is collimated by a fused - silica lens ( good for the monochromatic laser light ) . it passes through the first glan - taylor polarizer , the pockels cell , and the second ( crossed ) polarizer . then two more singlet lenses re - shape the beam and form the pupil image of 1.99 mm diameter on the lla , to get 10@xmath410 pupil sampling in the shack - hartmann wfs . the lla is placed directly in front of the ccd . its pitch , 0.192 mm , equals eight ccd pixels . the lla focal length of 6.5 mm ( at 633 nm ) leads to the pixel scale of 0375 , and the field of view of 28 , with 8 pixels per sub - aperture . the optics of sam allows the lgs distance to be set between 7 km and 14 km ( it is fixed presently at 7 km ) . when the wfs is refocused to a different distance ( it is mounted on a translation stage ) , the pupil size on the lla changes slightly . to compensate for this variation , another weak singlet lens is placed before the focus on a small translation stage . axial motion of this focal adjuster allows us to keep the constant pupil diameter at the lla . the @xmath7 wave plate is placed together with this lens to transform the elliptically polarized laser beam into linear polarization . the parameters of the lla apo - q - p192-f5.75 were chosen optimally for the ngs wfs , while the second lla with a shorter focal length of 3.17 mm was procured for the lgs wfs from the same vendor , a@xmath9s . both llas are made of fused silica and have excellent optical quality . however , we could not use the short - focus lla as planned because the distance from the ccd surface to its glass window exceeded the specification by 1 mm . to match the pixel scale and the size of the lgs spots , we bin the ccd 2@xmath42 , to 075 per binned pixel . the ccd-39 detector has 80@xmath480 pixels of 24@xmath9 m size and four output amplifiers for fast readout . the vendor , e2v , guaranteed quantum efficiency of @xmath100.4 at 355 nm . the detector is cooled thermoelectrically to about @xmath11c . ccd-39 works in sam with the sdsu - iii controller . with the 2@xmath42 binning , the loop frequency is 478hz and the readout noise is about 5.5 electrons . the digitized signal is transmitted by the optical fiber from the controller to the acquisition board inside the real - time computer ( rtc ) . the bias pattern is not very stable , requiring its re - calibration a few times per night , while slow drifts of the bias in each quadrant are tracked and subtracted using signals in the empty corners of the image . the duration of the range - gate pulse determines the vertical extent of the lgs and hence the maximum spot elongation in the peripheral sub - apertures . we normally use the 150-m range gate , or 1@xmath9s gate pulses . the amplitude of the driving pulses producing the @xmath12 phase shift in the pockels cell at the laser wavelength of 355 nm is 2.7kv . the kd*p crystal in the cell is piezoelectric , so the driving pulses excite its acoustic oscillations at 95khz . as a result , the shutter does not close completely and some after - pulses or `` ringing '' appear . the 1@xmath9s drive pulses produce a spurious pulse of @xmath810% transmission delayed by 3.6@xmath9s from the opening edge of the gate . the weak `` tails '' of the lgs spots caused by this ringing are truncated by the wfs entrance aperture , except the innermost sub - apertures where they pass through and slightly bias the centroids . the entrance aperture of the wfs is carefully aligned laterally to avoid `` spilling '' of light into adjacent sub - apertures . this aperture mask is mounted on a translation stage which , when moved , replaces it by a uv source used to measure the reference positions in each sub - aperture . at the same time the main beam is intercepted by a diagonal mirror that sends it to the acquisition camera . we use a simple gc650 ccd detector from prosilica , with a gigabit ethernet interface , to acquire `` live '' image of the un - gated lgs . it is strongly peaked and easy to center . any lgs ao system needs a tip - tilt ( tt ) correction using natural guide stars because the lgs does not provide valid tt signal . in the case of glao , the requirements on residual tt errors are relaxed in comparison to the classical ao , as we do not need to reach the diffraction - limited resolution . however , the sky coverage is essential . our calculations indicated that the faint limit of the tt probes should be no less than @xmath13 mag for reaching the full sky coverage . atmospheric tt errors increase with increasing angular distance from the tt guide star ( so - called tilt anisoplanatism ) . the resulting psf non - uniformity was a concern for sam . we planned to have three guide probes ( gps ) picking stars around the science field of view ( fov ) . further study showed that with just two gps the correction uniformity is almost as good , so sam has two gps , each patrolling half of the field . the technical fov for picking the stars is @xmath14 square . we take advantage of the soar fast tertiary mirror for the tt correction . the gps are placed at the telescope focus upstream from the optical relay . to reduce vignetting , the width of the gps is minimized to 6 mm . the optical scheme of the gps is shown in figure [ fig : gp ] . the 2@xmath42 array of 0.5-mm acrylic micro - lenses , cut from a larger sample , is optically coupled to four multimode fibers with a core diameter of 100@xmath9 m . the soar pupil image on the fiber ends has a diameter of 90@xmath9 m , allowing some tolerance for alignment . each fiber , glued into a steel ferrule , is aligned with respect to its micro - lens to coincidence between images of all fibers formed by the micro - lenses . after alignment , the lenslets and ferrules are potted by transparent epoxy to form a solid module . the ferrules with fibers are slightly tilted , so that the source at the center is well within their acceptance cone ( numerical aperture ) which , together with the lenslet size , define the field of view , 15 per quadrant . the four fibers terminate by the fc connectors plugged into the module of four photon - counting avalanche photo - diodes ( apds ) , model spcm - a q4c from perkin elmer . the measured optical transmission is 0.7 . the acrylic lenslets do not transmit the uv light , so the probes are not affected by the lgs even without any protective filters . the right - hand part of figure [ fig : gp ] shows the actual guide probe without cover . the light from the telescope passes through a 3-mm hole and is deflected by a 3-mm right - angle prism before being focused on the lenslets and injected into the fibers . tilts of the prism allow co - alignment of the optical axis with the direction of the incoming light , i.e. with the soar pupil . the back side of the probe facing sam can project a point source created by a single - mode fiber and a red laser diode . this fiducial source is approximately coincident with the star . it proved very useful for a number of tasks , such as mapping the guide - probe coordinates on the ccd ( to ease guide - star acquisition ) and testing the optics of the aom internally . the apds are cooled thermoelectrically within their modules to reduce the dark count which varies from 200 to 1500hz , depending on the detector . the high voltage ( hv ) enabling the avalanche amplification is turned on only after positioning the probes . to protect the apds from damage by over - light , we installed a circuit that switches off the hv as soon as the current exceeds some threshold . this hardware protection has saved the apds when a room light was turned on accidentally during sam integration . guide stars ( gss ) brighter than @xmath15 mag sometimes trigger the over - light protection . comparison of the @xmath16 magnitudes of the gs with the actual counting rate shows a scatter of about 1 mag , mostly because of unreliable photometry in the usno-2 catalog . the mean flux @xmath17 ( sum of counting rates in the four quadrants ) follows the expected trend @xmath18 } = 10^ { 0.4(v_0 -v ) } , \label{eq : flux}\ ] ] with the zero points @xmath19 of 19.9 and 19.7 mag for gp1 and gp2 , respectively . these zero points match the estimated sensitivity . a star of @xmath20 mag gives a flux of 5.75 kel / s ( or 57 counts per 10ms loop cycle ) in gp1 . a gaussian star of 1 full width at half maximum , fwhm ( rms width 042 ) will have a centroid error on the order of 002 in the 10-hz servo bandwidth . we do guide on @xmath20 mag stars when no better choice is left . the 3 aperture of the gps is smaller than the pointing error of soar . to acquire a gs , we first take the pointing exposure with sami and determine the offset by measuring position of a star with known coordinates in the image . the gp is then positioned on the selected gs automatically , the hv on the apds is turned on . if the star is detected , the gp can be centered on it manually ; however , by closing the tt loop the star is centered automatically . manual centering of at least one gp is required if both gps are used , so that both gps work near null . if the star is not detected , we use a software tool that modulates the tertiary mirror circularly with increasing radii of 1 , 3 , and 5 . this effectively enlarges the capture zone of the gps . if the modulated signal is detected , the gp is moved towards the star . otherwise , we revise the pointing offset or select another star ( some `` stars '' in the usno-2 catalog are galaxies not suitable for guiding ) . the aom contains a built - in artificial uv light source than can be injected in the optical path instead of the lgs . the focus and pupil of the simulated beam mimic the actual lgs ; the beam can be artificially distorted , simulating the turbulence . this device , called tursim , is essential for the sam operation ; it is used for flattening the dm and testing the health of the ao loop during daytime . the 365-nm uv light in tursim is emitted by the photo - diode from nicia , with a diffuser and a circular aperture to emulate a 1 star . two dc motors with cam mechanisms translate the source laterally and serve for centering its image in the wfs . the beam is collimated by a custom uv triplet lens , passes through the pupil mask simulating central obscuration , and is re - imaged by another lens group with the telescope s focal ratio , at a fixed axial position corresponding to the lgs at 7 km . one or two phase screens can be inserted in the parallel beam near the pupil mask to simulate turbulence of variable strength and speed ; they are driven by regulated dc motors . the phase screens , made in - house @xcite , do not match exactly the expected kolmogorov spectrum , but this is not critical for testing sam . tursim has a deployable arm for projecting the simulated lgs on - axis . with a total of seven mechanisms , it is a rather complex subsystem . to transmit the tursim light through the wfs , its pockels cell is triggered by internally generated pulses rather than by the laser . all elements of sam are housed in a box - like aluminum structure , called bench ( figure [ fig : mech ] ) . it provides stiffness against gravity change in one direction as sam rotates at the nasmyth focus . the total range of flexure , measured by displacement of the image of the light source in the gp on the science ccd , is only @xmath21035 . to prevent image blurring by flexure , avoid long exposures near the zenith associated with large rotation during the exposure . the sam bench is made of aluminum plates bolted together . they have large rectangular openings to access the optics and mechanisms of sam . the covers are hermetized to prevent light and dust from entering sam . for the same reason , the entrance opening of sam is covered by an environmental shutter , opened only during observations . the total mass of sam ( including sami ) is 330 kg . an instrument up to 70 kg can be mounted on the visitor port . the mechanical design of sam was complicated because of the space restrictions at the nasmyth focus . the aom has 23 remotely controlled motions ( e.g. 6 in the guiders , 7 in tursim , 3 in the adc , etc . ) . to reduce heat dissipation inside the instrument , all motor controllers are housed in two glycol - cooled boxes attached outside the bench . however , the apd modules and the driver of the pockels cell are located inside sam , as well as the wfs detectors ccd-39 and gc650 . .parameters of sami [ cols="<,<",options="header " , ] the laser ( model q301-hd from jdsu ) is a frequency - tripled pulsed nd : yag with a nominal average power of 10w . it is located on the telescope truss , in a thermal box maintained at @xmath22c temperature and flushed with dry air . in the real operation ( under variable gravity and at the altitude of cerro pachn ) the internal power meter of the laser indicates its output power in the range from 7.5w to 7.8w . the beam quality measured in the laboratory is very good , @xmath23 . we operate the laser at its nominal pulse frequency of 10khz . the pulse duration is about 34ns . the jdsu company fabricated hundreds of such lasers , mostly for photo - lithography , so it is a rugged industrial product . so far , we had no problems with the laser ; it works without any servicing , apart from the coolant refreshment every year or so . the narrow beam emerging from the laser is expanded to the diameter of @xmath88 mm ( at @xmath24 intensity level ) and directed to the llt with one reflection from the lgs - m4 mirror at the soar top ring . the beam path to the llt is enclosed in a 1-inch aluminum tube for safety and dust protection . the decision to expand the beam was taken to reduce its intensity on the path to the llt ( less coating damage and more tolerant to dust on the optics ) and to have extra flexibility for adjusting the diameter . the laser box has a fast safety shutter . during alignment , a mirror can be inserted in the light path to substitute the uv beam with a green laser pointer . tilts of the lgs - m4 mirror are controlled remotely for centering the beam on the llt . four uv photo - diodes around the llt primary mirror and the 5th diode on - axis serve for the beam diagnostic . the tilts of the lgs - m4 are regulated to center the beam by balancing the four photo - diode signals approximately , while the laser beam is propagated in the dome during daytime . signals of these photo - diodes serve as a useful diagnostic of normal lgs operation . the size of the beam on the llt mirror is adjusted by the magnification and/or focus of the beam expander in the laser box . it is measured by the signal ratio of the peripheral and central photo - diodes and/or by the shape of the uv beam at the llt exit , visualized by luminescence of a standard white paper and photographed ( this beam is not very intense , hence safe ) . the beam on the llt mirror has fwhm diameter of 125 mm . the design of the llt ( figure [ fig : llt ] ) is constrained by the small space available behind the soar m2 . its 25-cm aperture was chosen to minimize the lgs spot size , balancing between the diffraction and the atmospheric blur . for sodium lasers , the llts have typical diameters of @xmath80.5 m , but for the uv the optimum llt diameter is reduced in proportion to the wavelength . in hindsight , the sam llt could be even smaller because in reality the atmospheric blur of the lgs dominates over diffraction . the llt optics consists of the spherical secondary mirror m2 and the fast elliptical primary m1 ( a dall - kirkham telescope ) . spherical m2 is easier in fabrication and alignment , but such optics has only a small fov , limited by the off - axis coma aberration . although the lgs beam is on - axis , the fov of @xmath25 is needed for co - alignment between llt and soar , compensating the flexure of both telescopes . this requirement is fulfilled by mounting the m1 on three flexure rods that point to its focus . lateral motions of m1 are thus converted into its rotations around the focal point . as m1 is actuated to steer the llt beam on the sky , it stays perpendicular to the beam , thus avoiding the off - axis coma . the flexures also serve as a kinematic m1 mount . the llt m1 was fabricated by s. potanin using a 5-cm thick astro - sitall ( zerodur analogue ) blank , light - weighted by removing most of the material from the back side and leaving the rib structure . three invar inserts are glued into the circular pucks machined in the blank to connect mirror with the flexures . although the thermal expansion coefficients of zerodur and invar are both about @xmath26 , the 0.1-mm layer of the epoxy glue around the inserts produced thermal deformation of m1 . the problem was fixed by replacing the solid inserts with flexible equivalents that do not stress m1 . the m2 mirror is mounted above m1 on a column , attached to it by a light - weight truss and flexures . this allows its axial motion for focusing . the distance between m1 and m2 is fixed by an invar rod , actuated for focusing by a motor near m1 . in this way , the llt focus is not very sensitive to temperature changes , despite its aluminum structure . lateral position of m1 is defined by two motors that act on its flexures laterally , in a kinematic arrangement . the llt internal structure is very stiff and light - weight . it is connected to soar via three mounting points . the box - shaped llt enclosure is attached to the same points independently of the main opto - mechanical assembly . the environmental shutter of the enclosure is a thin steel sheet that slides to the side when opened . the 8-mm parallel beam arriving at the llt after reflection from the lgs - m4 is deflected towards lgs - m2 by the tertiary mirror lgs - m3 , mounted on the piezo - electric tip - tilt platform ( s-330 from physik instrumente ) for fast beam steering . the platform is driven by the average spot centroids in the sam wfs and keeps the spots centered in the wfs in closed loop . remember that the soar m3 is actuated in tt to stabilize the science image ; this also displaces the lgs , so the llt tt servo has to compensate for this disturbance and for the intrinsic beam wander of the llt itself , to keep the spots centered . the range of the llt fast tilt servo is @xmath2114 on one axis and @xmath2110 on the other axis , reduced by oblique reflection from the lgs - m3 . the lgs can be steered over a larger range by the m1 lateral actuators , while the piezo platform compensates only small and fast residual tilts . the lgs - m3 with a dielectric uv coating is almost transparent to the visible light . below it , a small `` telescope '' is located to capture the collimated light of a bright star , for co - alignment between llt and soar and for testing the llt image quality . it uses the gc650 ccd camera , as in the wfs acquisition channel . this device can also back - project a parallel beam from a red laser diode through the llt . we mount a 20-cm flat mirror in front of the llt to reflect the emerging red beam back , and test the llt optics in auto - collimation . this functionality proved critical for _ in situ _ testing and alignment of the llt . on 2015 september 16 , the strong earthquake caused an acceleration up to @xmath85 g at the soar m2 . the llt was slightly misaligned , but could be re - aligned quickly by tilting the spherical m2 , using the auto - collimation mirror and the red diode beam . the size of the lgs spots in the wfs often exceeded 2 , well above the expected seeing blur . partly it was caused by the heat dissipated by the llt electronics located just beneath it . the electronics was modified and is now powered only when the llt motors are actually used . as a result , the lgs spots became smaller . however , they are still enlarged sporadically by internal seeing in the soar dome . the @xmath7 wave plate located above the lgs - m3 transforms elliptical polarization of the laser beam into circular polarization propagated in the sky . there were several incidents when insects , attracted by the uv light , got inside the llt and were burned , damaging either the @xmath7 plate or the window of the laser box . these elements were easily replaced , without need for realignment . to prevent further damage , we covered the llt aperture with a metallic mesh ( 3-mm pitch and 0.1-mm thickness ) . it causes only a minor light loss of @xmath812% through blocking and diffraction . the sam real - time computer ( rtc ) is a standard pc with an intel@xmath27 core duo processor e7500 running at 2.93ghz . it works under centos release 5.8 operational system ( linux kernel version 2.6.25 # 5 smp ) with a real - time patch rtai 3.8 . the rtc is located in the soar computer room . it contains the acquisition board of the sdsu ccd controller receiving the wfs signal and the waveform generator board for the m3 control . other interface modules of the rtc are housed in the pxi chassis in the sam electronic rack . the fiber connection between rtc and pxi effectively brings the computer bus towards the instrument . the pxi chassis must be powered when the rtc is booted . the pxi contains two 32-channel digital - to - analog converters that generate the dm drive signals and the timer / counter boards for acquisition of counts from the tt apd detectors . another timing board generates the range - gate pulses of programmable delay and duration . the _ real - time core _ ( rtcore ) is at the heart of the ao control . each control loop cycle is triggered by an interrupt from the acquisition board when the fresh ccd frame arrives . the spot centroids are computed using a standard center - of - gravity algorithm . several alternative centroid algorithms such as correlation or weighted centroid @xcite are available and have been tested , but we found that with the 2@xmath42 binning and the small wfs aperture the standard centroid gives the smallest noise . the wave - front slopes are the differences between the current centroids and the reference spot positions . the 144 slopes from 72 sub - apertures are multiplied by the 60@xmath4144 reconstructor matrix , and the resulting 60 dm drive voltages are filtered temporarily by the loop controller ( integrator with a gain of 0.25 ) . the reconstructor matrix typically contains 38 to 40 system modes ( weak modes are rejected in the reconstructor calculation ) . to prevent growth of unseen modes , a leak of 0.01 is introduced . a somewhat faster second - order temporal controller ( so - called smith predictor ) is implemented , but it gives no sizable improvement over the simple integrator . when using the lgs , the standard ao control scheme is modified . slopes averaged over all subapertures are subtracted from the high - order control signals . those average slopes instead are driving the tip - tilt platform in the llt to keep the spots centered in the wfs . the two control voltages for the llt are generated together with the 60 dm voltages . occasionally , the lgs spots become blurred or contain too few photons because of clouds . some spot centroids then remain indeterminate . those `` missing slopes '' do not participate in the calculation of the average tilt and are replaced by zeros in the high - order loop control . the ao loop is robust against subapertures with missing information . the rtcore communicates with the non - real - time applications written mostly in labview . they include the graphic user interface ( gui ) and various service modules . a bank of shared memory containing slopes and dm voltages is used for the analysis of the ao loop performance ( e.g. temporal power spectra of zernike modes ) , estimation of atmospheric parameters and residual aberrations . some of those diagnostic data are saved in the log file and in the image headers . samples of ao data ( slopes , dm voltages , tt signals , and wfs frames ) can be saved by the sam operator for further off - line analysis and archiving . apart from the real - time software , the sam instrument is controlled by four other software modules running in the instrument computer ( ic ) , also operated under linux . the ic is housed in a cabinet on the soar mount platform . all instrument control software is written in labview ; each module has its own gui . the software is accessed by vnc connections either from the soar control room or remotely . two modules ( one for aom and one for lgs ) take care of the motion control and other hardware . the motor controllers , switches , and telemetry boards of the aom are all connected to the serial rs-485 line acting as a bus and accessed by a communication board in the ic . the laser subsystems are controlled in a similar way , with an addition of the direct serial communication with the laser . each controlled motion ( e.g. the wfs focusing stage ) has to be initialized ( search for the zero - point mark ) after powering the instrument . then the controllers `` know '' the current mechanism positions and can move them using incremental encoders . to speed up the start - up process and to avoid unnecessary motions , all current mechanism positions are saved by the software and re - loaded in the controllers at the start , thus by - passing the initialization . this is particularly important for the llt mechanisms that are powered only when actually used and hence must `` wake up '' quickly . the exception is made for the guiders : their six motors are initialized after each power - up . this is done to avoid the risk of collision between the gps ( their trajectories can overlap ) . collision is avoided by the motion control software , while an additional hardware anti - collision sensor is implemented as an extra precaution . the instrument control software ( icsoft ) of sam is a high - level labview application used to operate the whole instrument . it communicates with the lower - level modules ( motion control and rtsoft ) and the telescope through sockets . in normal use , sam is controlled only through the icsoft gui and there is no need to access other modules . icsoft has a scripting capability to code sequences of actions , thus simplifying sam operation . for example , search for guide stars by the soar m3 modulation with increasing amplitude , their detection and re - positioning of gp when the star is found is done by a script activated by a single button . there are scripts for preparing sam for observations and shutting it down at the end of the night . a software tool for visualization of the wfs acquisition images and a star catalog software for selecting guide stars are run together with the icsoft during sam operation . sam operation is subject to restrictions on laser propagation imposed by the us space command and managed by the laser clearing house ( lch ) . lists of target coordinates are sent to the lch several days prior to the observations , and the authorized laser propagation windows for each target are received on the day of operation , or a day before . these files are loaded in the software module that automatically shutters the laser when the telescope coordinates differ from the target coordinates by more than 120 or when the propagation is not authorized . this software receives information from the telescope control system . both the internal laser shutter and the fast safety shutter in the laser box are closed . typical lch interrupts last a few tens of seconds , while their frequency varies from a few per hour to one or none per night , depending on the target location . blanket laser closures occasionally announced by the lch are introduced manually in the software and then enacted automatically . the sam laser does not present any hazards to the airplanes . its uv radiation is not visible and does not penetrate the eye , being absorbed by the cornea . as a result , the safe illumination level is much higher compared to the visible - light lasers . an airplane going through the sam beam receives only one laser pulse , as the next pulse falls at a different non - overlapping location ; the resulting illumination is 100 times less than the safe threshold for eye and skin established by the us safety standard ansi z136.6 - 200 . the narrow beam of the sam laser and even the expanded 8-mm beam are hazardous to people . work with those beams involves standard laser safety precautions and can be done only be qualified personnel . in normal operation these beams are encapsulated inside boxes and beam conduits . the beam emerging from the llt is sufficiently diluted to be safe ; moreover , it is not accessible from anywhere inside the dome . consequently , the sam laser in normal operation does not require any safety measures . nevertheless , a luminous sign in the telescope control room warns when the laser diodes are energized . a safety circuit and several emergency stop buttons are implemented to shut down the laser in abnormal situations . the basic requirements follow from the intended use of sam : ( i ) provide a non - negligible improvement of the image resolution , with a fwhm of 2/3 compared to the uncorrected seeing , ( ii ) fwhm variation of @xmath28% over the field , and ( iii ) sky coverage of no less than 90% . the remaining requirements relate to the instrument use ( e.g. one - person operation ) , operating conditions such as temperature range , etc . a large number of technical requirements were formulated to guide the instrument design . the quality of ao compensation is normally characterized by the strehl ratio , which translates to the residual wavefront error . in glao the goal is to reduce the psf size , so the strehl ratio is not an adequate metric . the expected performance of sam was first evaluated for an ideal instrument , considering the finite compensation order , the geometry of the lgs , and the turbulence profiles measured at cerro pachn @xcite . the psf is computed from the structure function of residual wavefront errors ( rsf ) which , for sam , grows almost linearly with the distance . the corresponding psf has a sharper peak and stronger `` wings '' compared to the seeing profile ; it can be well approximated by a moffat function with @xmath29 ( * ? ? ? * see equation 1 below ) . in the real instrument , the errors of the spot centroids , tt residuals , and servo lag degrade the performance somewhat . the combined errors of the sam ao were required not to degrade the ideal resolution by more than 10% , meaning that their relative contribution to the rsf at the characteristic baseline of 0.5 m should not exceed 10% . in this way , the error budget and requirements to various ao subsystems were developed @xcite . unlike the standard strehl - based approach to the ao design , it constrains only the difference between the real and `` perfect '' sam and does not depend on the fixed instrument parameteres and turbulence profile . when sam was first installed at soar in 2009 , we found that the axis of the nasmyth rotator did not point to the center of the pupil . as a result , rotation displaced the pupil image on the lla . this has been adjusted by tuning the angle of the soar m3 , so that rotation - induced pupil motion became less than 0.2 of the sub - aperture size . the diameter of the pupil image on the lla was measured by recording sky through the wfs , measuring flux in each sub - aperture , and fitting a model with three parameters : pupil diameter and its lateral shifts . the optimum pupil radius is 5.1 sub - apertures , the actually measured one is 4.8 sub - apertures . this causes a loss of flux in the peripheral sub - apertures and a mismatch between the interaction matrix measured with tursim and the actual lgs . with a typical range gate of 0.15 km ( maximum spot elongation of 125 ) and the loop frequency of 478hz the return lgs flux varies between 200 and 800 electrons per sub - aperture per loop cycle . the flux depends on the spot size ( large spots are truncated by the wfs aperture ) , atmospheric conditions ( more photons are scattered by aerosols or less photons due to additional absorption by thin clouds ) and the rotator angle ( the polarization adjustment in the wfs is not perfect ) . light losses in the up - link and down - link lgs path lead to the estimated ( and measured ) total efficiency of 5% . this number , combined with the lidar equation , implies the return flux of 600 photons per sub - aperture per loop cycle , as actually observed under favorable circumstances . the fwhm size of the lgs spots is measured by stacking many wfs frames and fitting gaussians to several spots around the pupil center ( the elongation due to the finite range gate is thus small ) . it ranges from 13 to 23 , depending mostly on the local seeing ( figure [ fig : spots ] ) . the rms centroid noise is normally between 0.15 and 0.25 pixels . figure [ fig : loop ] gives an example of the ao loop performance on a night with mediocre atmospheric conditions . the tt loop uses signals of one or two gps to control the angles of soar m3 . the temporal controller accounts for the frequency response of m3 , which eventually restricts the tt bandwidth to @xmath810hz . as the tt probes are quad - cells , the loop gain depends on the seeing . with a very good seeing and a gain too large , the 15-hz oscillation appears , so the nominal tt gain is set to avoid such situations . the optical axis of soar vibrates with a frequency of 50hz and a typical rms amplitude between 20 and 50 mas . the cause of this vibration is not yet established , but it is clearly seen in the spectra of the tt signal . this vibration is too fast to be compensated by the tt loop . fortunately , its contribution to the fwhm resolution is only marginal under normal use of sam and long exposures . the 50-hz vibration is a very serious factor in speckle interferometry , forcing to use exposures shorter than 10ms and affecting the limiting magnitude . the gain over natural seeing brought by sam is quantified by several metrics : by the fwhm resolution and its uniformity over the field , by the psf profile and its variation , and by the increased energy concentration . the normalized rotationally - symmetric psf @xmath30 is frequently approximated by the moffat function @xmath31^{-\beta } \label{eq : moffat}\ ] ] with two parameters @xmath32 and @xmath33 ; the fwhm is @xmath34 . the case @xmath35 corresponds to the lorentz profile , whereas the seeing profile under kolmogorov turbulence is well modeled by @xmath36 . the sam psf under good compensation corresponds to @xmath37 , in agreement with its predicted shape ( see [ sec : req ] above ) . compared to the seeing - limited psf , it has a sharper core and stronger wings . the gain of sam depends on the metrics used and can not be expressed by a single number . here we quantify it by the fwhm resolution . if sam improves it by a factor of 2 , the gain in the central psf intensity is about 2.8 times , while the half - energy diameter is reduced only by @xmath81.5 times compared to the seeing . the fwhm resolution at long wavelengths ( e.g. in the @xmath1 band ) closely follows the free - atmosphere ( fa ) seeing measured by the mass instrument ( figure [ fig : fa ] ) . according to @xcite , the median fa seeing at cerro pachn is 040 . at shorter wavelengths , the gain in resolution over seeing is less because the compensation order of sam is not high enough . still , the resolution gain in the @xmath38 and @xmath16 bands remains substantial . however , when the total seeing is dominated by the high - altitude turbulence , sam brings no improvement because it does not sense this high turbulence . turbulence sensing by sam decreases with distance progressively and turbulence at , say , 0.5 km is compensated to some extent , but the compensation is always partial because of the finite ao order , servo lag , and noise . so , the resulting sam performance depends on a number of factors . however , the simplistic rule of thumb that sam resolution is almost always close to the fa seeing turned out to be remarkably good in practice , as demonstrated by plots similar to figure [ fig : fa ] that appear in @xcite . figure [ fig : diq ] shows cumulative distributions of the fwhm in several filters in eight nights during three sam runs in 2013 , under a variety of turbulence conditions . each point is a median fwhm of all stellar sources in one closed - loop exposure , not corrected to zenith . the histogram of the fa sesing at zenith measured by mass , matched in time to the sam exposures , is plotted for comparison . the fwhm in the @xmath1 or @xmath39 filters ( 214 estimates ) equals the fa seeing in @xmath880% of the exposures , with the same median of 052 . the remaining 20% are affected by very poor total seeing and , for deep exposures , by semi - resolved distant galaxies which bias the median fwhm estimates to larger values . on the other hand , the 53 exposures in h@xmath40 taken under a median fa seeing of 032 have a median fwhm of 035 . the uniformity of correction over the @xmath0 field depends on the atmospheric conditions . both the uncorrected high turbulence and the well - corrected ground layer do not cause psf variation in the field , but the partially corrected turbulence in the `` gray zone '' ( at a few hundred meters ) leads to the non - uniformity , with a better resolution on - axis @xcite . this phenomenon is actually observed , . ] but in general the fwhm is very uniform over the field and rarely degrades by more than a few percent towards the edge ( figure [ fig : m5 ] ) . for the @xmath1-band data presented in figure [ fig : diq ] , we modeled the fwhm dependence on the distance from the ccd center by a linear function and found that the median linear coefficient is zero , within errors . the uniformity becomes even better at shorter wavelengths where the resolution gain becomes smaller . in principle , the resolution should depend on the distance from the tt star(s ) , but this effect has not been studied on the real data . sam has been fully commissioned by the end of 2013 . some scientific results have already been obtained by that time using commissioning data @xcite . since the 2013b semester , sam was offered in a shared - risk mode , and its science verification ( sv ) program was executed in 2014 . regular observing proposals for sam are accepted since the 2014b semester . like other soar instruments , sam can be operated remotely from la serena or from other location with a fast enough network connection . sam is scheduled classically for specific nights , preferably grouped in runs ( the instrument is switched off between the runs but stays on the telescope ) . before each run , the sam scientist starts the software and makes daytime tests following the standard checklist . this procedure has been successful in detecting hardware or software problems , leaving time for their fixes . for example , failures of the line transformer in the high - voltage dm driver , failed motor - control boards , or a stuck motor shaft . when possible , sam is tested during engineering nights a few days before the science runs . fortunately , we have not yet lost any significant night time to sam failures . the filters in sami are installed by the telescope operators during daytime , as in other soar instruments . sami can use either 4@xmath44 inch square filters in a wheel with 5 positions or 3@xmath43 inch filters in another 7-slot wheel . the limited number of filters restricts sometimes the number of science programs that can be executed with sami on any given night . the overhead time , from the start of the telescope slew on a new target to closing all four loops , can be as short as 5 minutes , with 7 minutes being typical . most of this time is spent for acquisition of guide stars . owing to the small 3 field of the gps , the exact telescope pointing is determined by pre - imaging with sami ; a star with known coordinates must be identified in this image to compute the pointing offset . _ all _ science targets observed so far had guide stars down to the @xmath13 limit , i.e. the sam has a complete sky coverage . such a star gives @xmath860 counts per 10-ms cycle of the tt loop . the sam _ observing tool _ ( ot ) is a custom idl software that helps formatting the target lists for the lch , but also allows pre - selection of guide stars and loading dss2 images in the 5@xmath45 arcmin field to assist in field identification . guide stars pre - selected in the ot can be fed to the sam software and used alongside with the catalogs . when the tt loop is closed , the sam operator opens the laser shutter and acquires the lgs . for lgs centering , the un - gated lgs image is captured by the wfs acquisition camera . the llt electronics is powered on and the lgs is centered , compensating for differential flexure between the llt and soar optical axes ( typically less than 10 ) . then the lgs tilt loop is closed , and the high - order loop is closed as well ( the llt electronics is powered off automatically at this time by the script ) . the lgs acquisition is straightforward and takes less than a minute . if the telescope slew was small , the laser photons reach the wfs without re - centering the llt , in which case the two lgs loops are closed within a few seconds . experience shows that sam often remains operational with a light cirrus clouds because the lgs is located below the clouds . time remaining until the next lch interrupt is displayed in the icsoft . a few seconds before the interrupt , the lgs loops are opened automatically , while the science exposure is normally paused by the observer . the sam operator closes the laser loops manually after the interrupt . the first target of the night is usually used to check the relative focus between sam wfs and sami . the focus might change because of temperature or because of sami filters of non - standard thickness , although it was found to be quite stable . to test the focus , two short exposures are taken with closed ao loop and wfs defocused by @xmath214 mm from the nominal . the sam psf should be enlarged by the same amount if the nominal focus is correct . otherwise , the correct focus setting of the wfs stage is found by linear interpolation between bracketing exposures . sam can operate in open loop , without laser . in such case , the dm is flattened passively . the lgs , if available , still helps to focus the telescope in open loop by measuring the focus term and nulling it by adjusting the telescope focus . open - loop operation happens when the image quality is not critical or for targets that are not in the lch list ( e.g. photometric standards ) . during 2014 , sam received only a small number of proposals , but its use has steadily increased in 2015 , making it the second most popular instrument at soar . by concentrating the light of stars in a smaller number of pixels , sam reduces photometric errors in dense stellar fields dominated by confusion @xcite . the gain in sensitivity in such case is proportional to the central intensity of the psf . the first science paper produced by sam presented the color - magnitude diagram of the globular cluster ngc 6496 @xcite . the fwhm resolution in these early commissioning data of 2012 may was excellent , reaching 025 in the @xmath1 band . however , it was not uniform over the field , showing an asymmetry likely related to the wind direction ( the sam loop worked then with a low frequency of 233hz ) . despite this , accurate photometry was possible by fitting field - dependent psf with daophot . presently , j. santos ( brazil ) is conducting a large survey of globular clusters in the magellanic clouds , exploiting the potential of sam in crowded fields ; eight nights in the 2015b semester were used successfully . @xcite discovered new variable stars in globular clusters with sam . sam can help in resolving close binary stars , especially the faint ones beyond the reach of speckle interferometry . the survey of faint low - mass companions to nearby stars has revealed one new 02 binary @xcite . however , for binary stars the `` detection power '' of sam in terms of resolution and dynamic range is inferior to the classical ao in the infrared or to speckle interferometry . combination of speckle interferometry with laser - enhanced seeing is a promising technique . it was used in 2015 may for the survey of _ field @xcite . sam worked with the laser , but without guiding . series of short - exposure images were re - centered and combined offline ; the median fwhm resolution of the re - centered images was 043 . binarity of targets as faint as @xmath41 mag could be probed down to 015 separation by the speckle image processing . this technique was used again in 2016 january to discover young low - mass binaries in orion , in a project led by c. briceo . the fa seeing was better than in may and the fwhm resolution of re - centered images reached 025 ( figure [ fig : fa ] ) . several programs on emission - line objects are planned using sami with narrow - band filters . the herbig - haro jet hh 46/47 in the s[ii ] narrow - band filter was observed with sam on 2015 february 14 , yielding a fwhm resolution of 045 . continued observations with sam use a combination of custom narrow - band filters to probe physical conditions in the jets . the team of a. ardila observed with sam nearby galaxies ngc 1232 and ngc 1482 in the h@xmath40 band to study star - formation regions with high spatial resolution . the data are still being analyzed , while nice images resulting from this effort were generated with a resolution twice better than in the image of ngc 1232 from the vlt press release . during three nights in 2015 sam worked with a fabry - perot ( f - p ) etalon inserted in its collimated beam after the adc , on a translation stage @xcite . the existing f - p etalon with a spectral resolution of about 11000 just fits in the available space , with 1-mm gaps on both sides . selection of one f - p order is done by the interference filters in the sami wheel . this project is led by c. mendez de oliveira and p. amram . to overcome the sami readout noise , the binning of 4@xmath44 ( pixels of 018 ) was used . some science verification programs intended to use sam for detecting faint nebulae around stars . in this area , sam has no advantage over other ground - based imagers ; it concentrates the light in the core of the psf but does not reduce its wings . sam is well suited for high - resolution imaging of faint deep - space objects . here its complete sky coverage and a moderately wide 3 field play an essential role . d. murphy used sam in 2013 for two nights to get a very deep high - resolution image of a peculiar galaxy in a cluster , apparently affected by recent interaction @xcite . spectacular images of thin gravitational arcs produced by lensing distant galaxies were obtained with sam during its commissioning ( figure [ fig : arcs ] ) . no proposals for continuing this effort have been accepted so far . however , the program by v. motta was started in 2015 december and continues in 2016 . her team took high - resolution images of lensed quasar candidates revealed in a wide - field survey . this is an excellent example of the future use of sam , where it will complement wide - angle surveys by studying interesting objects with higher spatial and/or spectral resolution . yet another promising application of sam will be to study globular clusters ( gcs ) in other galaxies . even a small gain in resolution over seeing might be critical here to distinguish gcs from stars . results in this area are reported by @xcite . the crucial role of the hubble space telescope ( hst ) in astronomy , despite its moderate 2.4-m size , demonstrates the need of high angular resolution at optical wavelengths , so far not provided by ground - based ao instruments . the resolution delivered by sam is still much inferior to the resolution of the hst , which continues to be the facility of choice . the advantages of sam compared to the hst are its accessibility and the larger telescope aperture , hence the larger number of collected photons . when very faint objects are observed in narrow spectral bands , the minimum size of the resolution element may be set by the need to collect enough photons . in this situation , sam is preferable to the hst . its recent use with narrow - band filters and fabry - perot nicely illustrates this aspect . under good seeing , current ao instruments designed to work in the ir could provide substantial resolution gain in the red part of the visible spectrum , e.g. at h@xmath40 . however , their use at optical wavelengths is not envisioned and sometimes is impossible because of dichroics in the science path or the lack of optical instruments . this situation , presently to the advantage of sam , is slowly evolving . the muse instrument at the vlt @xcite will operate with partial seeing correction at visible wavelengths over a 1 field using four sodium lasers . the argos ao system at the lbt @xcite will implement glao with green rayleigh lasers and will work in the ir . the potential of sam will be fully exploited when its high spatial resolution is combined with spectroscopy . this is already happening when sam is used with f - p . a multislit spectrometer for sam is being developed by @xcite . this instrument , samos , will use a commercial digital micromirror device ( dmd ) as a software - configurable multislit mask . the spectral resolution between 2000 and 8000 is envisioned . the samos science targets are faint extragalactic and stellar objects where the increased spatial resolution brings a gain in sensitivity against the sky background , in addition to probing small spatial details and reducing source confusion . the sam instrument designed more than 10 years ago is technologically obsolete . using modern electron multiplication ccd detectors with a fast gating , a higher - order wavefront sensing is feasible with the same laser power . if the dm is replaced by a new device with a larger actuator count , the increased compensation order will improve image correction at shorter wavelengths , in the green and blue spectral regions , bringing it closer to the sam performance in the @xmath1 band . future upgrade of sam along these lines will be decided based on its popularity and availability of funds . many people working at ctio and soar have contributed to the creation of sam . apart from the authors , the non - exhaustive list of major participants in this project includes b. gregory , s. heathcote , e. mondaca , a. montan , w. naudy corts , f. delgado , o. estay . f. collao , r. rivera , d. sprayberry . the initial optical design of sam by v. terebizh and sharing of the ao code by ch . keller helped us to start the sam project in 2002 . consultations with m. hart and t. stalcup are gratefully acknowledged . the efforts of other experts in astronomical ao who reviewed the sam project must not be forgotten . s. potanin helped at a critical moment by fabricating the llt mirror while major optical vendors refused to bid . the soar crew ( e. serrano , g. gomez , g. dubo and others ) supported sam installation , commissioning , and operation . referee s comments helped to improve this article . rigaut , f. , salmon , d. , arsenault , r. et al . 1998 , pasp , 110 , 152 rigaut , f. , 2002 , in : beyond conventional adaptive optics , eds . e. vernet , r. ragazzoni , s. esposito , n. hubin , eso conf . workshop proc . 58 , eso , garching bei mnchen , p. 11 thomas , s. 2004 , proc . spie , 5490 , 766 thomas , s. , fusco , t. , tokovinin , a. et al . 2006 , mnras , 371 , 323 thompson , l. a. & teare , s. w. 2002 , pasp , 114 , 1029 tighe , r. , tokovinin , a. , schurter , p. et al . 2016 , proc . spie , 9908 , 99803b ; doi:10.1117/12.2233681 tokovinin , a. , tighe , r. , schurter , p. et al . 2008 , proc . spie 7015 , 157 tokovinin , a. , tighe , r. , schurter , p. et al . 2010 , proc . spie , 7736 , 132 tokovinin , a. , tighe , r. , schurter , p. et al . 2012 , proc . spie , 8477 , 166 tokovinin , a. , 2011 , in : second international conference on adaptive optics for extremely large telescopes , http://ao4elt2.lesia.obspm.fr

we examine the correlation between globular cluster ( gc ) color and magnitude using hst / acs imaging for a sample of 79 early - type galaxies ( mag ) with accurate surface - brightness fluctuation distances from the acs virgo cluster survey . using the mixture modeling algorithm , we find a highly significant correlation , , between color and magnitude for the subpopulation of blue gcs in the co - added gc color - magnitude diagram of the three brightest virgo cluster galaxies ( m49 , m87 and m60 ) . the sense of the correlation is such that brighter gcs are redder than their fainter counterparts . for the single gc systems of m87 and m60 , we find similar correlations ; m49 does not appear to show a significant trend . there is no correlation between and for gcs belonging to the red subpopulation . the correlation for the blue subpopulation is much weaker than . using monte carlo simulations , we attribute this finding to the fact that the blue subpopulation in extends to higher luminosities than does the red subpopulation , which biases the fit results . the highly significant correlation between color and , however , is a real effect : this conclusion is supported by biweight fits to the same color distributions . we identify two environmental dependencies which influence the derived color - magnitude relation : ( 1 ) the slope of the color - magnitude relation decreases in significance with decreasing galaxy luminosity , although it remains detectable over the full luminosity range of our sample ; and ( 2 ) the slope is stronger for gc populations located at smaller galactocentric distances . these characteristics suggest that the observed trend is , at least partially , shaped by external agents . we examine several physical mechanisms that might give rise to the observed color - magnitude relation including : ( 1 ) presence of contaminants like super - clusters , stripped galactic nuclei , or ultra - compact dwarfs ; ( 2 ) accretion of gcs from low - mass galaxies ; ( 3 ) stochastic effects ; ( 4 ) the capture of field stars by individual gcs ; and ( 5 ) gc self - enrichment . although none of these scenarios offers a fully satisfactory explanation of the observations , we conclude that self - enrichment and field - star capture , or a combination of these processes , offer the most promising means of explaining our observations .

astro-ph0609079

it is by now an accepted paradigm that the formation of globular clusters ( gcs ) is closely linked to the formation of their host galaxies ( e.g. searle & zinn 1978 , ashman & zepf 1992 , kissler - patig 1997 , hilker _ et al . _ 1999b , forbes 1997 , ct _ et al _ 1998 , beasley 2002 ) . in recent years , observational data for extragalactic gc systems have made a huge leap forward in quality , thanks mainly to high - resolution imaging from the hubble space telescope , and to wide field imaging and multi - object spectroscopy from 8m - class ground based telescopes ( see west 2004 and references therein ) . as a rule , studies of extragalactic gcs have tended to focus on correlations between the global photometric properties of the gc systems ( gcss ) and those of their host galaxy . for instance , peng @xcite , examined the color distributions for gcs belonging to 100 galaxies observed in the acs virgo cluster survey ( acsvcs ; ct 2004 ; see also below ) , finding nearly all of these galaxies ( @xmath7 mag ) to possess bimodal , or at least asymmetric , gc color distributions . consistent with previous findings ( e.g. , gebhardt & kissler - patig 1999 ; kundu 2001 ; larsen 2001 ) , the mean gc color was found to correlate with galaxy luminosity in the same sense as the well - known color - magnitude relation for early type galaxies ( e.g. bower , lucey & ellis 1992 ; karick 2003 ; hilker 2003 ; ferrarese 2006 ) . a corresponding gc color - galaxy color relation was also found by peng @xcite , confirming previous findings ( e.g. , burgarella 2001 ; larsen 2001 ; lotz 2004 ) . at the same time , analyses of the color magnitude diagrams for the gcss of early - type galaxies revealed some tentative evidence for a dependence of _ individual _ gc color on luminosity . ostrov @xcite found from washington photometry of ngc 1399 that the color peaks of the gc color distribution seem to merge at the highest luminosities . dirsch et al . @xcite confirmed this finding for the same galaxy based on an expanded sample of gcs . there have also been hints of such a behavior in other galaxies such as ngc 5128 ( harris 2004 ) and m87 ( whitmore 1995 ; larsen 2001 ) . with the improved imaging capabilities of hst provided by the advanced camera for surveys ( acs ; ford 1998 ) , it has become possible to investigate this issue in more detail . harris @xcite used acs imaging to investigate the distribution of gcs in the @xmath8 color magnitude plane for eight brightest cluster galaxies in the range @xmath9 km s@xmath10 . for the four galaxies with the largest number of gcs and also the joint sample of eight gc systems they found a significant trend towards redder colors for increasing luminosity for the subpopulation of blue gcs . this finding led them to suggest a mass - metallicity relation of the form @xmath11 for this subpopulation . curiously , no color - magnitude relation was found for the red gcs . strader @xcite used the publicly available data from the acsvcs to look for color - magnitude trends in the three brightest virgo cluster galaxies m49 , m87 and m60 . for the blue gc populations of m87 and m60 , they found a significant trend of redder color with brighter gc @xmath12-band luminosity similar to that found by harris @xcite . they too interpreted this trend as evidence for a mass - metallicity relation among the blue ( metal - poor ) gcs and suggested self - enrichment as a possible explanation . no significant color - magnitude trend was found for m49 , the brightest member of the virgo cluster . both the harris @xcite and strader @xcite studies used heteroscedastic mixture models to fit the colors of an assumed double - gaussian color distribution to the observed gc color - magnitude diagrams ( cmds ) as a function of magnitude . et al . _ use ( e.g. ashman , bird & zepf 1994 ) , while strader _ et al . _ use nmix ( richardson & green 1997 ) . two important issues that were not addressed by these studies are the subject of this paper : first , we will investigate the color - magnitude correlations not only in the red ( i.e. , @xmath12 ) , but also in the blue band ( i.e. , @xmath13 ) ; second , we will investigate how the blue peak slope varies as a function of host galaxy luminosity and galactocentric distance . note that the harris @xcite study targeted only brightest cluster galaxies , so it spanned a relatively limited range in galaxy luminosity . we finally point out that there is recent evidence for the existence of a gc color - magnitude relation also in spiral galaxies . based on acs imaging , spitler _ et al . _ @xcite found a significant color - magnitude trend for the sombrero galaxy ( ngc 4594 ) implying @xmath14 , which is a somewhat weaker trend than suggested by harris @xcite and strader @xcite for giant ellipticals . the aim of this paper is to investigate the color - magnitude trend of gcs over the entire host galaxy luminosity range covered by the acsvcs . special emphasis will be put on environmental dependencies of this trend and checks of the fitting technique . the paper is structured as follows : in [ gcsel ] we outline how gc candidates are selected in the acsvcs . in [ realvssim ] we compare fits to simulated cmds with fits to observed cmds . in [ environment ] we investigate the environmental dependence of the color - magnitude trend . in [ discussion ] we examine a variety of physical effects that may create these trends . we finish this paper with the summary and conclusions in [ conclusions ] . the acsvcs represents the most complete and homogeneous study of extragalactic gcs performed to date . the sample consists of 100 early - type ( e , s0 , de , de , n , ds0 ) members of the virgo cluster . each galaxy was imaged in the f475w and f850lp filters ( @xmath15 and @xmath16 , respectively ) for a total of 750 seconds and 1210 seconds , respectively . this filter combination gives roughly a factor - of - two improvement in wavelength baseline and metallicity sensitivity compared to the `` canonical '' @xmath17 color index . the identification of bona - fide gcs from these images is performed in the size - magnitude plane as described in peng @xcite . this selection procedure which is possible because the half - light radii of gcs are marginally resolved ( jordn @xcite ) at the distance of virgo ( @xmath18 mpc ; tonry @xcite ; mei @xcite ) greatly reduces contamination from foreground stars and background galaxies . the acs images had been reduced using a dedicated pipeline described in detail by jordn ( 2004ab ; papers ii and iii ) . in brief , the reduction steps are : image combination , galaxy modeling and model subtraction , rejection of obvious background galaxies , and the measurement of magnitudes and sizes for candidate gcs using the program kingphot ( jordn 2005 ; paper x ) . the result is a catalog of integrated @xmath13 and @xmath12 magnitudes , @xmath2 colors and half - light radii , @xmath19 , for each candidate gc . magnitudes and colors are corrected for foreground extinction using the reddening maps of schlegel @xcite . to estimate the contamination by background galaxies , an identical reduction procedure was applied to 17 blank high - latitude fields observed by acs in the @xmath13 and @xmath12 filters . because background galaxies are typically larger at comparable luminosities ( or , conversely , fainter at comparable sizes ) , a statistical decontamination is performed in size - magnitude space ( see peng 2006 and jordn 2006 , in prep . for details ) : to this end , first a non - parametric density model is fitted to the size - magnitude distribution of sources detected in background fields . for the gc candidate magnitude distribution , a gaussian luminosity function is assumed . the gc candidate size distribution is fitted by a non - parametric kernel estimate plus a power law tail . the resulting surface densities in the size - magnitude plane for background sources and gcs then make it possible to assign a gc probability , @xmath20 , to every source detected in the galaxy image . all candidates with probabilities @xmath21 are considered gcs . the remaining contamination by background galaxies that are assigned @xmath21 is negligible for the sake of this study . we have tested this by statistically cleaning the gc cmds for each galaxy , using the color - magnitude distribution of sources with @xmath22 in the 17 blank background fields . figure [ cmd1316 ] shows @xmath6-@xmath2 and @xmath3-@xmath2 cmds for the gcs of m87 (= ngc4486 = vcc1316 ) , the central galaxy of the virgo cluster and the galaxy with the singlest largest population of gcs . for comparison , figure [ cmdclasses ] shows the co - added cmds for the 79 acsvcs galaxies that have reliable sbf distances ( mei 2005 , 2006 ) and which have a discernible gc population ( see peng 2006 ) . we use the sbf distance moduli from mei ( 2005 , 2006 ) , which have an internal precision of @xmath23 mag , to assign absolute magnitudes to the associated gc candidates . to better visualize trends across host galaxy magnitude , the 79 galaxies are subdivided into four broad luminosity groups such that the number of gcs per group falls in the range @xmath24 2400 to 3300 . luminosity group 1 consists of the three brightest avcvcs galaxies m49 (= ngc4472 = vcc1226 ; @xmath25 mag ) , m87 (= ngc4486 = vcc1316 ; @xmath26 mag ) , and m60 (= ngc4649 = vcc1978 ; @xmath27 mag ) . group 2 consists of six acsvcs galaxies with @xmath28 mag . group 3 consists of 17 acsvcs galaxies with @xmath29 mag . the remaining 53 acsvcs galaxies with @xmath30 mag make up group 4 . the three galaxies in group 1 are all so - called `` core - srsic galaxies '' ( ferrarese 2006 ) , while those in groups 2 and 3 are a mixture of core - srsic and srsic elliptical and lenticular galaxies . the 53 galaxies in group 4 consist mainly , but not exclusively , of early - type dwarfs ( e.g. , de , de , n , ds0 and ds0,n ) . we have applied to the combined cmds for these four groups , as well as to the individual cmds for the three galaxies in group 1 . to this end , we subdivided the cmds into luminosity bins , each bin containing the same number of data points . for the cmds of m49 and m60 , the size of each luminosity bin is 100 , for m87 the size is 150 , and for the combined cmds the bin size is 200 gcs . was then run on the gc color distribution in each luminosity bin . as in the paper by harris _ et al . _ @xcite , we use the heteroscedastic option of kmm : we allow for gaussians of different widths to be fitted to the blue and red gc populations . the fitted mean positions are plotted over the respective cmds in figures [ cmd1316 ] and [ cmdclasses ] . we show the results for two different initial guesses for the mean colors of the blue and red gc populations : one for the pair @xmath31 and 1.2 mag , another one for @xmath32 and 1.4 mag . these different guesses were chosen to cover : ( 1 ) the range of peak positions between cmds of the various luminosity groups ; and ( 2 ) the range caused by the color - magnitude trend itself ( see figure [ cmdclasses ] ) . linear slopes between color and absolute magnitude are obtained from ordinary least square fits to the peak positions for both pairs of initial guesses , fitting color as a function of magnitude . gcs fainter than @xmath33 mag and @xmath34 mag were excluded from the fits , to avoid effects of photometric incompleteness on the measured slopes . the mean of both fits is adopted as the final best - fit slope . the fit errors of the slope are derived from re - sampling the points using as dispersion the observed scatter around the fitted relation . we note that this scatter is for all luminosity groups larger ( by 30% for luminosity group 1 and by 100% for group 4 ) than the average statistical error estimate for the peak position , given by @xmath35 . here , @xmath36 is the width of the gaussian peak and @xmath37 the average number of gcs per magnitude bin associated to the peak . we are therefore confident that our error estimates are not too optimistic . for the few cases where the difference between the results obtained using the two initial guesses was larger than the formal fit errors , we adopted the difference as the error . the resulting slopes of those fits are shown in table [ fitresults ] . we first consider @xmath3 vs. @xmath2 for group 1 . in this case , we detect a very significant slope @xmath38 for the blue peak position . when restricting the fit to @xmath39 mag ( i.e. , rejecting the two brightest magnitude bins in figure [ cmdclasses ] ) , the result is @xmath40 . this is significant at the 4.5@xmath36 level , indicating that the color - magnitude trend is _ not _ restricted to just the brightest gcs . for m87 and m60 alone , a comparable slope is derived , albeit at somewhat lower ( @xmath41 ) significance . the slope for m49 is consistent with zero , in agreement with the findings of strader @xcite . when excluding m49 from group 1 , the slope for the blue population rises slightly to @xmath42 . the non - detection of a color - magnitude trend for m49 is intriguing , since this galaxy has a similar number of gcs as m60 and is similar to m60 and m87 in terms of luminosity , color , surface brightness and gc dynamics ( peng 2006 ; ferrarese 2006 ; mei 2005 ; ct 2001 ; 2003 ) . the fact that m49 is located at the center of its own sub - cluster being offset by @xmath43 from the virgo cluster center may hint at some environmental effect on the strength of the observed color - magnitude trend . in contrast to the strong correlation @xmath5 exhibited by the blue gcs in group 1 , @xmath5 is insignificant for the red gcs in this group . however , the situation changes when using @xmath6 instead of @xmath3 as magnitude in the color - magnitude diagram ( see table [ fitresults ] ) . the slope @xmath44 for the blue population is more than 50% smaller than @xmath5 and , as a result , has a lower level of significance ( 1.7@xmath36 . ) considered individually , @xmath45 of the blue population becomes insignificant for all three galaxies in this group . in turn , the slope @xmath45 of the red population becomes marginally positive i.e. , opposite in sense to that defined by @xmath5 for the blue population . [ cols="<,>,>,>,>",options="header " , ] notes : analogous to table [ fitresults ] . difference is that peak positions are now defined by the biweight of the color distribution blue- and redwards of a luminosity independent limiting color , see fig . [ normal ] . [ slopes ] graphically compares the fits to and biweight peak positions for all four luminosity groups . the difference in slopes between @xmath12 and @xmath13 raises the question as to what the `` true '' slope actually is . to investigate this issue in more detail , we compare in figure [ cmd3brightsim ] the actual measurements for group 1 with two sets of 20 monte - carlo simulations for the m87 cmd . the first set of cmds has an implemented slope @xmath46 for the blue gcs corresponding to @xmath47 is of course not identical to @xmath5 , but this difference is negligible for the sake of this study : @xmath48 and @xmath49 for the red ones ; the other set has @xmath49 implemented for either gc population . these artificial cmds were created by assuming the average peak widths derived by for the blue and red gcs , taking into account also the photometric errors on the colors . the luminosity function is sampled from a kernel estimator of the original cmd of m87 , separately for each color peak . figure [ compsim ] compares the real and simulated m87 cmd with an implemented slope . note an important feature of those cmds : the red peak is populated to brighter absolute luminosities in @xmath3 than in @xmath6 , while the blue peak does not show such a difference . this is because , relative to gcs in the blue population , gcs in the red population are brighter in @xmath3 than in @xmath6 . table [ fitresults ] gives the slopes fitted to the mean colors from for the two simulated sets of m87 cmds , together with the slopes for the real data . this comparison shows that both the cmd of m87 alone , and that for the co - added cmd of group 1 , are inconsistent with a magnitude independent color for the blue gcs . the data are consistent with a simulated slope of 0.040 for the blue gcs and no slope for the red gcs . note the positive slope found for the red gcs in both the real and simulated data in @xmath6 . averaging over the 20 simulated cmds , the mean red slope of @xmath500.03 is different from zero at the 5.7@xmath36 level . that finding is noteworthy since the simulations do _ not _ implement a slope for the red peak . this exercise therefore demonstrates an important caveat of the fitting : at a given magnitude , as one of the two peaks becomes de - populated ( in this case the red peak in @xmath6 ) , its position will artificially be dragged towards the position of the more prominent peak . as a consequence , the mean color of the blue population will be artificially shifted towards bluer colors . this results in a weaker slope for the blue peak in @xmath6 than in @xmath3 . in @xmath3 , both peaks are populated up to comparable absolute luminosities , and therefore the fitted slopes for both red and blue peaks correspond to the simulated ones . that is to say , only the _ weaker _ slope in @xmath6 is caused by a bias . does not artificially introduce the very significant negative slope in the blue @xmath3 peak : this is clearly a real trend that is intrinsic to the data . when interpreting the slope as a mass - metallicity relation for gcs in the blue peak , we derive a scaling @xmath51 by combining the slope @xmath52 - 0.037 found in the previous section with @xmath53}}{d(g - z ) } = 5.14~{\rm dex~mag^{-1}}\ ] ] from peng @xcite and assuming a constant mass - to - light ratio over the luminosity range investigated . this is in good agreement with the scaling relation @xmath54 found by harris @xcite . as an alternative to kmm , we have calculated the biweight location ( beers 1990 ) for the gc subpopulations blue - wards and red - wards of a magnitude independent limiting color . this limiting color is the mean of the position of the blue and red populations , weighted by the average width of either peak . the results for this alternative method are shown in table [ fitresultsnaive ] . as an example , we show in fig . [ normal ] the peak positions for groups 1 and 4 in @xmath3 . generally speaking , the errors for this biweight method decrease significantly compared to the data points , and the results confirm the significant slope found for the blue gc population . they also confirm that no significant correlation is found for m49 . table [ fitresultsnaive ] and figure [ slopes ] show that the difference between fits in @xmath6 and @xmath3 is considerably smaller than in the case of kmm . nevertheless , this alternative method is itself not immune to biases , as is clear when it is applied it to the simulated cmds ( see table [ fitresultsnaive ] and figure [ slopes ] ) . in the case of the blue population exhibiting a negative slope , a fixed limiting magnitude between both populations will systematically decrease the absolute slope of the blue population , and introduce an artificial positive slope for the red . having examined how specific features of the cmd may influence the fitted color - magnitude relations for the blue and red gc populations , we now investigate the extent to which possible environmental effects may influence the observed trends . figure [ cmdclasses ] shows cmds for groups 14 with the peaks over - plotted . derived slopes for the color - magnitude relations in each group are presented in tables [ fitresults ] and [ fitresultsnaive ] . the dependence of slope on host galaxy magnitude is illustrated in figure [ slopes ] . the fits in @xmath3 show that the slope of the blue gcs in groups 2 and 3 are indistinguishable from that for group 1 ; in all three groups , the slope is inconsistent with zero at about the 3@xmath36 level . this indicates that the color - magnitude trend among metal - poor gcs is _ not _ restricted to just the brightest galaxies . only for group 4 does the measured slope of the blue population become consistent with zero . the biweight fits in the lower panels of figure [ slopes ] paint a similar picture : the slope decreases only gradually from group 1 to 4 , and apparently does not drop to zero . given that the biweight fits systematically underestimate the blue slope , this is an indication that the color magnitude trend persists over the entire magnitude range of the acsvcs galaxies . the trend for the red gcs determined using becomes progressively more uncertain for fainter galaxy magnitudes . this is because of the decreasing fraction of red ( metal - rich ) gcs with decreasing host galaxy luminosity ( peng 2006 ) . the firmest result from figure [ slopes ] regarding the color - magnitude relation for red gcs is that there is no strong @xmath3 slope for the two brightest groups ; the slopes for the two faintest galaxy groups have large uncertainties . the large positive slopes found in @xmath6 are particularly prone to biases , a conclusion that is confirmed by looking at the more robust biweight fit results . the comparison between the biweight fits to the real data and the simulations shows that the slope of the red population is consistent with zero over the full luminosity range . an interesting feature of the observed color - magnitude relations is illustrated in figure [ radial ] . here we subdivide the gc samples of groups 1 and 2 by galactocentric radius into equally - sized , inner " and outer " subsamples . as the dividing radius , we choose @xmath55 , which translates to a physical distance of @xmath56 kpc at the distance of the virgo cluster . in terms of the effective radii , @xmath57 , of the host galaxies in groups 1 and 2 , this fixed dividing radius corresponds to @xmath58 0.30.6 @xmath57 ( ferrarese 2006 ) . tables [ fitresults ] and [ fitresultsnaive ] report the slopes for the different subsamples . for group 1 , the slope found for the inner gcs is a factor of @xmath58 23 higher than that for the outer gcs . both the and biweight methods support this conclusion , at significance levels of 2.2@xmath36 and 2.5@xmath36 , respectively . for the combined group 1 + 2 samples , the fits still obtain a difference of about a factor of three ( at 2.9@xmath36 significance ) , while the differences are almost insignificant for the biweight fits . in figure [ sizecolor ] , we plot color vs. half - light radius for gcs with @xmath59 mag , separated into inner and outer samples . there is a small size offset between the inner and outer sample , in the sense that outer gcs are larger by about 10% . this dependence of @xmath19 on galactocentric radius has already been investigated by jordn @xcite . in figure [ colorhist ] , we plot the color histograms for two luminosity ranges : in the left panel , we show gcs in the same luminosity range as in figure [ sizecolor ] ; in the right panel , we show gcs in the range @xmath60 mag . clearly , the gap between the two peaks is larger among the fainter sample of gcs , a consequence of the color - magnitude trend . furthermore , for the brighter gcs , the inner and outer color histograms appear to differ more than for the fainter gcs . the ks probability with which the inner and outer sample of the bright gcs have the same underlying distribution is 1% , while it is 14% for the faint sample . indeed , while the bimodality feature is still notable for the outer bright gcs , it is slightly smeared out for the bright , inner gcs this suggests that the dependence of the color - magnitude relation on galactocentric distance is driven mainly by the bright gcs ( @xmath61 mag ) . figure [ colorhist ] also shows that the fraction of blue - to - red gcs decreases slightly for smaller galactocentric distance ( see also , e.g. , dirsch 2003 ) . the corresponding fractions of blue - to - red gcs are 0.45:0.55 and 0.50:0.50 for the inner and outer gcs , respectively . it is apparent from the color histograms in figure [ colorhist ] that this small effect can not be responsible for the steeper color - magnitude relation found for the inner sample of gcs . before proceeding , we pause to compare the distribution of gcs in the cmd with those of other stellar systems : i.e. , the nuclei of acsvcs galaxies , the acsvcs galaxies themselves , and some candidate ultra - compact dwarf galaxies ( ucds ) identified in the course of the acsvcs . figure [ cmdallobjects ] compares the combined cmd of gcs ( red circles ) , compact stellar nuclei ( asterisks ; ct 2006 ) , acsvcs galaxies ( green circles ; ferrarese 2006 ) and ucds ( open squares ; haegan 2005 ) . in this figure , we have converted from @xmath3 to stellar masses using the population synthesis models of bruzual & charlot @xcite under the assumption of a 10-gyr old population and a chabrier imf ( chabrier 2003 ) . for comparison , we also show the color - magnitude relation for early - type galaxies in the fornax cluster from hilker @xcite , after converting from @xmath17 to @xmath2 using the transformation of fukugita @xcite . the majority of the blue gcs are not consistent with an extrapolation of the galaxy color - magnitude relation , nor with that of the nuclei . moreover , the color - magnitude relations of the nuclei and the galaxies are steeper than that of the gcs . only the faintest nuclei appear to be roughly consistent with the color - magnitude relation observed for the blue gcs . in any case , we point out that the existence of a color - magnitude trend for the blue gcs results in a paucity of gcs at high luminosities and blue colors : i.e. , @xmath62 mag and @xmath63 mag . it is difficult to understand how contamination of the observed cmds by some unrecognized stellar system stripped nuclei , ultra - compact dwarfs ( ucds ) , or otherwise can lead to an underpopulation of this region of the cmd . as a starting point for this discussion , we summarize the two main findings thus far , both of which concern dependencies of the measured color - magnitude slope on external factors : ( 1 ) the slope for the blue gcs decreases gradually towards fainter host galaxy luminosities , whereas there is no significant color - magnitude relation for the red gcs ; and ( 2 ) the slope is steeper for gcs which lie closer to the centers of their host galaxies . we now discuss a number of scenarios that that might help explain the observed trend : ( 1 ) contamination by super star clusters , stripped nuclei , or ucds ; ( 2 ) accretion of gcs from low - mass galaxies ; ( 3 ) stochastic effects ; ( 4 ) capture of field stars by individual gcs ; and ( 5 ) gc self enrichment . in [ metallicity ] , we briefly discuss how the observed color - magnitude trends appear after transforming into the metallicity - mass plane . [ discussion ] one idea that may explain the slope difference between inner and outer gcs is the presence of overluminous star clusters of intermediate color that may lie close to the galaxy centers . such a putative population would populate the region between the two color sequences at high luminosities . merged super star clusters ( e.g. , fellhauer & kroupa 2002 , 2005 ; kissler - patig 2006 ) could , in principle , be those objects . they are expected to be more extended than their monolithic counterparts of equal mass ( e.g. , bekki 2004 ; fellhauer & kroupa 2002 ; bastian 2005 ) . however , in the size vs. color plot shown in figure [ sizecolor ] it is the _ outer _ gcs that are slightly larger . more importantly , there is no offset in the size distribution for the inner gcs with intermediate colors ( i.e. , @xmath64 mag ) . this makes the presence of a large number of merged star clusters unlikely . another possibility is the presence of naked " or tidally stripped de , n nuclei , as suggested by harris @xcite ( see also bassino _ et al . _ 1994 ) . the de , n nuclei of virgo cluster galaxies do exhibit a color - magnitude trend and have colors comparable to those of the blue gcs ( see , e.g. , lotz 2004 ; ct 2006 ; figure [ cmdallobjects ] ) . ct @xcite find a median half - light radius of @xmath65 pc for nuclei with @xmath59 mag . this is comparable to the sizes of gcs in this same luminosity range ( figure [ sizecolor ] ) . stripped nuclei might therefore be confused with normal " blue gcs at the highest luminosities . however , as pointed out in [ comparison ] , the overall distribution of gcs with the cmd shows a _ lack _ of bright , blue objects . very generally , this dearth is not explained by the addition of an extra " population of contaminators , but rather requires an underlying trend for most of the objects . stripped nuclei would have to dominate the population of blue gcs . this seems unlikely to us for several reasons . first , the nuclei luminosity function peaks @xmath58 24 mag brighter than that of gcs ( see lotz 2004 ; , ct 2006 , see also figure [ cmdallobjects ] ) . more seriously , though , is the shear number of nuclei that would be required in this scenario . observationally , there is an upper limit on the number of naked nuclei brighter than @xmath66 mag from jones _ et al . _ these authors found nine compact objects ucds in their notation with @xmath67 mag in a survey that is almost complete within the central degree of the virgo cluster ( excluding the 1 - 2 surrounding m87 ) . their nine compact objects are distributed over an area that is @xmath68 100@xmath69 the acs field of view , in qualitative agreement with the predictions by bekki _ et al . _ @xcite for the spatial distribution of tidally stripped de , n nuclei . therefore , a presence of dozens , or possibly hundreds , of naked nuclei in our acs fields for the most luminous galaxies seems unlikely . a final note regarding the possibility of ucds as contaminators : the two most discussed formation channels for these compact stellar systems are stripped de , n nuclei and merged stellar super - clusters ( e.g. minniti _ et al . 1998 , hilker 1999a , drinkwater _ et al . _ 2003 , haegan 2005 , kissler - patig 2006 , mieske 2006 ) . since these two formation channels correspond to the two kinds of contaminators discussed in this section , we conclude that also ucds are unlikely to account for the observed trends . a population of gcs accreted from galaxies with a range of masses ( e.g. ct 1998 or hilker 1999b ) will exhibit a color - magnitude relation if : ( 1 ) the mean gc color of a single galaxy scales with galaxy mass ; and ( 2 ) more massive gcs preferentially form in more massive galaxies . indeed , both conditions are satisfied in actual early - type galaxies ( see , e.g. , peng 2005 ; kundu & whitmore 2001 ; jordn _ et al . _ 2006 in prep . and references therein ) . moreover , this simple scenario may offer , at least in principal , a way to qualitatively understand the slope differences for individual galaxies ( e.g. , the non - existence of a correlation for m49 ) as a consequence of different accretion / merger histories . in order to quantify the tilt in the cmd created by the accretion of gcs from low - mass galaxies that do not themselves have color - magnitude relations , we consider the combined gc sample of groups 24 ( see figure [ cmdall ] ) . we first calculate the mean host galaxy magnitude for each bin , and then calculate the color in each gc bin according to the relation between host galaxy magnitude and mean color of the blue and red gc subpopulations ( table 3 of peng 2006 ) . the outcome of this exercise is shown in figure [ cmdall ] . as expected , there is a ( weak ) trend between mean color with luminosity for the gc subpopulations . this trend is compared to the fits for group 1 . there is a substantial offset in the sense that the fainter luminosity groups have bluer gc colors than group 1 . in addition , the simulated slope is substantially smaller than measured for the ensemble of galaxies in group 1 . all in all , we conclude that the observed color - magnitude relations are unlikely to have been created by accretion of gcs from low - mass galaxies , although we can not rule out the possibility that accretion has contributed to the observed trend . the light of old gcs is in the red optical to near - infrared pass - bands dominated by red giant stars . generally , the number of red giants is only a small fraction of the total number of stars in a given stellar population . it is therefore worth discussing to which extent stochastic effects due to small numbers of red giants can skew the probability distribution of the integrated gc luminosity . this may cause a trend of gc color with integrated luminosity , because any stochastic effect is weaker for higher gc luminosity ( larger total number of stars ) and for bluer bands ( lower average stellar luminosity ) . in the works by cervio ( 2002 ) , and cervio & luridiana ( 2004 ; 2006 ) , the effect of discrete isochrone sampling on integrated photometric properties is discussed in detail . in cervio & luridiana ( 2006 ) , the width , skewness and kurtosis for the probability distribution of integrated luminosity of a stellar population are calculated as a function of star number and age . the faint magnitude limit of our gcs ( @xmath70 mag ) corresponds to about 10@xmath71 , and hence about @xmath72 stars . in the @xmath73-band taken to represent the @xmath12 band due to lack of estimates for the latter band , the photometric uncertainty arising from sampling effects amounts to about 0.06 mag for such a population ( cervio & luridiana ( 2006 ) ; and cervio , private communications ) . the skewness of the distribution in integrated luminosity is about 0.25 . the product of these two numbers about 0.015 mag is a measure of the difference between mode and mean of the probability distribution in integrated luminosity . given that the colour - magnitude trend in our data requires an almost ten times larger color shift , stochastic effects do probably not have a significant contribution to the observed trend . there is also a phenomenological counterargument against the importance of stochastic effects : the color - magnitude trend is shaped more by brighter gcs ( see figure [ colorhist ] ) , while stochastic effects are stronger for fainter gcs . furthermore , the environmental dependencies of the trend can not be explained by purely intrinsic stochastic effects . given that the strength of the color - magnitude trend depends on local environment , it is reasonable to examine the extent to which capture of field stars may help to explain the observed slopes . the capture of field stars has previously been considered by bica @xcite for the case of a 10@xmath74m@xmath75 star cluster orbiting within the galactic bulge at a distance of 1 kpc . they found that such a cluster could absorb a substantial fraction of its total mass during its lifetime . similarly , kroupa @xcite has shown that stellar super - clusters created in disks like that of the milky way can capture several tens of percent of their total mass in form of disk stars . field stars of the brightest early - type galaxies are significantly redder than their blue gc population , typically by about 0.5 to 0.6 mag ( peng 2006 ) . therefore , there will be a color shift towards the red by about 0.10 - 0.15 for a blue gc by the time that @xmath58 20 - 25% of the gc light originates from field stars . this effect will be notable mainly for the blue gc population , since the color of red gcs is much closer to that of field stars . as will become clear from the subsequent discussion , the capture rate per unit gc mass is higher for more massive gcs , which will lead to a slope in color - magnitude space . this qualitatively explains finding ( 1 ) from above . in addition , the effect will be stronger in the inner part of the host galaxy because of the higher field star density in such regions , consistent with finding ( 2 ) . thus , field star capture seems qualitatively consistent with the observed trend . we now quantify this scenario by considering a gc with a mass of 10@xmath76 , a mass that roughly corresponds to the brightest clusters in groups 1 and 2 in figure [ cmdclasses ] . for an assumed m / l=2.5 , this corresponds to a total of 4@xmath77 solar luminosities . we assume the gc to be on a circular orbit around m87 . this is certainly an extreme assumption , given that ct @xcite rule out such tangentially biased orbits for the gc system _ as a whole_. we nevertheless use it for the sake of this simple estimate . we adopt a radius of 3 kpc , representative for the sample within 5 kpc projected galactocentric distance . we may estimate the number of captured stars by multiplying the total number of stars within the volume of influence of the gc with the fraction of stars having relative velocities below the gc escape velocity . as the volume of influence , we define a torus around the orbit of the gc with an inner radius of the gc s tidal radius with respect to its host galaxy m87 . for calculating the tidal radius , we use the mass profile of m87 derived by romanowsky & kochanek @xcite . they quote the integrated mass at a distance of 3 kpc to be 1.1@xmath6910@xmath78 @xmath79 , which is comparable to the value found by ct @xcite . together with the 10@xmath76 mass of the gc , this gives a tidal radius of 9 pc . the total volume of the torus then is 4.8@xmath80 pc@xmath81 . we use the same paper by romanowsky & kochanek to get the 3-dimensional stellar density of m87 at 3 kpc , which is about 0.06 l@xmath82 pc@xmath83 . that is , along one orbit there will be a total of 2.9 @xmath77 solar luminosities within the volume of influence of our test gc . however , only a small fraction of those will be captured : the average escape velocity @xmath84 of the gc within its tidal radius of 9pc is about 37 km s@xmath10 , while the stellar velocity dispersion in m87 is about 275 km s@xmath10 ( romanowsky & kochanek 2001 ) . the gc population in m87 has a velocity dispersion of @xmath68 400 km s@xmath10 . the probability distribution for the relative velocity between the field stars and a gc is given by the convolution of both distributions , which is a gaussian of width @xmath36 = 485 km s@xmath10 . the fraction of stars ( in solar luminosities ) that can be captured in the 1d case is given by the area below the gaussian between + /- @xmath85 , which is 0.035 . for the full 3d case , this is @xmath86 , or one out of 2.3 @xmath87 stars . given that in total 2.9 @xmath77 stars are in the volume of influence in each orbit , this implies a capture rate of one dozen stars per orbit - body simulations , which is beyond the scope of this paper . ] . for a keplerian orbit at 3 kpc , one revolution lasts about 2.5@xmath88 years . in a hubble time , one can therefore capture about 6000 stars . this is 1.5% of the total assumed gc luminosity , not sufficient to fully explain the observed effects . only when considering a more extreme orbital distance of 1 kpc does the total number of captured stars increase to about 4@xmath87 , i.e. @xmath8910% of all stars . we now briefly examine the opposite effect of field star capture : namely cluster star evaporation . as is well known , mass segregation in gcs ( gunn & griffin 1979 ; jordn 2004 ) will result in evaporation affecting mainly the lowest mass stars . lamers et al . @xcite find for clusters of solar metallicity that evaporation will cause the overall gc color to become bluer with time ( over most of their lifetime ) due to the preferential loss of red , low - mass stars . could such a color change help to explain our observations ? two effects are important : ( 1 ) the _ maximum _ blueward shift due to evaporation is larger for more massive clusters ( lamers et al . @xcite ) , reaching up to @xmath90 0.10 mag ; and ( 2 ) the dissolution time , @xmath91 , is longer for more massive clusters . to qualitatively explain our observations one would require the massive gcs to be in an early stage of evolution , so that evaporation has not yet had a significant effect the fainter gcs would mainly define the color magnitude trend . however , figure [ colorhist ] shows that the slope changes between inner and outer gcs are mainly driven by the _ brightest _ gcs . that is , evaporation is unlikely to account fully for the observed color - magnitude trend . even more serious , evaporation can not naturally explain the simultaneous presence of a trend for the blue gcs and its absence for the red . generally speaking , self enrichment will occur in any stellar system , gcs included , if the system is able to retain a fraction of the gas expelled by the first generation of supernovae and subsequently form a new generation of stars . to create a metallicity - mass ( or , equivalently , a color - magnitude ) trend via self - enrichment , one would therefore require that more massive gcs are more efficiently self - enriched . the extent to which gcs in the milky way may have self - enriched , however , is a controversial subject : e.g. , frank & gisler @xcite , smith @xcite , gnedin @xcite , parmentier & gilmore @xcite , dopita & smith @xcite , morgan & lake @xcite , thoul @xcite . in this context , it is interesting to note that the most massive gc in the galaxy , @xmath92 cen , is known to have a metallicity distribution function that shows a long tail towards high metallicities ( e.g. , norris , freeman & mighell 1996 ; stanford _ various studies have found this high - metallicity population to be either coeval , or perhaps 24 gyr younger , than the cluster s primary ( metal - poor ) population ( e.g. , ferraro 2004 ; sollima et al . 2005 ; hughes & wallenstein 2000 ; hilker & richtler 2000 ; rey 2004 , stanford 2006 ) . this is consistent with @xmath92 cen having been at least partially self - enriched . note , however , that because of its peculiar star formation history , numerous investigators have suggested that the cluster may be the stripped nucleus of a now - dissolved dwarf galaxy ( e.g. , freeman 1993 ; hilker & richtler 2000 ; bekki & freeman 2003 ) . if true , then it may be incorrect to consider @xmath92 cen to be a typical example of a self enriched gc . this caveat aside , if self enrichment is common to the formation of the most massive gcs , then it may indeed help to explain our observations . most significantly , the lack of a correlation among the red gc subpopulation is naturally explained in this scenario : i.e. , the redder gcs are already metal - enriched , and a fixed amount _ d_z of supernovae ejecta will cause a relatively weaker metallicity increase , _ d_[fe / h ] , for metal - rich gcs . for instance , for a system with [ fe / h ] = @xmath932 ( z = 0.0002 ) , adding a fixed amount of metals _ d_z = 0.0002 will result in _ d_[fe / h]=@xmath940.3 dex . according to peng @xcite this corresponds to a shift in @xmath2 of about 0.06 mag and is about the observed amount of the color magnitude trend . however , for a system with [ fe / h ] = @xmath930.5 and the same _ d_z , we only have _ d_[fe / h]=@xmath940.013 dex , with a corresponding color change in @xmath2 well below 0.01 mag . a possible shortcoming of this self - enrichment scenario is that is does not explicitly account for environmental effects such as a dependence on galactocentric radius . on the other hand , feedback induced by pressure - confinement from the surrounding medium may be able to help reconcile this scenario with observations . to conclude this discussion , we show in figure [ met ] a metallicity - luminosity plot for the co - added sample of gcs in group 1 . to estimate gc metallicities , we have applied a quartic color - metallicity transformation based on the empirical calibration data from peng @xcite . note that the resulting metallicity distribution is _ not _ bimodal , but rather shows a single pronounced peak at [ fe / h]@xmath89 @xmath930.4 dex and a long , extended tail towards lower metallicities . this confirms recent warnings ( e.g. , richtler 2005 ; yoon , yi & lee 2006 ) that bimodality in color does not necessarily imply bimodality in metallicity . using the data in figure [ met ] , we can compare the metallicity distributions of gcs brighter and fainter than @xmath95 mag . for the metal - poor population , ( [ fe / h]@xmath96 - 0.6 ) , the fainter gcs have a significantly broader tail towards lower metallicities than do the bright ones . a ks - test gives a 0.1% probability for a common underlying distribution between both samples . in the metal - rich part of the diagram , the agreement is very good , with a ks probability of 91% . in other words , the empirical finding of a color - magnitude trend for the blue gcs can hence be re - phrased in terms of metallicity : the gc systems appear to have a skewed , but unimodal , metallicity distribution function whose metal - poor tail becomes increasingly extended at lower luminosities / masses . in this paper we have analysed the color - magnitude relation for the gc subpopulations in early - type galaxies belonging to the virgo cluster . our sample consists of 79 galaxies , spanning the range @xmath0 mag , drawn from the acs virgo cluster survey ( ct 2004 ) . our principal findings can be summarized as follows : 1 . based on fits , we find a highly significant correlation @xmath1 for the blue gcs in the co - added cmd of the three brightest virgo cluster galaxies ( m49 , m87 , m60 ) . the sense of the correlation is such that brightest gcs in this subpopulation are redder than their faint counterparts . for m87 and m60 alone , we find comparable correlations ; by contrast , the blue gcs in m49 do not appear to follow such a trend . in no galaxy do we find a significant correlation between @xmath2 and @xmath3 for the gcs associated with the red subpopulation . 2 . based on fits , we find the slope @xmath97 within the blue subpopulation to be much weaker than @xmath5 . in turn , we find a mild _ positive _ correlation @xmath45 for the red subpopulation , in the opposite sense to what is found for the blue subpopulation . 3 . we test on artificial cmds with , and without , implemented slopes for the blue subpopulation , and no slope for the red population . correctly reproduces the implemented slopes when using the cmd @xmath2 vs. @xmath3 . for @xmath2 vs. @xmath6 , gives a biased estimate for the color of the blue population and yields a positive slope for the red subpopulation , although no slope had been implemented . we trace this bias to the fact that in @xmath6 unlike in @xmath3 the luminosity function of blue gcs is populated to significantly brighter luminosities than for the red gcs . we conclude that for the real data , the strong slope @xmath5 is _ not _ an artifact of the fitting method , while the weaker slope @xmath45 is an artificial feature of the fit . we derive correlations independent of by using a biweight estimator for gcs divided on the basis of a luminosity independent color . the results confirm the significant slope found by in @xmath3 and reveal no significant discrepancies between the slope measured using @xmath3 and @xmath6 . 5 . the slope measured from the blue subpopulation decreases with host galaxy luminosity . however , even among the faintest galaxies ( @xmath30 mag ) the slope is nonzero . the color - magnitude relation exhibited by the blue gcs is stronger for clusters with smaller projected galactocentric distances . this difference is driven mainly by gcs with @xmath61 mag . we examine a number of physical mechanisms that might give rise to a color - magnitude relation with these characteristics : ( 1 ) contamination by super - star - clusters , stripped galactic nuclei , or ultra - compact dwarf galaxies ; ( 2 ) accretion of gcs from low - mass galaxies ; ( 3 ) stochastic effects ; ( 4 ) capture of galactic field stars by gcs ; and ( 5 ) self - enrichment of individual gcs . neither contaminants nor accretion of low - mass galaxies appear likely to fully explain our observations . self - enrichment in more massive gcs may be able to account for the observed color - magnitude relation , although more work is needed to determine if the dependence on galactocentric radius can be explained within the context of the self - enrichment scenario . although the capture of field stars _ does _ lead to a dependence on galactocentric radius that is at least qualitatively in agreement with our observations , this mechanism does not appear efficient enough to explain the overall size of the observed trend . while none of these scenarios can fully account for the full set of observations in a straightforward manner , self - enrichment and capture of field stars seem the most promising approaches . there is , of course , no _ a priori _ reason that a single mechanism is responsible for the observed color - magnitude relation , and conceptually , the combined effects of self - enrichment and field star capture may be able to explain the observations : i.e. , field star capture could act as a second order effect , producing environmental variations of a global mass - metallicity relation induced by self - enrichment . this would alleviate the discrepancy between real and required field star capture rate estimated in this work . dedicated n - body simulations going beyond these simple estimates will be required to accurately address the efficiency of field star capture in different environments . [ conclusions ] we would like to thank pavel kroupa , mike fellhauer , holger baumgardt , h.j.g.l.m . lamers and miguel cervio for fruitful discussions . we thank the anonymous referee for her / his comments . support for program go-9401 was provided through a grant from the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , under nasa contract nas5 - 26555 . p.c . acknowledges support provided by nasa ltsa grant nag5 - 11714 . m.j.w . acknowledges support through nsf grant ast-0205960 . s.m . acknowledges additional support from nasa grant nag5 - 7697 to the acs team . d.m . acknowledges support provided by nsf grants ast-0071099 , ast-0206031 , ast-0420920 and ast-0437519 , by nasa grant nng04gj48 g , and by grant hst - ar-09519.01-a from stsci . this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . ashman , k. m. , zepf , s. e. 1992 , apj , 384 , 50 ashman , k. m. , bird , c. m. , zepf , s. e. 1994 , aj , 108 , 2348 bastian , n. , emsellem , e. , kissler - patig , m. , maraston , c. 2006 , a&a , 445 , 471 bassino , l. p. , muzzio , j. c. , & rabolli , m. 1994 , , 431 , 634 beasley , m. a. _ et al . _ 2002 , mnras , 333 , 383 beers , t. c. , flynn , k. , gebhardt , k. 1990 , aj , 100 , 32 bekki , k. , couch , w.j . , drinkwater , m.j . , shioya , y. , 2003a , mnras , 344 , 399 bekki , k. , & freeman , k. c. 2003 , mnras , 346 , l11 bekki , k. , couch , w.j . , drinkwater , m.j . , shioya , y. , 2004 , apjl , 610 , 13 bica , e. _ et al . _ 1997 , apj , 482 , 49 bower , r.g . , lucey , j.r . , ellis , r.s . 1992 , mnras , 254 , 589 bruzual , g. , charlot , s. 2003 , mnras , 344 , 1000 burgarella , d. , kissler - patig , m. , buat , v. 2001 , aj , 121 , 2647 cervio , m. , valls - gabaud , d. , luridiana , v. , & mas - hesse , j. m. 2002 , a&a , 381 , 51 cervio , m. , luridiana , v. 2004 , a&a , 413 , 145 cervio , m. , luridiana , v. 2006 , a&a , 451 , 475 chabrier , g. , 2003 , pasp , 115 , 763 ct , p. , marzke , r. o. , west , m. j. 1998 , apj , 501 , 554 ct , p. _ et al . _ 2001 , apj , 559 , 828 ct , p. , mclaughlin , d. e. , cohen , j. g. , & blakeslee , j. p. 2003 , apj , 591 , 850 ct , p. _ et al . _ 2004 , apjs , 153 , 223 ( paper i ) ct , p. _ et al . _ 2006 , apjs , accepted , astro - ph/0603252 ( paper viii ) dirsch , b. _ et al . _ 2003 , aj , 125 , 1908 dopita , m. a. , smith , g. h. 1986 , apj , 304 , 283 drinkwater , m.j . , gregg , m.d . , hilker , m. et al . , 2003 , nature , 423 , 519 fellhauer , m. , kroupa , p. , 2002 , mnras , 330 , 642 fellhauer , m. ; kroupa , p. , 2005 , mnras , 359 , 223 ferraro , f.r . _ et al . _ 2004 , apj , 603 , l81 ferrarese , l. _ et al . _ 2006 , apjs , 164 , 334 ( paper vi ) freeman , k. c. 1993 , in asp conf . 48 , the globular cluster - galaxy connection , ed . g. h. smith & j. p. brodie ( san francisco : asp ) , 27 forbes , d. a. , brodie , j. p. , grillmair , c. j. 1997 , aj , 113 , 1652 ford , h. c. , bartko , f. , bely , p. y. _ et al . _ 1998 , proc . 3356 , p. 234 - 248 , space telescopes and instruments v , pierre y. bely ; james b. breckinridge ; eds . frank , j. , gisler , g. 1976 , mnras , 176 , 533 fukugita , m. , ichikawa , t. , gunn , j.e . 1996 , aj , 111 , 1748 gebhardt , k. , kissler - patig , m. 1999 , aj , 118 , 1526 gnedin , o. y. _ et al . _ 2002 , apj , 568 , 23 gunn , j. e. , griffin , r. f. 1979 , aj , 84 , 752 harris , g. l. h. , harris , w. e. , geisler , d. 2004 , aj , 128 , 723 harris , w. e. _ et al . _ 2006 , apj , 636 , 90 hasegan , m. , jordn , a. , ct , p. et al . 2005 , apj , 627 , 203 hilker , m. , infante , l. , vieira , g. , kissler - patig , m. , richtler , t. 1999 , a&as , 134 , 75 hilker , m. , infante , l. , richtler , t. 1999 , a&as , 138 , 55 hilker , m. , & richtler , t. 2000 , a&a , 362 , 895 hilker , m. , mieske , s. , & infante , l. 2003 , a&al , 397 , l9 hughes , j. , & wallerstein , g. 2000 , aj , 119 , 1225 jones , j. b. , _ _ 2006 , aj , 131 , 312 jordn , a. , blakeslee , j. p. , peng , e. w. _ et al . _ 2004a , apjs , 154 , 509 ( paper ii ) jordn , a. , ct , p. , ferrarese , l. _ et al . _ 2004b , apj , 613 , 279 ( paper iii ) jordn , a. 2004 , apj , 613 , l117 jordn , a. , ct , p. , blakeslee , j. _ et al . _ 2005 , apj , 634 , 1002 ( paper x ) karick , a. , drinkwater , m.j . , gregg , m.d . 2003 , mnras , 344 , 188 kissler - patig , m. 1997 , a&a , 319 , 83 kissler - patig , m. , jordn , a. , bastian , n. 2006 , a&a , 448 , 1031 kravtsov , a. v. , gnedin , o. y. 2005 , apj , 623 , 650 kroupa , p. 1998 , mnras , 300 , 200 kundu , a. , whitmore , b. 2001 , aj , 121 , 2950 lamers , h.j.g.l.m , anders , p. , de grijs , r. 2006 , a&a , 452 , 131 larsen , s. s. , brodie , j. p. , huchra , j. p. , forbes , d. a. , grillmair , c. j 2001 , aj , 121 , 2974 lotz , j. m. , miller , b. w. , ferguson , h. c. 2004 , apj , 613 , 262 mei , s. _ et al . _ 2005 , apj , 625 , 121 ( paper v ) mei , s. _ et al . _ 2006 , submitted to apj ( paper xiii ) mieske , s. , hilker , m. , infante , l. & jordn , a. 2006 , aj 131 , 2442 minniti , d. , kissler - patig , m. , goudfrooij , p. , meylan , g. 1998 , aj , 115 , 121 morgan , s. , lake , g. 1989 , apj , 339 , 171 norris , j. e. , freeman , k.c . , mighell , k.j . 1996 , apj , 462 , 241 odenkirchen , m. , grebel , e. k. , dehnen , w. 2003 , aj , 126 , 2385 ostrov , p. g. , forte , j. c. , geisler , d. 1998 , aj , 116 , 2854 parmentier , g. , gilmore , g. 2001 , a&a , 378 , 97 peng , e. , _ et al . _ 2006 , apj , 639 , 95 ( paper ix ) pryor , c. , mcclure , r. d. , fletcher , j. m. , hesser , j. e. 1989 , aj , 98 , 596 rey , s. c. _ et al . _ 2004 , aj , 127 , 958 richardson , s. , & green , p. g. 1997 , jr . b , 1997 , 59 , 731 richtler , t. 2005 , in press at bull . india , astro - ph/0512545 romanowsky , a. , & kochanek , c.s . 2001 , apj 553 , 722 schlegel , d.j . , finkbeiner , d.p . , & davis , m. 1998 , apj , 500 , 525 searle , l , zinn , r. 1978 , apj , 225 , 357 smith , g.h . 1996 , pasp , 108 , 176 sollima , a. _ et al . _ 2005 , apj , 634 , 332 spitler , l. r. , _ et al . _ 2006 , aj accepted , astro - ph/0606337 stanford , l. m. , da costa , g. s. , norris , j. e. & cannon , r. d. 2006 , apj , in press , astro - ph/0605612 strader , j. , brodie , j.p . , spitler , l. , beasley , m.a . 2006 , aj submitted , astro - ph/0508001 thoul , a. _ et al . _ 2002 , a&a , 383 , 491 tonry , j.l . , dressler , a. , blakeslee , j.p . 2001 , apj , 546 , 681 west , m. j. , ct , p. , marzke , r. o. , jordn , a. 2004 , nature , 427 , 31 whitmore , b. c. _ et al . _ 1995 , apj , 454l , 73 yoon , s. , yi , s. k. , lee , y. 2006 , science , 311 , 1129

little higgs theories are an attempt to address the `` little hierarchy problem , '' _ i.e. _ , the tension between the naturalness of the electroweak scale and the precision electroweak measurements showing no evidence for new physics up to 5 10 tev . in little higgs theories , the higgs mass - squareds are protected at one - loop order from the quadratic divergences . this allows the cutoff of the theory to be raised up to tev , beyond the scales probed by the current precision data . however , strong constraints can still arise from the contributions of the new tev scale particles which cancel the one - loop quadratic divergences from the standard model fields , and hence re - introduces the fine - tuning problem . in this paper we show that a new symmetry , denoted as-parity , under which all heavy gauge bosons and scalar triplets are odd , can remove all the tree - level contributions to the electroweak observables and therefore makes the little higgs theories completely natural . the-parity can be manifestly implemented in a majority of little higgs models by following the most general construction of the low energy effective theory la callan , coleman , wess and zumino . in particular , we discuss in detail how to implement the-parity in the littlest higgs model based on . the symmetry breaking scale can be even lower than 500 gev if the contributions from the higher dimensional operators due to the unknown uv physics at the cutoff are somewhat small . the existence of-parity has drastic impacts on the phenomenology of the little higgs theories . the-odd particles need to be pair - produced and will cascade down to the lightest-odd particle ( ltp ) which is stable . a neutral ltp gives rise to missing energy signals at the colliders which can mimic supersymmetry . it can also serve as a good dark matter candidate .

hep-ph0405243

the origin of the electroweak symmetry breaking , a fundamental mystery of the weak interactions , will be probed directly by high energy experiments in the coming decade . in the standard model ( sm ) the symmetry breaking is triggered by the vacuum expectation value ( vev ) of a scalar higgs field . the quantum correction to the mass of the higgs particle is , however , very sensitive to the ultraviolet ( uv ) physics . to produce the observed electroweak breaking scale , the radiative corrections to the higgs mass - squared need to be fine - tuned , unless new particles are introduced at around the 1 tev scale to cut off the quadratically divergent contributions . generally speaking , if the scale of the electroweak symmetry breaking is to be stabilized naturally , we expect new physics to show up at the scale @xmath4 1 tev or below . at energies below the scale of new physics , new particles can be integrated out to obtain a set of higher dimensional operators involving the standard model fields only @xcite . these higher dimensional operators can contribute to experimental observables even at energies beneath the scales of new physics , and the size of these operators are constrained by precision measurements @xcite . the strongest bounds are on the operators violating the ( approximate ) symmetries of the standard model , such as baryon number , flavor and cp symmetries . the dimensionful parameters suppressing these higher dimensional operators typically need to be above 100 tev due to the absence or rareness of events which violate these symmetries . this implies that the new physics at the tev scale should also respect these symmetries ( at least approximately ) . on the other hand , operators preserving the symmetries of the standard model are expected to be induced by new physics at the tev scale . these operators can be probed by the electroweak precision measurements . the advances of the electroweak precision measurements in the past decade put significant constraints on many of these operators . the most constrained operators need to be suppressed by the energy scales @xmath5 tev @xcite , assuming that all dimensionless coefficients are @xmath6 . taken at face value , it seems to indicate that there is no new physics up to @xmath0 tev , which creates a tension , known as the little hierarchy problem , with the naturalness requirement that new physics should appear at @xmath7 tev or below in order for the electroweak symmetry breaking to be stabilized . this discrepancy might be accidental and the electroweak scale is fine - tuned . however , it is also quite possible that the little hierarchy problem is in fact giving us an important hint of the new physics at the tev scale and ought to be taken more seriously . in the supersymmetric extension of the standard model , a conserved @xmath8-parity can be imposed to eliminate all the tree - level contributions to the electroweak observables from the superpartners so that the coefficients of the higher dimensional operators are naturally suppressed by a loop factor . this would relieve the little hierarchy problem . however , there is still a related fine - tuning problem in supersymmetric theories . in the standard model the higgs particle has a special status as the only scalar in the theory . in supersymmetric theories there are , nevertheless , many superpartners which are also scalars . generically superpartner masses are expected to be in the same order as the higgs mass since the higgs is not special . however , the loop contributions to the electroweak observables from some superpartners as light as @xmath9(100 gev ) could still be dangerous ( _ e.g. _ , @xmath10 from top squark loops ) . moreover , in the minimal supersymmetric standard model ( mssm ) , raising the lightest higgs mass above the experimental bound requires large top squark masses . on the other hand , the radiative corrections to the higgs mass from the top squark masses are at least as large as the top squark masses themselves if one requires the theory to be valid up to the energy scale of the gauge coupling unification . as a result , such heavy top squarks , though more phenomenologically desirable from above discussion , introduce fine - tuning at the level of a few percent or less . this naturalness problem in supersymmetry has been discussed in many places before @xcite . the fact that so far we have found neither the higgs , the superpartners , nor deviations from the electroweak precision data , indicates that the electroweak scale still appears to be somewhat fine - tuned for the simplest supersymmetric extensions of the standard model , even if supersymmetry is the answer to the physics beyond the standard model . little higgs theories @xcite provide an alternative approach to stabilize the electroweak scale based on a non - linear chiral lagrangian of a coset space @xmath11 . the higgs is light because it is a pseudo - nambu - goldstone boson ( pngb ) with the unique property that its mass is protected at one - loop order from the quadratic divergences . the cutoff can therefore be raised above @xmath0 tev , beyond the scales probed by the current electroweak precision measurements . consequently the contributions from the uv physics above the cutoff are in general safe from the electroweak constraints , as long as the unbroken group @xmath12 contains a custodial @xmath13 symmetry . the one - loop quadratically divergent contributions to the higgs mass from the sm top quark , gauge fields and higgs itself are cancelled by contributions from new particles at the @xmath7 tev scale with the same spins as the sm particles . in contrast to supersymmetry , there is a natural separation in the mass scale of the higgs and the new tev particles , since the higgs is the only pngb whose mass is suppressed by a two - loop factor . the electroweak observables also receive contributions from the new tev particles in little higgs theories , and there have been several studies on this issue for various little higgs models @xcite . the results are quite model and parameter dependent , although the general impression is that the electroweak precision data impose strong constraints on the symmetry breaking scale of @xmath14 in most of the parameter space , and hence the tev particles which cancel the quadratic divergences are required to be heavy . this in turn re - introduces fine - tuning into the little higgs models . a closer look shows that the strongest constraints come from the direct couplings of the sm fields to the new gauge bosons , as well as the vev of any @xmath13 scalar triplet which arises from coupling to the sm higgs . these couplings , however , are not an essential part of the little higgs theories ; they do not participate in the cancellation of the quadratic divergences of the higgs mass corrections . if there is a natural way , _ e.g. _ , by imposing a new symmetry , to suppress these dangerous couplings in a little higgs model , it would be completely consistent with the precision electroweak constraints without any fine tuning . indeed this was achieved in ref . @xcite for a three - site moose model , where a @xmath1-parity forbids all these dangerous couplings . in this paper , we discuss how the dangerous couplings in little higgs theories can be removed in a more general context . the @xmath1-parity is usually easy to be incorporated into the scalar and the gauge sectors . however , it was not respected by the fermion sectors in the conventional little higgs models . this is due to the fact that in the original models and the follow - up phenomenological studies , the sm fermions are assigned to particular representations under the broken group @xmath14 . from a low energy effective theory point of view , however , only the symmetry transformation property of a particle under @xmath12 is unambiguous since only the unbroken group @xmath12 is manifest . as callan , coleman , wess and zumino @xcite ( ccwz ) showed long time ago , by starting with a lagrangian manifestly invariant under the unbroken group @xmath12 and matter content furnishing linear representations of @xmath12 , one can construct a lagrangian non - linearly realizing the full symmetry @xmath14 with the help of nambu - goldstone fields which parametrize the coset space @xmath11 . from the low energy effective theory perspective , this is a more appropriate way to construct the effective lagrangian and it is not necessary to introduce linear representations of the full group @xmath14 at all . the advantage of this approach in our context is that there is a natural separation of the model - dependent dangerous interactions from those essential interactions which engineer the little higgs mechanism . in a majority of the little higgs models , one can easily identify a little symmetry the @xmath1-parity , which can be imposed to remove all the unwanted dangerous couplings , rendering these models completely natural . in the next section , we start with the minimal moose model of ref . it is an ideal example for the illustrative purpose because of its similarity to the very familiar qcd chiral lagrangian . in section [ sec : coset ] we move on to the most popular @xmath2 littlest higgs model @xcite and show how the @xmath1-parity can be implemented to remove the strong constraints from the electroweak precision data . the existence of the @xmath1-parity has a drastic impact on the little higgs phenomenology , which will be briefly discussed in section [ sec : pheno ] . we will comment on general models in which @xmath1-parity can or can not be naturally implemented and draw the conclusions in section [ sec : con ] . we start our discussion with the minimal moose model @xcite , a two - site model based on @xmath15 global symmetry . this model does not have a custodial @xmath13 symmetry , and a large contribution to the @xmath10 parameter is already present in the non - linear sigma model , which is independent of the contributions from the tev scale particles . however , this can be easily cured by replacing the @xmath15 group with another group , such as @xmath16 , containing a custodial @xmath13 . despite of the strong electroweak constraints on this model , we will nevertheless use it to illustrate our main point because the model shares the same global symmetry breaking pattern , @xmath17 , with the very familiar qcd chiral lagrangian . the purpose here is to show how a @xmath1-parity can be naturally imposed to get rid of the dangerous tree - level couplings between sm fermions and the heavy gauge bosons . the qcd chiral lagrangian has a well - known @xmath18 symmetry : the parity which exchanges the chirality @xmath19 . baryons , transforming as octets under the unbroken @xmath20 , can be incorporated in the chiral lagrangian in many different ways . the freedom lies in the assignments of the full @xmath21 representations under which baryons transform , since there are many representations containing an octet of the unbroken @xmath20 after chiral symmetry breaking . not every assignment respects the parity symmetry in a manifest way . alternatively , we could also not worry about how baryons transform under the full chiral group and contend ourselves with the fact that baryons form an octet under the unbroken group la ccwz . in any case , parity is a good symmetry of qcd and can be imposed in the chiral lagrangian @xcite ; the parity transformation is non - linear and involves the goldstone bosons if baryons are assigned to a parity non - invariant representation of @xmath21 . we run into a completely similar situation in the minimal moose model . there is an incarnation of the parity in qcd : the @xmath18 reflection which interchanges the two sites . however , a nave assignment of the sm fermion under the full symmetry group does not respect the reflection symmetry , resulting in direct couplings between the standard model fermions and the heavy gauge bosons . nevertheless , as in the qcd chiral lagrangian , a linear @xmath18 parity can be consistently implemented to forbid these unwanted couplings . the parallel problem in the qcd is how to write down a chiral lagrangian including fermions which linearly realizes the parity . this problem has been explicitly worked out in ref . @xcite , which is an adaption of ccwz to the chiral symmetry breaking . we will see that , by following this approach , a @xmath1-parity which removes the electroweak constraints from the new tev particles in the minimal moose model will become manifest . the minimal moose model is a two - site @xmath15 model with four links @xmath22 . we gauge the same @xmath23 subgroup within both @xmath15 groups with equal strength . on one site and @xmath24 on the other . the reason is we want the gauge sector to respect the @xmath18 reflection symmetry . this minor difference does not affect the little higgs mechanism . ] the goldstone boson matrices , @xmath25 , each contains a triplet , a doublet and a singlet under the unbroken diagonal @xmath26 gauge group , which we take to be the electroweak @xmath27 . this model has a large global symmetry @xmath28 ^ 4 $ ] spontaneously broken down to @xmath29 ^ 4 $ ] , with the non - linear realization @xmath30 the effective lagrangian is written as @xmath31 here @xmath32 contains the conventional non - linear sigma model kinetic terms and gauge interactions , whereas @xmath33 includes the plaquette operators which give rise to higgs quartic interactions : @xmath34 + { \rm h.c . } , \label{plaquette}\ ] ] where @xmath35 tev sets the cutoff of the theory to be at @xmath36 tev . the gauge and scalar sectors obviously respect the parity , @xmath37_{l , r}$,}\\ x_i & \leftrightarrow & x_i^\dagger , \end{aligned}\ ] ] under which the sm gauge bosons are even , and all the goldstone bosons and heavy gauge bosons are odd . in order to make the standard model higgs even under the parity , we note that the matrix @xmath38 commutes with all the gauge generators and can be included in the parity transformation to flip the parity of the higgs doublet without affecting anything else . ( this is in fact equivalent to a hypercharge rotation . ) we therefore define the @xmath1-parity for the gauge and the scalar sectors as , @xmath39 under the @xmath1-parity , all sm gauge bosons and higgs doublets are even , and the heavy gauge bosons and scalar triplets and singlets are odd . one can immediately see that with the @xmath1-parity , there is no mixing between the sm gauge bosons and the heavy gauge bosons , and the coupling of the scalar triplets to two higgs doublets is forbidden . so there is no tadpole term to induce the triplet vevs after electroweak symmetry breaking . in the fermionic sector @xmath40 , the right - handed ( electroweak singlet ) fermions can be easily incorporated to respect the @xmath1-parity : we assign them to be charged only under the diagonal @xmath41 , ( _ i.e. , _ they carry equal charges under @xmath42 and @xmath43 , ) @xmath44 where @xmath45 and @xmath46 is the sm hypercharge gauge field which is a linear combination of the @xmath42 gauge field @xmath47 and @xmath43 gauge field @xmath48 . to realize the @xmath1-parity linearly for the left - handed ( electroweak doublet ) fermions , we define the matrix @xmath49 , so that @xmath50 is the square root of the link field , @xmath51 @xcite . under the global @xmath21 transformations , @xmath52 where @xmath53 belongs to the unbroken @xmath20 and is a non - linear function of @xmath54 and the goldstone bosons @xmath55 . now for each electroweak doublet fermion in the standard model , @xmath56 , we take it to transform under @xmath57 as @xmath58 if @xmath59 were a complete multiplet of the unbroken @xmath20 , eq . ( [ assign ] ) would define a a non - linear representation of the @xmath57 group . to write down the kinetic term for the fermions , we need to compute the maurer - cartan one form @xcite @xmath60 where @xmath61 are the unbroken and broken gauge fields , respectively , and @xmath62 are the unbroken and broken generators . the maurer - cartan one form takes value in the lie algebra of the full symmetry group @xmath14 , therefore it can be written as linear combinations of @xmath63 and @xmath64 . the components of @xmath65 in the direction of unbroken and broken generators are defined as @xmath66 and @xmath67 respectively . under the @xmath53 rotation , @xmath68 transforms like a gauge field and @xmath69 transforms covariantly , @xmath70 these objects can be used to write down lagrangians invariant under the full broken group @xmath14 . for example , a fermion kinetic term is given by @xmath71 where @xmath72 . apart from the gauge interactions it realizes the full @xmath73 non - linearly . the additional term involving @xmath46 serves to give the correct hypercharge to the sm fermion @xmath74 . there is a compact way to write down @xmath75 by noting that , under the parity which interchanges @xmath76 , the generators @xmath77 and @xmath78 . then eq . ( [ oneform ] ) turns into @xmath79 the fermion kinetic term , eq . ( [ fermionkinetic ] ) , can now be written as @xmath80 where , to be more concrete , we have inserted @xmath81 , the square root of the fourth link @xmath82 , although it could be any of the links . ( [ fermion ] ) is manifestly invariant under @xmath1-parity , since the sm fermions are taken to be @xmath1-even . by expanding @xmath66 , one can check that at lowest order ( with no goldstone field ) , it only contains the standard model gauge interactions for the fermions , but not the direct couplings between the fermions and the heavy gauge bosons @xmath83 . another term invariant under the full @xmath57 symmetry which we can write down is @xmath84 with an _ arbitrary _ coefficient @xmath85 . it contains the dangerous direct couplings between the sm fermions and the heavy gauge bosons , which is the origin of the strong electroweak precision constraints on the little higgs models . however , one can easily see that this term is @xmath1-odd , since the @xmath1-parity interchanges @xmath86 and @xmath87 in eq . ( [ t - violating ] ) , , and @xmath88 since @xmath59 only has the upper two components . ] and can be forbidden by imposing the @xmath1-parity . the nice thing about following the approach of ref . @xcite is that the essential @xmath1-preserving interactions and the model - dependent @xmath1-violating interactions are naturally separated , thus allowing for an easy identification of the @xmath1-parity which removes the strong electroweak constraints on the little higgs models . in terms of gauge fields @xmath89 and @xmath83 , goldstone fields @xmath25 , and the fermion @xmath59 , the @xmath1-parity defined as @xmath90 is an exact symmetry of @xmath91 . the remaining task is to write down the standard model yukawa coupling @xmath92 , which was worked out in ref . @xcite , except that we need to @xmath1-symmetrize the yukawa interactions . for the top yukawa coupling we introduce a vector - like pair of @xmath1-even , colored weyl fermions @xmath93 and @xmath94 : @xmath95 \left ( \begin{array}{c } 0 \\ 0 \\ u_3^c \end{array } \right ) + \lambda^\prime f { u'}^c u ' , \label{top}\ ] ] so that the quadratically divergent contribution to the higgs mass - squared induced by the large top yukawa coupling is cancelled by the additional fermion @xmath96 . for the light up - type quarks , the yukawa coupling has the same form as eq . ( [ top ] ) except that @xmath93 and @xmath94 are not needed , whereas for the down - type quarks and charged leptons , the yukawa coupling can be written as @xmath97 \left ( \begin{array}{c } { q}\\ 0 \end{array } \right ) + \frac{\lambda_e}2 f ( 0\ 0\ e^c ) \left[\xi_1 \xi_2^\dagger + \omega_x \xi_1^\dagger \xi_2 \omega_x \right ] \left ( \begin{array}{c } { l } \\ 0 \end{array } \right ) .\ ] ] with these yukawa interactions , the @xmath1-parity defined in eq . ( [ tparity ] ) is also an exact symmetry of the yukawa sector @xmath92 , and hence a symmetry of the full effective lagrangian @xmath98 . _ the two - loop diagram with the quartic divergence._,width=226 ] the matter content of the @xmath1-invariant @xmath15 minimal moose model is very similar to the original minimal moose model in ref . @xcite . because the gauge group is @xmath99 ^ 2 $ ] , only two scalar triplets and singlets are eaten through the higgs mechanism after symmetry breaking . navely it seems that in the @xmath1-invariant model there are three doublets remaining light since only one becomes heavy due to the plaquette operators eq . ( [ plaquette ] ) . an additional doublet would have been eaten by the heavy gauge bosons if the gauge group were @xmath100 as in the original minimal moose model . however , the fermions in eq . ( [ assign ] ) do not form a complete multiplet of the global symmetry . the kinetic term eq . ( [ fermion ] ) breaks all chiral symmetries associated with @xmath81 . the scalar doublet residing in @xmath81 is no longer protected and picks up a mass of the order of @xmath101 tev . diagrammatically , a two loop _ quartic - divergent _ contribution to the doublet mass - squared is indeed generated through the two - loop diagram in fig . [ figureone ] , @xmath102 therefore below 1 tev , the matter content of the @xmath1-invariant minimal moose model is simply the two - higgs - doublet standard model , the same as the original @xmath15 minimal moose . at around 1 tev , there are three sets of scalar triplets and singlets , two doublets , as well as massive @xmath13 gauge bosons @xmath103 , @xmath104 and @xmath105 , and one vector - like colored fermion responsible for cancelling the quadratic divergence from the top loop . before moving on to other models , we note that for the minimal moose model the quartic divergence from fig . [ figureone ] , which gives the doublet a heavy mass , is not problematic because there are extra electroweak doublets in the model to spare . for little higgs models which do not contain any extra electroweak doublet , it would give a contribution to the higgs mass - squared in the order of @xmath106 if the cutoff is @xmath0 tev , and hence destabilizing the electroweak scale . it is therefore desirable to see if there is a way to cancel the two - loop quartic divergences . as we discussed above , the divergence arises due to the incomplete fermion multiplet in the kinetic term ( [ fermionkinetic ] ) which breaks all chiral symmetries . the remedy is thus to complete the fermion multiplet by introducing a vector - like pair of weyl fermions @xmath107 and @xmath108 so that the fermion @xmath59 now transforms as a complete fundamental under the @xmath20 : @xmath109 in this way , under the global @xmath21 symmetry , @xmath110 transforms like the fermion living on the @xmath111-site , while @xmath112 transforms like the fermion living on the @xmath8-site . then each interaction in the kinetic term eq . ( [ fermion ] ) preserves either the @xmath113 or the @xmath114 which keeps the doublet in @xmath81 from getting a mass . in terms of feynman diagrams the quartic divergence in fig . [ figureone ] is cancelled by the diagrams involving the @xmath107 fermion and the massive gauge bosons , as shown in fig . [ figuretwo ] . the fermions @xmath107 and @xmath108 are @xmath1-odd because @xmath115 under @xmath1-parity . the dirac mass @xmath116 , acting as the cutoff of the divergence in fig [ figureone ] , needs to be below @xmath117 for naturalness , which implies @xmath118 only need to be as low as @xmath119 tev . _ the diagrams responsible for cancelling the quartic divergence . diagram @xmath120 is there only for the @xmath121 gauge boson since @xmath107 is an @xmath27 singlet and does nt couple to @xmath122 and @xmath123 bosons directly . _ , width=415 ] in ref . @xcite a three - site model with @xmath1-parity was constructed . the gauge couplings on the third site are large and the particles associated with the third site are heavy , being close to the cutoff scale , so it is appropriate to integrate them out below their masses . the low energy effective theory after integrating out the third site is then described by the @xmath1-invariant two - site model discussed here , except for the trivial choices of the @xmath15 or @xmath16 groups . in this way , the three - site model may be viewed as a possible uv extension of the two - site model , in which all fermions transform linearly under the full symmetry , although in principle there could be other possibilities . having the experience with the qcd - like minimal moose model , we can now generalize this construction to other little higgs models . in particular , we will focus on the littlest higgs model based on an @xmath2 non - linear sigma model @xcite , which is the most extensively studied little higgs model in the literature . it has been claimed that , in order to be compatible with the precision data , the original model needs to be fine - tuned and the viable parameter space is thus severely constrained @xcite . from the discussion in the previous sections , however , we can see that the strong constraints can be removed completely , hence providing a natural model for electroweak symmetry breaking if the @xmath1-parity can be implemented . in this section we show that this is indeed possible . a few more new states are needed to complete the construction , which will be explained along the way . following the notation and the basis in ref . @xcite , we consider a non - linear sigma model arising from a @xmath124 symmetric matrix @xmath125 , transforming under the global @xmath126 symmetry as @xmath127 , with a vacuum expectation value @xmath128 which breaks @xmath129 . the unbroken generators @xmath63 and broken generators @xmath64 satisfy @xmath130 and @xmath131 , respectively . same as @xmath132/su(3)$ ] , @xmath2 is also a symmetric space in which the unbroken and broken generators satisfy the following schematic commutation relations , @xmath133 \sim t^c , \quad [ t^a , x^b ] \sim x^c , \quad [ x^a , x^b ] \sim t^c.\ ] ] the lie algebra of a symmetric space therefore has an automorphism : @xmath134 and @xmath135 , which in the case of @xmath2 can be expressed as @xmath136 for any generator @xmath137 . it is this automorphism that allows us to define the @xmath1-parity consistently . the goldstone fields , @xmath138 , are parametrized as @xmath139 two different @xmath23 subgroups of @xmath126 are gauged with equal strength : @xmath140 the unbroken gauge group is identified with the electroweak @xmath141 , under which the goldstone bosons decompose into a doublet , taken to be the higgs doublet , and a triplet : @xmath142 where we have omitted goldstone bosons eaten by the higgs mechanism . in components the higgs scalars are @xmath143 . a higgs vev , @xmath144 triggers the electroweak symmetry breaking . the gauge interactions break the global @xmath126 symmetry , but they are chosen in such a way that , when the gauge coupling of @xmath99_1 $ ] is turned off , there is an enhanced global @xmath145 living in the upper - left @xmath146 block of the @xmath126 group . conversely , when the coupling of @xmath99_2 $ ] is turned off , there is an @xmath147 living in the lower - right @xmath146 block . in this way it is guaranteed that the higgs doublet does not receive one - loop quadratic divergences from the gauge boson loops the effective lagrangian of the model is written as @xmath148 where @xmath149 contains the kinetic terms for the gauge and goldstone bosons ; @xmath40 includes the kinetic terms for the fermions ; @xmath150 generates the yukawa coupling for the top quark ; and @xmath151 generates the yukawa couplings for the light quarks . the gauge and scalar kinetic terms @xmath149 of the littlest higgs model is invariant under the @xmath1-parity which is defined as @xmath152 where @xmath153 commutes with all the gauge generators and @xmath154 . similar to the previous section , @xmath155 is included in the definition of the @xmath1-parity to flip the parity of the higgs doublet . in this way , the light fields including the standard model gauge fields and the higgs doublet are even under @xmath1-parity , while the heavy gauge bosons and the scalar triplet , which picks up a mass @xmath7 tev at one loop , are @xmath1-odd . the kinetic term for the @xmath13 singlet fermions can be written down as in eq . ( [ eq : singlet ] ) , then they only couple to the standard model gauge bosons . to write down a kinetic term for the @xmath13 doublet fermion , we introduce the object @xmath156 , just like in the @xmath1-invariant minimal moose model . since @xmath157 , the transformation property of @xmath50 under @xmath126 can be deduced from that of @xmath158 : @xmath159 where @xmath160 belongs to the unbroken @xmath16 and is a non - linear representation of the @xmath126 ; the function @xmath161 depends on the goldstone fields @xmath162 and the @xmath126 rotation @xmath163 in a non - linear way . for a set of fermions furnishing a complete multiplet of the unbroken @xmath16 , the kinetic term is written down as in eq . ( [ fermion ] ) . here we just need to know how to embed an electroweak doublet in an @xmath16 representation . since it is easy to embed an electroweak doublet in an @xmath126 representation , we begin by observing that , in the basis we choose , a fundamental representation of @xmath126 decomposes into two copies of fundamental representation of the unbroken @xmath16 : @xmath164 where @xmath165 , @xmath166 are doublets and @xmath107 is a singlet under @xmath26 . now we can simply put an electroweak doublet fermion @xmath74 of the standard model in the position of @xmath165 . however , without any additional fermions , the kinetic term breaks all global symmetries that protect the higgs mass , and a two - loop quartic - divergent contribution to the higgs mass - squared will be generated as discussed in the previous section . there are two ways to deal with this problem . the first is to enlarge the scalar sector and introduce additional electroweak doublet , just like in the moose model . for example , we can double the goldstone bosons by having two @xmath167 fields . then there are two electroweak doublets . one doublet can get a tev mass from the fermion kinetic terms by the two - loop quartic divergences , but the other can remain light and be the sm higgs doublet . other additional goldstone bosons such as triplets and singlets can be lifted by either the one - loop contributions or appropriate ( @xmath155-dependent ) plaquette operators . alternatively , we can cancel the quartic divergences by introducing new vector - pair of singlet and doublet fermions , @xmath107 , @xmath108 and @xmath168 , @xmath169 , respectively , for each electroweak doublet fermion in the standard model . put @xmath74 , @xmath107 , @xmath168 into the complete multiplet of the @xmath16 : @xmath170 and introduce dirac mass terms , @xmath116 and @xmath171 , for the new fermions . the kinetic term is now written as @xmath172 where @xmath66 is defined as in eq . ( [ oneform ] ) . the kinetic term is invariant under the @xmath1-parity with @xmath59 transforming as @xmath173 , which implies that @xmath107 , @xmath108 are @xmath1-odd and @xmath168 , @xmath169 are @xmath1-even . the kinetic term does not contain any direct couplings between standard model fermions and the heavy gauge bosons . they are contained in the @xmath1-odd term @xmath174 which is forbidden by @xmath1-parity . there is still a question about how to embed the doublet @xmath169 which marries with @xmath168 . in the low energy theory , a possibility is to embed @xmath169 , @xmath108 , together with an additional singlet @xmath175 into a spinor representation of @xmath16 . ( @xmath175 can get a mass @xmath176 with another singlet @xmath177 . ) then its kinetic term can be written down following ccwz . alternatively , one can extend the theory with a third @xmath13 gauge group , which is neutral under the @xmath1-parity , and have @xmath169 be a doublet of this extra @xmath13 . this extra @xmath13 is broken together with the two @xmath13 gauge groups inside the @xmath126 down to the diagonal sm @xmath13 . the dirac mass term @xmath178 can then arise from the yukawa type interactions after the breaking . the gauge coupling of the third @xmath13 can be large so that the mass of the extra gauge bosons is near the cutoff , and the sm gauge fields are mostly made of the gauge fields of the two @xmath13 s inside the @xmath126 . this is in the similar spirit as the three - site moose model of ref . in fact , such extensions also allow us to construct uv extensions of the @xmath1-invariant @xmath2 model , which contain only linear representations , i.e. , no ccwz construction is needed @xcite . more generally , a little higgs model based on a symmetric space @xmath11 can be made @xmath1-invariant by extending it to @xmath179 . in this way , all fermions can be assigned to linear representations under the full group @xmath180 without invoking ccwz . it can be explicitly checked that , in the presence of @xmath107 and @xmath168 , the quartic divergence in fig . 1 is cancelled by diagrams involving @xmath107 and @xmath168 similar to those in fig . 2 . because we have the complete fermion multiplet , the only symmetry breaking sources in the kinetic term come from the gauge generators @xmath181 . we already know that each gauge generator preserves either @xmath145 or @xmath147 in the global @xmath126 group , so for each interaction @xmath182 there is an unbroken @xmath15 symmetry protecting the higgs doublet from getting a mass . a higgs mass can be induced only in the presence of both @xmath183 and @xmath184 , which requires going to higher loops or a mass insertion between @xmath185 and @xmath186 . either way the contribution to the higgs mass - squared can be no bigger than the two loop _ quadratic _ divergences and hence is safe . because the quartic divergence in fig 1 is cut off by the masses of the extra fermions , @xmath118 , @xmath187 , for naturalness they need to be below @xmath117 . next step is to write down a top yukawa coupling without introducing the associated quadratically divergent contribution to the higgs mass , which can be achieved by adding the usual pair of @xmath1-even colored fermions @xmath188 and @xmath189 , as well as three additional colored , fermionic weak singlet fermions , @xmath190 , and @xmath53 . grouping the third generation quark doublet @xmath191 with @xmath188 and @xmath192 into a row vector @xmath193 , the top yukawa coupling @xmath92 can be written as @xmath194 \nonumber \\ & & + \lambda_2 f { t'}{t'}^c + \frac{\lambda_3 f}{\sqrt{2 } } u ( u^c - u^{c}_t ) + \mbox{h.c . } , \label{topyukawa}\end{aligned}\ ] ] where @xmath195 , @xmath196 , and @xmath197 . this is nothing but the @xmath1-symmetrized version of the top yukawa interaction written down in ref . @xcite , which can be easily recognized by noting that @xmath198 . we define @xmath199 to be the image of @xmath200 under @xmath1-parity , so that the first term in the square bracket in eq . ( [ topyukawa ] ) is mapped into the second under the @xmath1-parity . the singlet fermion @xmath53 is taken to be @xmath1-odd and marries the @xmath1-odd linear combination @xmath201 . any @xmath1-odd interactions have been projected out in eq . ( [ topyukawa ] ) . one way to show that there are no one - loop quadratic divergences generated from eq . ( [ topyukawa ] ) is to perform a spurion analysis as in ref . @xcite . for that purpose we can regard @xmath92 as having four independent interactions with coupling constants @xmath202 and @xmath203 , where @xmath204 is the coefficient of @xmath205 and forced by @xmath1-parity to be equal to @xmath206 . each interaction in @xmath207 preserves enough @xmath15 global symmetries to individually protect the higgs mass from getting one - loop quadratic divergences ; it takes more than one @xmath208 to get a quadratically divergent mass . moreover , with the introduction of new fermions , each @xmath208 has a global @xmath209 rephasing symmetry associated with redefining the phases of @xmath210 and @xmath53 . therefore these spurions can only enter as @xmath211 and the quadratic divergence does not come in until two - loop order . without the rephasing symmetries , there could be divergences proportional to , for instance , @xmath212 which would be too big a contribution to the higgs mass - squared . expanding @xmath92 to leading order in the higgs particle gives @xmath213 where @xmath214 is the @xmath1-even linear combination of @xmath200 and @xmath215 . as usual , a linear combination @xmath216 marries with @xmath217 ( plus a small mixture from @xmath218 ) and becomes heavy with a mass @xmath219 . after integrating out the heavy top partner @xmath1 , the orthogonal combination @xmath220 has the desired yukawa coupling to @xmath221 with the strength @xmath222 . little higgs models without @xmath1-parity are in general strongly constrained by the precision electroweak measurements . the leading corrections to the electroweak observables in these models come from several sources . first , the heavy gauge bosons mix with the sm gauge bosons through vertices like @xmath223 after electroweak symmetry breaking , when the higgs doublet gets a vev . this shifts the masses of @xmath122 and @xmath123 and also modifies the couplings of @xmath122 and @xmath123 to the sm fermions . secondly , integrating out heavy gauge bosons gives additional contributions to 4-fermion interactions at low energies if the heavy gauge bosons couple to sm fermions directly . in particular , the contribution to the fermi constant @xmath224 from muon decays feeds into all precision measurements since @xmath224 is used as an input parameter . finally , for models containing scalar triplets , a vev of the triplet @xmath225 will be induced by the higgs vev if there is a coupling @xmath226 . the triplet vev also shifts the @xmath122 boson mass . these corrections are all of the order @xmath227 . because many electroweak observables such as @xmath122 mass and @xmath123 couplings to fermions are very accurately measured , the scale of @xmath3 is in general constrained to be somewhat larger than 1 tev as most phenomenological studies on little higgs theories found . this in turn re - introduces fine - tuning in the higgs mass . for example , in the original littlest higgs model , the @xmath2 model without @xmath1-parity , the scale @xmath3 needs to be above 4 tev or so @xcite in order to be compatible with precision data . on the other hand , with @xmath1-parity all the above tree - level corrections to the electroweak observables are absent because they violate the @xmath1-parity . apart from the unknown contributions from the cutoff physics , the corrections to the electroweak observables within the effective theory are loop - suppressed . one example of such constraints is the correction to @xmath10 parameter due to the loop of the heavy top quark partner , which is responsible for cancelling the quadratic divergence from the top yukawa coupling . this correction is in the order of @xmath228 and very small . the correction to the @xmath229 vertex also gives comparable effect . in addition , the heavy gauge bosons could also contribute to the @xmath230 parameter at one loop level . all these constraints are down by a loop factor and allow the scale @xmath3 to be well below 1 tev . @xmath1-parity also drastically affects the collider phenomenology of the little higgs theories . the @xmath1-odd particles can not be singly produced due to @xmath1-parity , which implies that direct searches must rely on pair - productions . when the @xmath1-odd particles are produced , they will eventually decay into the ltp , the lightest @xmath1-odd particle , which is stable since @xmath1-parity forbids it from decaying into lighter particles which are all @xmath1-even . a charged ltp is not preferred as it can cause cosmological problems . in @xmath1-invariant models the ltp is most likely to be the @xmath231 gauge boson because its mass is suppressed by the smallness of the @xmath121 coupling and also often by the normalization factor of the @xmath209 generator , although some other neutral scalar is also possible . for a neutral ltp , the typical collider signals will be the @xmath1-odd particles decaying into jets and/or leptons plus missing energies . in this respect the phenomenology is similar to those of the supersymmetric theories with @xmath8-parity and kk - parity conserving universal extra - dimensions ( ued ) @xcite , where the lightest supersymmetric particle ( lsp ) and the lightest kk - odd particle ( lkp ) are also stable and would escape detection in collider experiments if they are neutral . moreover , the ltp could serve as a viable dark matter candidate , similar to lsp and lkp . one distinct feature of the @xmath1-invariant little higgs theories is that the top sector still contains a @xmath1-even heavy top quark partner , whose phenomenology is not affected by the @xmath1-parity . it can be used to test the little higgs mechanism in colliders , as shown in ref . the @xmath1-even top partner can be singly produced and its decay products do not necessarily give missing energies . a recent study show that it can be observed up to mass of order 2.5 tev via its decay to @xmath232 at the lhc @xcite . on the other hand , for the @xmath1-odd particles such as heavy gauge bosons and scalar triplets , all the previous phenomenological studies without assuming the @xmath1-parity do not apply . the details of their phenomenology depend on the particle content and the spectrum of each individual model . in the rest of this section , we will describe in a little more detail the phenomenology of the @xmath1-invariant @xmath2 model as an example . compared with the original littlest higgs model , the @xmath1-invariant model has an extended scalar sector or fermion sector ( or both ) in the tev scale . here we consider the model with an extended fermion sector . then , in the scalar and gauge sectors , the @xmath1-invariant @xmath2 model has the same matter / field content as the littlest higgs model . there is one higgs doublet @xmath12 which is light , one heavy triplet scalar @xmath233 , and a set of massive @xmath23 gauge bosons @xmath234 and @xmath231 . the triplet scalar , responsible for cancelling the quadratic divergence of the higgs quartic coupling , can not get a vev here because it is @xmath1-odd and forbidden to have a tadpole term from coupling to the higgs doublet after electroweak symmetry breaking . the gauge couplings of the two @xmath23 groups are required to be equal by @xmath1-parity , implying @xmath235 and @xmath236 , where @xmath237 and @xmath238 are the gauge coupling constants for the electroweak @xmath239 . the mixing angles in the gauge couplings are set by the @xmath1-parity to be @xmath240 , instead of being free parameters as in the littlest higgs model . the masses for various gauge bosons are given by @xmath241 in the limit @xmath242 , these formulae reduce to the familiar ones : @xmath243 in the fermion sector , in addition to the top partners which cancels the quadratic divergence from the top yukawa coupling , for every sm doublet fermion there is also a vector - like @xmath1-odd @xmath27 singlet fermion and a vector - like @xmath1-even doublet fermion . if the masses of these fermions are not degenerate among different generations or aligned with the sm fermion masses , they can induce flavor - changing effects , such as @xmath244 mixing , @xmath245 and so on . however , their couplings to sm fermions always involve the higgs or the scalar triplet , _ e.g. _ , @xmath246 , @xmath247 , @xmath248 . their contributions to the flavor - changing neutral currents are therefore suppressed either by powers of @xmath249 , or by higher loops if the physical higgs boson or the scalar triplet is involved . for @xmath250 1 tev , the induced flavor - changing effects are already safe even without degeneracies or alignments . as we discussed earlier in this section , within the effective theory the corrections to the electroweak observables are loop - suppressed . the strongest constraint comes from the @xmath10 correction due to heavy gauge boson loops . one can see from eq.([exact_mass ] ) that even if @xmath238 is turned off , the heavy @xmath251 and @xmath252 gauge bosons are not degenerate and the mass splitting is @xmath253 this mass splitting induces at one - loop level a custodial @xmath254 breaking effect and the correction to @xmath10 is given by @xmath255 where @xmath256 , and @xmath257 is the weinberg angle . the logarithmic divergence means that a counter term in the non - linear sigma model of the form @xmath258\ ] ] is required to cancel the divergence , where the numerical coefficient @xmath259 is expected to be of order unity . indeed , after turning on the higgs vev , this interaction gives a mass - splitting within the standard model @xmath13 gauge bosons @xmath260 this is due to the fact that gauging two @xmath13 s in this model is a custodial @xmath254 violating effect . there are further @xmath254 breaking effects from the @xmath209 gauge couplings , but they are expected to be smaller due to the smallness of @xmath238 . assuming there is no large counter term at the cutoff , the contribution to @xmath10 from @xmath104 loop between the @xmath261 and the cutoff @xmath262 is negative and enhanced by a @xmath263 factor , @xmath264 if one requires that @xmath265 , this only constrains the symmetry breaking scale @xmath3 to be greater than @xmath266 gev which corresponds to @xmath267 where @xmath268 , @xmath105 represent the mass eigenstates and are very close to the gauge eigenstates @xmath103 , @xmath252 . the @xmath10 parameter also receives contribution from the top partner at the one - loop level , @xmath269,\ ] ] where @xmath270 is the mixing angle between the top quark and the top partner in the left - handed sector , and @xmath271 is the mass of the heavy top partner . the top yukawa coupling is given by @xmath272 . it does not provide any significant constraint as long as @xmath273 is somewhat small ( @xmath274 ) . the correction to the @xmath229 vertex from the top partner loop has the same leading logarithm as in eq . ( [ toprho ] ) and gives comparable constraint to @xmath275 . the contributions to the @xmath276 parameter of peskin - takeuchi @xcite are also small . in fact , the top partner contribution to @xmath10 is positive so it is possible to partially cancel the negative contribution from the @xmath104 loop , allowing @xmath3 to be even lower . ( the direct search limit is @xmath277 gev from lep ii which corresponds to @xmath278 gev . ) of course , a very low @xmath3 undermines the purpose of the little higgs theories and one expects that the higher dimensional operators at the cutoff @xmath279 will give too big contributions to the electroweak observables . indeed , as the constraints from the calculable effects within the effective theory are rather weak in this model , the most stringent constraints are expected to come from the higher dimensional operators arising from unknown uv physics at the cutoff scale , which would prefer @xmath280 tev if these unknown contributions are not suppressed . in terms of collider phenomenology , the @xmath1-invariant model escapes conclusions reached by most studies in the literature so far , except for those on the @xmath1-even top partners @xmath281 . the @xmath1-odd particles in the model are the heavy gauge bosons , a triplet scalar , and the singlet fermions @xmath282 and @xmath283 ; they are produced in pairs through the sm gauge bosons and eventually decay into the ltp . it can be seen from eq . ( [ wprimemass ] ) that the gauge bosons are likely to be lighter than the triplet scalar and the heavy fermions . the relatively light @xmath231 boson is most likely to be the ltp , as will be assumed in the following studies . moreover , electroweak symmetry breaking induces corrections to the mass of the heavy gauge bosons , as well as a small mixing between @xmath284 and @xmath231 , which causes the mass eigenstate @xmath285 to be slightly heavier than the @xmath234 . the mass splitting can be calculated from eq.([exact_mass ] ) and is of the order of @xmath286 . the triplet scalar @xmath233 can be slightly heavier than 1 tev , since it is only responsible for cancelling the quadratic divergence from the higgs quartic coupling . the singlet fermions @xmath107 and doublet fermions @xmath287 can be 3 tev or higher , because they are there only to cancel the quartic divergence from the fermion kinetic term . the additional fermions in the top sector , on the other hand , is at around 1 tev or so . due to the @xmath1-parity , all the @xmath1-odd particles need to decay into another @xmath1-odd particle plus something , and will eventually cascade down to the ltp , resulting in the missing energy in the detector . @xmath104 is likely to be the next - to - lightest @xmath1-odd particle . it decays into @xmath122 plus @xmath268 through the small @xmath284 component in @xmath288 due to the mixing . @xmath105 decays into @xmath289 through @xmath290 ( one of them needs to be virtual because of small mass splitting between @xmath104 and @xmath105 ) . other heavy particles can decay through the interactions , @xmath291 , @xmath292 , @xmath293 , @xmath294 , @xmath295 , and @xmath296 , where @xmath12 can be replaced by its vev or the physical higgs boson . .[decay]_the major decay channels of the heavy particles in the @xmath1-invariant @xmath2 model . the @xmath231 is assumed to be the ltp . the particles produced in the decays may be replaced by the ones in the parentheses . _ [ cols="^,^,^",options="header " , ] in table [ decay ] we list the major decay modes for all the heavy particles in the @xmath1-invariant model . for concreteness we have assumed the following spectrum , @xmath297 the decays of the lighter ( and hence more accessible ) states ( @xmath298 ) often produce @xmath122 , @xmath123 bosons associated with the missing energy , which could be the first things to look for in collider experiments , though separating them from sm backgrounds would be a challenge . with tevatron run ii currently running and lhc to start in 2007 , the tev scale physics will be fully explored in the coming decade . the mystery of the origin of the electroweak symmetry breaking is expected to be unveiled . so far the precision electroweak measurements have not provided any evidence for new physics beyond the standard model , which is quite puzzling . perhaps the fact that there is no sign of new physics in the precision electroweak data itself is one of the biggest hints on the possible new physics that will show up at the tev scale . in this paper , we show that the little higgs theories , supplemented with the @xmath1-parity , can be a solution to this little hierarchy problem . the large quadratically divergent corrections to the higgs mass - squared from the standard model particles are cancelled by those from new tev particles , which stabilizes the electroweak scale . at the same time the @xmath1-parity forbids the tree - level contributions to the electroweak observables from the tev particles , and makes the little higgs theories consistent with the electroweak precision data without fine - tuning . a subtlety for imposing @xmath1-parity on little higgs theories is how to incorporate the standard model fermions . however , here we show that by following ccwz , which is the most general way to construct the low energy effective theory for a broken symmetry , this problem is readily resolved . the essential @xmath1-even interactions and the troublesome @xmath1-odd interactions can be naturally separated , which allows for an obvious implementation of @xmath1-parity . although we only discussed two examples , the minimal moose model and the littlest higgs model , from the discussion it should be apparent that the @xmath1-parity can be introduced in many other models . as long as a model is based on a symmetric space with the unbroken and broken generators satisfying ( [ eq : symmetric ] ) , the automorphism @xmath299 , @xmath300 , allows us to define a @xmath1-parity consistently . for some models one needs to worry about the 2-loop quartic - divergent contributions to the higgs mass - squared , but it can be easily addressed by completing fermions into complete multiplets of the unbroken group @xmath12 or enlarging the global symmetry . there are also little higgs models which do not live in a symmetric coset space . one example is the little higgs model based on a `` simple group '' @xcite . in the simplest version ( which all other variants based on ) , an @xmath15 gauge symmetry is broken down to @xmath13 by the vevs of two triplets . in this case , the broken generator @xmath301 does not satisfy ( [ eq : symmetric ] ) , and the coset space is not a symmetric space . one can not find a consistent definition of @xmath1-parity under which all heavy gauge bosons are odd . nevertheless , one can still follow the approach of ccwz . the model - independent couplings of the sm fermions to the sm gauge bosons and the model - dependent couplings of the sm fermions to the heavy gauge bosons are still naturally separated into different terms . the difference is that in this case there is no symmetry to forbid the couplings to the heavy gauge bosons . in fact , they will be radiatively generated even if one does not include them in the first place . for such models , the electroweak precision constraints turn into constraints on these model dependent couplings . how much fine - tuning , if any , is required depends on the model and needs to be examined individually . the existence of an exact @xmath1-parity obviously has a big impact on the little higgs phenomenology . the @xmath1-odd particles need to be pair - produced and the lightest @xmath1-odd particle is stable . a neutral ltp provides a viable dark matter candidate and gives rise to missing energy signals in collider experiments . to a first approximation , the collider signals for this class of models are similar to those of the supersymmetric standard model . the details of what can be observed at the colliders depend on each individual model and its spectrum . they certainly deserve more thorough investigations . in order to distinguish the little higgs theories with @xmath1-parity from supersymmetry , the information on the spins of the new particles participating in the cancellation of quadratic divergences will be crucial . in this respect , a linear collider with a high enough energy could be very helpful . we would like to thank nima arkani - hamed who provided many important insights to this work . we would also like to acknowledge helpful conversations with jose espinosa , markus luty , jay wacker and martin schmaltz . this work is supported in part by the national science foundation under grant phy-0244821 . s. weinberg , `` phenomenological lagrangians , '' physicaa * 96 * , 327 ( 1979 ) . w. buchmuller and d. wyler , `` effective lagrangian analysis of new interactions and flavor conservation , '' nucl . phys . b * 268 * , 621 ( 1986 ) . for a recent analysis , see r. barbieri and a. strumia , `` what is the limit on the higgs mass ? , '' phys . b * 462 * , 144 ( 1999 ) [ arxiv : hep - ph/9905281 ] . p. batra , a. delgado , d. e. kaplan and t. m. p. tait , the higgs mass bound in gauge extensions of the minimal supersymmetric jhep * 0402 * , 043 ( 2004 ) [ arxiv : hep - ph/0309149 ] . r. harnik , g. d. kribs , d. t. larson and h. murayama , `` the minimal supersymmetric fat higgs model , '' arxiv : hep - ph/0311349 . r. barbieri , electroweak symmetry breaking as of 2003 , on the way to the large hadron arxiv : hep - ph/0312253 . n. arkani - hamed , talk at the kitp conference : collider physics ( jan 12 - 16 , 2004 ) , http://online.kitp.ucsb.edu / online / collider_c04/arkanihamed/. n. arkani - hamed , a. g. cohen and h. georgi , `` electroweak symmetry breaking from dimensional deconstruction , '' phys . b * 513 * , 232 ( 2001 ) [ hep - ph/0105239 ] . n. arkani - hamed , a. g. cohen , t. gregoire and j. g. wacker , `` phenomenology of electroweak symmetry breaking from theory space , '' jhep * 0208 * , 020 ( 2002 ) [ arxiv : hep - ph/0202089 ] . n. arkani - hamed , a. g. cohen , e. katz , a. e. nelson , t. gregoire and j. g. wacker , `` the minimal moose for a little higgs , '' jhep * 0208 * , 021 ( 2002 ) [ arxiv : hep - ph/0206020 ] . n. arkani - hamed , a. g. cohen , e. katz and a. e. nelson , `` the littlest higgs , '' jhep * 0207 * , 034 ( 2002 ) [ arxiv : hep - ph/0206021 ] . i. low , w. skiba and d. smith , `` little higgses from an antisymmetric condensate , '' phys . rev . d * 66 * , 072001 ( 2002 ) [ arxiv : hep - ph/0207243 ] . d. e. kaplan and m. schmaltz , `` the little higgs from a simple group , '' arxiv : hep - ph/0302049 . s. chang and j. g. wacker , `` little higgs and custodial su(2 ) , '' arxiv : hep - ph/0303001 . s. chang , `` a littlest higgs model with custodial su(2 ) symmetry , '' arxiv : hep - ph/0306034 . w. skiba and j. terning , `` a simple model of two little higgses , '' arxiv : hep - ph/0305302 . j. l. hewett , f. j. petriello and t. g. rizzo , `` constraining the littlest higgs , '' arxiv : hep - ph/0211218 . c. csaki , j. hubisz , g. d. kribs , p. meade and j. terning , `` big corrections from a little higgs , '' phys . d * 67 * , 115002 ( 2003 ) [ arxiv : hep - ph/0211124 ] . c. csaki , j. hubisz , g. d. kribs , p. meade and j. terning , `` variations of little higgs models and their electroweak constraints , '' arxiv : hep - ph/0303236 . t. gregoire , d. r. smith and j. g. wacker , `` what precision electroweak physics says about the su(6)/sp(6 ) little higgs , '' arxiv : hep - ph/0305275 . c. kilic and r. mahbubani , `` precision electroweak observables in the minimal moose little higgs model , '' arxiv : hep - ph/0312053 . m. c. chen and s. dawson , `` one - loop radiative corrections to the rho parameter in the littlest higgs model , '' arxiv : hep - ph/0311032 . r. casalbuoni , a. deandrea and m. oertel , jhep * 0402 * , 032 ( 2004 ) [ arxiv : hep - ph/0311038 ] . w. kilian and j. reuter , phys . d * 70 * , 015004 ( 2004 ) [ arxiv : hep - ph/0311095 ] . cheng and i. low , `` tev symmetry and the little hierarchy problem , '' jhep * 0309 * , 051 ( 2003 ) [ arxiv : hep - ph/0308199 ] . s. r. coleman , j. wess and b. zumino , `` structure of phenomenological lagrangians . 1 , '' phys . * 177 * ( 1969 ) 2239 . c. g. . callan , s. r. coleman , j. wess and b. zumino , `` structure of phenomenological lagrangians . 2 , '' * 177 * ( 1969 ) 2247 . a. manohar and h. georgi , `` chiral quarks and the nonrelativistic quark model , '' nucl . b * 234 * , 189 ( 1984 ) . h. georgi , _ weak interactions and modern particle physics _ , the benjamin / cummings publishing company , 1984 . cheng , k. t. matchev and m. schmaltz , `` bosonic supersymmetry ? getting fooled at the lhc , '' phys . d * 66 * , 056006 ( 2002 ) [ arxiv : hep - ph/0205314 ] . m. perelstein , m. e. peskin and a. pierce , `` top quarks and electroweak symmetry breaking in little higgs models , '' arxiv : hep - ph/0310039 . g. azuelos _ et al . _ , `` exploring little higgs models with atlas at the lhc , '' arxiv : hep - ph/0402037 . m. e. peskin and t. takeuchi , `` a new constraint on a strongly interacting higgs sector , '' phys . lett . * 65 * , 964 ( 1990 ) .

we introduce and study graphic lambda calculus , a visual language which can be used for representing untyped lambda calculus , but it can also be used for computations in emergent algebras or for representing reidemeister moves of locally planar tangle diagrams . [ section ] [ theorem]proposition [ theorem]lemma [ theorem]corollary [ theorem]definition [ theorem]remark [ section ]

1305.5786

graphic lambda calculus consists of a class of graphs endowed with moves between them . it might be considered a visual language in the sense of erwig @xcite . the name `` graphic lambda calculus '' comes from the fact that it can be used for representing terms and reductions from untyped lambda calculus . it s main move is called `` graphic beta move '' for it s relation to the beta reduction in lambda calculus . however , the graphic beta move can be applied outside the `` sector '' of untyped lambda calculus , and the graphic lambda calculus can be used for other purposes than the one of visual representing lambda calculus . for other visual , diagrammatic representation of lambda calculus see the vex language @xcite , or david keenan s @xcite . the motivation for introducing graphic lambda calculus comes from the study of emergent algebras . in fact , my goal is to build eventually a logic system which can be used for the formalization of certain `` computations '' in emergent algebras , which can be applied then for a discrete differential calculus which exists for metric spaces with dilations , comprising riemannian manifolds and sub - riemannian spaces with very low regularity . emergent algebras are a generalization of quandles , namely an emergent algebra is a family of idempotent right quasigroups indexed by the elements of an abelian group , while quandles are self - distributive idempotent right quasigroups . tangle diagrams decorated by quandles or racks are a well known tool in knot theory @xcite @xcite . it is notable to mention the work of kauffman @xcite , where the author uses knot diagrams for representing combinatory logic , thus untyped lambda calculus . also meredith and snyder@xcite associate to any knot diagram a process in pi - calculus , is there any common ground between these three apparently separated field , namely differential calculus , logic and tangle diagrams ? as a first attempt for understanding this , i proposed @xmath0-scale calculus @xcite , which is a formalism which contains both untyped lambda calculus and emergent algebras . also , in the paper @xcite i proposed a formalism of decorated tangle diagrams for emergent algebras and i called `` computing with space '' the various manipulations of these diagrams with geometric content . nevertheless , in that paper i was not able to give a precise sense of the use of the word `` computing '' . i speculated , by using analogies from studies of the visual system , especially the `` brain a geometry engine '' paradigm of koenderink @xcite , that , in order for the visual front end of the brain to reconstruct the visual space in the brain , there should be a kind of `` geometrical computation '' in the neural network of the brain akin to the manipulation of decorated tangle diagrams described in our paper . i hope to convince the reader that graphic lambda calculus gives a rigorous answer to this question , being a formalism which contains , in a sense , lambda calculus , emergent algebras and tangle diagrams formalisms . [ [ acknowledgement . ] ] acknowledgement . + + + + + + + + + + + + + + + + this work was supported by a grant of the romanian national authority for scientific research , cncs uefiscdi , project number pn - ii - id - pce-2011 - 3 - 0383 . an oriented graph is a pair @xmath1 , with @xmath2 the set of nodes and @xmath3 the set of edges . let us denote by @xmath4 the map which associates to any node @xmath5 the set of adjacent edges @xmath6 . in this paper we work with locally planar graphs with decorated nodes , i.e. we shall attach to a graph @xmath1 supplementary information : 1 . a function @xmath7 which associates to any node @xmath5 an element of the `` graphical alphabet '' @xmath8 ( see definition [ defalp ] ) , 2 . a cyclic order of @xmath6 for any @xmath5 , which is equivalent to giving a local embedding of the node @xmath9 and edges adjacent to it into the plane . we shall construct a set of locally planar graphs with decorated nodes , starting from a graphical alphabet of elementary graphs . on the set of graphs we shall define local transformations , or moves . global moves or conditions will be then introduced . the graphical alphabet contains the elementary graphs , or gates , denoted by @xmath0 , @xmath10 , @xmath11 , @xmath12 , and for any element @xmath13 of the commutative group @xmath14 , a graph denoted by @xmath15 . here are the elements of the graphical alphabet : 1 . @xmath0 graph , @xmath10 graph , 2 . @xmath11 graph , @xmath15 graph , 3 . @xmath12 graph . with the exception of the @xmath12 , all other elementary graphs have three edges . the graph @xmath12 has only one edge . [ defalp ] there are two types of `` fork '' graphs , the @xmath0 graph and the @xmath10 graph , and two types of `` join '' graphs , the @xmath11 graph and the @xmath15 graph . further i briefly explain what are they supposed to represent and why they are needed in this graphic formalism . the @xmath0 gate corresponds to the lambda abstraction operation from untyped lambda calculus . this gate has one input ( the entry arrow ) and two outputs ( the exit arrows ) , therefore , at first view , it can not be a graphical representation of an operation . in untyped lambda calculus the @xmath0 abstraction operation has two inputs , namely a variable name @xmath16 and a term @xmath8 , and one output , the term @xmath17 . there is an algorithm , presented in section [ constru ] , which transforms a lambda calculus term into a graph made by elementary gates , such that to any lambda abstraction which appears in the term corresponds a @xmath0 gate . the @xmath10 gate corresponds to a fan - out gate . it is needed because the graphic lambda calculus described in this article does not have variable names . @xmath10 gates appear in the process of elimination of variable names from lambda terms , in the algorithm previously mentioned . another justification for the existence of two fork graphs is that they are subjected to different moves : the @xmath0 gate appears in the graphic beta move , together with the @xmath11 gate , while the @xmath10 gate appears in the fan - out moves . thus , the @xmath0 and @xmath10 gates , even if they have the same topology , they are subjected to different moves , which in fact characterize their `` lambda abstraction''-ness and the `` fan - out''-ness of the respective gates . the alternative , which consists into using only one , generic , fork gate , leads to the identification , in a sense , of lambda abstraction with fan - out , which would be confusing . the @xmath11 gate corresponds to the application operation from lambda calculus . the algorithm from section [ constru ] associates a @xmath11 gate to any application operation used in a lambda calculus term . the @xmath15 gate corresponds to an idempotent right quasigroup operation , which appears in emergent algebras , as an abstractization of the geometrical operation of taking a dilation ( of coefficient @xmath13 ) , based at a point and applied to another point . as previously , the existence of two join gates , with the same topology , is justified by the fact that they appear in different moves . [ [ the - set - graph . ] ] 1 . the set graph . + + + + + + + + + + + + + + + + + we construct the set of graphs @xmath18 over the graphical alphabet by grafting edges of a finite number of copies of the elements of the graphical alphabet . @xmath18 is the set of graphs obtained by grafting edges of a finite number of copies of the elements of the graphical alphabet . during the grafting procedure , we start from a set of gates and we add , one by one , a finite number of gates , such that , at any step , any edge of any elementary graph is grafted on any other free edge ( i.e. not already grafted to other edge ) of the graph , with the condition that they have the same orientation . for any node of the graph , the local embedding into the plane is given by the element of the graphical alphabet which decorates it . the set of free edges of a graph @xmath19 is named the set of leaves @xmath20 . technically , one may imagine that we complete the graph @xmath19 by adding to the free extremity of any free edge a decorated node , called `` leaf '' , with decoration `` in '' or `` out '' , depending on the orientation of the respective free edge . the set of leaves @xmath20 thus decomposes into a disjoint union @xmath21 of in or out leaves . moreover , we admit into @xmath22 arrows without nodes , , called wires or lines , and loops ( without nodes from the elementary graphs , nor leaves ) graphs in @xmath18 can be disconnected . any graph which is a finite reunion of lines , loops and assemblies of the elementary graphs is in @xmath22 . [ [ local - moves . ] ] 2 . local moves . + + + + + + + + + + + + + + + these are transformations of graphs in @xmath18 which are local , in the sense that any of the moves apply to a limited part of a graph , keeping the rest of the graph unchanged . we may define a local move as a rule of transformation of a graph into another of the following form . first , a subgraph of a graph @xmath23 in @xmath18 is any collection of nodes and/or edges of @xmath23 . it is not supposed that the mentioned subgraph must be in @xmath18 . also , a collection of some edges of @xmath23 , without any node , count as a subgraph of @xmath23 . thus , a subgraph of @xmath23 might be imagined as a subset of the reunion of nodes and edges of @xmath23 . for any natural number @xmath9 and any graph @xmath23 in @xmath18 , let @xmath24 be the collection of subgraphs @xmath25 of the graph @xmath23 which have the sum of the number of edges and nodes less than or equal to @xmath9 . a local move has the following form : there is a number @xmath9 and a condition @xmath26 which is formulated in terms of graphs which have the sum of the number of edges and nodes less than or equal to @xmath9 , such that for any graph @xmath23 in @xmath18 and for any @xmath27 , if @xmath26 is true for @xmath25 then transform @xmath25 into @xmath28 , where @xmath28 is also a graph which have the sum of the number of edges and nodes less than or equal to @xmath9 . graphically we may group the elements of the subgraph , subjected to the application of the local rule , into a region encircled with a dashed closed , simple curve . the edges which cross the curve ( thus connecting the subgraph @xmath25 with the rest of the graph ) will be numbered clockwise . the transformation will affect only the part of the graph which is inside the dashed curve ( inside meaning the bounded connected part of the plane which is bounded by the dashed curve ) and , after the transformation is performed , the edges of the transformed graph will connect to the graph outside the dashed curve by respecting the numbering of the edges which cross the dashed line . however , the grouping of the elements of the subgraph has no intrinsic meaning in graphic lambda calculus . it is just a visual help and it is not a part of the formalism . as a visual help , i shall use sometimes colors in the figures . the colors , as well , do nt have any intrinsic meaning in the graphic lambda calculus . [ [ graphic - beta - move . ] ] 2.1 . graphic @xmath29 move . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + this is the most important move , inspired by the @xmath29-reduction from lambda calculus , see theorem [ lambdathm ] , part ( d ) . the labels `` 1 , 2 , 3 , 4 '' are used only as guides for gluing correctly the new pattern , after removing the old one . as with the encircling dashed curve , they have no intrinsic meaning in graphic lambda calculus . this `` sewing braids '' move will be used also in contexts outside of lambda calculus ! it is the most powerful move in this graphic calculus . a primitive form of this move appears as the re - wiring move ( w1 ) ( section 3.3 , p. 20 and the last paragraph and figure from section 3.4 , p. 21 in @xcite ) . an alternative notation for this move is the following : a move which looks very much alike the graphic beta move is the unzip operation from the formalism of knotted trivalent graphs , see for example the paper @xcite section 3 . in order to see this , let s draw again the graphic beta move , this time without labeling the arrows : the unzip operation acts only from left to right in the following figure . remarkably , it acts on trivalent graphs ( but not oriented ) . let us go back to the graphic beta move and remark that it does not depend on the particular embedding in the plane . for example , the intersection of the `` 1,3 '' arrow with the `` 4,2 '' arrow is an artifact of the embedding , there is no node there . intersections of arrows have no meaning , remember that we work with graphs which are locally planar , not globally planar . the graphic beta move goes into both directions . in order to apply the move , we may pick a pair of arrows and label them with `` 1,2,3,4 '' , such that , according to the orientation of the arrows , `` 1 '' points to `` 3 '' and `` 4 '' points to `` 2 '' , without any node or label between `` 1 '' and `` 3 '' and between `` 4 '' and `` 2 '' respectively . then , by a graphic beta move , we may replace the portions of the two arrows which are between `` 1 '' and `` 3 '' , respectively between `` 4 '' and `` 2 '' , by the pattern from the lhs of the figure . the graphic beta move may be applied even to a single arrow , or to a loop . in the next figure we see three applications of the graphic beta move . they illustrate the need for considering loops and wires as members of @xmath22 . also , we can apply in different ways a graphic beta move , to the same graph and in the same place , simply by using different labels `` 1 '' , ... `` 4 '' ( here @xmath8 , @xmath30 , @xmath26 , @xmath31 are graphs in @xmath18 ) : a particular case of the previous figure is yet another justification for having loops as elements in @xmath22 . these two applications of the graphic beta move may be represented alternatively like this : [ [ co - assoc - move . ] ] 2.2 . ( co - assoc ) move . + + + + + + + + + + + + + + + + + + + + + this is the `` co - associativity '' move involving the @xmath10 graphs . we think about the @xmath10 graph as corresponding to a fan - out gate . by using co - assoc moves , we can move between any two binary trees formed only with @xmath32 gates , with the same number of output leaves . [ [ co - comm - move . ] ] 2.3 . ( co - comm ) move . + + + + + + + + + + + + + + + + + + + + this is the `` co - commutativity '' move involving the @xmath10 gate . it will be not used until the section [ secbraid ] concerning knot diagrams . [ [ a - r1a - move . ] ] 2.3.a ( r1a ) move . + + + + + + + + + + + + + + + + + this move is imported from emergent algebras . explanations are given in section [ semer ] . it involves an @xmath10 graph and a @xmath15 graph , with @xmath33 . [ [ b - r1b - move . ] ] 2.3.b ( r1b ) move . + + + + + + + + + + + + + + + + + the move r1b ( also related to emergent algebras ) is this : [ [ r2-move . ] ] 2.4 . ( r2 ) move . + + + + + + + + + + + + + + + this corresponds to the reidemeister ii move for emergent algebras . it involves an @xmath10 graph and two other : a @xmath15 and a @xmath34 graph , with @xmath35 . this move appears in section 3.4 , p. 21 @xcite , with the supplementary name `` triangle move '' . [ [ ext2-move . ] ] 2.5 . ( ext2 ) move . + + + + + + + + + + + + + + + + + this corresponds to the rule ( ext2 ) from @xmath0-scale calculus , it expresses the fact that in emergent algebras the operation indexed with the neutral element @xmath36 of the group @xmath14 has the property @xmath37 . [ [ local - pruning . ] ] 2.6 . local pruning . + + + + + + + + + + + + + + + + + + + local pruning moves are local moves which eliminate `` dead '' edges . notice that , unlike the previous moves , these are one - way ( you can eliminate dead edges , but not add them to graphs ) . [ [ global - moves - or - conditions . ] ] global moves or conditions . + + + + + + + + + + + + + + + + + + + + + + + + + + + global moves are those which are not local , either because the condition @xmath26 applies to parts of the graph which may have an arbitrary large sum or edges plus nodes , or because after the move the graph @xmath28 which replaces the graph @xmath25 has an arbitrary large sum or edges plus nodes . [ [ ext1-move . ] ] 2.7 . ( ext1 ) move . + + + + + + + + + + + + + + + + + this corresponds to the rule ( ext1 ) from @xmath0-scale calculus , or to @xmath38-reduction in lambda calculus ( see theorem [ lambdathm ] , part ( e ) for details ) . it involves a @xmath0 graph ( think about the @xmath0 abstraction operation in lambda calculus ) and a @xmath11 graph ( think about the application operation in lambda calculus ) . the rule is : if there is no oriented path from `` 2 '' to `` 1 '' , then the following move can be performed . [ [ global - fan - out - move . ] ] 2.8 . ( global fan - out ) move . + + + + + + + + + + + + + + + + + + + + + + + + + + + this is a global move , because it consists in replacing ( under certain circumstances ) a graph by two copies of that graph . the rule is : if a graph in @xmath19 has a @xmath10 bottleneck , that is if we can find a sub - graph @xmath39 connected to the rest of the graph @xmath23 only through a @xmath10 gate , then we can perform the move explained in the next figure , from the left to the right . conversely , if in the graph @xmath23 we can find two identical subgraphs ( denoted by @xmath8 ) , which are in @xmath18 , which have no edge connecting one with another and which are connected to the rest of @xmath23 only through one edge , as in the rhs of the figure , then we can perform the move from the right to the left . remark that ( global fan - out ) trivially implies ( co - comm ) . ( as an local rule alternative to the global fan - out , we might consider the following . fix a number @xmath40 and consider only graphs @xmath41 which have at most @xmath40 ( nodes + arrows ) . the @xmath40 local fan - out move is the same as the global fan - out move , only it applies only to such graphs @xmath41 . this local fan - out move does not imply co - comm . ) [ [ global - pruning . ] ] 2.9 . global pruning . + + + + + + + + + + + + + + + + + + + + this a global move which eliminates `` dead '' edges . the rule is : if a graph in @xmath19 has a @xmath12 ending , that is if we can find a sub - graph @xmath39 connected only to a @xmath12 gate , with no edges connecting to the rest of @xmath23 , then we can erase this graph and the respective @xmath12 gate . the global pruning may be needed because of the @xmath0 gates , which can not be removed only by local pruning . [ [ elimination - of - loops . ] ] 2.10 . elimination of loops . + + + + + + + + + + + + + + + + + + + + + + + + + + + it is possible that , after using a local or global move , we obtain a graph with an arrow which closes itself , without being connected to any node . here is an example , concerning the application of the graphic @xmath29 move . we may erase any such loop , or add one . [ [ lambdagraphs . ] ] @xmath0graphs . + + + + + + + + + + + + + + + + + + + + + + + + + + + + the edges of an elementary graph @xmath0 can be numbered unambiguously , clockwise , by 1 , 2 , 3 , such that 1 is the number of the entrant edge . a graph @xmath19 is a @xmath0-graph , notation @xmath42 , if : 1 . it does not have @xmath15 gates , 2 . for any node @xmath0 any oriented path in @xmath23 starting at the edge 2 of this node can be completed to a path which either terminates in a graph @xmath12 , or else terminates at the edge 1 of this node . the condition @xmath42 is global , in the sense that in order to decide if @xmath42 we have to examine parts of the graph which may have an arbitrary large sum or edges plus nodes . here i show how to associate to a lambda term a graph in @xmath18 , then i use this to show that @xmath29-reduction in lambda calculus transforms into the @xmath29 rule for @xmath18 . ( thanks to morita yasuaki for some corrections . ) indeed , to any term @xmath43 ( where @xmath44 is the set of lambda terms over the variable set @xmath45 ) we associate its syntactic tree . the syntactic tree of any lambda term is constructed by using two gates , one corresponding to the @xmath0 abstraction and the other corresponding to the application . we draw syntactic trees with the leaves ( elements of @xmath45 ) at the bottom and the root at the top . we shall use the following notation for the two gates : at the left is the gate for the @xmath0 abstraction and at the right is the gate for the application . remark that these two gates are from the graphical alphabet of @xmath18 , but the syntactic tree is decorated : at the bottom we have leaves from @xmath45 . also , remark the peculiar orientation of the edge from the left ( in tree notation convention ) of the @xmath0 gate . for the moment , this orientation is in contradiction with the implicit orientation ( from down - up ) of edges of the syntactic tree , but soon this matter will become clear . we shall remove all leaves decorations , with the price of introducing new gates , namely @xmath10 and @xmath12 gates . this will be done in a sequence of steps , detailed further . take the syntactic tree of @xmath43 , drawn with the mentioned conventions ( concerning gates and the positioning of leaves and root respectively ) . we take as examples the following five lambda terms : @xmath46 , @xmath47 , @xmath48 , @xmath49 and @xmath50 . [ [ step-1 . ] ] step 1 . + + + + + + + elimination of bound variables , part i. any leaf of the tree is connected to the root by an unique path . start from the leftmost leaf , perform the algorithm explained further , then go to the right and repeat until all leaves are exhausted . we initialize also a list @xmath51 of bound variables . take a leaf , say decorated with @xmath52 . to this leaf is associated a word ( a list ) which is formed by the symbols of gates which are on the path which connects ( from the bottom - up ) the leaf with the root , together with information about which way , left ( l ) or right ( r ) , the path passes through the gates . such a word is formed by the letters @xmath53 , @xmath54 , @xmath55 , @xmath56 . if the first letter is @xmath53 then add to the list @xmath30 the pair @xmath57 formed by the variable name @xmath16 , and the associated word ( describing the path to follow from the respective leaf to the root ) . then pass to a new leaf . else continue along the path to the roof . if we arrive at a @xmath0 gate , this can happen only coming from the right leg of the @xmath0 gate , thus we can find only the letter @xmath54 . in such a case look at the variable @xmath58 which decorates the left leg of the same @xmath0 gate . if @xmath59 then add to the syntactic tree a new edge , from @xmath58 to @xmath16 and proceed further along the path , else proceed further . if the root is attained then pass to next leaf . examples : the graphs associated to the mentioned lambda terms , together with the list of bound variables , are the following . 1 . @xmath46 has @xmath60 , @xmath47 has @xmath61 , @xmath62 has @xmath63 . + 2 . @xmath49 has @xmath64 , @xmath50 has @xmath64 . + [ [ step-2 . ] ] step 2 . + + + + + + + elimination of bound variables , part ii . we have now a list @xmath30 of bound variables . if the list is empty then go to the next step . else , do the following , starting from the first element of the list , until the list is finished . an element , say @xmath57 , of the list , is either connected to other leaves by one or more edges added at step 1 , or not . if is not connected then erase the variable name with the associated path @xmath65 and replace it by a @xmath12 gate . if it is connected then erase it , replace it by a tree formed by @xmath10 gates , which starts at the place where the element of the list were before the erasure and stops at the leaves which were connected to @xmath16 . erase all decorations which were joined to @xmath16 and also erase all edges which were added at step 1 to the leave @xmath16 from the list . examples : after the step 2 , the graphs associated to the mentioned lambda terms are the following . 1 . the graphs of @xmath46 , @xmath47 , @xmath62 are + 2 . the graphs of @xmath49 , @xmath50 are + remark that at this step the necessity of having the peculiar orientation of the left leg of the @xmath0 gate becomes clear . remark also that there may be more than one possible tree of gates @xmath10 , at each elimination of a bound variable ( in case a bound variable has at least tree occurrences ) . one may use any tree of @xmath10 which is fit . the problem of multiple possibilities is the reason of introducing the ( co - assoc ) move . [ [ step-3 . ] ] step 3 . + + + + + + + we may still have leaves decorated by free variables . starting from the left to the right , group them together in case some of them occur in multiple places , then replace the multiple occurrences of a free variable by a tree of @xmath10 gates with a free root , which ends exactly where the occurrences of the respective variable are . again , there are multiple ways of doing this , but we may pass from one to another by a sequence of ( co - assoc ) moves . examples : after the step 3 , all the graphs associated to the mentioned lambda terms , excepting the last one , are left unchanged . the graph of the last term , changes . 1 . as an illustration , i figure the graphs of @xmath49 , left unchanged by step 3 , and the graph of @xmath50 : + let @xmath66 $ ] be a transformation of a lambda term @xmath8 into a graph @xmath67 $ ] as described previously ( multiple transformations are possible because of the choice of @xmath10 trees ) . then : 1 . for any term @xmath8 the graph @xmath67 $ ] is in @xmath68 , 2 . if @xmath67'$ ] and @xmath67"$ ] are transformations of the term @xmath8 then we may pass from @xmath67'$ ] to @xmath67"$ ] by using a finite number ( exponential in the number of leaves of the syntactic tree of @xmath8 ) of ( co - assoc ) moves , 3 . if @xmath30 is obtained from @xmath8 by @xmath69-conversion then we may pass from @xmath67 $ ] to @xmath70 $ ] by a finite sequence of ( co - assoc ) moves , 4 . let @xmath71 be two terms and @xmath52 be a variable . consider the terms @xmath17 and @xmath72 $ ] , where @xmath72 $ ] is the term obtained by substituting in @xmath8 the free occurrences of @xmath16 by @xmath30 . we know that @xmath29 reduction in lambda calculus consists in passing from @xmath73 to @xmath72 $ ] . then , by one @xmath29 move in @xmath18 applied to @xmath74 $ ] we pass to a graph which can be further transformed into one of @xmath72 $ ] , via ( global fan - out ) moves , ( co - assoc ) moves and pruning moves , 5 . with the notations from ( d ) , consider the terms @xmath8 and @xmath75 with @xmath76 ; then the @xmath38 reduction , consisting in passing from @xmath77 to @xmath8 , corresponds to the ext1 move applied to the graphs @xmath78 $ ] and @xmath67 $ ] . [ lambdathm ] [ [ proof . ] ] proof . + + + + + + \(a ) we have to prove that for any node @xmath0 any oriented path in @xmath67 $ ] starting at the left exiting edge of this node can be completed to a path which either terminates in a graph @xmath12 , or else terminates at the entry peg of this node , but this is clear . indeed , either the bound variable ( of this @xmath0 node in the syntactic tree of @xmath8 ) is fresh , then the bound variable is replaced by a @xmath12 gate , or else , the bound variable is replaced by a tree of @xmath10 gates . no matter which path we choose , we may complete it to a cycle passing by the said @xmath0 node . \(b ) clear also , because the ( co - assoc ) move is designed for passing from a tree of @xmath10 gates to another tree with the same number of leaves . \(c ) indeed , the names of bound variables of @xmath8 do not affect the construction of @xmath67 $ ] , therefore if @xmath30 is obtained by @xmath69-conversion of @xmath8 , then @xmath70 $ ] differs from @xmath67 $ ] only by the particular choice of trees of @xmath10 gates . but this is solved by ( co - assoc ) moves . \(d ) this is the surprising , maybe , part of the theorem . there are two cases : @xmath16 is fresh for @xmath8 or not . if @xmath16 is fresh for @xmath8 then in the graph @xmath79 $ ] the name variable @xmath16 is replaced by a @xmath12 gate . if not , then all the occurrences of @xmath16 in @xmath8 are connected by a @xmath10 tree with root at the left peg of the @xmath0 gate where @xmath16 appears as a bound variable . in the case when @xmath16 is not fresh for @xmath8 , we see in the lhs of the figure the graph @xmath74 $ ] ( with a remanent decoration of `` x '' ) . we perform a graphic ( @xmath29 ) move and we obtain the graph from the right . this graph can be transformed into a graph of @xmath72 $ ] via ( global fan - out ) and ( co - assoc ) moves . the case when @xmath16 is fresh for @xmath8 is figured next . we see that the graph obtained by performing the graphic ( @xmath29 ) move is the union of the graph of @xmath8 and the graph of @xmath30 with a @xmath12 gate added at the root . by pruning we are left with the graph of @xmath8 , which is consistent to the fact that when @xmath16 is fresh for @xmath8 then @xmath73 transforms by @xmath29 reduction into @xmath8 . \(e ) in the next figure we see at the lhs the graph @xmath80 $ ] and at the rhs the graph @xmath67 $ ] . the red asterisk marks the arrow which appears in the construction @xmath80 $ ] from the variable @xmath16 , taking into account the hypothesis @xmath76 . we have a pattern where we can apply the ext1 move and we obtain @xmath67 $ ] , as claimed . @xmath81 as an example , let us manipulate the graph of @xmath49 : we can pass from the lhs figure to the rhs figure by using a graphic ( @xmath29 ) move . conversely , we can pass from the rhs figure to the lhs figure by using a ( global fan - out ) move . these manipulations correspond to the well known fact that @xmath82 is left unchanged after @xmath29 reduction : let @xmath83 , then @xmath84 . [ [ s - k - and - i - combinators - in - graph . ] ] @xmath85 , @xmath86 and @xmath87 combinators in @xmath18 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the combinators @xmath46 , @xmath88 and @xmath62 have the following correspondents in @xmath18 , denoted by the same letters : \(a ) by one graphic ( @xmath29 ) move @xmath89 transforms into @xmath8 , for any @xmath39 with one output . \(b ) by two graphic ( @xmath29 ) moves , followed by a global pruning , for any @xmath90 with one output , the graph @xmath91 transforms into @xmath8 . \(c ) by five graphic ( @xmath29 ) moves , followed by one local pruning move , the graph @xmath92 transforms into @xmath87 . \(d ) by three graphic ( @xmath29 ) moves followed by a ( global fan - out ) move , for any @xmath93 with one output , the graph @xmath94 transforms into the graph @xmath95 . [ pcombi ] [ [ proof.-1 ] ] proof . + + + + + + the proof of ( b ) is given in the next figure . the proof of ( c ) is given in the following figure . \(a ) and ( d ) are left to the interested reader . the manipulations of graphs presented in this section can be applied for graphs which represent lambda terms . however , they can also be applied for graphs which do not represent lambda terms . [ [ fixed - points . ] ] fixed points . + + + + + + + + + + + + + let s start with a graph @xmath39 , which has one distinguished input and one distinguished output.i represent this as follows . for any graph @xmath96 with one output , we denote by @xmath97 the graph obtained by grafting the output of @xmath96 to the input of @xmath98 . i want to find @xmath96 such that @xmath99 , where @xmath100 means any finite sequence of moves in graphic lambda calculus . i call such a graph @xmath30 a fixed point of @xmath8 . the solution of this problem is the same as in usual lambda calculus . we start from the following succession of moves : this is very close to the solution , we only need a small modification : [ [ grafting - application - or - abstraction ] ] grafting , application or abstraction ? + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + if the @xmath8 , @xmath30 from the previous paragraph were representing lambda terms , then the natural operation between them is not grafting , but the application . or , in graphic lambda calculus the application it s represented by an elementary graph , therefore @xmath101 ( seen as the term in lambda calculus which is obtained as the application of @xmath8 to @xmath30 ) is not represented as a grafting of the output of @xmath30 to the input of @xmath8 . we can easily transform grafting into the application operation . suppose that @xmath8 and @xmath30 are graphs representing lambda terms , more precisely suppose that @xmath8 is representing a term ( denoted by @xmath8 too ) and it s input represents a free variable @xmath16 of the term @xmath8 . then the grafting of @xmath30 to @xmath8 is the term @xmath72 $ ] and the graph from the right is representing @xmath102 , therefore both graphs are representing terms from lambda calculus . we can transform grafting into something else : this has no meaning in lambda calculus , but excepting the orientation of one of the arrows of the graph from the right , it looks like if the abstraction gate ( the @xmath0 gate ) plays the role of an application operation . [ [ zippers - and - combinators - as - half - zippers . ] ] zippers and combinators as half - zippers . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + let s take @xmath103 a natural number and let s consider the following graph in @xmath104 , called the n - zipper : at the left is the n - zipper graph ; at the right is a notation for it , or a `` macro '' . the zipper graph is interesting because it allows to perform ( nontrivial ) graphic beta moves in a fixed order . in the following picture is figured in red the place where the first graphic beta move is applied . in terms of zipper notation this graphic beta move has the following appearance : we see that a n - zipper transforms into a ( n-1)-zipper plus an arrow . we may repeat this move , as long as we can . this procedure defines a `` zipper move '' : we may see the 1-zipper move as the graphic beta move , which transforms the 1-zipper into two arrows . the combinator @xmath105 satisfies the relation @xmath106 . in the next figure it is shown that @xmath107 ( figured in green ) , when applied to @xmath8 , is just a half of the 1-zipper , with an arrow added ( figured in blue ) . by opening the zipper we obtain @xmath108 , as it should . the combinator @xmath109 satisfies @xmath110 . in the next figure the combinator @xmath111 ( in green ) appears as half of the 2-zipper , with one arrow and one termination gate added ( in blue ) . after opening the zipper we obtain a pair made by @xmath108 and @xmath112 which gets the termination gate on top of it . a global pruning move sends @xmath30 to the trash bin . finally , the combinator @xmath113 satisfies @xmath114 . the combinator @xmath115 ( in green ) appears to be made by half of the 3-zipper , with some arrows and also with a `` diamond '' added ( all in blue ) . interestingly , the diamond looks alike the ones from the emergent algebra sector , definition [ defemers ] . expressed with the help of zippers , the relation @xmath116 appears like this . [ [ lists - and - currying . ] ] lists and currying . + + + + + + + + + + + + + + + + + + + with the help of zippers , we may enhance the procedure of turning grafting into the application operation . we have a graph @xmath117 which has one output and several inputs . we use an n - zipper in order to clip the inputs with the output . this graph is , in fact , the following one . we may interpret the graph inside the green dotted rectangle as the currying of @xmath108 , let s call him @xmath118 . this graph has only one output and no inputs . the graph inside the red dotted rectangle is almost a list . we shall transform it into a list by using again a zipper and one graphic beta move . [ [ packing - arrows . ] ] packing arrows . + + + + + + + + + + + + + + + we may pack several arrows into one . i describe first the case of two arrows . we start from the following sequence of three graphic beta moves . with words , this figure means : we pack the 1 , 2 , entries into a list , we pass it trough one arrow then we unpack the list into the outputs 3 , 4 . this packing - unpacking trick may be used of course for more than a pair of arrows , in obvious ways , therefore it is not a restriction of generality to write only about two arrows . we may apply the trick to a pair of graphs @xmath108 and @xmath112 , which are connected by a pair of arrows , like in the following figure . with the added packing and unpacking triples of gates , the graphs @xmath108 , @xmath112 are interacting only by the intermediary of one arrow . in particular , we may use this trick for the elementary gates of abstraction and application , transforming them into graphs with one input and one output , like this : if we use the elementary gates transformed into graphs with one input and one output , the graphic beta move becomes this almost algebraic , 1d rule : with such procedures , we may transform any graph in @xmath18 into a 1d string of graphs , consisting of transformed elementary graphs and packers and un - packers of arrows , which could be used , in principle , for transforming graphic lambda calculus into a text programming language . emergent algebras @xcite @xcite are a distillation of differential calculus in metric spaces with dilations @xcite . this class of metric spaces contain the `` classical '' riemannian manifolds , as well as fractal like spaces as carnot groups or , more general , sub - riemannian or carnot - carathodory spaces , bellache @xcite , gromov @xcite , endowed with an intrinsic differential calculus based on some variant of the pansu derivative @xcite . in @xcite section 4 `` binary decorated trees and dilatations '' , i propose a formalism for making easy various calculations with dilation structures . this formalism works with moves acting on binary decorated trees , with the leaves decorated with elements of a metric space . here is an example of the formalism . the moves are ( with same names as those used in graphic lambda calculus , see the explanation further ) : define the following graph ( and think about it as being the graphical representation of an operation @xmath119 with respect to the basepoint @xmath16 ) : then , in the binary trees formalism i can prove , by using the moves r1a , r2a , the following `` approximate '' associativity relation ( it is approximate because there appear a basepoint which is different from @xmath16 , but which , in the geometric context of spaces with dilations , is close to @xmath16 ) : it was puzzling that in fact the formalism worked without needing to know which metric space is used . moreover , reasoning with moves acting on binary trees gave proofs of generalizations of results from sub - riemannian geometry , while classical proofs involve elaborate calculations with pseudo - differential operators . at a close inspection it looked like somewhere in the background there is an abstract nonsense machine which is just applied to the particular case of sub - riemannian spaces . in this paper i shall take the following pure algebraic definition of an emergent algebra ( compare with definition 5.1 @xcite ) , which is a stronger version of the definition 4.2 @xcite of a @xmath14 idempotent right quasigroup , in the sense that here i define a @xmath14 idempotent quasigroup . let @xmath14 be a commutative group with neutral element denoted by @xmath36 and operation denoted multiplicatively . a @xmath14 idempotent quasigroup is a set @xmath45 endowed with a family of operations @xmath120 , indexed by @xmath33 , such that : 1 . for any @xmath121 the pair @xmath122 is an idempotent quasigroup , i.e. for any @xmath123 the equations @xmath124 and @xmath125 have unique solutions and moreover @xmath126 for any @xmath52 , 2 . the operation @xmath127 is trivial : for any @xmath128 we have @xmath129 , 3 . for any @xmath130 and any @xmath35 we have : @xmath131 . [ defirq ] here are some examples of @xmath14 idempotent quasigroups . [ [ example-1 . ] ] example 1 . + + + + + + + + + + real ( or complex ) vector spaces : let @xmath45 be a real ( complex ) vector space , @xmath132 ( or @xmath133 ) , with multiplication as operation . we define , for any @xmath33 the following quasigroup operation : @xmath134 . these operations give to @xmath45 the structure of a @xmath14 idempotent quasigroup . notice that @xmath135 is the dilation based at @xmath16 , of coefficient @xmath13 , applied to @xmath58 . [ [ example-2 . ] ] example 2 . + + + + + + + + + + contractible groups : let @xmath23 be a group endowed with a group morphism @xmath136 . let @xmath137 with the operation of addition of integers ( thus we may adapt definition [ defirq ] to this example by using `` @xmath138 '' instead of `` @xmath139 '' and `` @xmath140 '' instead of `` @xmath36 '' as the name of the neutral element of @xmath14 ) . for any @xmath141 let @xmath142 . this a @xmath143 idempotent quasigroup . the most interesting case is the one when @xmath144 is an uniformly contractive automorphism of a topological group @xmath23 . the structure of these groups is an active exploration area , see for example @xcite and the bibliography therein . a fundamental result here is siebert @xcite , which gives a characterization of topological connected contractive locally compact groups as being nilpotent lie groups endowed with a one parameter family of dilations , i.e. almost carnot groups . [ [ example-3 . ] ] example 3 . + + + + + + + + + + a group with an invertible self - mapping @xmath136 such that @xmath145 , where @xmath146 is the identity of the group @xmath23 . it looks like example 2 but it shows that there is no need for @xmath144 to be a group morphism . [ [ local - versions . ] ] local versions . + + + + + + + + + + + + + + + we may accept that there is a way ( definitely needing care to well formulate , but intuitively clear ) to define a local version of the notion of a @xmath14 idempotent quasigroup . with such a definition , for example , a convex subset of a real vector space gives a local @xmath147 idempotent quasigroup ( as in example 1 ) and a neighbourhood of the identity of a topological group @xmath23 , with an identity preserving , locally defined invertible self map ( as in example 3 ) gives a @xmath143 local idempotent quasigroup . [ [ example-4 . ] ] example 4 . + + + + + + + + + + a particular case of example 3 is a lie group @xmath23 with the operations defined for any @xmath148 by @xmath149 . [ [ example-5 . ] ] example 5 . + + + + + + + + + + a less symmetric example is the one of @xmath45 being a riemannian manifold , with associated operations defined for any @xmath148 by @xmath150 , where @xmath151 is the metric exponential . [ [ example-6 . ] ] example 6 . + + + + + + + + + + more generally , any metric space with dilations is a local idempotent ( right ) quasigroup . [ [ example-7 . ] ] example 7 . + + + + + + + + + + one parameter deformations of quandles . a quandle is a self - distributive quasigroup . take now a one - parameter family of quandles ( indexed by @xmath33 ) which satisfies moreover points 2 . and 3 . from definition [ defirq ] . what is interesting about this example is that quandles appear as decorations of knot diagrams @xcite @xcite , which are preserved by the reidemeister moves ( more on this in the section [ secbraid ] ) . at closer examination , examples 1 , 2 are particular cases of one parameter quandle deformations ! i define now the operations of approximate sum and approximate difference associated to a @xmath14 idempotent quasigroup . for any @xmath152 we give the following names to several combinations of operations of emergent algebras : 1 . the approximate sum operation is @xmath153 @xmath154 , 2 . the approximate difference operation is @xmath155 , 3 . the approximate inverse operation is @xmath156 . let s see what the approximate sum operation is , for example 1 . @xmath157 it is clear that , as @xmath13 converges to @xmath140 , this becomes the operation of addition in the vector space with @xmath16 as neutral element , so it might be said that is the operation of addition of vectors in the tangent space at @xmath16 , where @xmath16 is seen as an element of the affine space constructed over the vector space from example 1 . this is a general phenomenon , which becomes really interesting in non - commutative situations , i.e. when applied to examples from the end of the provided list . these approximate operations have many algebraic properties which can be found by the abstract nonsense of manipulating binary trees . another construction which can be done in emergent algebras is the one of taking finite differences ( at a high level of generality , not bonded to vector spaces ) . let @xmath158 be a function ( from @xmath45 to itself , for simplicity ) . the finite difference function associated to @xmath8 , with respect to the emergent algebra over @xmath45 , at a point @xmath52 is the following . @xmath159 [ defpan ] for example 1 , the finite difference has the expression : @xmath160 which is a finite difference indeed . in more generality , for example 2 this definition leads to the pansu derivative @xcite . finite differences as defined here behave like discrete versions of derivatives . again , the proofs consist in manipulating well chosen binary trees . all this can be formalized in graphic lambda calculus , thus transforming the proofs into computations inside graphic lambda calculus . i shall not insist more on this , with the exception of describing the emergent algebra sector of graphic lambda calculus . for any @xmath161 , the following graphs in @xmath104 are introduced : 1 . the approximate sum graph @xmath162 1 . the approximate difference graph @xmath163 1 . the approximate inverse graph @xmath164 [ defemers ] let @xmath108 be a set of symbols @xmath165 . ( these symbols will play the role of scale parameters going to @xmath166 . ) with @xmath108 and with the abelian group @xmath167 we construct a larger abelian group , call it @xmath168 , which is generated by @xmath108 and by @xmath167 . now we introduce the emergent algebra sector ( over the set @xmath108 ) . @xmath169 is the subset of @xmath104 ( over the group @xmath168 ) which is generated by the following list of gates : 1 . arrows and loops , 2 . @xmath170 gate and the gates @xmath171 for any @xmath161 , 3 . the approximate sum gate @xmath172 and the approximate difference gate @xmath173 , for any @xmath174 , with the operations of linking output to input arrows and with the following list of moves : 1 . fan - out moves 2 . emergent algebra moves for the group @xmath168 , 3 . < pruning moves . the set @xmath169 with the given list of moves is called the emergent algebra sector over the set @xmath108 . [ defemerse ] the approximate inverse is not included into the list of generating gates . that is because we can prove easily that for any @xmath174 we have @xmath175 . ( if @xmath161 then we trivially have @xmath176 because it is constructed from emergent algebra gates decorated by elements in @xmath167 , which are on the list of generating gates . ) here is the proof : we start with the approximate difference @xmath173 and with an @xmath170 gate and we arrive to the approximate inverse @xmath177 by a sequence of moves , as follows . we proved the following relation for emergent algebras : @xmath178 . this relation appears as a computation in graphic lambda calculus . as for the finite differences , we may proceed as this . a graph @xmath39 , with one input and one output distinguished , is computable with respect to the group @xmath179 if the following graph can be transformed by the moves from graphic lambda calculus into a graph which is made by assembling : 1 . graphs from @xmath180 , 2 . gates @xmath0 , @xmath11 and @xmath12 . [ defemercomp ] it would be interesting to mix the emergent algebra sector with the lambda calculus sector ( in a sense this is already suggested in definition [ defemercomp ] ) . at first view , it seems that the emergent algebra gates @xmath181 are operations which are added to the lambda calculus operations , the latter being more basic than the former . i think this is not the case . in @xcite theorem 3.4 , in the formalism of lambda - scale calculus ( graphic lambda calculus is a visual variant of this ) , i show on the contrary that the emergent algebra gates could be applied to lambda terms and the result is a collection , or hierarchy of lambda calculi , organized into an emergent algebra structure . this is surprising , at least for the author , because the initial goal of introducing lambda - scale calculus was to mimic lambda calculus with emergent algebra operations . in this section we discuss about tangle diagrams and graphic lambda calculus . an oriented tangle is a collection of wired in 3d space , more precisely it is an embedding of a oriented one dimensional manifold in 3d space . two tangles are the same up to topological deformation of the 3d space . an oriented tangle diagram is , visually , a projection of a tangle , in general position , on a plane . more specifically , an oriented tangle diagram is a globally planar oriented graph with 4-valent nodes which represent crossings of wires ( as seen in the projection ) , along with supplementary information about which wire passes over the respective crossing . a locally planar tangle diagram is an oriented graph which satisfies the previous description , with the exception that it is only locally planar . visually , a locally planar tangle diagram looks like an ordinary one , excepting that there may be crossings of edges of the graph which are not tangle crossings ( i.e. nodes of the graph ) . the purpose of this section is to show that we can `` simulate '' tangle diagrams with graphic lambda calculus . this can be expressed more precisely in two ways . the first way is that we can define `` crossing macros '' in graphic lambda calculus , which are certain graphs which play the role of crossings in a tangle diagram ( i.e. we can express the reidemeister moves , described further , as compositions of moves from graphic lambda calculus between such graphs ) . the second way is to say that to any tangle diagram we can associate a graph in @xmath18 such that to any reidemeister move is associated a certain composition of moves from graphic lambda calculus . meredith ad snyder @xcite achieve this goal with the pi - calculus instead of graphic lambda calculus . kauffman , in the second part of @xcite , associates tangle diagrams to combinators and writes about `` knotlogic '' . two tangles are the same , up to topological equivalence , if and only if any tangle diagram of one tangle can be transformed by a finite sequence of reidemeister moves into a tangle diagram of the second tangle . the oriented reidemeister moves are the following ( i shall use the same names as polyak @xcite , but with the letter @xmath182 replaced by the letter @xmath183 ) : in section [ semer ] , example 7 , it is mentioned that there is a connection between tangle diagrams and emergent algebras , via the notion of a quandle . quandles are self - distributive idempotent quasigroups , which were invented as decorations of the arrows of a tangle diagram , which are invariant with respect to the reidemeister moves . let us define the emergent algebra crossing macros . ( we can choose to neglect the @xmath13 decorations of the crossings , or , on the contrary , we can choose to do like in definition [ defemerse ] of the emergent algebra sector , namely to add a set @xmath8 to the group @xmath14 and use even more nuanced decorations for the crossings . ) in @xcite , sections 3 - 6 are dedicated to the use of these crossings for exploring emergent algebras and spaces with dilations . all constructions and reasonings from there can be put into the graphic lambda calculus formalism . here i shall explain only some introductory facts . let us associate to any locally planar tangle diagram @xmath184 a graph in @xmath185 \in graph$ ] , called the translation of @xmath184 , which is obtained by replacing the crossings with the emergent crossing macros ( for a fixed @xmath13 ) . also , to any reidemeister move we associate it s translation in graphic lambda calculus , consisting in a local move between the translations of the lhs and rhs tangles which appear in the respective move . ( note : these translations are not added to the moves which define graphic lambda calculus . ) the translations of all oriented reidemeister moves of type 1 and 2 can be realized as sequences of the following moves from graphic lambda calculus : emergent algebra moves ( r1a , r1b , r2 , ext2 ) , fan - out moves ( i.e. co - comm , co - assoc , global fan - out ) and pruning moves . more precisely the translations of the reidemeister moves r1a , r1b are , respectively , the graphic lambda calculus moves r1a , r1b , modulo fan - out moves . moreover , all translations of reidemeister moves of type 2 can be expressed in graphic lambda calculus with the move r2 , fan - out and pruning moves . [ temercross ] the fact that the reidemeister moves of type 3 are not true for ( the algebraic version of ) the emergent algebras , i.e. that the translations of those can not be expressed as a sequence of moves from graphic lambda calculus , is a feature of the formalism and not a weakness . this is explained in detail in sections 5 , 6 @xcite , but unfortunately at the moment of the writing that article the graphic lambda calculus was not available . it is an interesting goal the one of expressing the constructions from the mentioned sections as statements about the computability in the sense of definition [ defemercomp ] of the translations of certain tangle diagrams . as a justification for this point of view , let us remark that all tangle diagrams which appear in the reidemeister moves of type 3 have translations which are related to the approximate difference or approximate sum graphs from definition [ defemers ] . for example , let s take the translation of the graph from the rhs of the move r3d and call it @xmath31 . this graph has three inputs and three outputs . let s then consider a graph formed by grafting three graphs @xmath8 , @xmath30 , @xmath26 at the inputs of @xmath31 , such that @xmath8 , @xmath30 , @xmath26 are not otherwise connected . then we can perform the following sequence of moves . the graph from the left lower side is formed by an approximate difference , a @xmath15 gate and several @xmath10 gates . therefore , if @xmath8 , @xmath30 , @xmath26 are computable in the sense of definition [ defemers ] then the initial graph ( the translation of the lhs of r3d with @xmath8 , @xmath30 , @xmath26 grafted at the inputs ) is computable too . as previously , we define translations of ( locally planar ) tangle diagrams into graphs in @xmath18 . the class of locally planar tangle diagrams is out in a one - to one correspondence with a class of graphs in @xmath18 , let us call this class @xmath186 . we could proceed in the inverse direction , namely consider the class of graphs @xmath186 , along with the moves : graphic beta move and elimination of loops . then we make the ( inverse ) translation of graphs in @xmath186 into locally planar tangle diagrams and the ( inverse ) translation of the graphic beta move and the elimination of loops . the following proposition explains what we obtain . the class of graphs @xmath186 is closed with respect to the application of the graphic beta move and of the elimination of loops . the translations of the graphic beta and elimination of loops moves are the following splice 1 , 2 ( translation of the graphic beta move ) and loop 1 , 2 ( translation of the elimination of loops ) moves . [ ptrans ] the proposition becomes obvious if we find the translation of the graphic beta move . there is one translation for each crossing . ( likewise , there are two translations for elimination of loops , depending on the orientation of the loop which is added / erased . ) @xmath81 the following theorem clarifies which are the oriented reidemeister moves which can be expressed as sequences of graphic lambda calculus moves applied to graphs in @xmath186 . among these moves , some are more powerful than others , as witnessed by the following all the translations of the oriented reidemeister move into moves between graphs in @xmath186 , excepting r2c , r2d , r3a , r3h , can be realized as sequences of graphic beta moves and elimination of loops . moreover , the translations of moves r2c , r2d , r3a , r3h are equivalent up to graphic beta moves and elimination of loops ( i.e. any of these moves , together with the graphic beta move and elimination of loops , generates the other moves from this list ) . [ threidgen ] it is easy , but tedious , to verify that all the mentioned moves can be realized as sequences of splice and loop moves . it is as well easy to verify that the moves r2c , r2d , r3a , r3h are equivalent up to splice and loop moves . it is not obvious that the moves r2c , r2d , r3a , r3h ca nt be realized as a sequence of splice and loop moves . in order to do this , we prove that r2d ca nt be generated by splice and loop . thanks are due to peter kravchuk for the idea of the proof , given in an answer to a question i asked on mathoverflow @xcite , where i described the moves splice and loop . to any locally planar tangle diagram a associate it s reduced diagram r(a ) , which is obtained by the following procedure : first use splice 1,2 from left to right for all crossings , then use loop 1,2 from right to left in order to eliminate all loops which are present at this stage . notice that : 1 . the order of application of the splice moves does not matter , because they are applied only once per crossing . there is a finite number of splices , equal to the number of crossings . define the bag of splices splice(a ) to be the set of splice moves applied . the same is true for the order of eliminations of loops by loop 1 , 2 . there is a finite number of loop eliminations , because the number of loops ( at this stage ) can not be bigger than the number of edges of the initial diagram . define the bag of loops loop(a ) to be the set of all loops which are present after all splices are done . apply a splice 1,2 move to the initial diagram a , from left to right , and get b. then splice(b ) is what is left in the bag splice(a ) after taking out the respective splice . also loop(b ) = loop(a ) because of the definition of bags of loops . therefore r(a ) = r(b ) . apply a loop1 , 2 from left to right to a and get b. the new loop introduced in the diagram does not participate to any crossing ( therefore splice(a ) = splice(b ) ) , so we find it in the bag of loops of b , which is made by all the elements of loop(a ) and this new loop . therefore r(a ) same goes for loop1 , 2 applied from right to left . finally , remark that the reduced diagram of the lhs of the move r2d is different than the reduced diagram of the rhs of the move r2d , therefore the move r2d can not be achieved with a sequence of splices and loops addition / elimination . @xmath81 m. buliga , braided spaces with dilations and sub - riemannian symmetric spaces , in : geometry . exploratory workshop on differential geometry and its applications , eds . d. andrica , s. moroianu , cluj - napoca 2011 , 21 - 35 , http://arxiv.org/abs/1005.5031[arxiv:1005.5031 ]

the precise quark mass dependence of the one - loop effective action in an instanton background has recently been computed . the result interpolates smoothly between the previously known extreme small and large mass limits . the computational method makes use of the fact that the single instanton background has radial symmetry , so that the computation can be reduced to a sum over partial waves of logarithms of radial determinants , each of which can be computed numerically in an efficient manner . the bare sum over partial waves is divergent and must be regulated and renormalized . in this paper we provide more details of this computation , including both the renormalization procedure and the numerical approach . we conclude with comparisons of our precise numerical results with a simple interpolating function that connects the small and large mass limits , and with the leading order of the derivative expansion . 2.5 cm

hep-th0502087

the computation of fermion determinants in nontrivial background fields is an important challenge for both continuum and lattice quantum field theory . explicit analytic results are known only for very simple backgrounds , and are essentially all variations on the original work of heisenberg and euler @xcite . for applications in quantum chromodynamics ( qcd ) , an important class of background gauge fields are instanton fields , as these minimize the euclidean gauge action within a given topological sector of the gauge field . furthermore , instanton physics has many important phenomenological consequences @xcite . thus , we are led to consider the fermion determinant , and the associated one - loop effective action , for quarks of mass @xmath0 in an instanton background . here , no exact results are known for the full mass dependence , although several terms have been computed analytically in the small mass @xcite and large mass @xcite limits . recently , in @xcite , the present authors presented a new computation which is numerical , but essentially exact , that evaluates the one - loop effective action in a single instanton background , for any value of the quark mass ( and for arbitrary instanton size parameter ) . the result is fully consistent with the known small and large mass limits , and interpolates smoothly between these limits . this could be of interest for the extrapolation of lattice results @xcite , obtained at unphysically large quark masses , to lower physical masses , and for various instanton - based phenomenology . our computational method is simple and efficient , and can be adapted to many other determinant computations in which the background is sufficiently symmetric so that the problem can be reduced to a product of one - dimensional radial determinants . while this is still a very restricted set of background field configurations , it contains many examples of interest , the single instanton being one of the most obvious . it is well known how to compute determinants of _ ordinary _ differential operators @xcite ; but in higher - dimensional problems with _ partial _ differential operators , one must confront the renormalization problem since there are now an infinite number of 1-d determinants to deal with ( even when the partial differential operator has a radial symmetry ) . in this paper we present more details of the results of @xcite . in section [ renormalized ] we define the renormalized effective action in the minimal subtraction scheme , as introduced by t hooft @xcite , and summarize what is known about the small and large mass limits . in section [ radial ] we review how the single instanton background reduces the spectral problem to a set of radial problems , and indicate how to regularize the effective action . this reduces the computation to two parts , one of which is analytic and the other is numerical . the analytic part concerns the renormalization of the effective action , and for this we use a wkb expansion as is developed in our earlier paper @xcite . we stress that this renormalization computation , which constitutes section [ wkbsection ] , is analytic and exact , even though we use a wkb expansion , since we show that only the first two orders of the wkb expansion contribute . section [ numerical ] presents details of the numerical part of the computation and shows how to combine the numerical part of the computation with the renormalization part to obtain the finite renormalized effective action , which is plotted in figure [ fig5 ] . in section [ comparison ] we present a simple interpolating function that has been fit to our results , and we also compare our precise mass dependence with the mass dependence of the leading derivative expansion approximation , which was computed previously in @xcite . the final section contains some concluding comments . in appendix a we give some details ( not fully given in @xcite ) which are needed in the approximate effective action calculation using the wkb phase - shift method . for this wkb analysis , the schwinger proper - time framework @xcite provides a natural way to implement the renormalization procedure consistently . appendix b confirms that the same result is obtained in the regular and singular gauges for the instanton background . an instanton background field is self - dual , and self - dual gauge fields have the remarkable property that the dirac and klein - gordon operators in such a background are isospectral ; that is , they have identical spectra , apart from an extra degeneracy factor of four in the spinor case and zero modes present in the spinor case @xcite . since the one - loop effective action is proportional to the logarithm of the determinant of the respective operator , this has the immediate consequence that it is sufficient to consider the scalar effective action to learn also about the corresponding fermionic effective action , for any mass value m. in particular , for a quark in a background instanton field , the renormalized one - loop effective action of a dirac spinor field of mass @xmath0 ( and isospin @xmath1 ) , @xmath2 , can be related to the corresponding scalar effective action , @xmath3 , for a complex scalar of mass @xmath0 ( and isospin @xmath1 ) by @xcite @xmath4 where @xmath5 is the renormalization scale . the ln term in ( [ susy ] ) corresponds to the existence of a zero eigenvalue in the spectrum of the dirac operator for a single instanton background . the one - loop effective action must be regularized . we choose pauli - villars regularization adapted to the schwinger proper - time formalism , and later we relate this to dimensional regularization , as in the work of t hooft @xcite . the pauli - villars regularized one - loop scalar effective action is @xcite @xmath6.\ ] ] where @xmath7 , with @xmath8 . in ( [ det ] ) , @xmath9 is a heavy regulator mass . we consider an su(2 ) single instanton in the regular gauge @xcite : @xmath10 where @xmath11 are the standard t hooft symbols @xcite . the regularized effective action ( [ det ] ) has the proper - time representation @xmath12 the renormalized effective action , in the minimal subtraction scheme , is defined as @xcite @xmath13 \nonumber \\ & \equiv & \lim_{\lambda\rightarrow\infty } \left[\gamma_{\lambda}^{s}(a;m)-\frac{1}{6 } \ln\left(\frac{\lambda}{\mu}\right)\right]\quad , \label{renaction}\end{aligned}\ ] ] where we have subtracted the charge renormalization counter - term , and @xmath5 is the renormalization scale . by dimensional considerations , we may introduce the modified scalar effective action @xmath14 , which is a function of @xmath15 only , defined by @xmath16 and concentrate on studying the @xmath15 dependence of @xmath14 . then there is no loss of generality in our setting the instanton scale @xmath17 henceforth . , from equation ( ) note the gap in the region @xmath18 , in which the two expansions do not match up .. [ fig1 ] ] it is known from previous works that in the small mass @xcite and large mass @xcite limits , @xmath19 behaves as @xmath20 where @xmath21 and @xmath22 is euler s constant @xcite . the leading behavior of the small mass limit in ( [ masslimit ] ) was first computed by t hooft @xcite , and the next corrections were computed in @xcite and @xcite . this small mass expansion is based on the fact that the massless propagators in an instanton background are known in closed - form @xcite . on the other hand , the large mass expansion in ( [ masslimit ] ) can be computed in several ways . the @xmath23 and @xmath24 terms were computed in @xcite , while the next two terms were computed in @xcite . a very direct approach is to use the small-@xmath25 behavior of @xmath26 , the proper - time function appearing in ( [ ptaction ] ) , as given by the schwinger - dewitt expansion . in our case this expansion reads @xcite @xmath27 and using this series with ( [ ptaction ] ) immediately leads to the large @xmath0 expansion for @xmath19 in ( [ masslimit ] ) . equation ( [ masslimit ] ) summarizes what is known analytically about the mass dependence of the renormalized one - loop effective action in an instanton background . this situation is represented in figure [ fig1 ] , which shows a distinct gap approximately in the region @xmath28 , where the small and large mass expansions do not match up . in this paper we present a technique which computes @xmath19 numerically for _ any _ value of the mass @xmath0 . our results interpolate smoothly between the analytic small and large mass limits depicted in figure [ fig1 ] . our computational approach makes use of the fact that the single instanton background ( [ instanton ] ) has radial symmetry @xcite . this has the important consequence that the computation of the regularized one - loop effective action ( [ ptaction ] ) can be reduced to a sum over partial waves of logarithms of determinants of radial ordinary differential operators . each such radial determinant can be computed by a simple numerical method , described in section [ numerical ] . the physical challenge is to renormalize the ( divergent ) sum over partial waves . in the instanton background ( [ instanton ] ) , with scale @xmath17 , the klein - gordon operator @xmath29 for isospin @xmath1 particles can be cast in the radial form @xcite @xmath30\ , , \label{insth}\ ] ] where @xmath31 , and @xmath32 , and there is a degeneracy factor of @xmath33 for each partial wave characterized by @xmath34-values . [ note that @xmath35 can be identified with the eigenvalue of @xmath36 for @xmath37 , and @xmath38 with the eigenvalue of @xmath39 for @xmath40 . in the absence of the instanton background , the free operator is @xmath41 . \label{freeh}\ ] ] this radial decomposition means that we can express the pauli - villars regularized effective action ( [ det ] ) also as @xmath42 here we have combined the radial determinants for @xmath43 and @xmath44 , which have the common degeneracy factor @xmath45 , so that the sum over @xmath46 and @xmath47 reduces to a single sum over @xmath46 . in our actual analysis , as explained in detail below , we need to consider only a truncated sum over @xmath46 with the expression ( [ pv ] ) , and hence possible ambiguities as regards effecting the infinite sum over @xmath46 become irrelevant . in section [ numerical ] we present a simple and efficient numerical technique for computing each of the radial determinants appearing in ( [ pv ] ) . but to extract the renormalized effective action we need to be able to consider the @xmath48 limit in conjunction with the infinite sum over @xmath46 . this can be achieved as follows . split the @xmath46 sum in ( [ pv ] ) into two parts as : @xmath49 where @xmath50 is a large but finite integer . in the first sum , which is finite , the cutoff @xmath9 may be safely removed since for any given finite @xmath46 @xcite , @xmath51 thus , the first sum in ( [ actionsplit ] ) may be written without the regulator @xmath9 as @xmath52 this sum can be computed numerically , and we find [ see section [ numerical ] ] that for any mass @xmath0 it is quadratically divergent as @xmath53 . this divergence is canceled by a divergence of the second sum in ( [ actionsplit ] ) in the large @xmath50 limit , as we show in the next section . in the second sum in ( [ actionsplit ] ) we can not take the large @xmath50 and large @xmath9 limits blindly , as each leads to a divergence . however , we show in this section that the @xmath9 divergence is precisely of the counterterm form in the renormalized action in ( [ renaction ] ) , and that the large @xmath50 divergence is such that it precisely cancels the large @xmath50 divergences from the large @xmath50 limit of the sum in ( [ lj ] ) . thus we obtain a finite renormalized effective action . the advantage of this technique is that the large @xmath9 and large @xmath50 divergences of the second sum in ( [ actionsplit ] ) can be computed _ analytically _ , using the wkb approximation for the corresponding determinants . the wkb approach to radial determinants was derived in @xcite up to third order in the wkb approximation , and the relevant results are reviewed below , and in the appendix . it turns out that in the large @xmath9 and large @xmath50 limits we only need up to the second order in wkb . the large @xmath50 limit of the second sum in ( [ actionsplit ] ) can then be analyzed using the euler - maclaurin summation formula @xcite . it is convenient to express the second piece by the proper - time representation @xmath54 , \label{second}\ ] ] with @xmath55 . \label{wkbfl}\end{aligned}\ ] ] here the term @xmath56 , obviously related to @xmath57 , can be found using the scattering phase shifts of the schrdinger problem with the radial hamiltonian in ( [ insth ] ) . also , in considering the above infinite sum over partial wave terms , it is now crucial to have the contribution from partial - wave @xmath58 and that from partial wave @xmath59 treated together as a package , as indicated in ( [ wkbfl ] ) . for details on this , readers may consult the appendix . we here only note that , for large @xmath46 , the wkb approximation becomes exact and so in the large @xmath50 limit we can use the wkb approximation ( to an appropriate order ) for @xmath60 . according to the wkb expressions derived in the appendix , it is found that , for each @xmath46 , @xmath61 has a local expansion in terms of the langer - modified @xcite potential , @xmath62 , and the corresponding langer - modified free potential , @xmath63 : @xmath64 specifically , the first order wkb result for @xmath61 is obtained by using the form @xmath65 - \frac{1}{2\sqrt{\pi s}}\ , \exp\left[-s \tilde{v}_{l}(r)\right ] \,\ , . \label{wkbfirst}\end{aligned}\ ] ] in the right hand side of ( [ wkbfl ] ) , and the second order wkb by using @xmath66\nonumber\\ & & -\frac{1}{2\sqrt{\pi s } } \left(\frac{s}{4 r^2}-\frac{s^2}{12 } \frac{d^2 \tilde{v}_{l}}{dr^2}\right ) \ , \exp\left[-s \tilde{v}_{l}(r)\right]\,\ , . \label{wkbsecond}\end{aligned}\ ] ] the corresponding expression for the third order of wkb is also given in the appendix , but this result is not needed for our present purposes . an important observation ( which holds true to any order in the wkb expansion ) is that the @xmath46 dependence in @xmath67 has the form of a polynomial in @xmath46 multiplied by an exponential in which @xmath46 appears quadratically . thus , if we use the the euler - maclaurin expansion @xcite for ( [ wkbem ] ) @xmath68 all terms in this expansion , including the integral term , can be computed analytically . the integral term yields an error function @xcite @xmath69=\sqrt{\frac{\pi}{4a}}\ , e^{b^2/a}\ , { \rm erfc}\left(\frac{a l+b}{\sqrt{a}}\right)\quad , \label{efrc}\end{aligned}\ ] ] where @xmath70 . we here remark that it is important to perform the @xmath46-sum prior to considering the @xmath71-integration . for more details on the issue of the integration order , see the appendix . to compute the sum in ( [ second ] ) we still need to perform the proper - time integral over @xmath25 as well as the radial integral over @xmath71 appearing in the wkb expression ( [ wkbem ] ) . to achieve this , we adopt the following procedure . first , we trade the regulator mass @xmath9 for a dimensional regularization parameter @xmath72 , by demanding that @xmath73 = \int_0^\infty ds \left[-\frac{1}{s}\left ( e^{-m^2 s}\,s^{\epsilon}\right ) f_l(s ) \right ] . \label{dimreg}\end{aligned}\ ] ] then , from the facts that @xmath74 as @xmath75 , and @xmath76 for small @xmath25 , we see that ( [ dimreg ] ) requires @xmath77 thus , the correspondence between @xmath72 and @xmath9 is @xmath78 we now proceed to do the proper - time and radial integrals as follows . first , for @xmath50 very large , it becomes convenient to rescale variables as @xmath79 and @xmath80 . second , we expand all terms ( except the @xmath81 factor ) in decreasing powers of large @xmath50 . then the @xmath82 integral can be performed in closed form , yielding incomplete gamma functions @xcite . these can be further expanded for large @xmath50 , after which the @xmath83 ( that is , the radial ) integral can be done . it is straightforward to perform these operations using mathematica . to zeroth order in @xmath72 , the results for the first and second order of wkb are given below . for the first order wkb term , the result is @xmath84 \left(\sum_{l = l+\frac{1}{2}}^\infty f_l^{(1)}(s , r)\right ) \nonumber \\ & = & \frac{1}{24\epsilon } + 2l^2 + 4l- \frac{\ln l}{2}\left(\frac{1}{6}+m^2\right ) + \frac{119}{72 } - \frac{\ln 2}{12}+ \frac{\psi(\frac{1}{2})}{24 } \nonumber \\ & & + \frac{m^2}{2 } \left(1 - 2\ln 2+\ln m\right ) -\frac{1 + 6\,m^2}{12\,l } + o\left(\frac{1}{l^2}\right ) , \label{wkb1}\end{aligned}\ ] ] where @xmath85 . for the second order wkb term , the result is @xmath84 \left ( \sum_{l = l+\frac{1}{2}}^\infty f_l^{(2)}(s , r)\right ) \nonumber \\ & = & \frac{1}{24\epsilon}-\frac{\ln l}{12 } + \frac{1}{9 } - \frac{\ln 2}{12}+ \frac{\psi(\frac{1}{2})}{24 } - \frac{1}{12\,l } + o\left(\frac{1}{l^2}\right ) . \label{wkb2}\end{aligned}\ ] ] the third order wkb term gives a contribution of at most @xmath86 , and has no @xmath72 pole . similarly , it can be shown that all higher order wkb terms have no @xmath72 pole , and vanish for large @xmath50 . we can thus compute the large @xmath50 limit by considering only the relevant parts of the first two wkb expressions in ( [ wkb1 ] ) and ( [ wkb2 ] ) . inserting the identification between @xmath72 and @xmath9 in ( [ corr ] ) , we obtain @xmath87+o\left(\frac{1}{l}\right ) \label{div}\end{aligned}\ ] ] we can now identify the physical role of the various terms in ( [ div ] ) . the first term is the expected logarithmic counterterm which is subtracted in ( [ renaction ] ) , and explains the origin of the @xmath88 term in ( [ modaction ] ) . the next three terms give quadratic , linear and logarithmic divergences in @xmath50 . we shall show in the next section that these divergences cancel corresponding divergences in the first sum in ( [ actionsplit ] ) , which were found in our numerical data . it is a highly nontrivial check on this wkb computation that these divergent terms have the correct coefficients to cancel these divergences . note that the @xmath89 coefficient , and the finite term , are mass dependent . thus , the minimally subtracted renormalized effective action @xmath90 , defined in ( [ modaction ] ) , is @xmath91\right\}\quad , \label{answer}\end{aligned}\ ] ] where the first sum is to be computed numerically from the partial wave expansion in ( [ lj ] ) . in this section we describe the numerical technique for computing the radial determinants which enter the partial wave expansion in ( [ lj ] ) . these one - dimensional determinants can be computed efficiently using the following result @xcite . suppose @xmath92 and @xmath93 are two second order ordinary differential operators on the interval @xmath94 , with dirichlet boundary conditions assumed . then the ratio of the determinants of @xmath95 and @xmath93 is given by @xmath96 where @xmath97 ( for @xmath98 ) satisfies the initial value problem @xmath99 since an initial value problem is very simple to solve numerically , this theorem provides an efficient way to compute the determinant of an ordinary differential operator . note in particular that no direct information about the spectrum ( either bound or continuum states , or phase shifts ) is required in order to compute the determinant . we can simplify the numerical computation further . note that for the free massive klein - gordon partial - wave operator , @xmath100 ( with @xmath101 given in ( [ freeh ] ) ) , the solution to ( [ ode ] ) is the modified bessel function @xcite @xmath102 this solution grows exponentially fast at large @xmath71 , as do the numerical solutions to ( [ ode ] ) for the operators @xmath103 , with @xmath104 specified in ( [ insth ] ) . thus , it is numerically better to consider the ode satisfied by the _ ratio _ of the two functions @xmath105 this quantity has a finite value in the large @xmath71 limit , which is just the ratio of the determinants as in ( [ theorem ] ) . the boundary conditions for the ratio function are @xmath106 a similar idea of considering the ratio function was used by baacke and lavrelashvili in their analysis of metastable vacuum decay @xcite . it is worthwhile making a brief side comment about the boundary conditions at @xmath107 . the two functions on the right hand side of ( [ ratio ] ) do not necessarily satisfy the boundary conditions at @xmath107 in ( [ ode ] ) . for example , when @xmath108 , @xmath109 does not vanish at @xmath107 . and in the massless case this issue is more serious . however , since only the ratio is important , one can introduce an ultraviolet regulator by imposing the boundary conditions in ( [ ode ] ) at @xmath110 , for @xmath111 small but nonzero . then the free solution satisfying the boundary conditions is a linear combination of the two modified bessel functions @xmath112 and @xmath113 . it is straightforward to show that the differential equation governing the ratio function , and the asymptotic ( @xmath114 ) value of the ratio function are independent of @xmath111 as @xmath115 . in fact , since we are ultimately interested in the logarithm of the determinant , it is more convenient ( and more stable numerically ) to consider the logarithm of the ratio , i.e. , @xmath116 which satisfies the differential equation @xmath117 with boundary conditions @xmath118 the ` potential ' term @xmath119 in ( [ nonlinear ] ) is given by @xmath120 to illustrate the computational method , in figure [ fig2 ] we plot @xmath121 and @xmath122 for various values of @xmath46 , with mass value @xmath123 ( which is in the region in which neither the large nor small mass expansions is accurate ) . note that the curves quickly reach an asymptotic large-@xmath71 constant value , and also notice that the contributions from @xmath124 and @xmath125 almost cancel one another when summed . this behavior is generic for all values of mass @xmath0 . dependence of @xmath121 and @xmath122 , solutions of the nonlinear differential equation in ( ) , for @xmath123 , and for @xmath126 . the upper curves are for @xmath121 , while the lower ones are for @xmath122 . note that the curves quickly reach an asymptotic large-@xmath71 constant value , and also notice that the contributions from @xmath124 and @xmath125 almost cancel one another when summed . [ fig2 ] ] to obtain very high precision for @xmath127 in the numerical computation , it proves useful to make a further numerical modification . for large @xmath71 , a good first approximation to @xmath128 is provided by neglecting the first two terms on the left - hand - side of the differential equation in ( [ nonlinear ] ) . thus we define a new function @xmath129 by @xmath130 this new function @xmath129 satisfies the modified equation @xmath131 with the boundary conditions : @xmath132 . numerical values for the quantities defined in ( [ stot ] ) provide greater accuracy at large @xmath71 , and are also better for large @xmath46 . in fact , we can iterate this type of transformation as many times as we wish . for our computation we achieved excellent numerical precision by iterating this transformation twice . , @xmath125 , and their sum @xmath133 , defined in ( ) , for @xmath123 . note that @xmath124 and @xmath125 almost cancel , with their sum @xmath133 vanishing at large @xmath46 . see also figure . ] dependence of @xmath133 , for @xmath123 . ( this is a blow - up of the @xmath133 data from figure ) . @xmath133 behaves like @xmath134 for large @xmath46 . note that this implies that the sum over @xmath46 in ( ) has quadratic divergences in the large @xmath50 limit . [ fig4 ] ] the large @xmath71 values of @xmath121 and @xmath122 can be extracted with very good precision ( we integrated out to @xmath135 ) . notice that the asymptotic values of @xmath121 and @xmath122 very nearly cancel one another , as illustrated in figure [ fig3 ] . this behavior occurs for all @xmath0 , and becomes more accurate as @xmath46 increases . in fact , for a given mass , it is found that , as a function of @xmath46 , @xmath124 grows like @xmath136 while @xmath125 decreases like @xmath137 . this divergence cancels in the sum , resulting in the behavior : @xmath138 this behavior is illustrated in figure [ fig4 ] , which is a blow - up of the @xmath133 data in figure [ fig3 ] . recall from ( [ lj ] ) that the first numerical sum in ( [ answer ] ) is in fact a sum over @xmath133 , with degeneracy factors : @xmath139 thus , we can rewrite our final expression ( [ answer ] ) for the minimally subtracted renormalized effective action @xmath90 as @xmath140\right\}. \label{answer2}\end{aligned}\ ] ] the first sum is over terms that are computed numerically , as described above . the rest represents renormalization terms which have been computed using minimal subtraction and wkb . since the degeneracy factor @xmath45 is quadratic , the large @xmath46 behavior of @xmath133 indicated in ( [ pl ] ) [ and plotted in figure [ fig4 ] ] shows that in the large @xmath50 limit , the sum ( [ lj1 ] ) has potentially divergent terms going like @xmath141 , @xmath50 and @xmath89 , as well as terms finite and vanishing for large @xmath50 . remarkably , we find that these divergent terms are exactly canceled by the divergent large @xmath50 terms found in the previous section for the second sum in ( [ actionsplit ] ) . thus , the renormalized effective action @xmath90 calculated using ( [ answer2 ] ) is finite , converges for large @xmath50 , and can be computed for any mass @xmath0 . we found excellent convergence with @xmath142 in our numerical data , combined with richardson extrapolation @xcite . in figure [ fig5 ] we plot these results for @xmath19 , and compare them with the analytic small and large mass expansions in ( [ masslimit ] ) . the agreement is spectacular . thus , our expression ( [ answer2 ] ) provides a simple and numerically precise interpolation between the large mass and small mass regimes . from ( ) , compared with the analytic extreme small and large mass limits [ dashed curves ] from ( ) . the dots denote numerical data points from ( ) , and the solid line is a fit through these points . the agreement with the analytic small and large mass limits is very precise . [ fig5 ] ] as an interesting check , our formula ( [ answer2 ] ) provides a very simple computation of t hooft s leading small mass result . in the @xmath143 , ( [ nonlinear ] ) becomes @xmath144 one can find the solution of this equation in analytic form : @xmath145 + \ln\left[\sqrt{1+r^2}+\frac{1}{2l+1}\frac{1}{\sqrt{1+r^2}}\right ] \\ s_{(l+\frac{1}{2},l)}(r)&=&-\ln\left[\sqrt{1+r^2}\right ] . \label{masslesssol}\end{aligned}\ ] ] as @xmath114 , each quantity diverges but the sum , @xmath133 , has a finite value @xmath146 . \label{exact}\end{aligned}\ ] ] then it follows that @xmath147 which agrees precisely with the leading term ( [ alphahalf ] ) in the small mass limit in ( [ masslimit ] ) . since this is the first computation of the full mass dependence of the one - loop effective action in an instanton background , there is not much with which we can compare , except the small and large mass limits ( [ masslimit ] ) , which agree very well . there are , however , two other comparisons worth making . the first is with a modified pad interpolating fit proposed in @xcite , which is consistent with the two leading terms in each of the known analytic small and large mass limits given in ( [ masslimit ] ) : @xmath148 with @xmath149 and @xmath150 . based on the numerical data found in the present paper , we can fit the exact mass dependence in figure [ fig5 ] with an expanded form of this interpolating function . let us assume the form @xmath151 one may easily check that the leading two terms of the small mass expansion of this expression ( [ interpol2 ] ) is the same as the small mass expansion in ( [ masslimit ] ) . then , comparing the four leading terms of the large mass expansion of ( [ interpol2 ] ) with the large mass expansion of ( [ masslimit ] ) fixes the coefficients @xmath152 to be : @xmath153 there remain two free parameters , @xmath154 and @xmath155 , unfixed in ( [ interpol2 ] ) . we can choose them so that the pad approximant in ( [ interpol2 ] ) best fits the numerical data found in the previous section . this is a straightforward numerical exercise , and we find the best fit is given by @xmath156 in fig . [ fig6 ] , we compare these approximations with the precise numerical data . note that the fit based on ( [ interpol2 ] ) [ solid line ] is extremely precise , so we can use ( [ interpol2 ] ) as a simple analytic expression approximating the full mass dependence of the effective action , over the entire range of mass values . this is analogous to modified pad fits used in chiral extrapolation of lattice data @xcite , and has also been explored for heisenberg - euler effective actions @xcite . this form will also be useful if one wishes to capture the full @xmath157-dependence ( for a given quark mass value ) of the scalar effective action @xmath158 via ( [ modaction ] ) ( and then also of the fermion effective action @xmath159 via ( [ susy ] ) ) . ) proposed in , while the solid line is the approximation in ( [ interpol2 ] ) which is an interpolating function fit to the exact mass dependence found in this paper . the solid dots are the exact numerical data . ] another comparison we can make is to the derivative expansion approximation . this approximation was already studied in @xcite , where it was noted that it was remarkably close to the extreme small and large mass limits in ( [ masslimit ] ) . now that we have computed the full mass dependence of @xmath90 , it is worth revisiting this comparison . recall that the philosophy of the derivative expansion is to compute the one - loop effective action for a covariantly constant background field , which can be done exactly , and then perturb around this constant background solution . the leading order derivative expansion for the effective action is obtained by first taking the ( exact ) expression for the effective lagrangian in a covariantly constant background , substituting the space - time dependent background , and then integrating over space - time . for an instanton background , which is self - dual , we base our derivative expansion approximation on a covariantly constant and self - dual background @xcite . this leads to the following simple integral representation for the leading derivative expansion approximation to the effective action @xcite @xmath160_{\rm de } & = & -\frac{1}{14}\int_0^\infty \frac{dx\ , x}{e^{2\pi x}-1}\left\ { -84 + 14 \ln\left(1+\frac{48 x^2}{m^4}\right ) + 7\sqrt{3}\,\frac{m^2}{x } { \rm arctan}\left(\frac{4\sqrt{3}\ , x}{m^2}\right ) \right.\nn\\ & & \left . \hskip 3 cm + 768\frac{x^2}{m^4}~_2 f_1\left(1,\frac{7}{4},\frac{11}{4};-\frac{48x^2}{m^4}\right)\right\ } -\frac{1}{6}\ln m \label{dm}\end{aligned}\ ] ] figure [ fig7 ] shows a comparison of this leading derivative expansion expression with the exact numerical data . in the range covered , the agreement is surprisingly good for such a crude approximation . , comparing the leading derivative expansion approximation ( solid line ) with the precise numerical answers ( dots ) . the dashed lines show the small and large mass limits from ( ) . in this paper we have presented the details of a computation of the fermion determinant in an instanton background for all values of the quark mass . the agreement with the known analytic expressions in the small and large mass limits is excellent . as another application of our result , we can reinstate the dependence on @xmath157 , the instanton scale parameter , simply by replacing @xmath0 by @xmath161 . then , given a quark field of fixed mass @xmath0 , the fermion determinant as a function of instanton size @xmath157 can be studied . for phenomenological applications , this can now be simply described with better than 1% numerical accuracy by using the interpolating function ( [ interpol2 ] ) for @xmath162 , together with our formulas ( [ modaction ] ) and ( [ susy ] ) . the resulting interpolating expression for the fermion determinant in an instanton background is @xmath163 this expression assumes minimal subtraction for the renormalized coupling entering the tree - level contribution . to obtain the corresponding expression for @xmath164 in other renormalization schemes , one needs to perform additional finite renormalizations , as discussed in refs.@xcite . notice that such one - loop finite renormalization terms @xmath165 can have dependence on the quark mass @xmath0 ( but not on @xmath157 ) , and the lack of manifest decoupling for large quark mass in the expression ( [ fermiondet ] ) is a renormalization artifact @xcite . we also remark that in instanton - based qcd phenomenology one may well choose the quark mass value @xmath0 in ( [ fermiondet ] ) to be different from the lagrangian ( or current ) quark mass , taking instead some effective mass value @xcite . our formula ( [ fermiondet ] ) can be used for discussing instanton effects in gauge theories with compact extra dimensions as well @xcite . the computational method we described is versatile and can be adapted to a large class of previously insoluble computations of one - loop functional determinants in nontrivial backgrounds in various dimensions of space - time , as long as the spectral problem of the given system can be reduced to that of partial waves . one may especially consider using analogous methods for the computation of quantum corrections to the soliton energy in field theories . several examples along this direction are currently under investigation . .5 cm * acknowledgments : * gd thanks h. gies , v. khemani and k. kirsten for helpful comments , and the us doe for support through the grant de - fg02 - 92er40716 . the work of jh was supported by the brain korea 21 project . the work of cl was supported by the korea science foundation abrl program ( r14 - 2003 - 012 - 01002 - 0 ) . hm thanks the university of connecticut for hospitality during sabbatical , and the university of seoul for support . wkb theory is a powerful tool for obtaining a global approximation to the solution of a second - order ordinary differential equation @xcite . hence one expects that it can be utilized for the approximate calculation of a one - dimensional functional determinant @xcite . in the case of higher - dimensional functional determinants , which are usually needed in the one - loop effective action calculation of field theory , one can still try to use this wkb theory if the relevant partial differential operator becomes separable ( as is often the case with rotationally - invariant background fields ) . in the latter case , however , no useful result can be derived from such analysis if one does not have an unambiguous renormalization procedure that goes with the wkb theory . in fact , the usual leading - order wkb theory is not sufficient for the determinant calculation if a consistent renormalization demands the contribution from higher - order wkb approximation to be included . we shall see below that this is the case . the renormalization problem mentioned above has been solved in our earlier paper @xcite , by including needed higher - order wkb contributions within the schwinger proper - time representation @xcite for the effective action . also achieved there is a generalization of the schwinger - dewitt small proper - time expansion @xcite to the appropriate expression for arbitrary proper - time value ( in the case of rotationally - invariant background fields only ) so that one can have an approximation to the full effective action . in this appendix we shall briefly summarize this development and also provide further technical details on the formulas stated in @xcite . in the proper - time representation ( [ ptaction ] ) for the effective action , it is the function @xmath166 which contains the important information . given a rotationally invariant background field , we may utilize the phase shift analysis with scattering solutions of the ` hamiltonian ' @xmath167 to rewrite the expression ( [ appf ] ) . to that end we will put the system in a large spherical box of radius @xmath168 ( with a dirichlet or neumann boundary condition at @xmath169 ) , thus making the spectrum discrete @xcite . in a single instanton background in particular , we may then consider the quantum mechanical scattering solution for each partial wave , that is , @xmath170 the corresponding free schrdinger equation yields @xmath171 we are interested in the solution of ( [ appradial ] ) and ( [ appradialfree ] ) that vanish as @xmath172 for @xmath173 . then , for large @xmath71 , @xmath174 behaves as @xmath175,\ ] ] where @xmath176 ( for nonnegative integer @xmath177 ) , and @xmath111 is a certain constant which is not important . on the other hand , we may write the large-@xmath71 asymptotic behavior of the solution to ( [ appradial ] ) in the form @xmath178,\ ] ] where @xmath179 denotes the appropriate scattering phase shift [ the related scattering matrix is give by @xmath180 . here , because of the boundary condition at @xmath169 , we may demand the discretized momentum @xmath181 to be related to @xmath182 above according to @xmath183 from ( [ appkrel ] ) we conclude that @xmath184 if @xmath185 ^ 2 $ ] and @xmath186 ^ 2 $ ] denote the energy eigenvalues associated with the schrdinger equations ( [ appradial ] ) and ( [ appradialfree ] ) , respectively , the function @xmath26 give by ( [ appf ] ) can be represented as @xmath187 ^ 2 } - e^{-s [ k^l_0 ( n)]^2 } \right\},\ ] ] including the degeneracy factor @xmath33 for the @xmath188 partial wave . but , because of ( [ appkrel2 ] ) , we find for large @xmath168 @xmath189 ^ 2 } - e^{-s [ k^l_0 ( n)]^2 } \right\ } = e^{-s [ k^{l , j}(n)]^2 } \left\ { \frac{2k^{l , j}(n)\eta_{l , j}(k(n))}{r } s + o \left ( \frac{1}{r^2 } \right ) \right\}.\ ] ] using ( [ appkrel3 ] ) in ( [ appf2 ] ) gives rise to @xmath190 ^ 2 } \frac{2k^{l , j}(n)\eta_{l , j}(k(n))}{\pi } s,\ ] ] where @xmath191 , and then , replacing the sum @xmath192 by an integral for @xmath193 , we obtain the following formula : @xmath194 with ( [ appfinalf ] ) some caution must be exercised in dealing with the infinite partial wave sum . actually , in the instanton background we are considering , the nature of the scattering problem as defined by ( [ appradial ] ) and ( [ appradialfree ] ) does not allow us to consider the @xmath46-sum and @xmath47-sum in ( [ appfinalf ] ) in a completely independent manner . the point is that , as one looks at the given forms of @xmath195 and @xmath196 , their small-@xmath71 behaviors match for a given @xmath46-value ; but it is the @xmath47-value that governs the large-@xmath71 behavior of the potential entering @xmath195 , while @xmath47 does not appear in @xmath197 at all . to obtain a convergent expression from ( [ appfinalf ] ) , it is necessary @xcite to consider the @xmath58 and @xmath198 partial - wave contributions , both of which have the same degeneracy factor of @xmath45 , together as one package . with this understanding , the expression ( [ appfinalf ] ) can now be cast in the form @xmath199 . \label{fsum}\ ] ] if one has complete phase shifts for all partial waves at hand , one may use this formula ( [ fsum ] ) to calculate the function @xmath26 and then the one - loop effective action as well . but , in the massive case , the exact phase shifts can not be obtained analytically . therefore , in @xcite , we proposed to use the wkb expressions for the phase shifts , together with our formula ( [ appjsummedf ] ) . this method is elaborated below . first , we need the results of dunham @xcite for higher - order wkb approximations to the scattering phase shifts . if the schrdinger equation is written in the form @xmath200 the phase shift in the leading wkb approximation is given by @xmath201,\ ] ] where the integration path goes around the turning point @xmath202 ( i.e. , the point where @xmath203 vanishes ) in the complex plane , and crosses the real axis at @xmath204 ( with @xmath205 ) and @xmath206 ( with @xmath207 taken to positive infinity ) . the choice of @xmath208 has no effect on the value of the integral , and ( ` free ' ) in ( [ appeta1 ] ) represents the same integral but with @xmath203 of the free schrdinger equation . dunham also derived the formulas in the second and third order wkb approximations : @xmath209 , \\ \eta^{(3 ) } & = & \frac{1}{2 } \left [ \oint \left ( \frac{1}{768 } \frac{q^{(4)}(x)}{q(x)^{5/2 } } - \frac{7}{1536 } \frac{[q''(x)]^2}{q(x)^{7/2 } } \ dx \right ) - ( \text{`free ' } ) \right].\end{aligned}\ ] ] one can not use dunham s formula directly with the _ radial _ schrdinger equation in ( [ appradial ] ) and ( [ appradialfree ] ) . the latter should be transformed appropriately , following langer @xcite . thus , writing @xmath210 and introducing the function @xmath211 we recast ( [ appradial ] ) as @xmath212 since this is of the form ( [ appdunhameq ] ) , we can get the relevant scattering phase shifts simply by setting @xmath213 and also , in connection with the free equation ( [ appradialfree ] ) , @xmath214 then , using the original variable @xmath210 and integrating by parts , the desired wkb expressions for the phase shifts assume the form @xmath215 where @xmath216 is the so - called langer modification @xcite of the potential @xmath217 and the term referred to ( ` free ' ) denotes the integral expression appearing before but with @xmath216 replaced by @xmath218 note that @xmath219 , and @xmath220 . using the results ( [ appreta1])-([appreta3 ] ) in ( [ appjsummedf ] ) , we can simplify the expression by carrying out the @xmath221-integration . first , consider the leading order wkb . the first contour integral in ( [ appreta1 ] ) can be changed to the integral along the real axis over the interval @xmath222 , where @xmath223 is a turning point , i.e. , @xmath224 = 0 $ ] . on the other hand , @xmath221-integral in ( [ appjsummedf ] ) runs from 0 to @xmath225 . we may here change the order of integration , that is , perform the @xmath221-integral prior to considering the @xmath71-integration : in this case , the integration range for @xmath221 would be over the interval @xmath226 , where @xmath227 represents the value specified by the condition @xmath228 = 0 $ ] for a given value of @xmath71 . similar consideration may be given to the second contour integral in ( [ appreta2 ] ) , the ` free ' part . then , observing that we obtain from the first contour integral @xmath229 ( and similarly the form @xmath230 from the second contour integral ) , we are led to the following leading - order wkb expression for @xmath26 : @xmath231 , \label{appsmallf } \\ & & f^{(1)}_{(l , j)}(s , r ) = \frac{e^{-s \tilde{v}_{(l , j)}(r)}}{2 \sqrt{\pi } \sqrt{s } } - \frac{e^{-s \tilde{v}_l ( r)}}{2 \sqrt{\pi } % leeedit - \tilde{v}^0_l - > \tilde{v}_l \sqrt{s}}.\end{aligned}\ ] ] we use similar procedures to simplify the contributions coming from the second- and third- order wkb phase shifts in ( [ appreta2 ] ) and ( [ appreta3 ] ) . for this , a particularly useful relation is @xmath232^{n+\frac{1}{2 } } } = \frac{e^{-s \tilde{v}(r ) } s^{n+\frac{1}{2 } } \gamma(-n + \frac{1}{2})}{\pi } , \ ( n=0,1,2,\cdots).\ ] ] as a result , we obtain the higher order wkb expressions for @xmath26 , i.e. , @xmath233 and @xmath234 , which may be expressed again by the form ( [ appsmallf ] ) but with @xmath235 in connection with using the formula ( [ appsmallf ] ) , our discussion will not be complete without being clear about the _ order _ between executing the infinite series sum over @xmath46 and performing the ( improper ) radial integral . this is a subtle point , and one possible way to settle the issue unambiguously would be to check explicitly which order gives rise to the known small-@xmath25 behavior for the function @xmath26 correctly . as was asserted in @xcite , doing the @xmath46-sum before the @xmath71-integration yields the correct result . if instead one performs the @xmath71-integration first and then considers the @xmath46-sum , it gives a result differing from the correct small @xmath25 expression of @xmath26 by @xmath236 . [ note that , although we used the wkb series for the calculation , the thus - found difference is an exact result since the order - dependent ambiguity is purely a high - energy phenomenon and the wkb series can be trusted in the high energy limit ] . in view of this remark , the correct formula to be used in the wkb analysis of the effective action should read @xmath237.\ ] ] the result of using this formula for the fermion determinant in a single instanton background is presented in @xcite . in this appendix we address the question of the gauge invariance of the determinant or the effective action . the proper time representation of the effective action in ( [ ptaction ] ) is written in terms of covariant derivatives and is clearly gauge invariant . however , to formulate the partial wave expansion , as discussed in section [ radial ] , we chose a particular gauge ( [ instanton ] ) . thus the partial wave expansion used in the main text does not possess manifest gauge invariance . in this appendix we verify gauge independence by showing that we obtain precisely the same result in a different gauge . since our computational method relies on the radial symmetry of the background field , we are restricted in which gauge we can choose . the choice in ( [ instanton ] ) is often called the `` regular '' gauge . but an instanton background also has radial symmetry in the so - called `` singular '' gauge : @xmath238 in the singular gauge . in ( [ singular ] ) , @xmath239 differs from @xmath11 in ( [ instanton ] ) only by the sign in the components with @xmath5 or @xmath240 equal to 4 @xcite . the gauge field in the singular gauge has singular behavior in the vicinity of @xmath107 . one may worry about the validity of the radial approach in the main text in the singular gauge because of this . however , it turns out that it is not so singular . physically , the reason for this is the conformal invariance of the instanton background @xcite . the differential operator @xmath29 in the instanton background ( [ singular ] ) in the singular gauge can be written as , setting @xmath17 , @xmath241\ , , \label{singh}\ ] ] in the region of @xmath242 , the `` potential '' term in ( [ singh ] ) has @xmath243 singular behavior . but the last two singular terms proportional to @xmath244 combine into @xmath245 , which is regular . we decompose @xmath246 into @xmath247 ; i.e. , into singular and regular parts . the singular part , @xmath248 combines with the orbital part @xmath249 . we can cast ( [ singh ] ) in the form @xmath250\ , . \label{singh2}\ ] ] this should be compared with the corresponding operator in the regular gauge [ see ( [ insth ] ) ] : @xmath251\ , . \label{regular}\ ] ] note that these two have the same form except that @xmath252 and @xmath141 are interchanged . therefore . in the partial wave analysis the radial hamiltonian ( for isospin @xmath1 ) can be written as @xmath253\ , . \label{radhsing}\ ] ] there is a one - to - one correspondence between the eigenvalues for the sector @xmath254 in the singular gauge and those for the sector @xmath255 in the regular gauge , and they have the same multiplicities , @xmath256 . the regularized one - loop effective action in the singular gauge can be written as @xmath257 here we have combined the radial determinants for @xmath43 and @xmath258 , as in section [ radial ] , and we have arranged the free determinants appropriately . but from the above arguments , we know that @xmath259 hence the singular gauge expression in ( [ pvsingular ] ) is identical with the regular gauge expression in ( [ pv ] ) . so the pauli - villars regularized one loop effective action has the same value in the singular gauge and in the regular gauge , and therefore the renormalized effective action ( [ renaction ] ) , and the modified effective action ( [ modaction ] ) , has the same value in each gauge . g. v. dunne , j. hur , c. lee and h. min , `` precise quark mass dependence of instanton determinant , '' arxiv : hep - th/0410190 , _ phys . lett . _ in press . w. heisenberg and h. euler , `` consequences of dirac s theory of positrons , '' z. phys . * 98 * , 714 ( 1936 ) ; v. weisskopf , `` the electrodynamics of the vacuum based on the quantum theory of the electron , '' kong . math - fys . xiv no . 6 ( 1936 ) . j. s. schwinger , `` on gauge invariance and vacuum polarization , '' phys . rev . * 82 * , 664 ( 1951 ) . m. r. brown and m. j. duff , `` exact results for effective lagrangians , '' phys . d * 11 * , 2124 ( 1975 ) . for a recent review , see : g. v. dunne , `` heisenberg - euler effective lagrangians : basics and extensions , '' arxiv : hep - th/0406216 , in ian kogan memorial collection , _ from fields to strings : circumnavigating theoretical physics _ , m. shifman et al ( eds ) , ( world scientific , singapore , 2004 ) , volume i , pages 445 - 522 . g. t hooft , `` computation of the quantum effects due to a four - dimensional pseudoparticle , '' phys . d * 14 * , 3432 ( 1976 ) [ erratum - ibid . d * 18 * , 2199 ( 1978 ) ] . g. t hooft , `` how instantons solve the u(1 ) problem , '' phys . * 142 * , 357 ( 1986 ) . t. schafer and e. v. shuryak , `` instantons in qcd , '' rev . phys . * 70 * , 323 ( 1998 ) [ arxiv : hep - ph/9610451 ] . m. a. shifman , _ itep lectures on particle physics and field theory . vol . 1 and 2 _ , world sci . notes phys . * 62 * , 1 ( 1999 ) . d. diakonov , `` instantons at work , '' prog . phys . * 51 * , 173 ( 2003 ) [ arxiv : hep - ph/0212026 ] . r. d. carlitz and d. b. creamer , `` light quarks and instantons , '' annals phys . * 118 * , 429 ( 1979 ) . o. k. kwon , c. lee and h. min , `` massive field contributions to the qcd vacuum tunneling amplitude , '' phys . d * 62 * , 114022 ( 2000 ) [ arxiv : hep - ph/0008028 ] . v. a. novikov , m. a. shifman , a. i. vainshtein and v. i. zakharov , `` calculations in external fields in quantum chromodynamics '' , fortsch . phys . * 32 * , 585 ( 1985 ) . a. w. thomas , `` chiral extrapolation of hadronic observables '' , nucl . suppl . * 119 * , 50 ( 2003 ) [ arxiv : hep - lat/0208023 ] ; c. bernard et al , `` panel discussion on chiral extrapolation of physical observables '' , nucl . suppl . * 119 * , 170 ( 2003 ) [ arxiv : hep - lat/0209086 ] . s. levit and u. smilansky , `` a theorem on infinite products of eigenvalues of sturm - liouvill operators '' , proc . * 65 * , 299 ( 1977 ) , `` a new approach to gaussian path integrals and the evaluation of the semiclassical propagator , '' annals phys . * 103 * , 198 ( 1977 ) . s. coleman , _ the uses of instantons _ , erice lectures 1977 , reprinted in s. coleman , _ aspects of symmetry _ , ( cambridge university press , 1988 ) . k. kirsten and a. j. mckane , `` functional determinants by contour integration methods , '' ann . phys . ( ny ) * 308 * , 502 ( 2003 ) , [ arxiv : math - ph/0305010 ] ; `` functional determinants for general sturm - liouville problems , '' j. phys . a. * 37 * , 4649 ( 2004 ) , [ arxiv : math - ph/0403050 ] ; k. kirsten , _ spectral functions in mathematics and physics _ , ( chapman - hall , boca raton , 2002 ) . g. v. dunne , j. hur , c. lee and h. min , `` instanton determinant with arbitrary quark mass : wkb phase - shift method and derivative expansion , '' phys . b * 600 * , 302 ( 2004 ) [ arxiv : hep - th/0407222 ] . a. s. schwarz , `` on regular solutions of euclidean yang - mills equations , '' b * 67 * , 172 ( 1977 ) ; l. s. brown , r. d. carlitz and c .. lee , `` massless excitations in instanton fields , '' phys . d * 16 * , 417 ( 1977 ) ; r. jackiw and c. rebbi , `` spinor analysis of yang - mills theory , '' phys . d * 16 * , 1052 ( 1977 ) ; a. dadda and p. di vecchia , `` supersymmetry and instantons , '' phys . b * 73 * , 162 ( 1978 ) . l. s. brown and d. b. creamer , `` vacuum polarization about instantons , '' phys . d * 18 * , 3695 ( 1978 ) ; e. corrigan , p. goddard , h. osborn and s. templeton , `` zeta function regularization and multi - instanton determinants , '' nucl . b * 159 * , 469 ( 1979 ) ; b. berg and m. lscher , `` computation of quantum fluctuations of quark fields in an arbitrary yang - millian instanton background , '' nucl . b * 160 * , 281 ( 1979 ) ; h. osborn , `` semiclassical functional integrals for selfdual gauge fields , '' annals phys . * 135 * , 373 ( 1981 ) ; c. lee , h. w. lee and p. y. pac , `` calculation of one loop instanton determinants using propagators with space - time dependent mass , '' nucl . b * 201 * , 429 ( 1982 ) . a. a. belavin , a. m. polyakov , a. s. shvarts and y. s. tyupkin , `` pseudoparticle solutions of the yang - mills equations , '' phys . b * 59 * , 85 ( 1975 ) . m. abramowitz and i. stegun , `` handbook of mathematical functions '' , ( dover , new york , 1965 ) . l. s. brown , r. d. carlitz , d. b. creamer and c. lee , `` propagation functions in pseudoparticle fields , '' phys . rev . d * 17 * , 1583 ( 1978 ) . c. m. bender and s. a. orszag , _ advanced mathematical methods for scientists and engineers _ , ( mcgraw - hill inc . , new york , 1978 ) . r. e. langer , `` on the connection formulas and the solutions of the wave equation '' , phys . rev . * 51 * , 669 ( 1937 ) ; j. b. krieger and c. rosenzweig , `` application of a higher - order wkb approximation to radial problems '' , phys . rev . * 164 * , 171 ( 1967 ) . j. baacke and g. lavrelashvili , `` one - loop corrections to the metastable vacuum decay , '' phys . d * 69 * , 025009 ( 2004 ) [ arxiv : hep - th/0307202 ] . g. v. dunne , a. w. thomas and s. v. wright , `` chiral extrapolation : an analogy with effective field theory , '' phys . b * 531 * , 77 ( 2002 ) [ arxiv : hep - th/0110155 ] . g. v. dunne and c. schubert , `` two - loop self - dual euler - heisenberg lagrangians . i : real part and helicity amplitudes , '' jhep * 0208 * , 053 ( 2002 ) [ arxiv : hep - th/0205004 ] . j. l. dunham , `` the wentzel - brillouin - kramers method of solving the wave equation '' , phys . rev . * 41 * , 713 ( 1932 ) ; c. rosenzweig and j. b. krieger , `` exact quantization conditions '' , j. math . phys . * 9 * , 849 ( 1968 ) ; c. m. bender , k. olaussen and p. s. wang , `` numerological analysis of the wkb approximation in large order , '' phys . rev . d * 16 * , 1740 ( 1977 ) . b. s. dewitt , `` dynamical theory of groups and fields '' ( gordon and breach , new york , 1965 ) ; phys . rep . * 19 * , 295 ( 1975 ) . c. lee and c. rim , `` background fermi fields and schwinger - dewitt proper time method , '' nucl . b * 255 * , 439 ( 1985 ) ; i. jack and h. osborn , `` general background field calculations with fermion fields , '' nucl . b * 249 * , 472 ( 1985 ) ; r. d. ball , `` chiral gauge theory , '' phys . rept . * 182 * , 1 ( 1989 ) . d. a. wasson and s. e. koonin , `` local expressions for one loop calculations , '' phys . d * 43 * , 3400 ( 1991 ) ; c. l. y. lee , `` evaluation of the one loop effective action at zero and finite temperature , '' phys . d * 49 * , 4101 ( 1994 ) [ arxiv : hep - ph/9311369 ] . i. g. moss , `` soliton vacuum energies and the cp(1 ) model , '' phys . b * 460 * , 103 ( 1999 ) [ arxiv : hep - th/9905113 ] ; i. g. moss and w. naylor , `` effective action for bubble nucleation rates , '' nucl . b * 632 * , 173 ( 2002 ) [ arxiv : hep - ph/0102104 ] . a. hasenfratz and p. hasenfratz , `` the scales of euclidean and hamiltonian lattice qcd , '' nucl . b * 193 * , 210 ( 1981 ) . p. faccioli and e. v. shuryak , `` systematic study of the single instanton approximation in qcd , '' phys . d * 64 * , 114020 ( 2001 ) [ arxiv : hep - ph/0106019 ] . e. poppitz and y. shirman , `` the strength of small - instanton amplitudes in gauge theories with compact extra dimensions , '' jhep * 0207 * , 041 ( 2002 ) [ arxiv : hep - th/0204075 ] .

we present an analytic model for the sizes of voids in the galaxy distribution . peebles and others have recently emphasized the possibility that the observed characteristics of voids may point to a problem in galaxy formation models , but testing these claims has been difficult without any clear predictions for their properties . in order to address such questions , we build a model to describe the distribution of galaxy underdensities . our model is based on the `` excursion set formalism , '' the same technique used to predict the dark matter halo mass function . we find that , because of bias , galaxy voids are typically significantly larger than dark matter voids and should fill most of the universe . we show that voids selected from catalogs of luminous galaxies should be larger than those selected from faint galaxies : the characteristic radii range from for galaxies with absolute-band magnitudes to . these are reasonably close to , though somewhat smaller than , the observed sizes . the discrepancy may result from the void selection algorithm or from their internal structure . we also compute the halo populations inside voids . we expect small haloes ( ) to be up to a factor of two less underdense than the haloes of normal galaxies . within large voids , the mass function is nearly independent of the size of the underdensity , but finite - size effects play a significant role in small voids ( ) . cosmology : theory large - scale structure of the universe galaxies : luminosity functions

astro-ph0509148

one of the key predictions of any cosmological structure formation model is the distribution of matter on large scales . we now know that the cold dark matter ( cdm ) paradigm can account for many of the observed characteristics of the galaxy distribution . in this picture , the universe contained tiny ( gaussian ) density fluctuations at the time the cosmic microwave background was last scattered . bound structures assembled themselves through gravitational instability around these perturbations in the relatively recent past . in the past two decades , with the advent of high - resolution numerical simulations , it has become possible to follow this picture through the formation of massive galaxies and clusters . the cdm model accurately describes the abundance and clustering of collapsed objects from dwarf galaxies to rich galaxy clusters over a wide range of redshifts ( although the details of galaxy formation itself remain somewhat mysterious ) as well as the distribution of neutral gas in the intergalactic medium . these systems all correspond to density peaks in the matter distribution ( with the exception of the lowest column density ly@xmath7 forest absorbers ) . for a variety of reasons , the other end of the density distribution underdense voids has received considerably less attention , despite their long observational history @xcite and their place as the most visually striking features of the galaxy distribution . this is largely because voids subtend enormous volumes and so require large surveys to garner representative samples . although voids have been found in every redshift survey @xcite , the first statistically significant sample came only with the 2df redshift survey @xcite . now , with the deep2 redshift survey and the sloan digital sky survey , it is even possible to constrain the evolution of voids over the redshift interval @xmath8@xmath9 @xcite . @xcite presented the most complete search for voids to date . they found that voids with characteristic radii @xmath10 fill @xmath11 of the universe . but their search illustrates a second difficulty in studying voids : how to define and identify them precisely ( and meaningfully ) . @xcite proposed the _ voidfinder _ algorithm based on separating the observed galaxies into `` void '' and `` wall '' populations and building voids around gaps in the wall population ( see also @xcite ) . while clearly defined for any given observational sample , the results can nevertheless be difficult to interpret in relation to the underlying physical quantities of interest . for example , the distribution of observed sizes depends on the galaxy sample ( intrinsically brighter galaxies yield larger voids ) as well as the search algorithm ( @xcite restrict their search to radii greater than @xmath12 , for example ) . smaller voids are difficult to pinpoint because of confusion with random fluctuations in the galaxy distribution . we will nevertheless follow this approach and define a void to be any coherent region where the galaxy density falls below some threshold . note that this differs from many other studies ( e.g. , @xcite ) that require a void to be _ completely _ empty of galaxies . these voids , which are much smaller than the voids we consider here , do not correspond to the voids detected by eye in the galaxy distribution . the void phenomenon is also relatively difficult to study theoretically . the simplest model is a spherical tophat underdensity . the early evolution of such a system is well - described by spherical expansion ( the analog of the well - known spherical collapse model ; @xcite ) . underdensities expand in comoving units , gradually deepening , until they reach `` shell - crossing , '' when the center is evacuated and individual mass shells cross paths . at this stage , the initial spherical expansion model breaks down . later , the voids continue to expand relatively slowly in a self - similar fashion @xcite . unfortunately , these dark matter models are idealizations in that real voids are observed only through the galaxy distribution . because galaxies are biased relative to the dark matter , we must ask what kind of physical systems observed voids actually represent . they are in fact nearly empty of galaxies but does that require shell - crossing , or can they be at an earlier evolutionary stage ? another difficulty is that the large size of voids restricts the usefulness of numerical simulations . early efforts focused on understanding the dynamics of individual voids , which did not require particularly high resolution @xcite . but placing voids in their proper cosmological context demands both large volumes and high mass resolution the latter because we must resolve the galaxies from whose absence we identify voids ( e.g. , @xcite ) . only recently have @xmath13-body simulations of the required dynamic range become practical . this has allowed the first systematic studies of the structure of large voids @xcite as well as of simulated voids in the dark matter distribution @xcite and in the galaxy distribution @xcite . nevertheless , there are still no full hydrodynamic simulations of the void phenomenon : galaxy properties are currently determined through semi - analytic models @xcite . there are a number of reasons to study the void phenomenon . early attention focused on using the observed voids to constrain cosmological parameters . @xcite and @xcite argued that the scale of typical voids would depend on the matter power spectrum , just as the abundance of clusters does . it was quickly realized that the void distribution seemed to have much more large - scale power than collapsed objects most obviously , although voids with sizes @xmath14 are not uncommon , collapsed objects do not reach the same mass scales . this is puzzling because an underdense region reaches shell - crossing only after the equivalent overdensity would virialize . @xcite pointed out that a solution might lie in a proper treatment of the galaxies used to define voids : galaxy bias could lower the required dark matter underdensity , allowing voids to be larger for a given amount of cdm power . they argued that the observed population of large voids preferred a @xmath15cdm model . recently , interest in voids has focused on their role in galaxy formation . @xcite argued that ( in the cdm model ) voids should be populated by small dark matter haloes but that the observed voids appear to lack faint galaxies as well as bright ones . is this a fundamental problem for the cdm paradigm , or does it indicate that galaxy formation proceeds differently in voids ? one popular explanation is that photoheating during reionization may have suppressed the formation of dwarf galaxies in low - density environments @xcite , helping to clear faint galaxies from the voids . to study these problems , recent attention has focused on the variation of the galaxy luminosity function with the large - scale environment . @xcite and @xcite ( see also @xcite ) found that the characteristic galaxy luminosity and the galaxy density both decrease significantly in low - density environments but that the faint - end slope remains nearly constant . the latter may be surprising given the expected variation in the halo mass function with environment . before we can answer any of these questions , however , we require a proper understanding of voids in cdm models . @xcite and @xcite made important strides forward with their studies of voids in @xmath13-body simulations . they claimed that current semi - analytic galaxy formation models predict void properties similar to those observed and argued that the conflict pointed out by @xcite was illusory . but such studies are still limited by their dynamic range and by the ( necessarily complex ) galaxy formation prescriptions imposed on the dark matter haloes . the goal of this paper is to produce a straightforward analytic model of the void distribution and of galaxy populations within voids . such a model will sharpen our understanding of voids in a cdm model and generate a baseline prediction with which we can contrast their observed properties . we will build on @xcite , the most compelling theoretical model of voids to date . they used the excursion set formalism , which reproduces the abundance of collapsed haloes extremely well , to predict the distribution of void sizes . however , they required voids to reach shell - crossing and defined their properties in terms of the dark - matter underdensity . as a result , they predicted characteristic void radii much smaller than those observed . our main goal is to modify their approach so as to describe voids in the _ galaxy _ distribution . along the way , we will also be able to predict the halo populations within voids and quantify the claimed discrepancy with observational results . the remainder of this paper is organized as follows . in [ exp ] , we examine the nonlinear evolution of the void density . then , in [ defn ] , we show how to compute the linearized underdensity of voids with a specified galaxy underdensity , and we briefly discuss the expected galaxy populations inside voids . in [ abundance ] , we show how to compute the cosmological abundance of large _ galaxy _ voids and compare our results to observations . finally , we conclude in [ disc ] . in our calculations , we assume a cosmology with @xmath16 , @xmath17 , @xmath18 , @xmath19 ( with @xmath20 ) , @xmath21 , and @xmath22 , consistent with the most recent measurements @xcite . because the principal quantity of interest is the physical volume of voids , we first describe how they expand beyond their initial comoving size . we will follow the method of @xcite . we consider the expansion of a tophat density perturbation that begins with a ( negative ) density perturbation @xmath23 inside a physical radius @xmath24 at an initial time @xmath25 ( corresponding to @xmath26 ) . if we assume zero peculiar velocity perturbation and consider only the epoch before shell - crossing ( so that the mass inside of each shell is conserved ) , conservation of energy yields the equation of motion @xcite @xmath27 , \label{eq : exp}\end{aligned}\ ] ] where the first term on the right hand side is ( twice ) the initial total energy of the shell , the second term is ( twice ) the gravitational potential energy at time @xmath28 , and the last term is ( twice ) the excess energy from the cosmological constant . we solve this equation for the physical radius of the shell as a function of time or redshift . the solution is such that the void expands in comoving units once nonlinear effects set in . we define @xmath29 to be the ratio of the comoving size at redshift @xmath30 to its initial comoving size . the real fractional underdensity due to gravitational expansion is thus @xmath31 . note that this is independent of @xmath24 . equation ( [ eq : exp ] ) includes nonlinear expansion , but most of the following will be phrased in terms of the equivalent linear density @xmath32 extrapolated to the present day . ( for clarity , we will denote all linear - extrapolated densities with a superscript l " in the following . ) in order to transform to physical densities and scales , we require @xmath33 or equivalently @xmath34 . we therefore also compute the density as if linear theory were always accurate . any perturbation can be divided into growing and decaying modes ; in linear theory the growing mode ( which is of course the only component surviving to the present day ) obeys @xcite @xmath35 where @xmath36 in a flat universe . the constant @xmath37 comes from our choice of initial conditions : a perturbation with zero peculiar velocity has @xmath38 of its amplitude in the growing mode and the remainder in the decaying mode @xcite . we show the relation between the linearized and physical underdensities with the solid line in figure [ fig : dl ] . at first , while linear evolution is accurate , @xmath39 , but then the physical underdensity flattens out and even a small deepening of the void corresponds to a large increase in @xmath40 . the dashed line in figure [ fig : dl ] shows a fitting function for @xmath34 from eq . ( 18 ) of @xcite . although those authors were interested in @xmath41 , the formula is accurate to a few per cent in the underdense regime as well . for reference , shell - crossing occurs at @xmath42 , corresponding to a physical underdensity @xmath43 . beyond this point , our model for @xmath44 breaks down because the enclosed mass no longer remains constant . the expansion then proceeds self - similarly , with the comoving radius @xmath45 in an einstein - de sitter universe @xcite , significantly slower than the expansion before shell - crossing . as a result , the void volume fraction will be dominated by regions that have not yet ( or have just ) reached shell - crossing @xcite : deeper perturbations are rarer in the cdm model and can be neglected because they expand so slowly . we therefore do not consider the later evolution in any more detail . for completeness , we also note that figure [ fig : dl ] is nearly independent of the input cosmological parameters ; differences in the cosmology manifest themselves in the growth rate of @xmath46 rather than in the function @xmath47 @xcite . it may now seem straightforward to compute the void distribution : one simply chooses a physical density threshold @xmath48 , transforms to @xmath32 , and uses the excursion set formalism to determine the mass distribution of objects with the specified linearized underdensity @xcite . while such a procedure is perfectly well - defined , it has one crucial shortcoming : in observed samples , we do not define voids in terms of their dark matter density but in terms of their _ galaxy _ density . thus we actually want to relate the physical galaxy underdensity @xmath49 within some region to @xmath32 . that is the task of this section . [ cols="^,^,^,^,^",options="header " , ] + number densities are computed from the @xcite @xmath2-band and @xcite @xmath50-band luminosity functions . columns 4 and 5 give the minimum halo mass required to match these densities , with the @xcite @xmath51 , for the press - schechter and sheth - tormen mass functions , respectively . we will use the excursion set formalism to derive the dark matter halo mass function @xcite . in the simplest such model , a halo of mass @xmath52 forms whenever the smoothed density field first exceeds a linear - extrapolated density @xmath53 that is fixed by the physics of spherical halo collapse ( @xmath54 for objects collapsing at the present day ; @xcite ) . to solve such a problem , we consider diffusion in the @xmath55 plane , where @xmath56 is the variance of the density field smoothed on scale @xmath52 . trajectories begin with @xmath57 at @xmath58 ( i.e. , they must have the mean density on infinitely large scales ) . they then diffuse away from the origin as density modes are added on smaller scales . the problem is simply to compute the distribution of @xmath56 ( or equivalently @xmath52 ) at which these trajectories cross the absorbing barrier @xmath53 for the first time . from this first - crossing distribution , the mean comoving number density of haloes with masses @xmath59 is @xcite @xmath60 , \label{eq : nmh}\ ] ] where @xmath61 is the mean matter density . in a region with linearized underdensity @xmath62 and mass @xmath63 [ with @xmath64 for a void ] , the comoving number density of haloes is @xcite @xmath65 , \label{eq : nmcond}\end{aligned}\ ] ] where @xmath66 and of necessity @xmath67 . this conditional mass function follows from an identical diffusion problem , but in this case the trajectories begin at @xmath68 rather than the origin . although equation ( [ eq : nmh ] ) provides a reasonable match to halo abundances in cosmological simulations , it is by no means perfect . @xcite and @xcite provide more accurate fits to the simulation results . we will nevertheless use the standard press - schechter abundances in the following . the main reason is that , although the sheth - tormen mass function can also be motivated by a diffusion problem with the absorbing barrier fixed by ellipsoidal collapse , there is no analytic form for the conditional mass function because the barrier changes its effective shape with a shift of the origin . @xcite showed that the conditional mass function could be approximated through a taylor series , but in that case the corresponding unconditional mass function differs slightly from the usual sheth - tormen form ( which provides the best match to numerical simulations ) . @xcite showed how to derive the mass function corresponding to such barriers using a simple numerical scheme . but for our purposes the press - schechter form suffices . the most important property is the density dependence . @xcite and @xcite found that it is extremely similar to that contained in equation ( [ eq : nmcond ] ) , at least at high redshifts , and the former also showed that it fits simulations fairly well . equations ( [ eq : nmh ] ) and ( [ eq : nmcond ] ) give the dark matter halo abundance , which is not the same as the galaxy abundance . to connect the two we require the mean number of galaxies per halo , @xmath69 , the first moment of the halo occupation distribution . for simplicity , we will use the universal form of @xcite . they split the mean occupation number into two parts . the first , @xmath70 , represents the number of central galaxies in the halo and is unity if @xmath71 and zero otherwise . we will normally choose @xmath72 by comparison to some observational detection threshold ; note therefore that it need not reflect the _ actual _ minimum mass of a galaxy . the second part , @xmath73 , describes the number of satellite galaxies above the same detection limit . @xcite found that this has a simple form @xmath74 from fits to the subhalo distribution in @xmath13-body simulations , they estimate that @xmath75 and @xmath76 at @xmath77 over a broad range of halo masses . we will use these fiducial values in most of our calculations . in order to estimate the importance of the halo occupation distribution , we will also compare to a model with @xmath78 . the total comoving number density of ( observable ) galaxies within a region is therefore @xmath79 where the superscript `` c '' denotes a cumulative quantity . we wish to compare to voids selected through galaxy surveys . we will therefore set @xmath72 by comparison to the galaxy number densities in the sloan digital sky survey and the 2df galaxy redshift survey . specifically , we match to the number density of galaxies with @xmath2-band luminosity greater than a specified value , using the luminosity function of @xcite , and to the corresponding quantities from the @xmath50-band luminosity function of @xcite . the appropriate mass thresholds @xmath72 for the press - schechter and sheth - tormen mass functions are shown in table [ tab : numden ] . we will also show some results for @xmath80 in order to mimic an exceptionally deep survey . note that normalizing in this way to the cumulative number density of galaxies tends to wash out differences between mass functions , insulating us from much of the uncertainty in using the press - schechter mass function . the total ( observed ) galaxy underdensity in a void with physical size @xmath81 is @xmath82 given @xmath72 , we solve this equation to find the linearized underdensity required to produce a void of size @xmath81 and mean observed galaxy underdensity @xmath49 . figure [ fig : barriers ] shows the results of this inversion . the solid curves in panel _ ( a ) _ give the required underdensity for several different survey depths : @xmath83 , and @xmath84 , from top to bottom . ( here and throughout , we will suppress the @xmath85 " in measured magnitudes for the sake of brevity . ) we take @xmath86 as a fiducial value . we plot @xmath62 as a function of mass variance @xmath56 . for reference , @xmath87 , and @xmath88 correspond to comoving sizes @xmath89 , and @xmath90 , respectively . note that this is _ not _ the physical size of the void ( which is larger by a factor @xmath44 and varies with @xmath62 ) . we halt each curve at @xmath91 . the required underdensity becomes more extreme in deeper surveys ; this is because underdense regions are much more deficient in massive galaxies than small haloes in the extended press - schechter formalism . haloes with masses @xmath92 can not collapse because too few density modes are available between @xmath93 to pass the collapse threshold @xmath53 , but if @xmath94 ( for small haloes ) the loss of large - scale power becomes unimportant . in other words , the bias increases with galaxy mass , so the large galaxies are more sensitive to the underlying density field @xcite . this is why @xmath62 curves upward as @xmath56 increases , especially in the @xmath95 case . there is also a slight trend for @xmath62 to increase as @xmath96 ; this is because the shape of the mass function changes in underdense regions , with more mass in small haloes . as the void size approaches infinity , slightly more of the mass shifts into galaxy - mass haloes . + the two dashed curves in figure [ fig : barriers]_a _ show @xmath62 if we assume @xmath78 for @xmath97 and @xmath98 . this decreases the threshold , because it increases the weight of the smallest haloes ( which are least sensitive to the density ) . the value of @xmath62 becomes more sensitive to the halo occupation number for deeper surveys ; a larger minimum mass pushes the survey to the steep part of the mass function , where the extra satellites in extremely large ( but rare ) haloes make only a small difference . note as well that in this case the @xmath84 threshold lies significantly below shell - crossing , where our model breaks down . figure [ fig : barriers]_b _ varies @xmath49 for @xmath95 and @xmath99 ( top and bottom sets of curves ) . we let @xmath100 , and @xmath101 for the dot - dashed , solid , and dashed curves . this obviously has a strong effect on @xmath62 , because the massive galaxies are so sensitive to the underlying dark matter density . note that the differences are largest at small @xmath56 ( or large physical scales ) , because at smaller scales the finite size of the void limits the maximum halo mass . this is also why the differences are larger for the deeper survey . now that we have defined voids , we can look more closely at their resident halo populations . figure [ fig : barriers ] clearly shows that the intrinsic properties of voids depend sensitively on the mass limit of the survey used to detect them : deeper surveys require larger dark matter underdensities and hence will presumably find smaller voids . but , in a given void , will smaller haloes be more abundant relative to the ( brighter ) galaxy population used to define the void ? to answer this question , we let @xmath102 we emphasize that @xmath103 is the halo underdensity at mass @xmath104 , _ not _ the galaxy underdensity . the analogous galaxy underdensity at a particular mass is more difficult to calculate , because it requires the distribution of galaxy masses within each halo instead of simply the number of galaxies ( i.e. , how many satellite galaxies of mass @xmath104 are contained in any galaxy group ) . such models require considerably more machinery than is appropriate for our simple analytic model . we will therefore content ourselves with computing the halo distribution , which will illustrate the most important features anyway . we will comment on its relation to the galaxy population below . figure [ fig : nmh ] shows predictions for the halo underdensity in voids with @xmath105 . the parameters in the two panels are identical to those in figure [ fig : barriers ] . in panel _ ( a ) _ , we see that small haloes become less underdense if void selection is performed with larger galaxies . within any given void , we also see that small haloes are less underdense than large haloes . this is , of course , because larger haloes are more biased than small ones . the disparity can be quite large for the @xmath95 sample , with small haloes only half as underdense as those with @xmath106 . the dashed curves show that these conclusions are qualitatively independent of @xmath51 . in panel _ ( b ) _ , we see that varying @xmath49 also modifies the fractional overdensity . however , its effects on the shape of the curve are relatively modest except at the most massive end . + the horizontal bars on the right axis of figure [ fig : nmh ] show the physical dark matter underdensity within each void . interestingly , it takes a relatively narrow range of values in panel _ ( a ) _ , clustered around @xmath107@xmath108 ) , with @xmath109 decreasing as @xmath72 increases . the halo populations ( especially at large masses ) spread over a much broader range in @xmath110 . this is essentially a consequence of the shape of @xmath47 shown in figure [ fig : dl ] : a narrow range in physical underdensities spans a large range of @xmath46 , which is the quantity relevant to our model for halo abundances . in general , @xmath111 for haloes selected near @xmath72 . the density of haloes with @xmath112 is closer to the mean because their bias is less than unity . our model also predicts that @xmath110 at a fixed mass will vary with the size of the void , because @xmath62 is also a function of scale . we show the implications in figure [ fig : nmhr ] ; in this case we consider surveys with @xmath113 and @xmath99 ( top and bottom sets , respectively ) . for large voids , @xmath114 , we see only a slight steepening of the curves , which occurs because the finite mass of the void limits the maximum halo mass . thus , galaxy populations in voids with @xmath114 should be nearly independent of the void radius . this is a consequence of the flatness of @xmath62 at small @xmath56 in figure [ fig : barriers ] , which occurs because so little power exists on such large physical scales . however , when @xmath115 , the shapes change dramatically because the finite size of the void strongly limits the haloes inside . a comoving volume with radius @xmath116 contains @xmath117 , which is only @xmath118@xmath119 times larger than the mass limits of these surveys . thus the region requires only a modest underdensity to decrease the abundance of massive galaxies . such `` voids '' therefore sit near the mean density and smaller haloes ( for which finite size effects can be neglected ) will have nearly their mean density . this simplified model suggests that many properties of voids will depend on their spatial extents , at least if @xmath120 . + one consistent theme of this section is that large haloes are relatively less abundant than small haloes within voids . we can compare this expectation to observations , which suggest that the mean number density @xmath121 and the characteristic luminosity @xmath122 decrease significantly in voids but that the faint - end slope of the luminosity function remains nearly constant @xcite . unfortunately , the comparison is not trivial . these studies measured the galaxy environment through the mean density within spheres of size @xmath123 ( generally by comparing to the density field of a volume - limited sample of galaxies ) . this is just large enough that finite size effects should not be dramatic , but it is still in a regime in which poisson fluctuations in the galaxy distribution can not be neglected . as a result , some of the `` low - density '' regions may in fact be at the mean density , which would wash out the void effect . an ideal test would focus on those galaxies identified to be within large voids ( for example using the _ voidfinder _ algorithm ) . moreover , we must also bear in mind that @xmath124 . in particular , many small galaxies will be satellites inside large haloes ; such a population will be efficiently suppressed in just the same way as bright central galaxies . thus we expect the galaxy luminosity function in voids to be flatter than the halo mass function . with these caveats , our model naturally explains the decrease in @xmath121 and @xmath122 , especially because these regions are relatively small and finite size effects help suppress galaxies with @xmath125 . note that this probably explains the discrepancy between the observed @xmath122 and that predicted by @xcite , who let @xmath126 rather than including a cutoff at @xmath127 which would have quenched the formation of massive galaxies . on the other hand , our models robustly predict an excess ( by about a factor of two ) between the abundance of small and large haloes in voids . if there were a one - to - one correspondence between halo mass and galaxy luminosity , and if the luminosity function could be approximated by a power law @xmath128 over this interval , we would expect @xmath7 to be @xmath129 larger in voids than in mean density environments . the suppression of satellite galaxies would reduce the disparity . even so , our result is within the error bars of @xcite , though their best - fit values indicate no significant change . thus , there is currently only weak evidence , at best , for non - standard galaxy formation within voids . cdm models robustly predict an excess of small haloes inside of voids , but it is not particularly strong and requires care to interpret . @xcite used the excursion set formalism to estimate the number density of voids . they defined a void as a region with dark matter underdensity @xmath62 . at first sight , the void abundance seems to follow from equation ( [ eq : nmh ] ) , but with @xmath130 : this is the same problem as the halo abundance , except that the absorbing barrier is negative rather than positive . however , @xcite pointed out one crucial difference , which they called the void - in - cloud " problem . consider a small region with @xmath131 that is contained in a larger region with @xmath132 . equation ( [ eq : nmh ] ) would have assigned such a point to a small void ; however , physically we know that this void " lies inside of a collapsed object and hence has been crushed out of existence . ( this problem does not occur for @xmath133 because haloes are allowed to collapse inside voids . ) thus , the appropriate diffusion problem has _ two _ barriers : we wish to compute the first - crossing distribution for the void barrier @xmath62 while excluding trajectories that have already crossed a barrier with @xmath134 that describes void - crushing . the simplest assumption is to take both these barriers to be independent of @xmath56 ; in that case @xcite showed that the mass function is ( see also the alternate derivation in the appendix ) @xmath135 \right\ } , \label{eq : nvm}\end{aligned}\ ] ] where @xmath136 describes the relative importance of void - crushing ; a fraction @xmath137 of all matter lies inside voids in this model . note that @xmath62 and @xmath138 implicitly depend on redshift , because we evaluate @xmath139 at the present day . the @xcite model contains one crucial assumption that voids can be defined at a constant @xmath62 , independent of scale and two free parameters . in what follows we will usually set the void - crushing parameter @xmath140 , which is the linearized overdensity at turnaround according to the spherical collapse model . this is reasonable because it marks the point at which larger scale overdensities will begin to collapse around their resident voids , but we will also let @xmath138 take larger values for illustrative purposes . the second free parameter is @xmath62 ; @xcite argued that @xmath141 ( shell - crossing ) is an appropriate choice , because that marks the maximum point of efficient expansion . however , we have already seen that the dark matter underdensity is generally significantly deeper than the galaxy underdensity . instead , if one wishes to compare to voids observed in redshift surveys , @xmath62 should be chosen from the @xmath49 and @xmath72 appropriate to a given void search . the relevant absorbing barriers are precisely the curves shown in figure [ fig : barriers ] : only in voids selected from exceptionally deep galaxy surveys do we expect shell - crossing to have occurred . interestingly , the barriers are also relatively flat , at least for reasonably deep surveys and on large scales . we will therefore take @xmath142constant and use equation ( [ eq : nvm ] ) to compute the number density of voids . we evaluate @xmath62 at @xmath143 in the following ; we examine how @xmath144 varies with the barrier shape in [ assump ] below . figure [ fig : nvoid ] shows the void abundance for several different survey depths . a _ shows the fraction of volume filled by voids of a given radius ; note that we use the physical void radius here , including the extra gravitational expansion . b _ shows the volume filling fraction of voids , @xmath145 where the lower integration limit is the mass corresponding to a void of physical radius @xmath146 ; the prefactor @xmath147 accounts for the extra gravitational expansion . we halt each curve when the void mass is less than twice @xmath72 . + figure [ fig : nvoid ] makes clear several key characteristics of the void size distribution . first , @xmath148 has both large and small mass cutoffs . the former occurs where @xmath149 , above which voids are exponentially suppressed because of the lack of large - scale power ( the same mechanism that provides the large - scale cutoff in the halo mass function ) . the small mass cutoff occurs because of the void - crushing threshold @xmath138 . the resulting characteristic scale @xmath150 is quite sensitive to the galaxies used to select voids : it ranges from @xmath151 for @xmath99 to @xmath152 for @xmath153 , nearly an order of magnitude in volume . the shape , on the other hand , does not change much across the different selection thresholds . finally , we also see that @xmath154 : the voids apparently fill a volume larger than the universe . this occurs because we have let each and every void expand to its full size , even those surrounded by relatively weak overdensities just short of turnaround . these filling fractions should thus not be taken too seriously ; they suggest only that , when voids are selected from galaxy surveys , we expect empty regions with @xmath155 to fill a large fraction of the universe . the largest possible voids also provide an interesting benchmark for comparison to surveys @xcite . to estimate @xmath156 , we find where @xmath157 ; i.e. , the size for which we expect one void per hubble volume . this will likely overestimate the maximum size in a real survey , because voids were smaller in the past ( see [ redshift ] below ) . of course the maximum size will also vary with the magnitude limit of the survey ; for @xmath158 we find @xmath159 . these are compatible with existing redshift surveys ; the largest voids in the 2df survey have @xmath160 @xcite . we will make a more detailed comparison to observations in [ obs ] below . finally , we note some simple properties of @xmath161 . ignoring void - crushing , the void filling fraction becomes @xmath162 . \label{eq : fvapprox}\ ] ] this is the same expression found by @xcite , except that they assumed @xmath62 corresponded to shell crossing and included an extra factor of one - half ( which occurred because they ignored those trajectories that joined larger voids ) . void - crushing is unimportant for voids larger than the characteristic size ( see [ assump ] below ) , so equation ( [ eq : fvapprox ] ) provides a reasonable estimate for @xmath161 and shows explicitly how it depends on @xmath62 ( and hence on the survey characteristics ) . @xcite pointed out the interesting coincidence that @xmath163 at the present day ; this remains true in our model . only recently have large voids in the faint galaxy population come to dominate the geography of the universe . we have seen that one key quantity in defining voids is the mean enclosed galaxy underdensity @xmath49 . figure [ fig : nvdgal ] shows how the void sizes vary with this criterion . the curves correspond to the same cases shown in figure [ fig : barriers]_b_. we see that @xmath49 has an enormous effect on the characteristic void size @xmath150 : it changes by a factor of three over this range of galaxy density ( or a factor of nearly thirty in volume ! ) . moreover , in panel _ ( b _ ) , we see that large , deep voids ( @xmath164 ) only fill a small fraction of space . + just as with the halo mass function , the excursion set formalism also allows us to study how @xmath165 evolves with redshift . we show the predicted void distribution at @xmath166 in figure [ fig : nvz ] . in order to facilitate comparison with the @xmath77 results , we hold @xmath72 constant between @xmath166 and the present day ( as in table 1 ) of course , because the galaxy luminosity function evolves over this interval , the _ number density _ of these haloes evolves over this redshift interval . + the voids also evolve , with the characteristic radius increasing significantly ( by about a factor of two in each case ) and the volume filling fraction also increasing ( again by a factor of @xmath167@xmath168 ) . for example , only @xmath169 of the universe lies inside of voids with @xmath170 selected from @xmath171 galaxies ( as opposed to @xmath172 at @xmath77 ) . this implies that voids continue to grow , albeit relatively slowly , even after the cosmological constant dominates the energy density . our results are in qualitative agreement with @xcite , who find that voids are both smaller and rarer at @xmath8 . a precise comparison is difficult because they use statistical techniques ( specifically the void probability function ) rather than identifying individual voids . unfortunately , this model for @xmath165 contains a number of free parameters and simplifications , so it is not nearly as well - specified as the halo mass function @xmath173 . here we will examine how sensitive our results are to these assumptions . we begin with the halo occupation number @xmath51 . the solid and dashed curves in figure [ fig : nvdparam]_a _ contrast the @xcite value ( our fiducial model ) and a model with one galaxy per halo . we have previously seen that @xmath174 only significantly affects @xmath62 if @xmath72 is relatively small ; as a result , @xmath165 is only substantially affected for the @xmath175 survey . in this case the voids shrink by about @xmath176 in radius , with @xmath177 . thus , if we selected voids from the _ halo _ underdensity ( as , for example , may be possible by associating groups of galaxies with single haloes ) , we would expect slightly smaller voids . the difference is minimal for bright galaxies because the mass function is steep in that regime . + figure [ fig : nvdparam]_a _ also compares the void sizes computed from the press - schechter mass function with those from the sheth - tormen mass function . to perform this comparison , we have scaled the sheth - tormen mass function with density in the same way as the press - schechter mass function @xcite . thus the only real difference is the mass threshold @xmath72 , although that difference can be substantial . we see that @xmath165 remains nearly unchanged . thus while the void distribution may in detail depend on the exact form of @xmath173 , it is not likely to be substantially affected . another unknown parameter in the model is @xmath138 , which describes void - crushing . figure [ fig : nvdparam]_b _ shows its effects . @xcite argue that @xmath138 is unlikely to be smaller than @xmath178 , because voids should continue to increase in size until turnaround . we therefore consider @xmath179 ( where voids are only destroyed when they lie inside of collapsed haloes ) and @xmath180 ( where void crushing is negligible ) . this parameter clearly does have an effect on @xmath165 : ignoring void crushing dramatically increases the number of small voids . however , it has no effect on large scales , because these voids can not be inside of extremely dense regions anyway . it also has little effect on the characteristic size @xmath150 . thus , while @xmath138 may affect the distribution of small voids , it is unlikely to affect the much more striking population of large voids . perhaps the most important simplification we have made is to ignore the scale dependence of @xmath62 so that we could use equation ( [ eq : nvm ] ) . figure [ fig : barriers ] shows that the threshold actually varies with the mass of the void . in principle , we should find the exact size distribution by following a procedure similar to @xcite , but with two absorbing barriers . however , it is relatively easy to see that in most cases the constant barrier approximation suffices for our purposes , though more sophisticated comparisons in the future may require a more exact solution . one straightforward test is to take @xmath62 to be constant but to evaluate its amplitude at several different scales . in figures [ fig : nvoid][fig : nvdparam ] , we have used @xmath181 . in figure [ fig : nvdbarr]_a _ , we show the corresponding distributions for the threshold evaluated at @xmath182 , and @xmath183 ( solid , dashed , and dotted curves ) . we show results for @xmath184 , and @xmath4 ( left , middle , and right sets ) . clearly , for deep surveys @xmath165 is nearly independent of the scale at which we evaluate the barrier . this is , of course , because the corresponding thresholds in figure [ fig : barriers]_a _ are nearly independent of @xmath56 . + however , if @xmath95 , the barrier rises rapidly at small scales . from figure [ fig : nvdbarr]_a _ , the distribution is still nearly independent of scale so long as @xmath185 , but using @xmath186 gives entirely different results . of course , this choice is not self - consistent , because the distribution is dominated by much larger voids . to better estimate the effects of this `` moving '' barrier , we note that it is approximately a linear function in @xmath56 . we would therefore ideally like to solve for the mass function in the presence of a constant barrier @xmath138 and a linearly increasing barrier @xmath187 . unfortunately , we were unable to find an analytic solution for that distribution ( though a straightforward numerical solution may be possible by extending @xcite ) . we therefore instead turn to a model in which the void - crushing barrier is also linear with the same slope as the void barrier . in the appendix , we show that the appropriate mass function is @xmath188 ( note that this differs from the solution presented by @xcite , which corresponds to a case with two barriers of opposite slopes ; see the appendix for a more detailed discussion . ) treating @xmath138 as a linear barrier is probably not a bad approximation . for one , the amplitude of @xmath138 does not strongly affect @xmath165 , as shown by figure [ fig : nvdparam]_b_. moreover , most trajectories that cross @xmath62 would have crossed the void - crushing barrier at small @xmath56 , when @xmath189 . finally , the appropriate choice of @xmath138 is somewhat arbitrary anyway , so a linear model may actually be more accurate . the similarity of @xmath190 and @xmath165 suggests that the linear barrier will have only a small effect on the distribution . in the cases of interest , @xmath191 , so the first exponential factor will be larger than unity and more mass will lie inside of voids . this is intuitively obvious , because an increasing negative barrier is easier to cross . the second term affects the shape of the mass function , but it is only important for large @xmath192 ; in the regime we study , this factor is always smaller than unity . figure [ fig : nvdbarr]_b _ confirms these expectations . for @xmath95 , the shape remains nearly invariant ( except at the smallest scales ) but more mass lies inside of voids . most importantly , the characteristic scale @xmath150 is nearly identical in the two cases . the linear barrier makes no difference to fainter galaxy samples , because the corresponding barrier is so nearly flat . we are now in a position to compare our model to the observed distribution of voids . @xcite compiled the most complete sample to date from the 2df galaxy redshift survey . they used the _ voidfinder _ algorithm @xcite to identify voids with @xmath170 in a volume - limited sample of galaxies with @xmath193 . they found that @xmath11 of the universe is contained inside voids with @xmath170 ; the mean effective radius of their sample was @xmath194 , with a maximum of @xmath195 . their algorithm separates galaxies into `` wall '' and `` void '' populations ; the latter can sit inside of voids but the former are excluded . they measured a mean galaxy underdensity @xmath196 within voids . with this magnitude limit , the @xcite sample should find voids somewhere between the two dot - dashed curves in figure [ fig : nvoid ] or near the thin curves in figure [ fig : nvdgal ] . given the measured @xmath196 , the most naive comparison is to the @xmath164 curve in the latter , which has @xmath197 and @xmath198 . both of these are considerably smaller than the observed value , but it is unclear how significant the discrepancy is . unfortunately , identifying voids at the predicted characteristic scale is difficult , because poisson fluctuations in the galaxy distribution are substantial on mpc scales . as a result , it is impossible to know whether an observed galaxy underdensity with @xmath199 corresponds to a true dark matter underdensity or a statistical fluctuation . largely for this reason , @xcite restrict their search to voids with @xmath170 . we therefore do not necessarily consider the small @xmath150 predicted by our model to conflict with observations ; however , the discrepancy in @xmath161 for large voids is more worrying . we do remark that a large population of small voids have clear observational signatures , increasing the number of @xmath200 empty or nearly empty regions well above those expected in a random distribution . this would manifest itself in such statistical measures as the void probability function ( the probability that a region has zero galaxies ; @xcite ) or the underdensity probability function ( the probability that a region has a given underdensity ; @xcite ) , but such statistics are ( at least for the moment ) beyond the capabilities of our model . there are several more subtleties in a comparison to observations . most importantly , the _ voidfinder _ algorithm does not use a density threshold , so the appropriate choice of @xmath49 is not obvious . the only `` void '' galaxies allowed in this approach are those that are well - isolated . but not all galaxies inside real voids need be isolated : if some fraction of the wall galaxies are actually contained inside regions with @xmath201 , the true voids would be less underdense than allowed by the algorithm . indeed , the @xmath86 curves in figure [ fig : nvoid ] predict @xmath202@xmath203 and @xmath204@xmath205 , much closer to the observed values . there are solid physical reasons to expect confusion between genuine void and wall galaxies . spherical underdensities tend to evolve by evacuating their interiors , with mass accumulating in a shell near the edge ( see , e.g. , @xcite ) . @xcite showed that massive haloes pile up at the edges even more strongly . thus it is reasonable to expect that many of the bright galaxies belonging to a large - scale underdensity lie near its edges , where they are difficult to identify unambiguously . moreover , real surveys operate in redshift space . because voids are expanding more rapidly than the hubble flow , their observed volumes will be larger than the physical volumes used by our model . @xcite show that the enhancement can be @xmath206 near shell - crossing . an equally important consideration is the geometrical method used by _ voidfinder _ to identify voids . it fills the gaps between wall galaxies with spheres and then merges overlapping spheres into discrete entities . this procedure may join two neighboring voids that our approach would consider distinct , especially if the wall between them is weak or if we include poisson fluctuations . galaxies on void walls tend to escape along the walls when neighboring voids merge @xcite , making this picture even more likely . finally , voids are not uniform ; rather , they are composed of a patchwork of ( deeper ) voids separated by weak filaments @xcite . in the excursion set picture , this is the analog of the progenitor distribution of collapsed objects and could in principle be computed through similar techniques @xcite , with the important difference that `` sub - voids '' can more easily retain their distinct identity as deeper underdensities in the galaxy distribution . the interaction of this structure with void selection algorithms is not trivial . another clear prediction of our model is that voids become larger as @xmath72 increases . @xcite addressed this question with statistical techniques . specifically , they computed the void probability function and underdensity probability function for volume - limited samples with a range of magnitude thresholds ( from @xmath207 to @xmath208 ) . they found that both the fraction of the universe filled by voids and their characteristic size increase with the luminosity threshold , in qualitative agreement with our results . a more precise comparison requires an extension of our model to these statistics . in summary , although our model predicts a significant population of large voids , they are still somewhat smaller than the observed structures . this could be a real discrepancy in which case voids clearly require more sophisticated modeling or it could be a combination of the contrasting selection techniqes , the internal void structure , and redshift - space effects . however , our qualitative conclusions ( large voids , increasing in size as the underlying galaxy population brightens ) do match the observed trends , a significant step forward for analytic models . unfortunately , precise comparisons to observations will still require numerical simulations . another useful comparison is to voids inside cosmological simulations . in particular , @xcite used the _ voidfinder _ algorithm to identify voids inside an @xmath13-body simulation . they used a semi - analytic model to place galaxies within dark matter haloes and selected the voids based on the predicted properties of the galaxies . they presented the void size distributions for several different survey depths . our results compare favorably with theirs . like us , they found larger voids for brighter galaxies as well as larger voids for galaxies compared to the dark matter ( though below @xmath209 the two appear to converge ; this may be related to the increasing importance of poisson fluctuations ) . they also found that the size distribution only peaks at @xmath210 for the brightest galaxies ( @xmath211 and @xmath212 in their models ) ; for fainter galaxies their distributions continue rising to @xmath213 ( below which their selection technique becomes incomplete ) . finally , they find that voids typically have @xmath214 and @xmath215 , with both quantities decreasing rapidly toward the center of the void . the density increases sharply near , but slightly beyond , the nominal void radius . this provides a hint that the density structure may play a role in the _ voidfinder _ selection , making direct comparisons with our predictions somewhat more difficult . nevertheless , the good qualitative , and even reasonable quantitative , agreement of our model with these simulations provides strong support for the major features of our approach . @xcite recently examined voids in a series of large - volume @xmath13-body simulations . they found voids to be much smaller than our predictions , with @xmath216 of the volume contained in voids with @xmath217 . however , they selected voids based on a dark - matter density criterion @xmath218 ( corresponding to shell - crossing ) . this is much more restrictive than our galaxy - density criterion . the best comparison is to the @xmath219 survey in figure [ fig : nvoid ] , whose @xmath62 is close to shell - crossing ( see fig . [ fig : barriers ] ) . for this sample , we find @xmath220 , with much of the universe in smaller voids . the remaining discrepancy may result from the search algorithm . they began their void searches around local minima in the density field , extending them until the enclosed density exceeded @xmath221 . such a prescription may tend to find `` sub - void '' progenitors rather than the larger structures that correspond to the observed voids in the galaxy distribution . we have described a simple analytic model for voids in the galaxy distribution . our model is based on @xcite , who showed how to apply the excursion set formalism to underdensities . its most important parameter is @xmath62 , the linearized underdensity of a void . those authors originally set @xmath62 to be the density corresponding to shell - crossing . however , this condition produces voids much smaller than the observed structures , whose masses can exceed those of galaxy clusters @xcite . our major contribution has been to show how to define @xmath62 through the _ galaxy _ underdensity . because galaxies are biased relative to the dark matter , voids can be nearer to the mean density than the observations naively indicate . this significantly increases the characteristic void size and the volume filling fraction . our model predicts voids with characteristic radii @xmath222@xmath223 . this is similar to , though somewhat smaller than , the voids found in galaxy redshift surveys @xcite . it is a closer match to the void population in semi - analytic galaxy formation models @xcite . because bright galaxies are more highly biased , our model also predicts that voids selected from shallow surveys should be characteristically larger , in agreement with observations @xcite . however , we also predict larger galaxy densities inside voids than observed . the significance of these discrepancies is not clear : we argued in [ obs ] that the algorithms used to identify voids in surveys , redshift - space distortions , and the internal structure of voids are all important in detailed comparisons to the observations . the last of these is especially crucial , because simulations show that galaxies inside voids tend to congregate near their edges @xcite , where they are difficult to separate from genuine `` wall '' galaxies . two other subtleties may also affect the comparison . the first is the nonlinear evolution of substructure within the void ; we have used linear theory to describe the halo population inside voids , which likely breaks down in some regimes because power can be transferred between scales . the second is poisson variation in the galaxy number counts @xcite . this effect makes direct detection of small voids impossible , forcing existing surveys to search only for voids with @xmath224 . thus it is difficult to make a precise comparison between our simple model and observed voids . however , the qualitiative agreement is reasonable . our analytic approach is the first to contain voids with characteristic radii @xmath209 and hence to be in even qualitative agreement with the observations . our model shows that the enormous extent of observed voids is not particularly surprising and should not be perceived as a `` crisis '' for the cdm model . by using the excursion set approach , we have also self - consistently predicted the halo population inside of voids . in agreement with naive expectations , small haloes should be less underdense than massive haloes . but the discrepancy is by no means large typically smaller than a factor of two over three decades in halo mass . including satellite galaxies in this calculation ( which we have not done ) will decrease the difference . thus , the predicted steepening of the mass function in low - density environments is only modest and is not ruled out by existing observations of the galaxy luminosity function @xcite . we also emphasize that these observations determined the environmental density on relatively small scales ( @xmath225 ) , where random fluctuations in the galaxy field may mimic true underdensities . if so , differences between mean and low - density environments will be further washed out . moreover , on such scales , finite - size effects become important in setting the characteristic luminosity @xmath122 . the test can be sharpened by focusing on the luminosity function within large , easily identified voids although , even there , separating void and wall galaxies may be difficult . thus , we find no compelling reason to believe that galaxy formation differs inside and outside of voids although we have no evidence against such a possibility , either . statistical properties of the galaxy distribution can also be used to test our predictions , especially for small voids that can not be identified unambiguously . the void probability function quantifies the probability that a sphere of a given size contains no galaxies @xcite , and the underdensity probability function quantifies the probability that a sphere is more underdense than some specified value @xcite . these measures are free from bias in the void selection process , but they are harder to predict directly from the excursion set formalism and are less useful in identifying galaxies that reside in voids . our model is in qualitative agreement with the observed trends in these statistics @xcite , but more work is needed to connect our formalism to them . r. , loeb a. , 2004 , apj , 609 , 474 a. j. , hoyle f. , torres f. , vogeley m. s. , 2003 , mnras , 340 , 160 e. , 1985 , apjs , 58 , 1 m. r. , et al . , 2003 , apj , 592 , 819 g. r. , da costa l. n. , goldwirth d. s. , lecar m. , piran t. , 1992 , apj , 388 , 234 j. r. , cole s. , efstathiou g. , kaiser n. , 1991 , apj , 379 , 440 r. , mo h. j. , sheth r. k. , boerner g. , 2002 , mnras , 333 , 730 r. , 2005 , submitted to apj ( astro - ph/0507014 ) j. m. , sheth r. k. , diaferio a. , gao l. , yoshida n. , 2005 , mnras , 360 , 216 s. , kaiser n. , 1989 , mnras , 237 , 1127 c. , et al . , 2005 , submitted to apj ( astro - ph/0508250 ) a. , 2005 , submitted to mnras ( astro - ph/0505421 ) d. j. , et al . , 2005 , mnras , 356 , 1155 v. , geller m. j. , huchra j. p. , 1986 , apj , 302 , l1 j. , da costa l. n. , goldwirth d. s. , lecar m. , piran t. , 1993 , apj , 410 , 458 j. , einasto m. , gramann m. , 1989 , mnras , 238 , 155 g. , & frenk c. s. , & white s. d. m. , & davis m. , 1988 , mnras , 235 , 715 h. , piran t. , 1997 , apj , 491 , 421 h. , piran t. , 2000 , mnras , 313 , 553 h. , piran t. , dacosta l. n. , 1997 , mnras , 287 , 790 j. a. , goldreich p. , 1984 , apj , 281 , 9 y. , piran t. , 2001 , apj , 548 , 1 s. r. , mcquinn m. , hernquist l. , 2005 , mnras , in press ( astro - ph/0507524 ) d. m. , jones t. d. , hoyle f. , rojas r. r. , vogeley m. s. , blanton m. r. , 2005 , apj , 621 , 643 d. m. , vogeley m. s. , 2004 , apj , 605 , 1 s. , okas e. l. , klypin a. , hoffman y. , 2003 , mnras , 344 , 715 s. a. , thompson l. a. , 1978 , apj , 222 , 784 d. j. , 1977 , mnras , 179 , 351 f. , rojas r. r. , vogeley m. s. , brinkmann j. , 2005 , apj , 620 , 618 f. , vogeley m. s. , 2002 , apj , 566 , 641 f. , vogeley m. s. , 2004 , apj , 607 , 751 a. , frenk c. s. , white s. d. m. , colberg j. m. , cole s. , evrard a. e. , couchman h. m. p. , yoshida n. , 2001 , mnras , 321 , 372 r. p. , oemler a. , schechter p. l. , shectman s. a. , 1981 , apj , 248 , l57 a. v. , berlind a. a. , wechsler r. h. , klypin a. a. , gottl " ober s. , allgood b. , primack j. r. , 2004 , apj , 609 , 35 c. , cole s. , 1993 , mnras , 262 , 627 h. white s. d. m. , 1993 , mnras , 282 , 347 v. , arbabi - bidgoli s. , einasto j. , tucker d. , 2000 , mnras , 318 , 280 h. , white s. d. m. , 2002 , mnras , 337 , 1193 m. , furlanetto s. r. , hernquist l. , zahn o. , zaldarriaga m. , 2005 , apj , in press ( astro - ph/0504189 ) p. j. e. , 1980 , the large - scale structure of the universe . princeton : princeton university press p. j. e. , 2001 , apj , 557 , 495 t. , lecar m. , goldwirth d. s. , da costa l. n. , blumenthal g. r. , 1993 , mnras , 265 , 681 w. h. , schechter p. , 1974 , apj , 187 , 425 r. k. , lemson g. , 1999 , mnras , 304 , 767 r. k. , tormen g. , 1999 , mnras , 308 , 119 r. k. , tormen g. , 2002 , mnras , 329 , 61 r. k. , van de weygaert r. , 2004 , mnras , 350 , 517 d. n. , et al . , 2003 , apjs , 148 , 175 y. , sato k. , sato h. , 1984 , progress of theoretical physics , 71 , 938 r. b. , somerville r. s. , trentham n. , verheijen m. a. w. , 2002 , apj , 569 , 573 r. , van kampen e. , 1993 , mnras , 263 , 481 m. s. , geller m. j. , huchra j. p. , 1989 , baas , 21 , 1171 m. s. , geller m. j. , park c. , huchra j. p. , 1994 , aj , 108 , 745 s. d. m. , 1979 , mnras , 186 , 145 j. , hui l. , 2005 , submitted to apj ( astro - ph/0508384 ) @xcite showed how to derive equation ( [ eq : nvm ] ) using the laplace transform . here we present an alternate derivation ( similar to @xcite ) and generalize it to the case of two linear barriers with equal slope . we begin by considering two constant barriers @xmath226 and @xmath227 . the distribution of trajectories @xmath228 , where @xmath229 , obeys the diffusion equation @xmath230 the appropriate boundary conditions are @xmath231 and @xmath232 , where @xmath233 is the dirac delta function . we first make the simple transformation @xmath234 . we then assume @xmath235 ; equation ( [ eq : diff ] ) becomes @xmath236 where @xmath7 is a constant . the general solutions are @xmath237 and @xmath238 . the boundary conditions @xmath239 select the sine function for the latter and force its argument to take discrete values ; thus @xmath240 . \label{eq : q1}\ ] ] we fix the constants @xmath241 by matching to the function @xmath242 . thus @xmath243 , \label{eq : q2}\end{aligned}\ ] ] where @xmath244 . the total rate at which trajectories disappear from the permitted region is @xmath245_0^{b_p - b_v } , \label{eq : flow}\ ] ] where we have used equation ( [ eq : diff ] ) . we interpret the two terms on the right hand side as the rate at which trajectories flow across the two barriers . thus the first - crossing distribution for the void barrier is @xmath246 , \label{eq : fc - con}\end{aligned}\ ] ] which yields equation ( [ eq : nvm ] ) after converting to mass units . we now consider the case of two absorbing barriers linear in @xmath247 , @xmath248 and @xmath249 . we make the coordinate transformation @xmath250 ; then the boundary conditions on @xmath251 become @xmath252 and @xmath253 , where @xmath254 . the boundary condition on @xmath247 becomes @xmath255 the diffusion equation ( [ eq : diff ] ) becomes @xmath256 we again assume a separable solution @xmath257 . the general solutions are exponentials , and the boundary conditions on @xmath258 select the sine function with discrete arguments . thus the solution is again a fourier sine series in @xmath251 . the fourier coefficients @xmath241 can be fixed by matching the solution to equation ( [ eq : bcs ] ) . we then find that @xmath259 \ , \nonumber \\ & & \times \exp \left [ - b_1 b_v - \frac{b_1 ^ 2 s}{2 } - \frac{n^2 \pi^2 s}{2 ( b_p - b_v)^2 } \right ] ; \label{eq : ql1}\end{aligned}\ ] ] note the similarity to equation ( [ eq : q2 ] ) . the rate at which trajectories cross the barriers is @xmath260_0^{u_1 } - b_1 [ q]_0^{u_1}. \label{eq : flowlin}\ ] ] but @xmath261 ; thus we can again identify the first - crossing distribution of the linear void barrier @xmath262 with the @xmath263 component of the first part , so @xmath264 from which the mass function follows directly . note that this expression differs from equation ( c10 ) of @xcite . those authors tried to derive the same distribution through laplace transforms . their solution required the first - crossing distribution of @xmath138 and @xmath62 to be identical ( their equation c6 ) ; this is not the case if the barriers have equal slope @xmath265 , because then they are not symmetric about the @xmath57 axis . instead their solution applies to a case in which the slopes of the two barriers have opposite signs .

it is shown that the mixing of lepton doublets of the standard model can yield sizeable contributions to the lepton asymmetry , that is generated through the decays of right - handed neutrinos at finite temperature in the early universe . when calculating the flavour - mixing correlations , we account for the effects of yukawa as well as of gauge interactions . we compare the freeze - out asymmetry from lepton - doublet mixing to the standard contributions from the mixing and direct decays of right - handed neutrinos . the asymmetry from lepton mixing is considerably large when the mass ratio between the right - handed neutrinos is of order of a few , while it becomes maxwell - suppressed for larger hierarchies . for an intermediate range between the case of degenerate right - handed neutrinos ( resonant leptogenesis ) and the hierarchical case , lepton mixing can yield the main contribution to the lepton asymmetry . tum - hep-858 - 12 + ttk-12 - 40 * baryogenesis from mixing of lepton doublets * bjrn garbrecht + 0.2 cm _ physik department t70 , james - franck - strae , + technische universitt mnchen , 85748 garching , germany + and institut fr theoretische teilchenphysik und kosmologie , + rwth aachen university , 52056 aachen , germany _ + 1.4 cm

1210.0553

sources for @xmath0-violating effects are often categorised into contributions from mixing and from direct decays . this applies to leptogenesis @xcite as well , where usually the mixing @xcite and the direct decays of right - handed singlet neutrinos @xmath1 are accounted for . of particular interest is the source from mixing , because in the situation where the mass - difference of the right - handed neutrinos is small , it gives rise to a resonantly enhanced @xmath0-asymmetry @xcite . besides the right - handed neutrinos , higgs bosons and standard model lepton doublets directly take part in leptogenesis . direct decay asymmetries from higgs bosons and lepton doublets are usually not considered , because in the vacuum , these particles can not decay into heavy right - handed neutrinos . consequently , at zero temperature , there are no kinematic cuts through the propagators of a right - handed neutrino and a lepton- or higgs - doublet that can lead to @xmath0-violation . at finite temperature however , these cuts contribute to the asymmetry @xcite , since the plasma continuously absorbs and produces leptons and higgs bosons . concerning mixing , right - handed neutrinos , as these are sterile , are well suited for the efficient production of asymmetries since flavour off - diagonal ( where we define the flavour of these particles by their mass - eigenstates , in contrast to standard model leptons ) correlations are only damped by the yukawa couplings , which are small by the requirement of a substantial deviation of the right - handed neutrinos from equilibrium . gauge interactions , in particular in the standard model , are typically much larger than the yukawa couplings of the right - handed neutrinos . however , as these are flavour blind , they do not directly damp off - diagonal flavour correlations . therefore , also the gauged particles within the standard model or its extensions can lead to substantial @xmath0-violation from mixing . for multiple higgs doublets , it is shown in ref . @xcite , that their mixing can be a viable source for leptogenesis . in the present work , we show that the mixing of the lepton doublets of the standard model contributes a sizeable amount of @xmath0-violation to leptogenesis as well . ( 100,60)(0,0 ) ( -25,0 ) ( 50,25)(24.3,0,180 ) ( 50,25)(25.7,0,180 ) ( 75,25)(25,25)4 ( 25,25)3 ( 75,25)3 ( 75,25)(100,25 ) ( 50,0)[](a ) 1.5 cm ( 100,60)(0,0 ) ( -25,0 ) ( 50,25)(25,0,180 ) ( 75,24.3)(25,24.3)4 ( 51,25)(49,25 ) ( 75,25.7)(25,25.7)4 ( 25,25)3 ( 75,25)3 ( 75,25)(100,25 ) ( 50,0)[](b ) 1.5 cm ( 100,60)(0,0 ) ( -25,0 ) ( 50,25)(25,0,180 ) ( 75,25)(25,25)4 ( 25,25)3 ( 75,25)3 ( 75,24.3)(100,24.3 ) ( 87,25)(88,25 ) ( 75,25.7)(100,25.7 ) ( 50,0)[](c ) in order to calculate the generation of the asymmetries in the finite temperature background , we use the closed - time - path ( ctp ) method @xcite . the diagrammatic representation of the source terms for the asymmetry in figure [ figure : collison : source ] indicates how the mixing of the various species yields contributions to leptogenesis . when we assume that the right - handed neutrinos are heavier than the lepton doublets and higgs bosons , the imaginary parts of the self - energies that appear within the resummed propagators in figure [ figure : collison : source](b ) and [ figure : collison : source](c ) are purely thermal , as the corresponding @xmath2 processes are kinematically forbidden in the vacuum . this resembles the situation for leptogenesis from the mixing of right - handed neutrinos with masses of the @xmath3 scale or below @xcite , where the @xmath0-violating cut is dominated by purely thermal contributions @xcite . for the direct @xmath0-violation , the contribution of such non - standard cuts is calculated in ref . @xcite . in section [ sec : off : corr ] , we derive the off - diagonal correlations within the distributions of the left - handed lepton doublets . even though gauge interactions do not directly violate flavour and damp its correlations , they catalyse the decay of flavour correlations in the presence of flavour - dependent masses or interactions . for flavoured leptogenesis , the relevance of such effects is discussed in ref . @xcite . here , we show that gauge interactions effectively limit the maximal resonant enhancement of a @xmath0-asymmetry resulting from the mixing of gauged particles . for the mixing of higgs doublets , this aspect is not discussed in ref . @xcite , a shortcoming that is addressed in section [ sec : higgs : mix ] . an important aspect within the present context is the role of lepton - number violation and conservation . the spinor trace of diagram [ figure : collison : source](c ) , that quantifies the rate of @xmath0-violation from mixing of lepton doublets , contains no lepton - number violation that could be described by insertions of an odd number of majorana mass - terms . therefore , lepton number must in some sense be conserved . this , we show in section [ sec : cancellation ] , where it turns out that the source for the asymmetry in the lepton doublets cancels the helicity asymmetry within the right - handed neutrinos . the latter is however rapidly violated by the majorana masses , provided the right - handed neutrinos are non - relativistic , as it is the case in the strong washout regime . in section [ sec : example : strongwashout ] , we then calculate the freeze - out value of the asymmetry that results from mixing lepton doublets in the strong washout regime . this , we compare to the asymmetry from the standard sources , _ i.e. _ from the mixing and direct decays of right - handed neutrinos . for illustrative purposes , we choose a particular slice in parameter space and vary the mass ratio of the right - handed neutrinos between one ( degenerate case ) and a few . it is a characteristic feature of the asymmetry from lepton mixing , that it is exponentially small for large mass ratios , as the @xmath0-violating cut is purely thermal and becomes maxwell suppressed in that situation . we close with summarising and concluding remarks , that are given in section [ sec : sum : disc ] . the model that we consider here is the same that underlies flavoured leptogenesis @xcite and is given by the lagrangian @xmath4 the standard model higgs doublet is denoted by @xmath5 , the lepton doublets by @xmath6 , right - handed charged leptons by @xmath7 and right - handed neutrinos by @xmath1 . the associated spinors are denoted by @xmath8 with the subscript of the respective field . we use the definition @xmath9 , but suppress explicit indices of the @xmath10 gauge group in the following . we transform the @xmath1 to a basis where @xmath11 is diagonal and write @xmath12 . furthermore , we take @xmath13 to be diagonal , where @xmath14 . we are mainly interested in the dynamics within the temperature range @xmath15 , such that only the @xmath16-lepton yukawa couplings are in equilibrium . in that case , we can neglect @xmath17 and @xmath18 and then perform a unitary transformation of the lepton flavours @xmath19 and @xmath20 to @xmath21 and @xmath22 such that @xmath23 , see ref . @xcite for the explicit construction of such a transformation . ( after this transformation , @xmath24 is no longer diagonal within the block of the flavours @xmath25 and @xmath16 . as we may however neglect these yukawa couplings at temperatures above @xmath26 , @xmath24 can still be considered as effectively diagonal ) . then , only lepton flavour asymmetries within @xmath27 and @xmath22 are produced , while there is no asymmetry in the remaining linear combination @xmath21 , since @xmath21 couples only to one of the right - handed neutrinos , and no @xmath0-violating interference terms occur . while it is the salient feature of flavoured leptogenesis that all @xmath0-odd off - diagonal correlations between the flavours @xmath16 and @xmath28 are rapidly erased , because the interactions mediated by @xmath29 are in equilibrium , we find in the present work that the deviation of the right - handed neutrinos from equilibrium does generically sustain @xmath0-even [ _ ( [ lepcon : orderysqr ] ) below ] off - diagonal correlations of the leptons [ eq . ( [ sol : deltan ] ) below ] , which in turn enter the source term for the @xmath0-violating lepton asymmetry . we also note that the present discussion does not directly apply but can easily be generalised to temperatures below @xmath26 , where the flavours @xmath25 and @xmath30 are distinguishable . the present calculations based on the ctp formalism require numerous definitions that can be found within refs . @xcite . in particular @xmath31 denotes fermionic and @xmath32 scalar wightman functions . subscripts on these indicate the particular fields and flavour correlations . distribution functions for particles and anti - particles are denoted by @xmath33 and @xmath34 , number densities by @xmath35 and @xmath36 . again , fields and flavour correlations are indicated by subscripts . a prefix @xmath37 implies that the particular quantity is the deviation from the equilibrium propagator , distribution function or density . the dynamics of the out - of equilibrium distributions of left - handed lepton doublets can be described by the kinetic equation @xcite @xmath38_{ab } \,,\end{aligned}\ ] ] where @xmath39 is the conformal time in the expanding universe . the functions @xmath40 appear within the spinor components of the left - handed lepton propagator @xmath41 and factorise into statistical and spectral terms in the narrow width limit @xcite . when @xmath42 , where @xmath43 , we may approximate [ gell : onshell ] @xmath44 for @xmath45 , this approximation , in particular the relation between the sign of @xmath46 and the helicity , becomes inaccurate due to the presence of hole modes . however , this region of soft momenta covers only a small part of the phase - space , such that the approximations ( [ gell : onshell ] ) are sufficiently accurate for the present purposes . the flavour - dependent self - mass terms @xmath47 are given by @xmath48_{ab } t^2}{16|\mathbf k| } \,,\end{aligned}\ ] ] where we have neglected the contributions from the couplings @xmath49 . it is useful to decompose the collision term as @xmath50 into contributions mediated by the yukawa couplings @xmath49 and @xmath24 and the gauge couplings @xmath51 . within @xmath52 , we collect the @xmath0-conserving processes mediated by @xmath49 , which are to leading approximation quadratic in @xmath49 . the term @xmath53 is mediated by @xmath49 as well and accounts for @xmath0-violating processes . to leading approximation , it is fourth order in @xmath49 . the part of the lepton self - energy mediated by @xmath49 is @xmath54 where @xmath55 are ctp indices . the term @xmath56 is of importance for two reasons : first , as in the standard scenario of leptogenesis , its diagonal components describe the washout of the @xmath0-odd lepton asymmetries accumulated in the diagonal correlations of lepton flavours . ( for flavoured leptogenesis , the off - diagonal correlations are rapidly erased due to the yukawa coupling @xmath29 . ) second , its off - diagonal components correspond to the source of @xmath0-even , flavour off - diagonal correlations in the lepton doublets . as the @xmath0-odd off - diagonal correlations are suppressed by the yukawa interactions @xmath29 , and the @xmath0-even off - diagonal correlations are suppressed by an order of @xmath57 times the deviation of the right - handed neutrinos from equilibrium , when compared to the diagonal distribution functions , we can substitute the flavour diagonal components of @xmath58 for the purpose of calculating the contributions to @xmath56 that are of leading importance . we thus obtain @xmath59_{ab } p_{\rm r}\int\frac{d^4 p}{(2\pi)^4 } \left [ { \rm i}s^>_{nii}(p ) { \rm i}\delta^<_\phi(p - k){\rm i}s^<_{\ell bb}(k ) - { \rm i}s^<_{nii}(p ) { \rm i}\delta^>_\phi(p - k){\rm i}s^>_{\ell bb}(k ) \right ] \\\notag + & { \cal c}_{\ell ab}^{cp \rm v } \,,\end{aligned}\ ] ] where we take @xmath60 to be of equilibrium form . note that while the off - diagonal correlations within @xmath61 and @xmath62 , which are always out - of - equilibrium , can be neglected for calculating @xmath52 to second order in @xmath49 , within the @xmath0 violating source @xmath53 , these off - diagonal correlations are essential for the asymmetry that we calculate in eq . ( [ source : lepto ] ) below . for definiteness , we assume strong washout , _ i.e. _ @xmath63 at the time of leptogenesis . ( due to the exponential maxwell - suppression , masses of a factor of a few above the temperature are enough for sufficiently accurate approximations . ) this allows us to neglect quantum statistical blocking and enhancement factors on many occasions and to use maxwell distributions instead of fermi - dirac or bose - einstein distributions . we parametrise the deviation of the neutrino distribution from equilibrium by a pseudo - chemical potential @xmath64 as @xmath65 while in the present setup , there are no interactions that directly force kinetic equilibrium of the sterile neutrinos , it turns out that above parametrisation indeed gives a good approximation to the distribution function of sterile neutrinos in strong washout scenarios @xcite . we now aim to compute the off - diagonal distributions @xmath66 and @xmath67 , @xmath68 . in kinetic equilibrium , these can be inferred from the number densities of leptons and anti - leptons , @xmath69 ( note that @xmath70 on @xmath71 refers to the particle and anti - particle , whereas on @xmath72 , it refers to helicity . ) the prefix @xmath37 indicates that these quantities are deviations of the number densities from their equilibrium values . consequently , the charge density @xmath73 can deviate from zero . pair creation and annihilation processes tend to strongly suppress the combination @xmath74 . however , it turns out that for small yukawa couplings , the finite value of this combination limits the maximal resonant enhancement of the asymmetry , _ cf . _ ( [ sol : deltan ] ) below . for this reason , equations for @xmath71 rather than just @xmath75 need to be derived , as a generalisation of the methods introduced in ref . @xcite . in order to obtain these densities when the time - derivative in the kinetic equation ( [ kin : eq ] ) is small compared to the effective mass difference @xmath76 , it is useful to multiply these equations by @xmath77 and then to integrate over @xmath78 . doing so , we encounter on the right - hand side @xmath79 in the limit of strong washout , where the quantum statistical distributions can be approximated by maxwell distributions , we obtain @xmath80 clearly , this term is exponentially suppressed for large @xmath81 . it can be interpreted as the @xmath0-violating cut that appears in the resummed lepton propagator of figure [ figure : collison : source](c ) , and since the reaction @xmath82 is kinematically forbidden in the vacuum , this cut is purely thermal . the interactions mediated by the yukawa couplings @xmath24 lead to the decay of off - diagonal correlations of the lepton doublets . the relevant contribution to the collision term is given by @xcite @xmath83{\rm i}\delta s_{\ell cb}(k)\,.\end{aligned}\ ] ] in the integrated kinetic equations , then the following terms occur : @xmath84{\rm i}\delta s_{\ell cb}(k ) \\\notag = & -h^\dagger_{ac}h_{cd } \delta n^+_{\ell ab } \int\limits_0^\infty \frac{dk^0}{2\pi } \int\frac{d^3k}{(2\pi)^3 } \int\frac{d^4p}{(2\pi)^4 } \\\notag\times & { \rm tr}\left [ { \rm i}s^>_{\rm r cc}(p){\rm i}\delta^>_\phi(k - p ) -{\rm i}s^<_{\rm r cc}(p){\rm i}\delta^<_\phi(k - p ) \right ] 2 s^{\cal a}_{\ell cb}(k ) \frac{12\beta^3{\rm e}^{\beta k^0}}{({\rm e}^{\beta k^0}+1)^2 } \\\notag & = : -b_\ell^{\slashed{\rm fl } } \left[h^\dagger h \delta n^+_\ell\right]_{ab}\end{aligned}\ ] ] and likewise @xmath85_{ab}\,.\end{aligned}\ ] ] with these coefficients , the integrated kinetic equations are [ eff : kin ] @xmath86 in principle , there also occur terms @xmath87_{aa}+[y^\dagger y]_{bb})\delta n^\pm_{\ell ab}\end{aligned}\ ] ] on the right hand sides of these equations . however , we assume here that the dominant flavour - sensitive interaction is mediated by the lepton - yukawa couplings . in particular , @xmath29 dominates over the relevant elements of @xmath49 , which is valid provided leptogenesis occurs at temperatures below @xmath88 . the coefficient @xmath89 is the averaged rate for flavour - sensitive processes that are mediated by the coupling @xmath24 . as the higgs boson as well as the standard model leptons are massless in the symmetric electroweak phase , these reactions require the radiation of an additional gauge boson . a systematic calculation is currently on the way , but from the similar problem of the production of light right - handed neutrinos @xcite , we may estimate that @xmath90 , see the appendix [ app : rates ] . pair creation and annihilation processes may change @xmath91 . even though these processes are flavour blind , we will see below that in interplay with flavour - sensitive processes , they can be important for the suppression of flavour off - diagonal correlations . accounting for the large number of standard model degrees of freedom , these processes are dominated by the @xmath92-channel exchange of gauge bosons , and we estimate these within appendix [ app : rates ] as @xmath93 . when neglecting the time - derivative , eqs . ( [ eff : kin ] ) yield the induced off - diagonal ( @xmath68 ) correlations [ sol : deltan ] @xmath94 ^ 2+(h_{aa}^2+h_{bb}^2)b_\ell^{\slashed{\rm fl}}[2 b_\ell^g+(h_{aa}^2+h_{bb}^2)b_\ell^{\slashed{\rm fl } } ] } \\\notag & = : ( { \cal q}_{\ell ab}/t^2)\sum_i y_{ai}^\dagger y_{ib } b_i^y \ , , \\ \delta n_{\ell ab}^+ + \delta n_{\ell ab}^- & = - \frac { 2(h_{aa}^2+h_{bb}^2)b_\ell^{\slashed{\rm fl } } \sum_i y_{ai}^\dagger y_{ib } b_i^y } { \left[(h_{aa}^2-h_{bb}^2)t^2/16\right]^2+(h_{aa}^2+h_{bb}^2)b_\ell^{\slashed{\rm fl}}[2 b_\ell^g+(h_{aa}^2+h_{bb}^2)b_\ell^{\slashed{\rm fl } } ] } \,.\end{aligned}\ ] ] the general solution to eqs . ( [ eff : kin ] ) involves the damping rates @xmath95 ^ 2 } \right)\,,\end{aligned}\ ] ] and is given by @xmath96 this implies that it takes a finite time for these @xmath0-even off - diagonal correlations to build up . in the present case , the rate for this is given by @xmath97 . this in particular implies that in the regime of flavoured leptogenesis below @xmath88 , when the interactions mediated by @xmath29 are in equilibrium , flavour off - diagonal lepton correlations are present , and their magnitude is given by eqs . ( [ sol : deltan ] ) . we express the resonant enhancement of the asymmetries through the factor @xmath98 ^ 2 t^4+(h_{aa}^2+h_{bb}^2)b^{\slashed{\rm fl}}[2 b^g+(h_{aa}^2+h_{bb}^2)b^{\slashed{\rm fl } } ] } \,.\end{aligned}\ ] ] the fact that @xmath99 is suppressed but not vanishing implies that kinetic equilibrium mediated by pair creation and annihilation processes is only approximately established . processes of kinetic equilibration that do not mediate between @xmath100 and @xmath101 , _ i.e. _ which do not correspond to pair creation and annihilation , are however not conflicting with the solution ( [ sol : deltan ] ) . we therefore describe the distribution functions in terms of the number densities as [ lepdist : kineq ] @xmath102 notice also that from eqs . ( [ sol : deltan ] ) it follows that @xmath103 , @xmath104 and therefore @xmath105 . this implies that the flavour off - diagonal lepton correlations are @xmath0-even , @xmath106^t c^\dagger + { \cal o}(y^4)\,,\end{aligned}\ ] ] such that to this end , the lepton asymmetry is not yet broken . in this section , we calculate the @xmath0-odd charge asymmetries within the diagonal components of the distribution functions of lepton doublets as well as the helicity asymmetries within the singlet neutrinos , that are induced by the off - diagonal , @xmath0-even correlations of lepton doublets . in the strong washout scenario , helicity asymmetries within the right - handed neutrinos are quantitatively irrelevant , because of the majorana masses , which effectively erase such asymmetries when the right - handed neutrinos decay . however , helicity asymmetries can be of crucial importance in weak washout scenarios @xcite . in the present context , we are interested in the helicity asymmetries , because the lepton - doublet propagator and its self - energy correction do not encompass lepton - number violating combinations of operators , such that the total source for the lepton asymmetry plus the helicity asymmetry in the right - handed neutrinos should vanish . this constraint provides a useful consistency check that we perform in the present section . in spite of the vanishing of the total source , the asymmetry of the singlet neutrinos is completely washed out in the decays , that proceed via the majorana mass term within the strong washout regime . in contrast , the active leptons experience finite washout rates from the inverse decays , such that a total lepton asymmetry remains at freeze out . the self - energy for the right - handed neutrino is @xcite @xmath107^tc^\dagger{\rm i}\delta_\phi^{gf}(p - k ) \,,\end{aligned}\ ] ] where @xmath108 is the dimension of the @xmath10 representation of the lepton doublet . the self - energy enters into the collision term for the right - handed neutrino as @xmath109\,.\end{aligned}\ ] ] a @xmath0- and helicity - violating contribution @xmath110 is induced by the flavour off - diagonal correlations of the lepton doublets . we denote this as the source term @xmath111_{ij } \\\notag = & \frac12\sum\limits_{\underset{a\not = b}{a , b } } \int\frac{d^4 k}{(2\pi)^4 } \int\frac{d^4 p}{(2\pi)^4 } g_w y_{ia}y^\dagger_{bj } { \rm tr } \big [ { \rm i}\delta s_{\ell ab}(p ) \big ( { \rm i}\delta_\phi^>(k - p ) \left [ { \rm i}s^<_{nii}(k)+{\rm i}s^<_{njj}(k ) \right ] \\\notag & \hskip5.7 cm -{\rm i}\delta_\phi^<(k - p ) \left [ { \rm i}s^>_{nii}(k)+{\rm i}s^>_{njj}(k ) \right ] \big ) \big]\,.\end{aligned}\ ] ] we have made use here of the @xmath112 lepton number conservation , eq . ( [ lepcon : orderysqr ] ) , and assumed that there is no charge asymmetry within the higgs field , such that @xmath113 . moreover , the off - diagonal correlations within the right - handed neutrinos , which are important for the usually considered @xmath0-violating sources for leptogenesis , are neglected here . ( an example with a quantitative comparison of the source from mixing and decays of right - handed neutrinos and the source from mixing lepton doublets is presented in section [ sec : example : strongwashout ] . ) the source term can be expressed to enter the integrated kinetic equation for the right - handed neutrinos in the form @xmath114 = -{\cal s}_{n ij}\,.\end{aligned}\ ] ] the integral on the left hand side corresponds to the rate of producing a helicity asymmetry , which is in the limit of vanishing majorana masses the same as an asymmetry between right - handed neutrinos and anti - neutrinos . note that this integrated kinetic equation does not account for the total production of singlet neutrinos . in order to obtain these rates , the collision term is to be integrated from @xmath115 to @xmath116 rather than from @xmath117 to @xmath116 over @xmath118 @xcite . the factor @xmath119 on the right hand side is a suitable normalisation , because in writing the majorana field as a four component spinor , the number of degrees of freedom is doubled , though they are related by the majorana constraint . as for the lepton collision term @xmath52 , the approximation ( [ coll : y ] ) is used in order to calculate the @xmath0-even off - diagonal correlations in terms of the deviation of the right - handed neutrinos from equilibrium . for that purpose , we have taken @xmath120 to be of equilibrium form under the integral . now , we want to obtain the @xmath0-odd asymmetry in the lepton doublets that is induced by these off - diagonal correlations . it follows from integrating @xmath121 as @xmath122_{ab } \\\notag = & \frac12\sum\limits_{\underset{c\not = b}{i , c } } \int\frac{d^4 k}{(2\pi)^4 } \int\frac{d^4 p}{(2\pi)^4 } { \rm tr } \big [ \left ( p_{\rm r } { \rm i } s_{n ii}^>(p ) { \rm i}\delta_\phi^<(p - k ) -p_{\rm r } { \rm i } s_{n ii}^<(p ) { \rm i}\delta_\phi^>(p - k ) \right ) \\\notag & \hskip4.3 cm \times ( y^\dagger_{ai}y_{ic}{\rm i}\delta s_{\ell cb}(k ) + { \rm i}\delta s_{\ell ac}(k)y^\dagger_{ci}y_{ib } ) \big]\,.\end{aligned}\ ] ] besides this source , at fourth order in @xmath49 , there are also the standard contributions from direct @xmath0-violation in the decays and inverse decays of right - handed neutrinos as well as from their mixing @xcite . these add linearly to the asymmetry from lepton - doublet mixing , that is the main topic of the present work . for simplicity , we do not account for the standard sources of asymmetry in most of the following discussion and refer to refs . @xcite for the calculation of these contributions within the ctp formalism . in section [ sec : example : strongwashout ] however , we present a comparison of the asymmetry resulting from lepton mixing to the asymmetry that results from the standard sources . when expressing the kinetic equation for the lepton doublets as @xmath123 = -\int\frac{d^4 k}{(2\pi)^4}\frac12{\rm tr } \left [ { \cal c}_\ell(k ) + { \cal c}_\ell(k ) \right]_{ab}\end{aligned}\ ] ] and observing that @xmath124=-g_w{\rm tr}[{\cal s}_\ell]\,,\end{aligned}\ ] ] we see that the sum of the helicity asymmetry within the right - handed neutrinos and the charge asymmetry in the lepton - doublets is conserved , unless taking proper account of the helicity - flipping majorana masses @xmath11 . this is expected , since @xmath125 , the self - energy that leads to the mixing of the lepton doublets , does not encompass lepton - number violating operators in the form of odd powers of the majorana masses @xmath11 within the approximation represented by the diagram in figure [ figure : collison : source](c ) ( _ i.e. _ when accounting for lepton - doublet mixing only and not for the mixing and the direct decays of the right - handed neutrinos ) . now , since in the strong washout scenario , @xmath126 , the right - handed neutrinos are non - relativistic , and they decay at tree - level with equal likelihood into leptons and anti - leptons . this is different for relativistic right - handed neutrinos , that are converted to leptons via scattering processes with the particles in the plasma . these reactions are approximately helicity conserving . in the following , we calculate the asymmetry in the strong washout scenario , and for this reason , we only have to account for the production and the washout of the asymmetry within the leptons @xmath61 , because the helictiy asymmetry within the right - handed neutrinos is rapidly violated by the majorana masses . in the strong washout regime , all energies of particles that participate in decay and inverse decay processes are much larger than the temperature . one may therefore approximate eqs . ( [ lepdist : kineq ] ) by their maxwell forms [ lepdist : kineq : maxw ] @xmath127 we then substitute above approximations into eqs . ( [ sol : deltan ] ) , and these into the source term for the lepton asymmetry ( [ source : lepto ] ) . it enters the kinetic equations ( [ kin : eq ] ) , when integrating these over @xmath118 , @xmath128m_{ni}^3 t k_1(m_{ni}/t ) + \int\frac{d^4k}{(2\pi)^4}c_{\ell aa}^y(k ) \\\notag & = \sum_{\underset{c\not = a}{ijc } } \frac{3}{16\pi^3}\frac{m_{ni}^3}{t^4}k_1(m_{ni}/t ) b_j^y { \cal}{\cal q}_{\ell ca } { \rm i } \left ( y^\dagger_{ai}y_{ic}y^\dagger_{cj}y_{ja } -y^\dagger_{aj}y_{jc}y^\dagger_{ci}y_{ia } \right ) \\ \notag & + \int\frac{d^4k}{(2\pi)^4}c_{\ell aa}^y(k ) \,.\end{aligned}\ ] ] here , we have evaluated the source term ( [ source : lepto ] ) explicitly in the limit of strong washout , while the remaining integral corresponds to the washout term . we now calculate the freeze - out value of the asymmetry produced in the out - of - equilibrium decays of the individual @xmath1 in the strong washout regime . the total asymmetry is then obtained as the sum of the particular asymmetries . as usual , it is convenient to parametrise the evolution by @xmath129 the scale factor @xmath130 can be expressed through the comoving temperature @xmath131 as @xmath132 where @xmath133 is the number of relativistic degrees of freedom of the standard model at high temperatures . the physical temperature is given by @xmath134 . in order to take account of the expansion of the universe , in all kinetic equations written down to this end , the mass terms should be multiplied by the scale factor @xmath130 and all temperatures be replaced by the comoving temperature @xmath131 , _ cf . _ ref @xcite . the kinetic equations then take the form [ kineq : strongwashout ] @xmath135 where @xmath136 and @xmath137 , and @xmath92 is the entropy density . the various distributions , collision and source terms are ( _ cf . _ ref . @xcite ) @xmath138_{ii}}{[yy^\dagger]_{jj } } { \cal q}_{\ell ac } { \rm i } \left ( y^\dagger_{ai}y_{ic}y^\dagger_{cj}y_{ja } -y^\dagger_{aj}y_{jc}y^\dagger_{ci}y_{ia } \right ) \ , , \\ y_{nk}^{\rm eq}&= 2^{-1/2}\pi^{-3/2 } \left(\frac{m_{nk}}{m_{ni}}\right)^{3/2 } z_i^{3/2}a_{\rm r}^3{\rm e}^{-z_i m_{nk}/m_{ni}}/s\ , , \\ \bar{\cal c}_{nk}&=\frac{g_w}{16\pi}\sum\limits_ay_{ka}y^\dagger_{ak}a_{\rm r } z_i\frac{m_{nk}}{m_{ni}^2 } \ , , \\ \label{def : b : washout } \bar w^{ni}_{\ell a } & = -\sum\limits_k y^\dagger_{ak}y_{ka } \frac{3 z_i^{5/2}}{2^{9/2}\pi^{5/2 } } \left(\frac{m_{nk}}{m_{ni}}\right)^{5/2 } \frac{a_{\rm r}}{m_{ni}}{\rm e}^{-z_i m_{nk}/m_{ni } } = : \sum\limits_k b^{ni}_{\ell ak}z_i^{5/2}{\rm e}^{-z_i m_{nk}/m_{ni}}\,.\end{aligned}\ ] ] to the leading order in deviations from equilibrium , @xmath139 , the formal solution to eqs . ( [ kineq : strongwashout ] ) is given by @xcite @xmath140 where we define @xmath141_{jj } } \left(\frac{m_{nj}}{m_{ni}}\right)^\frac72 \\\notag \times & { \cal q}_{\ell ac } { \rm i } \left ( y^\dagger_{ai}y_{ic}y^\dagger_{cj}y_{ja } -y^\dagger_{aj}y_{jc}y^\dagger_{ci}y_{ia } \right)\,.\end{aligned}\ ] ] for simplicity , we now consider leptogenesis from two right - handed neutrinos @xmath142 . this is relevant in the situation when there are only two right - handed neutrinos in the theory , when additional right - handed neutrinos are much heavier than @xmath143 and @xmath144 or when these decouple due to the smallness of their yukawa couplings . the exponent in eq . ( [ yell : formal ] ) is then extremal for @xmath145 with @xmath146 which we solve numerically for @xmath147 , what determines the freeze - out temperature . steepest descent method then yields the flavoured freeze - out asymmetries ( _ cf . @xcite ) @xmath148 finally , the total lepton asymmetry is the sum @xmath149 in order to assess the relative importance of the asymmetry from mixing lepton doublets , we compare it to the lepton asymmetry that results from the wave function and vertex corrections to the decays of right - handed neutrinos @xmath62 . we can simply express @xmath150 where again , @xmath151 is the asymmetry that results form the decay of @xmath1 . the decay asymmetry is @xcite @xmath152_{ii}}\sum\limits_{\underset{j\not = i}{j , b } } \big\ { { \rm im}\left[y^\dagger_{ai}y^*_{ib}y^t_{bj}y_{ja}\right]\frac{\xi(x_j)}{\sqrt{x_j } } + { \rm im}\left[y^\dagger_{ai}y_{ib}y^\dagger_{bj}y_{ja}\right]\frac{2}{3(x_j-1 ) } \big\}\,,\end{aligned}\ ] ] where @xmath153 and @xmath154 \,.\end{aligned}\ ] ] the freeze - out temperature is now determined by @xmath155 where @xmath156 is given by eq . ( [ def : b : washout ] ) . the factor that appears in the formal solution ( [ yell : formal ] ) for @xmath157 is now @xmath158 we now pick some particular , but illustrative points in parameter space in order to compare the asymmetry from the mixing of lepton doublets to the asymmetry from the wave function and vertex corrections to the decay parameter of right - handed neutrinos . parameters consistent with neutrino oscillations can be constructed with the help of the relation @xcite @xmath159 where @xmath160 is the diagonal mass matrix of active leptons , @xmath161 and @xmath162 . the pmns matrix @xmath163 is @xmath164 with @xmath165 . the non - zero entries of the matrices @xmath166 are @xmath167 the @xmath168 are the mixing angles of the active neutrinos , and @xmath169 , @xmath170 and @xmath37 are @xmath0-violating phases . the matrix @xmath171 must satisfy @xmath172 . we parametrise it by the complex angles @xmath173 and the matrices @xmath174 with non - zero entries @xmath175 we take @xmath176 or alternatively , make the assumption that there is no right - handed neutrino @xmath177 or simply that the yukawa couplings of @xmath177 are negligibly small . the parameter space can then be restricted by the choice @xmath178 and @xmath179 . additional parameters that we choose are listed in table [ table : params ] . finally , @xmath180 is a unitary matrix that brings @xmath49 to a form , such that only asymmetries in two flavours @xmath16 and @xmath28 are produced , where @xmath28 is a linear combination of @xmath25 and @xmath30 , see ref . @xcite for the explicit construction . ( explicitly , this means that @xmath181 , while the remaining entries are generically non - zero , and @xmath182 is the linear flavour combination orthogonal to @xmath16 and @xmath28 . ) .[table : params ] parameters for the numerical example in figure [ fig : leplep ] . [ cols= " > , > " , ] the one remaining parameter is @xmath183 , that we vary for values larger than @xmath184 . in figure [ fig : leplep ] , we compare the resulting asymmetry from lepton - doublet mixing to the asymmetry from mixing and direct decays of right - handed neutrinos , where we take account of the decays of both , @xmath143 and @xmath144 . for @xmath185 ( degenerate regime ) , the asymmetry from the mixing of right - handed neutrinos is strongly enhanced , which is known as resonant leptogenesis . the asymmetry from lepton - doublet mixing also is strongest for @xmath186 . for larger values of @xmath183 , we clearly see the exponential suppression of the asymmetry , because the kinematic cuts that give rise to the @xmath0-violation are purely thermal and are maxwell - suppressed for large values of @xmath183 ( hierarchical regime ) , _ cf . _ eq . ( [ b : y ] ) . for intermediate values of @xmath183 , between the degenerate and the hierarchical regimes , the asymmetry from mixing lepton doublets dominates over the usually considered asymmetry from mixing and direct decays of right - handed neutrinos . this example should illustrate that for @xmath187 ( more precisely , when the right - handed neutrino masses are within a factor of a few ) , there generically is a sizeable contribution to the lepton asymmetry , that originates from lepton mixing . while we have considered here a particular slice of the parameter space , it would be interesting to perform a more systematic parametric study of leptogenesis from mixing lepton doublets in the future . in ref . @xcite , leptogenesis from mixing higgs doublets is introduced . the model is based on the lagrangian @xmath188 notice that the yukawa couplings @xmath49 here are different from those in the lagrangian ( [ lagrangian ] ) . using the same notations and conventions as in ref . @xcite , we obtain the off - diagonal correlations in the number densities of higgs and anti - higgs particles as @xmath189 where @xmath190 is the wightman function for the higgs field . the kinetic equations for the off - diagonal correlations of the higgs particles are then ( we choose a basis where @xmath191 , that may include thermal corrections , is diagonal ) @xmath192_{12}b_\phi^y -b_\phi^{{\slashed{\rm fl}}}\sum_{j}(y_{j1}^2+y_{j2}^2)\delta n_{\phi12}^\pm -b^g_\phi\left(\delta n_{\phi12}^+ + \delta n_{\phi12}^-\right)\,.\end{aligned}\ ] ] we have assumed here that the flavour - sensitive interactions are dominated by couplings @xmath193 of the higgs - fields @xmath194 to standard model fermions . note however that the couplings @xmath49 and the quartic interactions among the higgs fields are flavour sensitive as well and should be taken into account when the @xmath193 are not dominating . an estimate for the weighted annihilation rate @xmath195 is given by eq . ( [ bgphi : int ] ) in appendix [ app : rates ] . as for @xmath196 , we do not provide an estimate here , since it is model dependent . if one of the higgs fields @xmath5 is the standard model higgs field , this rate will be dominated by the top - quark yukawa couplings @xmath197 . because the scattering processes then also involve gluons , we may expect this rate to be enhanced by a factor of a few when compared to @xmath198 . ( [ kineq : higgs ] ) has the solution @xmath199_{12}b_\phi^y\ , , \\ { \cal r}_{\phi}= & \frac{m_{\phi11}^2-m_{\phi22}^2 } { \left(m_{\phi11}^2-m_{\phi22}^2\right)^2 + b_\phi^{{\slashed{\rm fl}}}\sum_{j}(y_{j1}^2+y_{j2}^2 ) \left [ b_\phi^{{\slashed{\rm fl}}}\sum_{j}(y_{j1}^2+y_{j2}^2 ) + 2 b_\phi^g \right ] } \ , , \\ b_\phi^y=&\frac{\mu_n m_n^{7/2 } t^{1/2}}{32\sqrt 2 \pi^{5/2 } } { \rm e}^{-m_n / t}\,.\end{aligned}\ ] ] the enhancement factor @xmath200 is analogous to the quantity ( [ qlab ] ) @xmath201 for mixing leptons , but we have choose a different dimensionality for convenience and the sake of comparison with ref . @xcite . following ref . @xcite , we then obtain for the freeze - out asymmetry @xmath202}{{\rm tr[yy^\dagger ] } } \frac{135\,m_n^2}{64\pi^6 g_\star}{\cal r}_\phi \sqrt{\frac{{\rm e}^{z_{{\rm f}i}}}{\sqrt{z_{{\rm f}i}}b_i } } { \rm e}^{-2z_{{\rm f}i}-\int\limits_{z_{{\rm f}i}}^\infty dz b_i z^{5/2}{\rm e}^{-z}}\ , , \,\\ b_i&=[y^\dagger y]_{ii } \frac{9}{32\pi^4}\sqrt{\frac{5}{2g_\star}}\frac{m_{\rm pl}}{m_n}\ , , \,\\ z_{{\rm f}i}&=-\frac{5}{2}w_{-1}\left((-{2}/{5})\times\left({2}/{b_i}\right)^{2/5}\right)\,.\end{aligned}\ ] ] ( note a missing factor @xmath203 in the expression for @xmath204 in ref . @xcite . ) we have calculated the mixing of standard model lepton doublets in the early universe and studied its impact on baryogenesis via leptogenesis . the mixing is driven by non - equilibrium right - handed neutrinos , as described in section [ sec : off : corr ] . with no deviation from equilibrium , there is no mixing of the lepton doublets . the amount of mixing is quantified by the the correlations of the off - diagonal number densities ( [ sol : deltan ] ) . substituting these off - diagonal flavour correlations back into the collision term for the leptons , we obtain the source for the lepton asymmetry , eqs . ( [ source : lepto],[source : q ] ) in section [ sec : cancellation ] . this asymmetry is opposite to the source for the helicity asymmetry in the right - handed neutrinos ( which is however rapidly violated through the majorana masses ) . by calculating the freeze - out value of the lepton asymmetry , we compare the source from lepton mixing to the standard source from mixing and direct decays of right - handed neutrinos in section [ sec : example : strongwashout ] . the standard result is enhanced for mass - degenerate right - handed neutrinos ( resonant leptogenesis ) , while the asymmetry from lepton mixing is maxwell - suppressed for larger mass ratios . in between these regimes , we find that lepton - doublet mixing may be the dominant source of @xmath0-asymmetry for leptogenesis . results for the impact of gauge interactions on the mixing of higgs - doublets and implications for the freeze - out value of the induced asymmetries are presented in section [ sec : higgs : mix ] . in the future , it would be interesting to perform a more systematic scan of parameter space , as the asymmetries presented in figure [ fig : leplep ] are from a particular slice only . for that purpose , a more systematic and accurate calculation of the rates of right - handed neutrino production and pair creation and annihilation of doublet leptons , that we estimate here in appendix [ app : rates ] , would be beneficial . the relevant techniques are currently under development , see refs . @xcite . the closed - time - path method so far has already proven useful in order to describe the emergence of the baryon asymmetry of the universe in a more systematic manner and in order to calculate certain finite - temperature corrections to the approximations that are most commonly applied @xcite . together with ref . @xcite , the present work shows that these methods can also be useful in order to identify sources for the @xmath0-asymmetry , that have not been considered before . the further development of non - equilibrium field theory techniques and their application to the comparably simple problem of leptogenesis will therefore prove helpful for progress on both , the qualitative and quantitative understanding of the emergence of the baryon - asymmetric universe . the author acknowledges support by the gottfried wilhelm leibniz programme of the deutsche forschungsgemeinschaft . pair creation and annihilation can be mediated either through the @xmath92-channel or the @xmath205-channel . individual @xmath92-channel processes are typically suppressed compared to @xmath205-channel processes by a factor of a few ( essentially the logarithm of the squared gauge coupling ) , because the latter may be infrared enhanced due to the exchange of soft particles . nonetheless , due to the large number of matter degrees of freedom in the standard model , it should be a sufficient approximation to consider @xmath92-channel processes for pair creation and annihilation of matter particles and higgs bosons only . technically , it should also be possible to account for the @xmath205-channel contributions , which would however be more complicated due to the infrared enhancement and its screening . a standard procedure for calculating cosmological interaction rates of non - relativistic particles relies on the reduced cross section @xmath206 , where @xmath92 is the mandelstam @xmath92-variable . the main simplification arises due to the use of maxwell statistics . for massless particles , this approximation incurs an order one inaccuracy . moreover , the @xmath205-channel contributions may leave the reduced cross section infrared divergent , a problem that should be accounted for by including the screening effects in the plasma . in conjunction with the comments above , we should expect that the following procedure yields an @xmath207 estimate of the weighted pair creation and annihilation rate @xmath208 . a more accurate calculation of this quantity may be subject of a future study . the total annihilation rate of massless particles @xmath209 into massless matter or higgs particles @xmath210 in maxwell approximation is given by @xmath211 the squared matrix element is understood to be a sum over all external polarisation and weak isospin states . the reduced cross section is defined as @xmath212 when @xmath216 and @xmath217 are @xmath10 doublets , @xmath218 , where @xmath219 denotes the @xmath220 weak hypercharge of the particles . for singlets , @xmath221 . for the doublets , we have to divide the rates by two in order to average over the weak isospin such that we obtain the production rate for a single isospin state . summing over all standard model degrees of freedom for @xmath210 , we obtain the averaged rate for a single flavour - sensitive process is therefore @xcite @xmath223 we have taken here for the numerical values @xmath224 and @xmath225 at the time of leptogenesis . now with the order one approximation that all lepton doublets experience the same reaction rate , regardless of their momentum , we express @xmath226 such that it follows that @xmath227 the difference when using a fermi - dirac instead of a maxwell distribution in the above integral ( [ bfl : int ] ) is , due to the factor @xmath77 that favours contributions with large momentum , only about 6% . using the same methods , we obtain for the weighted annihilation rate of higgs particles @xmath228 as for the flavour sensitive rates , we proceed similarly and make our present estimates as follows : ref . @xcite provides the number for the total production rate for sterile right - handed neutrinos . this gives an approximation for the flavour - sensitive scatterings that involve charged right - handed charged leptons . the main difference is that the charged right - handed leptons may also radiate @xmath220 gauge bosons , an effect that is therefore neglected when adopting the number @xmath229 for the total reaction rate from ref . the averaged rate for a single flavour - sensitive process then is @xmath230 in the approximation of the same reaction rate for all leptons in phase space , we obtain @xmath231 and therefore @xmath232 in case a more accurate calculation of the lepton asymmetries is desired in the future , the estimates for the rates made in this appendix , which should be of @xmath207 accuracy , need to be improved . the main challenges are then to account for infrared effects and to abandon the approximations of averaging the reaction rates over phase space . l. covi , e. roulet and f. vissani , `` cp violating decays in leptogenesis scenarios , '' phys . b * 384 * ( 1996 ) 169 [ hep - ph/9605319 ] . m. flanz , e. a. paschos , u. sarkar and j. weiss , `` baryogenesis through mixing of heavy majorana neutrinos , '' phys . b * 389 * ( 1996 ) 693 [ hep - ph/9607310 ] . a. pilaftsis , `` resonant cp violation induced by particle mixing in transition amplitudes , '' nucl . b * 504 * ( 1997 ) 61 [ hep - ph/9702393 ] . a. pilaftsis , `` cp violation and baryogenesis due to heavy majorana neutrinos , '' phys . d * 56 * ( 1997 ) 5431 [ hep - ph/9707235 ] . a. pilaftsis and t. e. j. underwood , `` resonant leptogenesis , '' nucl . b * 692 * ( 2004 ) 303 [ hep - ph/0309342 ] . a. pilaftsis and t. e. j. underwood , `` electroweak - scale resonant leptogenesis , '' phys . d * 72 * ( 2005 ) 113001 [ hep - ph/0506107 ] . b. garbrecht and m. herranen , `` effective theory of resonant leptogenesis in the closed - time - path approach , '' nucl . b * 861 * ( 2012 ) , 17 . [ arxiv:1112.5954 [ hep - ph ] ] . m. garny , a. kartavtsev and a. hohenegger , `` leptogenesis from first principles in the resonant regime , '' arxiv:1112.6428 [ hep - ph ] . m. drewes and b. garbrecht , `` leptogenesis from a gev seesaw without mass degeneracy , '' arxiv:1206.5537 [ hep - ph ] . l. covi , n. rius , e. roulet and f. vissani , `` finite temperature effects on cp violating asymmetries , '' phys . d * 57 * ( 1998 ) 93 [ hep - ph/9704366 ] . g. f. giudice , a. notari , m. raidal , a. riotto and a. strumia , `` towards a complete theory of thermal leptogenesis in the sm and mssm , '' nucl . b * 685 * ( 2004 ) 89 [ hep - ph/0310123 ] . b. garbrecht , `` leptogenesis from additional higgs doublets , '' phys . d * 85 * ( 2012 ) 123509 [ arxiv:1201.5126 [ hep - ph ] ] . j. s. schwinger , `` brownian motion of a quantum oscillator , '' j. math . * 2 * ( 1961 ) 407 . l. v. keldysh , `` diagram technique for nonequilibrium processes , '' zh . eksp . * 47 * ( 1964 ) 1515 [ sov . jetp * 20 * ( 1965 ) 1018 ] . e. calzetta and b. l. hu , `` nonequilibrium quantum fields : closed time path effective action , wigner function and boltzmann equation , '' phys . d * 37 * ( 1988 ) 2878 . t. prokopec , m. g. schmidt and s. weinstock , `` transport equations for chiral fermions to order h bar and electroweak baryogenesis . part 1 , '' annals phys . * 314 * ( 2004 ) 208 [ hep - ph/0312110 ] . t. prokopec , m. g. schmidt and s. weinstock , `` transport equations for chiral fermions to order h - bar and electroweak baryogenesis . part ii , '' annals phys . * 314 * ( 2004 ) 267 [ hep - ph/0406140 ] . w. buchmuller and s. fredenhagen , `` quantum mechanics of baryogenesis , '' phys . b * 483 * , 217 ( 2000 ) [ hep - ph/0004145 ] . a. de simone and a. riotto , `` quantum boltzmann equations and leptogenesis , '' jcap * 0708 * ( 2007 ) 002 [ hep - ph/0703175 ] . m. garny , a. hohenegger , a. kartavtsev and m. lindner , `` systematic approach to leptogenesis in nonequilibrium qft : vertex contribution to the cp - violating parameter , '' phys . rev . d * 80 * ( 2009 ) 125027 [ arxiv:0909.1559 [ hep - ph ] ] . m. garny , a. hohenegger , a. kartavtsev and m. lindner , `` systematic approach to leptogenesis in nonequilibrium qft : self - energy contribution to the cp - violating parameter , '' phys . d * 81 * ( 2010 ) 085027 [ arxiv:0911.4122 [ hep - ph ] ] . a. anisimov , w. buchmller , m. drewes and s. mendizabal , `` leptogenesis from quantum interference in a thermal bath , '' phys . * 104 * ( 2010 ) 121102 [ arxiv:1001.3856 [ hep - ph ] ] . m. garny , a. hohenegger , a. kartavtsev , `` medium corrections to the cp - violating parameter in leptogenesis , '' phys . * d81 * ( 2010 ) 085028 . [ arxiv:1002.0331 [ hep - ph ] ] . m. beneke , b. garbrecht , m. herranen and p. schwaller , `` finite number density corrections to leptogenesis , '' nucl . b * 838 * ( 2010 ) 1 [ arxiv:1002.1326 [ hep - ph ] ] . m. beneke , b. garbrecht , c. fidler , m. herranen and p. schwaller , `` flavoured leptogenesis in the ctp formalism , '' nucl . b * 843 * ( 2011 ) 177 [ arxiv:1007.4783 [ hep - ph ] ] . m. garny , a. hohenegger and a. kartavtsev , `` quantum corrections to leptogenesis from the gradient expansion , '' arxiv:1005.5385 [ hep - ph ] . a. anisimov , w. buchmuller , m. drewes and s. mendizabal , `` quantum leptogenesis i , '' annals phys . * 326 * ( 2011 ) 1998 [ arxiv:1012.5821 [ hep - ph ] ] . v. cirigliano , c. lee , m. j. ramsey - musolf and s. tulin , `` flavored quantum boltzmann equations , '' phys . d * 81 * ( 2010 ) 103503 [ arxiv:0912.3523 [ hep - ph ] ] . v. cirigliano , c. lee and s. tulin , `` resonant flavor oscillations in electroweak baryogenesis , '' phys . d * 84 * ( 2011 ) 056006 [ arxiv:1106.0747 [ hep - ph ] ] . e. k. akhmedov , v. a. rubakov and a. y. .smirnov , `` baryogenesis via neutrino oscillations , '' phys . * 81 * ( 1998 ) 1359 [ hep - ph/9803255 ] . t. asaka and m. shaposhnikov , `` the numsm , dark matter and baryon asymmetry of the universe , '' phys . b * 620 * ( 2005 ) 17 [ hep - ph/0505013 ] . l. canetti , m. drewes and m. shaposhnikov , `` sterile neutrinos as the origin of dark and baryonic matter , '' arxiv:1204.3902 [ hep - ph ] . t. endoh , t. morozumi and z. -h . xiong , `` primordial lepton family asymmetries in seesaw model , '' prog . theor . * 111 * ( 2004 ) 123 [ hep - ph/0308276 ] . a. abada , s. davidson , f. x. josse - michaux , m. losada and a. riotto , `` flavour issues in leptogenesis , '' jcap * 0604 * , 004 ( 2006 ) [ arxiv : hep - ph/0601083 ] . e. nardi , y. nir , e. roulet and j. racker , `` the importance of flavor in leptogenesis , '' jhep * 0601 * , 164 ( 2006 ) [ arxiv : hep - ph/0601084 ] . k. kainulainen , t. prokopec , m. g. schmidt and s. weinstock , `` first principle derivation of semiclassical force for electroweak baryogenesis , '' jhep * 0106 * ( 2001 ) 031 [ hep - ph/0105295 ] . b. garbrecht , t. prokopec and m. g. schmidt , `` particle number in kinetic theory , '' eur . j. c * 38 * ( 2004 ) 135 [ hep - th/0211219 ] . a. basboll and s. hannestad , `` decay of heavy majorana neutrinos using the full boltzmann equation including its implications for leptogenesis , '' jcap * 0701 * ( 2007 ) 003 [ hep - ph/0609025 ] . a. anisimov , d. besak and d. bodeker , `` thermal production of relativistic majorana neutrinos : strong enhancement by multiple soft scattering , '' jcap * 1103 * ( 2011 ) 042 [ arxiv:1012.3784 [ hep - ph ] ] . a. salvio , p. lodone and a. strumia , `` towards leptogenesis at nlo : the right - handed neutrino interaction rate , '' jhep * 1108 * ( 2011 ) 116 [ arxiv:1106.2814 [ hep - ph ] ] . m. laine , a. vuorinen and y. zhu , `` next - to - leading order thermal spectral functions in the perturbative domain , '' jhep * 1109 * ( 2011 ) 084 [ arxiv:1108.1259 [ hep - ph ] ] . m. laine and y. schroder , `` thermal right - handed neutrino production rate in the non - relativistic regime , '' jhep * 1202 * ( 2012 ) 068 [ arxiv:1112.1205 [ hep - ph ] ] . d. besak and d. bodeker , `` thermal production of ultrarelativistic right - handed neutrinos : complete leading - order results , '' jcap * 1203 * ( 2012 ) 029 [ arxiv:1202.1288 [ hep - ph ] ] . e. w. kolb and m. s. turner , `` grand unified theories and the origin of the baryon asymmetry , '' ann . nucl . part . sci . * 33 * ( 1983 ) 645 . w. buchmuller , p. di bari and m. plumacher , `` leptogenesis for pedestrians , '' annals phys . * 315 * ( 2005 ) 305 [ hep - ph/0401240 ] . s. antusch , p. di bari , d. a. jones and s. f. king , `` leptogenesis in the two right - handed neutrino model revisited , '' arxiv:1107.6002 [ hep - ph ] . j. a. casas and a. ibarra , `` oscillating neutrinos and muon to e , gamma , '' nucl . b * 618 * ( 2001 ) 171 [ hep - ph/0103065 ] .

this introductory review is addressed to beginning researchers . some of the distinguishing features of loop quantum gravity are illustrated through loop quantum cosmology of frw models . in particular , these examples illustrate : i ) how ` emergent time ' can arise ; ii ) how the technical issue of solving the hamiltonian constraint and constructing the _ physical _ sector of the theory can be handled ; iii ) how questions central to the planck scale physics can be answered using such a framework ; and , iv ) how quantum geometry effects can dramatically change physics near singularities and yet naturally turn themselves off and reproduce classical general relativity when space - time curvature is significantly weaker than the planck scale .

gr-qc0702030

my lectures at the first stckleberg symposium began with an introduction to the challenges of quantum gravity , then discussed key features of loop quantum gravity ( lqg ) and finally summarized applications to cosmology . however , in view of the page limit , in these proceedings i decided to present the material from a different perspective , through the lens of loop quantum cosmology ( lqc ) . thus , rather than starting with general considerations and then descending to applications , here i will follow an opposite approach , using applications to illustrate some of the key problems of quantum gravity and distinguishing features of lqg . the advantage of this reverse strategy is that students and other beginning researchers can appreciate the key issues , challenges and advances in a rather simple setting , without having to first grasp the technical intricacies of full lqg . the main drawback is that this approach leaves out some of the most interesting developments such as statistical mechanical calculations of black hole entropy , discussions of the issue of information loss , and spin foams , particularly the recent advances in calculating graviton propagators in the non - perturbative setting of lqg @xcite . with these caveats in mind , let us begin with a list of some of the long standing conceptual and technical issues of quantum gravity on which loop quantum cosmology has a bearing . in general relativity , gravity is encoded in the very geometry of space - time . the most dramatic features of general relativity can be traced back to this dual role of geometry : the expansion of the universe , the big - bang , formation of black holes and emergence of gravitational waves as ripples of space - time curvature . already in the classical theory , it took physicists several decades to truly appreciate the dynamical nature of geometry and to get used to the absence of a kinematic background geometry . in quantum gravity , this paradigm shift leads to a new level of conceptual and technical difficulties . * the absence of a background geometry implies that classical dynamics is generated by constraint equations . in the quantum theory , physical states are solutions to quantum constraints . all of physics , including the dynamical content of the theory , has to be extracted from these solutions . but there is no external time to phrase questions about evolution . therefore we are led to ask : can we extract , from the arguments of the wave function , one variable which can serve as _ emergent time _ with respect to which the other arguments ` evolve ' ? if not , how does one interpret the framework ? what are the physical ( i.e. , dirac ) observables ? in a pioneering work , dewitt proposed that the determinant of the 3-metric can be used as an ` internal ' time @xcite . consequently , in much of the literature on the wheeler - dewitt ( wdw ) approach to quantum cosmology , the scale factor is assumed to play the role of time , although often only implicitly . however , in closed models the scale factor fails to be monotonic due to classical recollapse and can not serve as a global time variable already in the classical theory . are there better alternatives at least in the simple setting of quantum cosmology ? if not , can we still make physical predictions ? * can one construct a framework that cures the short - distance difficulties faced by the classical theory near singularities , while maintaining an agreement with it at large scales ? by their very construction , perturbative and effective descriptions have no problem with the second requirement . however , physically their implications can not be trusted at the planck scale and mathematically they generally fail to provide a deterministic evolution across the putative singularity . in lqg the situation is just the opposite . quantum geometry gives rise to new discrete structures at the planck scale that modify the classical theory in such a way that , at least in simple models , space - like singularities of general relativity are resolved . however , since the emphasis is on background independence and non - perturbative methods , a priori it is not clear whether the theory also has a rich semi - classical sector . do the novel dynamical corrections unleashed by the underlying quantum geometry naturally fade away at macroscopic distances or do they have unforeseen implications that prevent the theory from reproducing general relativity at large scales ? some of such unforeseen problems are discussed in , e.g. , @xcite . next , the dual role of the space - time metric also implies that space - time itself ends when the gravitational field becomes infinite . this is in striking contrast with minkowskian physics , where singularity of one specific field has no bearing at all on the space - time structure or on the rest of physics . in general relativity singularities of the gravitational field represent an _ absolute boundary _ of space - time where _ all of physics _ comes to a halt . now , it is widely believed that the prediction of a singularity such as the big - bang in classical general relativity is primarily a signal that the theory has been pushed beyond the domain of its validity and can not therefore be trusted . one needs a quantum theory of gravity to analyze true physics . this expectation immediately leads to a host of questions which have been with us for several decades now : + * how close to the big - bang does a smooth space - time of gr make sense ? inflationary scenarios , for example , are based on a space - time continuum . can one show from some first principles that this is a safe approximation already at the onset of inflation ? * is the big - bang singularity naturally resolved by quantum gravity ? this possibility led to the development of the field of quantum cosmology in the late 1960s . the basic idea can be illustrated using an analogy to the theory of the hydrogen atom . in classical electrodynamics the ground state energy of this system is unbounded below . furthermore , because an accelerating electron radiates continuously , it must fall into the proton making the atom extremely unstable . quantum physics intervenes and , thanks to a non - zero planck s constant , the ground state energy is lifted to a finite value , @xmath0 . the hope was that a similar mechanism would resolve the big - bang and big crunch singularities in simple cosmological models . however , as we will see in some detail in section [ s2 ] , the quantization did not realize this hope . can the quantum nature of geometry underlying lqg make a fundamental difference to this status - quo ? * is a new principle/ boundary condition at the big bang or the big crunch essential to provide a deterministic evolution ? the most well known example of such a boundary condition is the ` no boundary proposal ' of hartle and hawking @xcite . or , do quantum einstein equations suffice by themselves even at the classical singularities ? * do quantum dynamical equations remain well - behaved even at these singularities ? if so , do they continue to provide a deterministic evolution ? the idea that there was a pre - big - bang branch to our universe has been advocated by several approaches , most notably by the pre - big - bang scenario in string theory @xcite and ekpyrotic and cyclic models inspired by brane world ideas @xcite . however , in such perturbative treatments there is always a smooth continuum in the background and hence the dynamical equations break down at the singularity . consequently , the pre - big - bang branch is not joined to the current post - big - bang branch by a deterministic evolution . the hope has been that non - perturbative effects would remedy this situation in the future . lqg , on the other hand , does not have a continuum geometry in the background and the treatment is non - perturbative . so , one is led to ask : can the lqc quantum einstein s equation provide a deterministic evolution across the big bang and the big - crunch ? * if there is a deterministic evolution , what is on the ` other side ' ? is there just a quantum foam from which the current post - big - bang - branch is born , say a ` planck time after the putative big - bang ' ? or , was there another classical universe as in the pre - big - bang and cyclic scenarios , joined to ours by deterministic equations ? _ the goal of this article is show that such questions can be answered satisfactorily in the simplest lqc models_. the first significant results appeared already five years ago in the pioneering work of bojowald @xcite . since then there has been a steady stream of papers by a dozen or so researchers and the field achieved a new degree of maturity over the last year . however , as discussed in section [ s4 ] this is not the end of the story because in lqc one first symmetry reduces the classical theory and then quantizes it . nonetheless , since the quantization procedure mimics that of full lqg , the lqc answers provide not only the much needed intuition but also a strategy for answering the larger questions . the article is organized as follows . section [ s2 ] introduces the reader to quantum cosmology and reviews the theory . while one can answer some of the questions posed above , the theory fails to resolve the big bang and the big crunch singularities . the situation is drastically different in lqc where not only are the singularities resolved but most of the questions have physically desired answers . the key ideas and the structure of lqc are summarized in section [ s3 ] , which also explains the sense in which it ` descends ' form full lqg and emphasizes the main differences between lqc the theory . for definiteness , i will use the k=1 frw models coupled to a massless scalar field @xcite to illustrate in some detail both the difficulties discussed above and the way they can be handled . for results on the k=0 model , see @xcite , for a first discussion of the k=1 and of the anisotropic models , see @xcite and for a detailed review of the developments in lqc till 2005 , see @xcite . i conclude in section [ s4 ] with a brief summary of these other developments and a discussion of the many challenges that still remain . both on the mathematical and phenomenological fronts , most of the work to date in classical cosmology involves spatially homogeneous models and perturbations thereof . in the late 60 s dewitt @xcite and misner @xcite began investigations in _ quantum _ cosmology with a similar viewpoint : first reduce general relativity by imposing homogeneity and then quantize the resulting system . since this system has only a finite number of degrees of freedom , the key field - theoretic difficulties are bypassed . however , since there is no classical space - time in the background , most of the conceptual problems of full quantum gravity still remain . therefore one can hope that resolution of these problems in this technically simpler context would provide valuable insights for the full theory . in particular , as mentioned above , an initial expectation was that quantum effects particularly the uncertainty principle would intervene and resolve the big bang and the big crunch singularities . while subsequent work in the seventies and eighties did shed light on several conceptual issues , this specific expectation was not met . i will begin by briefly explaining the strategy used in these analyses and point out its limitation . this summary will also serve to bring out the new elements of lqc . in this early work , the configuration space consisted of positive definite 3-metrics . in the simplest , spatially homogeneous , isotropic context , this metric is completely determined by the scale factor @xmath1 . since @xmath1 is restricted to be positive , one generally introduced a new variable @xmath2 and , as in quantum mechanics , considered wave functions @xmath3 , where @xmath4 denotes possible matter fields . as in quantum mechanics , the basic operators were defined via @xmath5 , and @xmath6 . quantization of the classical hamiltonian constraint , p_^2 + e^ = h _ , where @xmath7 is the matter hamiltonian , then led to a differential equation , called the equation . if the matter field is a zero rest mass scalar field the case i will focus on for simplicity the equation becomes [ wdwhc ] ^2 _ - e^4 = ^2_(for details , see , e.g. , @xcite ) . physical quantum states are solutions to this equation . to extract physics , one has to introduce an inner product and suitable observables on the space of these states . the task of finding these observables is not entirely straightforward because they must _ preserve the space of solutions to ( [ wdwhc ] ) _ , i.e. , they must be dirac observables . using these observables , one can then ask if classical singularities are resolved in quantum theory . the older literature does not appear to analyze this issue in any detail since its focus was on the wkb approximation which , unfortunately , becomes unreliable near the singularity . however , they were analyzed in some detail recently ( see e.g. @xcite ) . i will now provide a summary of the main results . @xmath8@xmath9 classical solutions . since @xmath10 is a constant of motion , a classical trajectory can be plotted in the @xmath11-@xmath4 plane , where @xmath11 is essentially the volume in planck units ( see eq ( [ v ] ) ) . @xmath9 expectation values ( and dispersions ) of @xmath12 for the wdw wave function and comparison with the classical trajectory . the wave function follows the classical trajectory into the big - bang and big - crunch singularities . ( in this simulation , the parameters were : @xmath13 , and @xmath14 . ) , title="fig:",width=307 ] classical solutions . since @xmath10 is a constant of motion , a classical trajectory can be plotted in the @xmath11-@xmath4 plane , where @xmath11 is essentially the volume in planck units ( see eq ( [ v ] ) ) . @xmath9 expectation values ( and dispersions ) of @xmath12 for the wdw wave function and comparison with the classical trajectory . the wave function follows the classical trajectory into the big - bang and big - crunch singularities . ( in this simulation , the parameters were : @xmath13 , and @xmath14 . ) , title="fig:",width=307 ] each classical trajectory begins with the big bang , undergoes a subsequent expansion and a recollapse and finally contracts into the big crunch singularity ( see fig . [ fig : wdw1 ] a ) . since there is no potential , the matter momentum @xmath10 is a constant of motion . consequently , @xmath4 is a monotonic function on each classical trajectory . therefore , it can be thought of as ` internal time ' in the classical theory . the form of the equation ( [ wdwhc ] ) suggests that the same is true in quantum theory . thus the equation can be regarded as an ` evolution equation ' with respect to this _ emergent time _ , @xmath4 . next , @xmath15 is a dirac observable in the quantum theory . given any value @xmath16 of @xmath4 , one can also define a second dirac observable , @xmath17 , corresponding to the volume of the universe at any given ` time ' @xmath16 @xcite . the pair @xmath18 constitutes a complete set of dirac observables . finally , the requirement that these observables be self - adjoint fixes the physical inner product ( which can also be determined by the more precise group averaging technique @xcite ) . thus , the first set of questions raised in section [ s1 ] can be answered satisfactorily in the theory . now to answer the remaining questions about singularity , one can proceed as follows . consider a point @xmath19 on a classical trajectory with @xmath20 , where the universe is large and the curvature is low compared to the planck scale . at ` time ' @xmath21 , one can construct an initial _ quantum _ state which is peaked at values @xmath22 of the dirac observables @xmath23 and @xmath24 . one can then evolve it using the equation . of ( [ wdwhc ] ) , where the positive definite self - adjoint operator @xmath25 is the negative of the operator on the right side of ( [ wdwhc ] ) . one uses this ` first order in time ' equation to ` evolve ' the wave function . for details , see @xcite . ] the solutions can be expressed in terms of modified bessel functions . a resulting plot of the expectation values and dispersions of @xmath26 as a function of @xmath4 is shown in fig . [ fig : wdw1 ] b. the good news is that , when the universe is large , the quantum state remains sharply peaked at the classical trajectory , showing that the theory has a good semi - classical limit . the bad news is that the quantum state continues to be peaked at this trajectory even as the classical trajectory hits the big bang and the big crunch singularities ! thus , although the theory is background independent and non - perturbative , its results are similar to that of background dependent , perturbative treatments : good semi - classical limit but no improvement on the classical short distance pathologies . * remark : * in the classical solution , the scalar field @xmath4 tends to @xmath27 at the big bang and to @xmath28 at the big crunch . therefore , the evolution is unitary on the entire real line of the ` emergent time ' . as we just saw , even semi - classical states simply track the classical solution and therefore to the extent classical solutions are singular singularities are reached in finite _ proper time_ so are the quantum solutions . thus , while the emergent time @xmath4 is useful in enabling us to think of ` evolution ' , unitary evolution over its entire range does not guarantee non - singular behavior . additional considerations , such as the relation between the emergent time and proper time in semi - classical settings are essential . since readers may not be familiar with the hamiltonian description used in lqc , i will summarize its key elements before proceeding with quantization . let us consider space - time manifolds which have the form @xmath29 , where @xmath30 has the topology of a 3-sphere , @xmath31 . as misner showed in the late 60 s @xcite , one can identify @xmath30 with the symmetry group @xmath32 ( which ensures spatial homogeneity and isotropy ) and endow it with a fixed fiducial basis of 1-forms @xmath33 and vectors @xmath34 . the resulting fiducial metric is _ ab:= _ a^i _ b^j k_ij , @xmath35 turns out to be the metric of the round 3-sphere with radius @xmath36 ( rather than @xmath37 ) . the volume of @xmath30 w.r.t . this fiducial metric @xmath35 is @xmath38 and the scalar curvature is @xmath39 . we shall set @xmath40 . in lqg , the dynamical variables are @xmath32 ( gravitational ) spin - connections @xmath41 . their conjugate momenta @xmath42 are the analogs of electric fields in yang - mills theory . however , since the ` internal ' group @xmath32 is now tied with the group @xmath43 of rotations in the tangent space at any point of @xmath30 , @xmath42 now acquire a direct geometric meaning : they represent spatial , orthonormal triads ( with density weight 1 ) . in the present isotropic , homogeneous setting , these canonically conjugate fields are parameterized by constants @xmath44 and @xmath45 respectively : [ ps ] a_a^i = c ^-1 _ a^i , e^a_i = p ^-2 ^a_i which are functions of time . the factors of @xmath46 facilitate comparison with the spatially flat , @xmath47=0 case @xcite . in both cases , @xmath44 is dimensionless while @xmath45 has dimensions of area . in terms of geometrodynamical variables , @xmath45 determines the scale factor @xmath1 ( which determines the spatial metric ) and @xmath44 , the canonically conjugate momentum , or , @xmath48 ( which determines the extrinsic curvature ) . more precisely , at the point @xmath49 of the phase space , the physical 3-metric @xmath50 and the extrinsic curvature @xmath51 are given by : q_ab = |p| ^-2 _ abk_ab = ( c- ) |p|^ ^-2 _ ab the corresponding physical volume of @xmath30 is @xmath52 . the scale factor @xmath53 associated with a physical metric @xmath50 is generally expressed via @xmath54 where @xmath55 is the _ unit _ 3-sphere metric . then , the scale factor is related to @xmath45 via @xmath56 . @xmath45 can take both positive and negative values , the change in sign corresponds to a flip in the orientation of the triad which leaves the physical metric @xmath50 invariant . using the fact that @xmath41 and @xmath42 are canonically conjugate , it is easy to calculate the fundamental poisson bracket between @xmath44 and @xmath45 : [ pb ] \{c , p } = where @xmath57 is the so called ` barbero - immirzi parameter ' which labels quantization ambiguity of lqg . ( black hole entropy considerations show that @xmath58 . ) finally , using the fact that the co - triad @xmath33 satisfies the cartan identity [ cartan ] ^k + _ ij^k ^i^j = 0 , it is straightforward to calculate the field strength @xmath59 of the connection @xmath41 on @xmath30 : [ f1 ] f_ab^k = ^-2 _ ij^k _ a^j _ b^k . because of our choice of parametrization of @xmath41 and @xmath42 , the only non - trivial constraint is the hamiltonian one . its gravitational part reduces to @xcite : c _ & = & - _ m ^3 x ^ij_k e^-1 e^a_ie^b_j [ ham2 ] + & = & - since the phase space is now parameterized by @xmath60 , as in the theory one s first impulse would be to follow standard quantum mechanics . in the case this strategy was natural because one did not have access to full quantum geometrodynamics to take guidance from . in lqc it is not : a more appropriate strategy would be to follow the procedures used in full lqg . one s first reaction may be : how can this make any difference ? after all the system has only a finite number of degrees of freedom and von neumann s theorem assures us that , under appropriate assumptions , the resulting quantum mechanics is unique . the only remaining freedom appears to be in the factor - ordering of the hamiltonian constraint and this is generally insufficient to lead to qualitatively different predictions . it turns out however that if one mimics full lqg , one naturally violates a key assumption of von neumann s uniqueness theorem and , in spite of the presence of only a finite number of degrees of freedom , one is led to _ new quantum mechanics_. to explain this point , let me begin with a quick summary of the relevant structure of full lqg . since novel elements of interest to us appear in the gravitational sector , in the remainder of this subsection i will ignore matter . to maintain gauge covariance , configuration variables are taken to be the ` wilson lines ' or the holonomies @xmath61 along arbitrary edges @xmath62 , determined by the @xmath32 connection @xmath41 . geometrically , being 1-forms , connections are naturally ` smeared ' along 1-dimensional curves . the conjugate momenta are density - weighted triads @xmath42 which ( being duals of a triplet of 2-forms ) are naturally smeared along 2-surfaces @xmath63 with test fields @xmath64 : @xmath65 . these are referred to as triad fluxes . note that these two sets of phase space functions @xmath66 have been constructed without any reference to a background metric or any other background field . they generate an algebra , rather analogous to that generated by functions @xmath67 and @xmath45 on the phase space of a particle . our job is to find a representation of this ` holonomy - flux ' algebra which preserves background independence of the theory . the surprising result is that the representation is unique @xcite ! more precisely , the algebra admits a unique ( internal , i.e. @xmath32 gauge and ) diffeomorphism invariant state and the representation is generated by the action of the algebra on this state .- algebra @xmath68 of operators is first constructed abstractly , without reference to a hilbert space . a state @xmath69 is a positive linear functional on the algebra , i.e. a linear functional which satisfies @xmath70 for all @xmath71 . such a state determines , via a standard construction due to gelfand , naimark and segal , a @xmath72 representation of @xmath68 on a hilbert space , in which the abstract state @xmath69 is now represented by a vector @xmath73 in the hilbert space and its action by the expectation value functional : @xmath74 . the vector @xmath75 is ` cyclic ' in the sense that a dense subspace of the full hilbert space is generated by repeated action of operators in @xmath68 on @xmath75 . finally , note that the uniqueness result refers to the holonomy - flux algebra whose elements are not diffeomorphism invariant . an algebra of diffeomorphism invariant operators would admit many diffeomorphism invariant states . ] while the uniqueness result is rather recent , the representation itself was constructed over a decade ago and is well understood @xcite . for our purposes a key feature of this representation is the following : while there are well - defined operators @xmath76 corresponding to holonomies , _ there is no operator corresponding to the connection itself_. in quantum cosmology one follows the same overall procedure . in the k=1 model now under consideration , holonomies along integral curves of the left invariant vector fields @xmath34 already suffice to separate homogeneous , isotropic connections . let us denote by @xmath77 the directed length ( w.r.t . the fiducial metric @xmath35 ) of the curve along @xmath34 ( so that @xmath78 is positive if the curve is oriented in the direction of @xmath34 and negative if has the opposite orientation ) . then the holonomy along the edge of length @xmath77 in the @xmath47th direction is given by h^k_()(c ) = ( c/2 ) + 2(c/2 ) ^kwhere @xmath79 is the identity @xmath80 matrix . ( the holonomy is of course independent of the background metric @xmath35 . ) the functions of @xmath44 which enter as coefficients are ` almost periodic ' functions of @xmath44 , i.e. are of the form @xmath81 . in the quantum theory , then , we are led to a representation in which operators @xmath82 and @xmath83 are well - defined , but there is _ no _ operator corresponding to the connection component @xmath44 . at first this seems surprising because our experience with quantum mechanics suggests that one should be able to obtain the operator analog of @xmath44 by differentiating @xmath84 with respect to the parameter @xmath78 . however , in the representation of the basic quantum algebra that descends to lqc from full lqg , although the @xmath84 provide a 1-parameter group of unitary transformations , the group fails to be weakly continuous in @xmath78 . therefore one can not differentiate and obtain the operator analog of @xmath44 . in quantum mechanics , this would be analogous to having well - defined ( weyl ) operators corresponding to the classical functions @xmath85 but no operator @xmath86 corresponding to @xmath87 itself . this violates one of the assumptions of the von - neumann uniqueness theorem . new representations then become available which are _ inequivalent _ to the standard schrdinger one . in quantum mechanics , these representations are not of direct physical interest because we need the operator @xmath86 . in lqc , on the other hand , full lqg naturally leads us to a new representation . _ this theory is inequivalent to the type theory already at a kinematical level . _ the structure of this theory can be summarized as follows . since @xmath83 is a self - adjoint operator , the gravitational part @xmath88 of the kinematic hilbert space can be described in terms of its eigenbasis . a general state @xmath89 has the form [ state ] |= _ i _ i @xmath90 _ i |_i|^2 < , where @xmath91 ranges over a countable set and @xmath92 is an orthonormal basis : [ ip ] _ i|p_j = _ ij . while the general form of these equations may seem familiar from quantum mechanics , there are key differences : the state is a _ countable _ sum of eigenstates of @xmath83 rather than a direct integral and the right side of ( [ ip ] ) is a kronecker delta rather than the dirac delta distribution . , the kinematical lqc hilbert space @xmath88 is non - separable . however , as we will see , quantum dynamics is governed by a discrete evolution equation . this leads to superselection sectors so that the final physical hilbert spaces are all separable . for details , see @xcite . ] consequently , the intersection of @xmath88 and @xmath93 consists only of the zero function . however , the action of the basic operators on wave functions @xmath94 is the expected one : ( p ) = p ( p ) _ ( ) ( p ) = ( p+ ) the forms of these operators are the same as in ordinary quantum mechanics . however , in the new hilbert space @xmath95 fails to be ( weakly ) continuous in @xmath78 because @xmath96 is _ orthogonal to _ @xmath97 for all @xmath98 . consequently , one can not take a derivative of @xmath84 w.r.t . @xmath78 and define an operator @xmath99 . finally , let us note a fact that will be useful in section [ s3.3 ] : from the discussion of the classical theory of section [ s3.1 ] it follows that the physical volume operator of @xmath30 is given by : @xmath100 . + * remark : * kinematics of lqc provides a useful , broad - brush picture of the kinematics of full lqg ( for details , see @xcite ) . we saw that the lqc kinematical hilbert space is _ not _ the space @xmath101 that one would use in the theory . but it can in fact be written as the space of square integrable functions , not on the classical configuration space @xmath102 , but on a certain completion , @xmath103 thereof , called the bohr compactification of the real line . this is an abelian group with a natural haar measure . the kinematical hilbert space can be written as @xmath104 . was introduced by the mathematician harold bohr , niels s brother , in his theory of almost periodic functions . in the connection representation , elements of @xmath88 are of the form @xmath105 where @xmath106 are complex numbers , @xmath107 real and the sum extends over a countable set . these @xmath108 are almost periodic functions of @xmath44 , where the word ` almost ' refers to the fact that @xmath107 are arbitrary real numbers . the scalar product given by ( [ ip ] ) now becomes : @xmath109 . ] thus , one way to understand the difference between the schrdinger and the new quantum mechanics is to realize that while in the schrdinger theory the ` quantum configuration space ' continues to be the classical configuration space , @xmath102 , in the new theory it is replaced by a certain completion @xmath110 thereof . the same phenomenon occurs in full lqg : while the classical configuration space @xmath111 is the space of smooth @xmath32 connections on a 3-manifold @xmath112 , the quantum configuration space @xmath113 is a certain completion thereof , containing also ` generalized connections ' which can be arbitrarily discontinuous . the haar measure on @xmath32 induces a natural , faithful , regular borel measure @xmath114 on @xmath113 . the kinematical hilbert space of lqg is @xmath115 . thus the completion @xmath110 of lqc is replaced by @xmath113 and the measure @xmath116 by the measure @xmath117 . technically both completions @xmath113 and @xmath103 arise as the ` gelfand spectrum ' of the holonomy algebras in the two theories . next , let us consider the algebraic approach . the @xmath72-algebra generated by @xmath118 is the lqc analog of the holonomy flux algebra of full lqg . the analog of the unique lqg diffeomorphism state on this @xmath72-algebra is the positive - linear functional @xmath69 which has the following action on the basic operators : @xmath119 and @xmath120 . as in full lqg , the gelfand , naimark , segal construction then leads us to the kinematical hilbert space @xmath88 . in this space , the abstract state is represented by the concrete vector @xmath121 , or equivalently @xmath122 , in which the triad vanishes and , heuristically , the connection is completely undetermined . as in lqg , the full hilbert space is generated by a repeated action of the algebra on this state . a natural basis in the lqg kinematical hilbert space is given by the so - called spin network functions . these are analogous to the states @xmath123 in the connection representation of lqc . just as @xmath123 diagonalize operators @xmath124 , ( appropriately chosen ) spin networks diagonalize triads and hence geometric operators . this is why a firm grasp of the lqc quantum kinematics can provide good intuition for full lqg . finally , a word of caution : the analogy does fail in a few but important respects because of the homogeneity assumption of lqc . so , it should be pursued as a guideline to aid intuition rather than a literal dictionary . to describe quantum dynamics , we have to first introduce a well - defined operator on @xmath88 representing the hamiltonian constraint @xmath125 . since there is no operator corresponding to @xmath44 itself , we can not directly use the final expression ( [ ham3 ] ) . rather , we return to the form ( [ ham2 ] ) from full general relativity and promote it to an operator . this procedure also serves to bring lqc closer to lqg . the general strategy is the same as in ordinary quantum mechanics : one first expresses various terms in ( [ ham2 ] ) in terms of ` elementary variables ' , namely the holonomies @xmath126 and momenta @xmath45 which have direct quantum analogs and then replaces them with operators @xmath127 and @xmath83 . quantization of the first term , @xmath128 , can then be carried out by directly following a procedure given by thiemann in the full theory @xcite . the second term is the field strength @xmath129 . as in gauge theories , one can obtain @xmath130 by first calculating the holonomy around a closed loop in the a - b plane , multiplying it by @xmath131 , dividing by the area and taking the limit as the area shrinks to zero . however , precisely because there is no connection operator @xmath132 in lqg , the limit can not exist in quantum theory . and @xmath45 . ] however , the same feature that prevents @xmath133 from existing leads to quantum riemannian geometry in lqg and in particular implies that there is a smallest non - zero eigenvalue , @xmath134 , of the area operator . therefore the non - existence of the limit is interpreted as telling us that the procedure of shrinking the area of the loop to zero is inappropriate in quantum theory . rather , we should shrink the loop till its area equals @xmath134 . now , given a point @xmath49 of the classical phase space , a ` square ' loop each of whose sides has length @xmath135 with respect to the fiducial metric @xmath35 has area @xmath136 with respect to the _ physical _ metric @xmath137 . therefore , to define @xmath138 , the holonomy is evaluated around a ` square ' loop whose edges are formed by the integral curves of right and left invariant vector fields on @xmath31 , with @xmath139 given by ^2 |p| = 2 ^2 , so that _ physical _ area of the surface enclosed by the loop is @xmath134 . ( because of homogeneity , the precise location of the loop is irrelevant . ) the resulting @xmath140 , given by [ fop ] _ ab^k = ( ^2(c - _o/2 ) -^2(/2 ) ) _ ij^k _ a^i _ b^j , is well - defined but , as in full lqg , has a built in fundamental non - locality . it was clear from the beginning that something like this must occur because there is no connection operator in lqg . recall that passage to quantum theory from the classical one _ always _ requires new physical inputs . quantization of @xmath130 is well motivated in the representation of the holonomy - flux algebra we are forced to use by the requirement of back ground independence . furthermore , one can show that @xmath140 does reduce , in a precise sense , to the standard local expression of curvature @xmath129 in the classical limit . ( for details , see @xcite . ) hence @xmath140 is a well - defined quantization of @xmath129 . however , the non - locality makes a crucial difference in the planck regime . in particular , as we will now see , it makes the gravitational part @xmath141 of the constraint a _ difference _ operator , rather than a differential operator as in the wheeler - dewitt theory . as one might expect , the ` step size ' is governed by the area gap @xmath134 . using ( [ fop ] ) and the the operator corresponding to @xmath128 given by the thiemann procedure it is straightforward to construct the gravitational part @xmath141 of the constraint operator corresponding to ( [ ham2 ] ) . it turns out @xcite that its expression is simplest if one considers wave functions which are naturally diagonal in the volume operator @xmath142 rather than in @xmath83 . ( mathematically , this is a rather trivial transformation since @xmath143 . ) it is convenient to parameterize the eigenvalues of the volume operator by @xmath11 such that : [ v ] @xmath144 = ( ) ^ ^3 @xmath144 then , the gravitational part of the constraint is given by : [ ham4 ] _ ( v ) = & & e^if c c e^-if ( v ) + & - & ( v ) here , @xmath69 is a simple function of @xmath11 , @xmath145 , and the operator @xmath146 , given by [ a ] a ( v ) = - |v| | |v - 1| - |v + 1| | ( v ) , is also diagonal in @xmath11 . to summarize , if we mimic full lqg , we are naturally led to a theory that is inequivalent to the theory even at the kinematical level . indeed the two hilbert spaces have no non - zero element in common . a key distinguishing feature of this theory is that while the holonomies of the connection are well defined operators , connections themselves are not . consequently , in the definition of the quantum hamiltonian constraint ( [ ham2 ] ) , we are led to define the field strength through holonomies along loops which enclose area @xmath134 , the smallest non - zero eigenvalue of the area operator . the gravitational part of the resulting hamiltonian constraint is a _ difference _ operator , rather than a differential operator that would have resulted had we worked in a schrdinger type representation used in the theory . however , in a well - defined sense it reduces to the differential operator for large @xmath11 @xcite . as in the theory , we will now assume that the only matter field is a massless scalar field ( although it is straightforward to allow additional matter fields ) . the form of the hamiltonian constraint is such that this field can serve as an internal clock also in lqc and we can again use relational dynamics . this simplifies the intermediate technical steps and makes the physical meaning of results more intuitive . to write the complete hamiltonian constraint we also need the matter part . for the massless scalar field , in the classical theory it is given by : c_matt = 8 g |p|^- p_^2 the non - trivial part in the passage to quantum theory is the function @xmath147 . however , we can define this operator again by using the method introduced by thiemann in the full theory @xcite . the final result is @xcite : [ inversevol ] ( v ) = ( ) ^3/2 b(v ) ( v ) where [ b ] b(v ) = ( ) ^3 k |v| | |v + 1|^1/3 - |v - 1|^1/3 |^3 . this operator is self - adjoint on @xmath88 and diagonal in the eigenstates of the volume operator . we can express the total constraint [ hc1 ] ( v ) = ( _ grav + _ matt ) ( v ) = 0 , as follows : [ hc3 ] ^2_(v , ) = & - & ( v , ) + = & - & _ o ( v , ) + [ b(v)]^-1 ( v , ) .here , @xmath148 is a difference operator , [ hc4 ] _ o ( v , ) = - [ b(v)]^-1 ( c^+(v ) ( v+4 , ) + c^o(v ) ( v , ) + c^-(v ) ( v-4 , ) ) , where the coefficients @xmath149 and @xmath150 are given by : [ c ] c^+(v ) & = & |v + 2| | |v + 1| - |v + 3| | + c^-(v ) & = & c^+(v - 4 ) + c^o(v ) & = & - c^+(v ) - c^-(v ) . for the k=0 ( i.e. , flat ) frw model the quantum constraint is obtained by simply replacing @xmath151 by @xmath148 in ( [ hc3 ] ) @xcite . thus , the @xmath47=0 quantum constraint has the same form as in the @xmath47=1 case . as one would expect from the classical expression ( [ hc1 ] ) , the difference @xmath152 is diagonal in the @xmath11-representation and vanishes when we set @xmath153 . @xmath8@xmath9 expectation values and dispersion of @xmath154 are compared with the classical trajectory and the trajectory from effective friedmann dynamics ( see ( [ eff ] ) ) . the classical trajectory deviates significantly from the quantum evolution at planck scale and evolves into singularities . the effective trajectory provides an excellent approximation to quantum evolution at all scales . @xmath9 zoom on the portion near the bounce point of comparison between the expectation values and dispersion of @xmath155 , the classical trajectory and the solution to effective dynamics . at large values of @xmath156 the classical trajectory approaches the quantum evolution . in this simulation @xmath157 , @xmath158 , and @xmath159.,title="fig:",width=307 ] expectation values and dispersion of @xmath154 are compared with the classical trajectory and the trajectory from effective friedmann dynamics ( see ( [ eff ] ) ) . the classical trajectory deviates significantly from the quantum evolution at planck scale and evolves into singularities . the effective trajectory provides an excellent approximation to quantum evolution at all scales . @xmath9 zoom on the portion near the bounce point of comparison between the expectation values and dispersion of @xmath155 , the classical trajectory and the solution to effective dynamics . at large values of @xmath156 the classical trajectory approaches the quantum evolution . in this simulation @xmath157 , @xmath158 , and @xmath159.,title="fig:",width=307 ] let us continue with the k=1 case . the form of the hamiltonian constraint is similar to that of a massless klein - gordon field in a static space - time , with a static potential . @xmath4 is the analog of the static time coordinate and the difference operator @xmath151 , of the spatial laplace - type operator plus the static potential . hence , the scalar field @xmath4 can again be used as ` emergent time ' in the quantum theory . as in the case , @xmath160 and @xmath161 provide a complete set of dirac observables . the physical inner product can again be fixed either by requiring that these operators be self - adjoint or by the group averaging method @xcite . this provides us with the _ physical sector _ of the theory . as in the case , one can construct states which are sharply peaked at given values @xmath162 and @xmath163 of @xmath160 and @xmath164 at an instant of ` time ' @xmath165 and use ( [ hc3 ] ) to ` evolve ' them . while the general structure of the resulting lqc theory is thus analogous to that of the theory , there is a dramatic difference the final results . numerical techniques play a vital role in this analysis . more precisely , numerical evolutions have been performed for several values of parameters and using two different techniques . they attest to the robustness of results . while solutions mimic the behavior when the universe is large , there is a radical departure in the strong curvature region : _ the big bang and the big crunch singularities are resolved and replaced by big - bounces . _ the main results can be summarized as follows . ( for details , see @xcite . ) consider a classical solution depicted in fig.[fig : wdw1 ] a which evolves from the big - bang to the big crunch , reaching a large maximum radius @xmath166 . fix a point on this trajectory where the universe has reached macroscopic size and consider a semi - classical state peaked at this point . _ such states remain sharply peaked throughout the given ` cycle ' , i.e. , from the quantum bounce near the classical big - bang to the quantum bounce near the classical big - crunch . _ note that the notion of semi - classicality used here is rather weak : these results hold even for universes with @xmath167 and the ` sharply peaked ' property improves greatly as @xmath168 grows . the trajectory defined by the expectation values of the dirac observable @xmath169 in the full quantum theory is in good agreement with the trajectory defined by the classical friedmann dynamics until the 4-d scalar curvature ( the only independent curvature invariant in isotropic , homogeneous models ) attains the value @xmath170 , or , equivalently , the energy density of the scalar field becomes comparable to a critical energy density @xmath171 . _ then the classical trajectory deviates from the quantum evolution . _ in the classical solution , scalar curvature and the matter energy density keeps increasing on further evolution , eventually leading to a big bang ( respectively , big crunch ) singularity in the backward ( respectively , forward ) evolution , when @xmath172 . the situation is very different with quantum evolution . now the universe bounces at @xmath173 , avoiding the past ( or the big bang ) and future ( or the big crunch ) singularities . the volume of the universe takes its minimum value @xmath174 at the bounce point . @xmath175 scales linearly with @xmath10 : has the same physical dimensions as @xmath176 and the numerical value of @xmath176 is @xmath177 v_min = ( ) ^ p _ ( 1.2810 ^ -33 ) p_consequently , @xmath175 can be _ much _ larger than the planck size . consider for example a quantum state describing a universe which attains a maximum radius of a megaparsec . then the quantum bounce occurs when the volume reaches the value @xmath178 , _ some @xmath179 times the planck volume . _ deviations from the classical behavior are triggered when the density or curvature reaches the planck scale , even when the volume is large . since @xmath175 _ is _ large at the bounce point of macroscopic universes , the so - called ` inverse volume effects ' i.e . the fact that @xmath180 are largely insignificant to the quantum dynamics of such universes . after the quantum bounce the energy density of the universe decreases and , when @xmath181 , the quantum evolution is well - approximated by the classical trajectory . on subsequent evolution , the universe recollapses both in classical and quantum theory at the value @xmath182 when energy density reaches a minimum value @xmath183 . @xmath184 scales as the 3/2-power of @xmath10 : [ vmax ] v_max = ( 16g/3 ^2)^ p_^ 0.6 p_^ quantum corrections to the classical friedmann formula @xmath185 are of the order @xmath186 . for a universe with @xmath187 , the correction is only one part in @xmath188 . for universes which grow to macroscopic sizes , classical general relativity is essentially exact near the recollapse . there is a hamiltonian analog of the ` effective action ' framework @xcite which provides a systematic procedure to obtain ` effective equations ' from quantum theory . while the classical friedmann equation is @xmath189 , the effective friedmann equation turns out to be [ eff ] ( ) ^2 = ( -_1(v ) ) where @xmath190 and @xmath191 are specific functions of @xmath11 . the term in the square bracket is the key quantum correction ; away from the planck regime , @xmath192 and @xmath193 . bounces occur when @xmath48 vanishes , i.e. at the value of @xmath11 at which the matter density equals @xmath194 or @xmath195 . @xmath196 at the classical recollapse while @xmath197 at the quantum bounce.=0 , i.e. open universes , the friedmann equation @xmath198 is replaced just by @xmath199 . since there is no classical recollapse , there is a single pre - big - bang contracting branch which is joined deterministically to the post - big - bang expanding branch . for details see @xcite . ] for quantum states under discussion , the density @xmath200 is well approximated by @xmath201 up to terms @xmath202 , independently of the details of the state and values of @xmath10 . ( for a universe with maximum radius of a megaparsec , @xmath203 . ) the density @xmath183 at the recollapse point also agrees with the value @xmath204 predicted by the classical evolution to terms of the order @xmath205 . furthermore the scale factor @xmath166 at which recollapse occurs in the quantum theory agrees to a very good precision with the one predicted by the classical dynamics . the trajectory obtained from effective friedmann dynamics is in excellent agreement with quantum dynamics _ throughout the evolution . _ in particular , the maximum and minimum energy densities predicted by the effective description agree with the corresponding expectation values of the density operator @xmath206 computed numerically . the state remains sharply peaked for a _ very large number of ` cycles ' . _ consider the example of a semi - classical state with an almost equal relative dispersion in @xmath10 and @xmath156 and peaked at a large classical universe of the size of a megaparsec . when evolved , it remains sharply peaked with relative dispersion in @xmath156 of the order of @xmath207 _ even after @xmath208 cycles of contraction and expansion ! _ any given quantum state eventually ceases to be sharply peaked in @xmath156 ( although it continues to be sharply peaked in @xmath10 ) . nonetheless , the quantum evolution continues to be deterministic and well - defined for an infinite number of cycles . this is in sharp contrast with the classical theory where the equations break down at singularities and there is no deterministic evolution from one cycle to the next . in this sense , in lqc the @xmath47=1 universe is _ cyclic _ , devoid of singularities . _ this non - singular evolution holds for all states , not just the ones which are semi - classical at late times . _ there is no fine tuning of initial conditions . also , there is no violation of energy conditions . indeed , quantum corrections to the matter hamiltonian do not play any role in the resolution of the singularity . the standard singularity theorems are evaded because the geometrical side of the classical einstein s equations is modified by the quantum geometry corrections of lqc . to summarize , the issues raised in section [ s1 ] have all been answered in the frw models . the main departures from the theory occur due to quantum geometry effects of lqg . these effects are small but dominate the planck scale physics by creating an effective repulsive force which can overwhelm gravitational attraction . while these effects are small outside the planck regime , in principle , they could have accumulated and led to departures from general relativity even in weak field regime on the very long time scales that are relevant to cosmology . _ this does not happen . _ while they dominate the planck regime , the quantum geometry effects die extremely quickly outside this regime so that in the weak field regime lqc is in excellent agreement with general relativity even on the very large cosmological time scales . in the last two section , i focused on a simple model to illustrate how one might hope to solve the long standing problems of quantum gravity using lqg . as i mentioned in the beginning of section [ s2 ] , several other cosmological models have been analyzed . roughly , work to date can be divided into three categories . * in the isotropic models with free scalar fields , with and without cosmological constant , there are detailed analytical as well as numerical frameworks to answer all questions of physical interest . in these models , there is only one non - trivial curvature invariant the scalar curvature @xmath209 and in the classical theory it is related very simply to energy density . therefore the onset of the quantum epoch which signals departures from classical general relativity can be described by critical values of either the scalar curvature or the density @xcite . * for the kasner model with anisotropies as well as models with physically interesting potentials for scalar fields , the physical sector of the theory can be constructed along the same lines . however , the analysis of effective equations is still somewhat incomplete and numerical analysis is still in infancy . in the anisotropic models the key features of the isotropic models resolution of singularities and emergence of semi - classical pre - big - bang branches persist . but new phenomena also emerge . with massless scalar fields as sources , the scalar curvature @xmath209 is again determined by the matter density but there are also other curvature invariants in particular because the weyl tensor is no longer zero . now the quantum epoch is reached when any one of these invariants reaches the planck regime , whence multiple bounces can occur @xcite . density is no longer the governing factor ; rather it is space - time curvature . in models with potentials , the scalar field does not always serve as internal time globally . however , it is still possible to construct the physical hilbert space using group averaging @xcite and introduce relational observables . a global emergent time aids our intuition immensely but is not essential in the construction of the physical sector of the theory . * in more complicated models black hole interiors @xcite and the so called midi - superspaces @xcite which are symmetry reduced but have an _ infinite _ number of degrees of freedom the hamiltonian constraint has been written down . it does not break down near the putative singularities signalling that they are resolved . however , the physical sector of the theory is yet to be constructed and numerical analysis has not yet been undertaken . but there do not appear to be any unsurmountable difficulties for these investigations to reach maturity in the near future . these advances are encouraging because they deal with the long standing questions i discussed in section [ s1 ] . furthermore , in contrast to string theory , space - like singularities that are resolved are of direct physical interest . however , the major downside is that these advances are based on symmetry reduction and the precise relation between these models and the full theory is still to be spelled out . while significant efforts are being made on this key problem @xcite , i think we are still at a preliminary stage largely because we do not have a clear candidate for full lqg . what lessons do these cosmological investigations have for full lqg ? by now , the kinematic structure of full lqg is well controlled . key open issues involve dynamical issues . work in this area has been primarily focused on mathematical problems in full generality . in the leading approaches , one first solves the diffeomorphism constraint and then imposes the hamiltonian constraint on the resulting diffeomorphism invariant states . the non - trivial achievement is that there are well - defined candidates for constraint operators @xcite . furthermore advances related to thiemann and dittrich s ` master constraint ' program @xcite strongly indicate that , for each admissible choice of the master constraint , one would be able to construct the physical sector of the theory . however , there is still a great deal of freedom in the definition of the constraint operators and , more importantly , the issue of existence of a sufficiently rich semi - classical sector has remained largely open . more recently , via their ` algebraic quantum gravity program , ' giesel and thiemann @xcite have introduced new strategies to address both these issues . here , the diffeomorphism and the hamiltonian constraints can be treated on an equal footing and imposed simultaneously . this enables one to address the long standing problem of recovering the classical constraint algebra and it is in fact possible to recover it using suitably defined semi - classical states @xcite . thus there is ongoing progress . nonetheless , fresh insights are needed to address key _ physical _ problems such as the fate of classical singularities in full lqg and the detailed recovery of einstein dynamics in the classical limit . since lqc has provided concrete solutions to these problems in simple models , a useful strategy would be to work ` from bottom up ' to less and less symmetric models . for example , in the symmetry reduced systems , one inevitably carries out a gauge fixing which provides a good control on individual space - time geometries rather than equivalence classes of them under diffeomorphisms . similarly , the standard description of the low energy world involves specific space - time metrics . while one can translate both these descriptions in a manifestly diffeomorphism invariant language , the result would be quite cumbersome . the procedures used in models can complement the more general and more systematic programs that are being pursued in full lqg . the idea would be to work one s way up by incorporating , at each step , lessons learned from the symmetry reduced models . these models suggest that , to address physically important issues , it may be essential to restrict oneself to interesting sectors of the theory finite , non - linear neighborhoods of the frw solutions , or of minkowski space - time , in the phase space of general relativity and exploit the additional structure such restrictions make available . secondly , in lqc one could get truly valuable insights by analyzing in detail states which are semi - classical in a suitable sense . by contrast , the discussion of _ dynamics _ in the full theory generally focuses on ` elementary states ' the spin networks . these are analogous to the ` n - photon states ' of the maxwell theory , while the semi - classical states are analogous to the coherent states . both span the full hilbert space in the maxwell theory and are convenient in different regimes . genuinely new insights could be gained in interesting sectors of lqg if one revisits the issue of constructing and solving the hamiltonian constraint from a new perspective . as in mini and midi superspace analyses , one could exploit an astute gauge - fixing of ( a part of ) the diffeomorphism constraint , and the extra structure provided by a basis of semi - classical states . results of lqc appear to provide valuable guidelines not only for constructing the physical sector of the theory along these lines but also for answering some of the most challenging physical questions . i have benefited from valuable discussions with many colleagues i would like to thank especially martin bojowald , alex corichi , james hartle , gary horowitz , veronika hubney , jerzy lewandowski , donald marolf , tomasz pawlowski , parampreet singh , thomas thiemann , kevin vandersloot and participants in the first stckleberg workshop . this work was supported in part by the nsf grants phy99 - 07949 and phy04 - 56913 , the alexander von humboldt foundation , the kramers chair program of the university of utrecht , and the eberly research funds of penn state . t. thiemann , _ introduction to modern canonical quantum general relativity _ ( cup , cambridge , at press ) . c. rovelli , graviton propagator from background - independent quantum gravity , phys . lett . * 97 * , 151301 ( 2006 ) , ` arxiv : gr - qc/0508124 ` ; + e. bianchi , l. modesto , c. rovelli , s. speziale , graviton propagator in loop quantum gravity , class . quantum grav . * 23 * , 6989 - 7028 ( 2006 ) , ` arxiv : gr - qc/0604044 ` ; + e. r. livine and s. speziale , group integral techniques for the spinfoam graviton propagator , ` arxiv : gr - qc/0608131 ` . m. bojowald , absence of singularity in loop quantum cosmology , phys . * 86 * , 5227 - 5230 ( 2001 ) , ` arxiv : gr - qc/0102069 ` , isotropic loop quantum cosmology , class . * 19 * , 2717 - 2741 ( 2002 ) , ` arxiv : gr - qc/0202077 ` . d - w chiou , loop quantum cosmology in bianchi type i models : analytical investigation , ` 0609029 ` + d - w chiou and k. vandersloot ( in preparation ) + m. bojowald , homogeneous loop quantum cosmology , class . quantum grav . * 20 * , 2595 - 2615 ( 2003 ) , ` arxiv : gr - qc/0303073 ` + m. bojowald , g. date and k. vandersloot , homogeneous loop quantum cosmology : the role of the spin connection , class . quantum grav . * 21 * , 1253 - 1278 ( 2004 ) , ` arxiv : gr - qc/0311004 ` . c. w. misner , mixmaster universe , phys . * 22 * , 1071 - 1074 ( 1969 ) ; + minisuperspace , in : _ magic without magic : john archibald wheeler ; a collection of essays in honor of his sixtieth birthday _ , edited by j. r. klauder ( w. h. freeman , san francisco , 1972 ) . d. marolf , refined algebraic quantization : systems with a single constraint ` arxives : gr - qc/9508015 ` ; + quantum observables and recollapsing dynamics , class . * 12 * , 1199 - 1220 ( 1995 ) ` arxiv : gr - qc:9404053 ` . j. lewandowski , a. okolow , h. sahlmann and t. thiemann , uniqueness of diffeomorphism invariant states on holonomy flux algebras , comm . phys . * 267 * , 703 - 733 ( 2006 ) ` arxiv : gr - qc/0504147 ` ; + c. fleishchack , representations of the weyl algebra in quantum geometry , ` arxiv : math - ph/0407006 ` . t. thiemann , anomaly - free formulation of non - perturbative , four - dimensional lorentzian quantum gravity , _ phys . lett . _ * b380 * , 257264 ( 1998 ) ; + quantum spin dynamics ( qsd ) , class . . grav . * 15 * 839873 ( 1998 ) ; + qsd v : quantum gravity as the natural regulator of matter quantum field theories , class . 15 * , 12811314 ( 1998 ) . j. willis , on the low energy ramifications and a mathematical extension of loop quantum gravity , ph.d . dissertation , the pennsylvaina state university ( 2004 ) ; + a. ashtekar , m. bojowald and j. willis , corrections to friedmann equations induced by quantum geometry , igpg preprint ( 2004 ) + v. taveras , igpg preprint ( 2006 ) . a. ashtekar and m. bojowald , quantum geometry and the schwarzschild singularity , class . * 23 * 391 - 411 ( 2006 ) , ` arxiv : gr - qc/0509075 ` + l. modesto , the kantowski - sachs space - time in loop quantum gravity ; ` arxiv : gr - qc/0411032 ` . m. bojowald , non - singular black holes and degrees of freedom in quantum gravity , phys . lett . * 95 * , 061301 ( 2005 ) ` arxiv : gr - qc/0506128 ` + m. bojowald and r. swiderski , spherically symmetric quantum geometry : hamiltonian constraint . class . quantum grav . * 23 * , 21292154 ( 2006 ) , ` arxiv : gr - qc/0511108 ` . j. brunnemann and t. thiemann , on ( cosmological ) singularity avoidance in loop quantum gravity , class . * 23 * , 1395 - 1428 ( 2006 ) ; ` arxiv : gr - qc/0505032 ` ; + j. engle , quantum field theory and its symmetry reduction . class.quant . * 23 * , 2861 - 2893 ( 2006 ) , ` arxiv : gr - qc/0511107 ` ; + on the physical interpretation of states in loop quantum cosmology , ` arxiv : gr - qc/0701132 ` ; + t. koslowski , reduction of a quantum theory , ` arxiv : gr - qc/0612138 ` . t. thiemann , the phoenix project : master constraint programme for loop quantum gravity , class . * 23 * , 2211 - 2248 ( 2006 ) , ` gr - qc/0305080 ` ; + quantum spin dynamics viii . the master constraint , class . * 23 * , 2249 - 2266 ( 2006 ) + b. dittrich and t. thiemann , testing the master constraint programme for loop quantum gravity i. general framework , class . * 23 * , 1025 - 1066 ( 2006 ) , ` gr - qc/0411138 ` ; + testing the master constraint programme for loop quantum gravity ii . finite dimensional systems , class . quant * 23 * , 1067 - 1088 ( 2006 ) , ` gr - qc/0411139 ` ; + testing the master constraint programme for loop quantum gravity iii . sl(2,r ) models , class . quant . grav . * 23 * , 1089 - 1120 ( 2006 ) , ` gr - qc/0411140 ` ; + testing the master constraint programme for loop quantum gravity iv . free field theories , class . quant . grav . * 23 * , 1121 - 1142 ( 2006 ) , ` gr - qc/0411141 ` . k. giesel and t. thiemann , algebraic quantum gravity ( aqg)i . conceptual setup , ` gr - qc/0607099 ` ; + algebraic quantum gravity ( aqg ) ii . semiclassical analysis , ` gr - qc/0607101 ` ; + algebraic quantum gravity ( aqg ) iii . semiclassical perturbation theory , ` gr - qc/0607101 ` . t. thiemann , gauge field theory coherent states ( gcs ) : i. general properties , class . quant . grav . * 18 * , 2025 - 2064 ( 2001 ) , ` gr - qc/0005233 ` ; + t. thiemann , o. winkler , gauge field theory coherent states : ii . peakedness properties , class . * 18 * , 2561 - 2636 ( 2001 ) , ` gr - qc/0005237 ` ; + gauge field theory coherent states : iii . ehrenfest theorems , class . . grav . * 18 * , 4629 - 4682 ( 2001 ) , ` gr - qc/0005234 ` ; + gauge field theory coherent states : iv . infinite tensor product and thermodynamical limit , class . quant . grav . * 18 * , 4997 - 5054 ( 2001 ) , ` gr - qc/0005237 ` ; + h. sahlmann , t. thiemann , o. winkler , coherent states for canonical quantum general relativity and the infinite tensor product extension , nucl . phys . * b606 * , 401 - 440 ( 2001 ) , ` gr - qc/012038 ` .

in this paper we continue the investigation reported by concerning the morphology of binary configurations obtained via the collapse of rotating parent gas structures with total masses in the range of m= 1 to 5 m . here we extend the mass range and consider the collapse of two uniform gas clumps of m= 50 and 400 m , so that they also rotates rigidly in such a way that its approximate virial parameter takes the values of 0.5 , 1.5 , and 2.5 and their collapse is induced initially by implementing an azimuthal mass perturbation . to assess the effects of the total mass of the parent gas structure on the nature of the resulting binary configurations , we also consider the collapse of two cores of m= 1 and 5 m . we calculate the collapse of all these parent gas structures using three values of the ratio of thermal energy to potential energy and for two values of the mass perturbation amplitude . we next calculate the binary separations , masses and integral properties of the binary fragments and present them in terms of the total mass of the parent structure . for most of our models , we finally calculate the extreme value , so that a model with a slightly higher value would no longer collapse .

1701.07397

stars are formed in gravitationally collapsed clouds made of molecular hydrogen gas . these gas clouds appear to be composed of many well - defined gas substructures , so that these assembled clouds can be very large and massive . thus , all gas structures have been characterized according to @xcite by means of their size and mass . for instance , a gas clump is a gas structure with mass and size in the range of 50 - 500 m@xmath1 and 0.3 - 3 pc , while for a gas core these values are in the range of 0.5 - 5 m@xmath1 and 0.03 - 0.2 pc . numerical simulations of the collapse of a rotating , self - gravitating , isolated cloud cores , began to be performed many years ago ; see @xcite , @xcite . one of the classic models of binary formation is based on the collapse of a rotating spherical core of 1 m@xmath1 , in which an azimuthal symmetric mass perturbation was initially implemented such that a binary system was formed by prompt fragmentation ; see the so - called `` standard isothermal test case '' calculated by @xcite , @xcite , @xcite , @xcite , @xcite , and @xcite , among others . it has always been very important to determine whether or not a gas structure is stable against gravitational collapse . as early as two decades ago , theorists proposed collapse and fragmentation criteria by constructing configuration diagrams , the axes of which are usually the ratio of thermal energy to potential energy , denoted by @xmath3 , versus the ratio of rotational energy to gravitational energy , denoted by @xmath2 ; see for instance @xcite , @xcite , @xcite , and @xcite . it must be emphasized that numerical simulations are still necessary to conclusively show the final configuration obtained from the collapse of a particular model of gas structure , as all these @xmath3 versus @xmath2 configuration diagrams mentioned are , above all , mostly indicative , due to the fact that the parameter space that determines the final simulation outcome is very large . we mention that the role played by the total mass of the parent core on the collapse results has been considered for a long time ago . for instance , @xcite chose initial conditions in order to study the collapse of a cloud of m=60 m@xmath4 and compared it to the collapse of three models of 1 m@xmath4 . @xcite investigated the collapse and fragmentation in four cloud models , of which the first three are for 1 m@xmath4 and the last one is a model of a very massive uniform cloud of 5000 m@xmath4 . furthermore , @xcite studied in depth the collapse of rotating , uniform non - isothermal clouds by considering a four dimensional parameter space composed by @xmath3 , @xmath2 , @xmath5 and the total mass m. the upper limit of m was 2 m@xmath1 while the range of @xmath3 and @xmath2 varied within 0.01 - 0.50 and 0 - 0.33 , respectively . he found four different types of configurations and noted that as the mass @xmath6 decreased , cloud fragmentation is less favored . more recently , @xcite reported numerical simulations focusing on only one type of binary configuration , consisting of a well - defined pair of mass condensations that approach each other , achieve rotational speed , swing past each other , and finally separate to form the desired binary system , in which the mass condensations orbit around one another at the end of the simulation . a schematic diagram was also presented in which this particular binary configuration can be located among others . similar configurations were also obtained by @xcite by increasing the external pressure on a rotating core . the configuration that interested @xcite corresponds to the disk - bar type fragmentation studied by @xcite and @xcite . the schematic diagram reported by @xcite was constructed with the total mass of the parent core m@xmath0 as the vertical axis in the range of 0.75 to 5 m@xmath1 , versus the dimensionless ratio of rotational energy to gravitational energy @xmath2 , as the horizontal axis in the range of 0.1 to 0.21 . the dimensionless ratio of thermal energy to potential energy , @xmath3 , was kept fixed for all simulations . it was also noted that the increase in the total mass of the parent core diminished the formation of the desired binary configuration and instead favored the formation of a single central mass condensation surrounded by a disk as the final result . observationally , much effort has been put forth for many years in order to determine some physical property that could be useful in deciding whether or not a particular gas structure will collapse . particularly , the virial parameter , denoted here by @xmath7 , has recently been measured in order to characterize the dynamical state of gas structures . for instance , @xcite recently compiled a catalog of 1325 molecular gas clouds of very different sizes , including estimates of their virial parameters . there exists a critical virial parameter , denoted here by @xmath8 , which comes from a stability study in which perturbations both in pressure and density gradients were considered ; it has been mathematically determined that @xmath9 . the common belief in the recent past has been that most gas structures have @xmath10 , so that either they do not collapse or will even be diffused into the interstellar medium . it is important to note that @xcite recently observed low values of @xmath7 for regions of high - mass star formation ; that is , these gas structures have @xmath11 , so they are expected to collapse . the virial parameter is defined as the ratio of virial mass m@xmath12 to the total mass m@xmath0 of a gas structure , so that in the case of a spherical model of radius @xmath13 , it is given by @xmath14 , where @xmath15 is the velocity dispersion and @xmath16 is the newton s gravitational constant . it should be noted that the velocity dispersion entering in the calculation of @xmath7 includes both the thermal and the non - thermal velocity components . when the velocity distribution of a particular model of a gas structure is considered to be composed only of a thermal component , it is possible to relate the @xmath7 to the @xmath2 ratio by means of @xmath17 , where @xmath18 is a form factor , which is empirically included to take into account modifications for non - homogeneous and non - spherical gas distributions . as it seems that the dimensionless critical virial parameter , which separates the gas clouds that collapse from those that do not , is close to 2 , we here prepare initial conditions of the sph particles to have a rigidly rotating gas structure , such that its @xmath2 is given by 0.1 , 0.3 , and 0.48 . a possible approximation would be that of using the average velocity @xmath19 of the particle distribution instead of the @xmath15 . thus , for a numerical simulation , we would have a virial parameter approximated by @xmath20 . so that the values of @xmath2 given above will correspond to simulations with their approximated @xmath7 values given by 0.5 , 1.5 , and 2.5 , respectively . thus , the question raised by the virial parameter in the observational field , of finding where is the transition between the cores that collapse to those that do not , in this paper we translated this question to the numerical simulations field , so that we here locate those models that are near to the non - collapsing regime , such that we call them `` the last collapsing configuration '' . as expected , the occurrence of this last collapsing configuration depends on initial values of @xmath3 and @xmath2 and on the total mass of the parent structure . with regard to the theoretical aspect , we mention that self - similar solutions for the collapse of isothermal spheres were first obtained by @xcite and @xcite , and later by @xcite . @xcite performed a detailed mathematical analysis of the collapse problem and re - discovered previous solutions and also obtained new solutions . other self - similar solutions were also found for the collapse of adiabatic spheres , in which a polytropic equation of state was assumed . all these solutions provide an accurate description of the early collapse stage , which mostly proceeds isothermally , as the gas is optically thin to its own radiation . a radial density profile was obtained from these self - similar solutions . in the case of a rotating collapse with pressure , @xcite was able to get also the mathematical expression of a radial density profile by using an approach based on the virial theorem . the physical parameters of the parent core are included in the density profile , so that it can be simply scaled for a different set of parameters of a second parent core . motivated by the work of @xcite mentioned above , in this paper we want to study the formation of binary systems via the collapse of a more massive gas structure ; thus , we carry out a fully three - dimensional set of numerical hydrodynamical simulations aimed to model the gravitational collapse of a m@xmath21 50 and 400 @xmath22 m@xmath23 clumps , by using the sph particle technique , such that a finite number of sph particles is used to sample the entire gas structure . the formation of two antipode mass seeds during the early collapse stage of the clumps is here enforced by implementing a mass perturbation with the same mathematical structure of the density perturbation successfully used in the classic collapse calculation of @xcite for an isothermal one solar mass core . when the peak density of the collapsing core is high enough , the central region of the core become opaque to its own radiation ; thus the isothermal regime breaks down so that an adiabatic regime begins . in our case , this transition is taken into account by using the barotropic equation of state , first proposed by @xcite , which is characterized by a critical density . now , the value of this critical density depends on how much mass surrounds and obscures the central region of the core . in general , it is expected that the more massive the parent cores are , the critical density at which the obscuration occurs may be lower . strictly speaking , this would imply to change the value of the critical density of the barotropic equation of state in our suite of simulations . in this paper , we were able to follow the collapse of the clumps up to three orders of magnitude in density within this adiabatic regime , where there are no scaling solutions known , out of which new solutions can be simply scaled from the initial mass of the parent core , as can be done in the isothermal collapse . because of this , the treatment of this advanced collapse stage must be investigated by means of numerical simulations . as expected , for the first collapse stage , no significance differences in the collapse of a higher mass structure are seen . however , as we will see later , the influence of a higher mass parent structure on the collapse get manifested only in the non - isothermal collapse stage ; see @xcite . in this paper we re - simulate two core models of 1 and 5 m@xmath1 with @xmath2 also given by 0.1 , 0.3 , and 0.48 , respectively . these models are similar but not identical to those already reported by @xcite . we consider here all those collapse models with values of @xmath3 given by 0.1 , 0.2 , and 0.3 , in order to make a comparison of our results with those of @xcite , who studied the thermal sensitivity of binary formation via the collapse of cores . to complement the qualitative study of binary configurations that we carry out in the first part of this paper , we next present a quantitative analysis of some of the resulting binary configurations , including the calculation of the binary separation and masses of the fragments of some particular binary configurations , and we present them in terms of the total mass of the parent structure . it should be mentioned that we have not considered here the collapse of a more massive gas structure , such as a gas cloud with mass and size of the order of 10@xmath24 - 10@xmath25 m@xmath1 and 2 - 15 pc . this is due to the fact that it has been suggested by several authors that the observed masses and core separations in regions of high - mass star formation can not be explained by invoking only fragmentation via gravitational instability ; see @xcite among others . thus , the formation process of very massive dense cores in high - mass star formation regions needs other mechanisms to be taken into account other than that of thermal fragmentation ( also called jeans fragmentation ) , for instance , turbulent fragmentation or magneto - hydrodynamical induced fragmentation . however , for both the cores and the intermediate mass clumps , such as those considered in this paper , it is expected that only thermal fragmentation is relevant to explain their fragmentation properties ; see @xcite , @xcite and @xcite . we finally mention that the collapse of cores to model isolated or binary star formation is an active field of research , for example , the evolution of rotating cloud cores including turbulent velocity initial distributions has been investigated by many authors ; see for instance : @xcite , @xcite , @xcite , @xcite , @xcite and @xcite . collapse calculations of an isothermal cloud core rotating in a uniform magnetic field have been presented , among others , by @xcite , @xcite , @xcite and @xcite . all the parent gas structures considered in this paper are uniform spheres , which are rigidly rotating around the @xmath26 axis with an angular velocity @xmath27 , so the initial velocity of the @xmath28 sph particle is given by @xmath29 . the time needed for a test particle to reach the center of a gas sphere , when gravity is the only force acting on it , is defined as the free fall time @xmath30 , and it is given by @xmath31 , where @xmath32 is the average density . we will use @xmath32 and @xmath30 to normalize the figures that appear in the coming sections . following @xcite , the dynamical state of a general gas structure is usually characterized by the values of the dimensionless ratios @xmath3 and @xmath2 , which are given by @xmath33 for a spherical model of a gas structure of total mass @xmath34 , the total gravitational potential energy is approximated by @xmath35 , where @xmath13 is the sphere radius . the average total thermal energy @xmath36 ( kinetic plus potential interaction terms of the molecules ) is @xmath37 , where @xmath38 is the boltzmann constant , @xmath5 is the equilibrium temperature , @xmath39 is the total number of molecules in the gas , and @xmath40 is the speed of sound . the rotational energy of the clump is approximately given by @xmath41 , where @xmath42 is the moment of inertia and @xmath43 is the total angular momentum . if we consider that the velocity dispersion @xmath15 of a gas structure is directly related only to the thermal velocity component , then @xmath44 where @xmath45 is the molecular mass of the main gas component and @xmath46 is the kinetic temperature . in this case , an approximation to be used in this paper would be using the average velocity @xmath19 of the simulation particles instead of the @xmath15 . thus , we would have a virial parameter approximated by @xmath20 . in this paper , we have carefully selected the values of @xmath40 and @xmath27 , so the energy ratios obtained by considering all the simulation particles of an initial configuration ( the snapshot zero of a simulation ) are initially given by a pairs of values taken from the following sets : @xmath47 so that the corresponding values of the virial parameter for all our models are approximately given , respectively , by @xmath48 we will consider the gravitational collapse of two cores : the first is a variant of the so - called `` standard isothermal test case , '' which was first calculated by @xcite and later calculated by @xcite and @xcite ; the main outcome of this classic model was a protostellar binary system . in this paper , the core radius is r=4.99 @xmath49 cm @xmath50 0.016 pc and its mass is m@xmath0=1 m@xmath1 . thus , the average density and the corresponding free fall time of this core are @xmath513.8 @xmath52 g @xmath53 and @xmath541.0 @xmath55 s @xmath56 34077 yr , respectively . the radius and mass of the second core are r=6.16 @xmath57 cm @xmath560.2 pc and m@xmath0=5 m@xmath1 , respectively . this core has an average density of @xmath511.0 @xmath58 g @xmath53 and a @xmath542.0 @xmath59 s @xmath60 yr , respectively . despite the size and mass differences of these gas structures , both of them are still cores , as defined statistically by @xcite . the first clump structure considered in this paper has a radius and mass given by r=0.3 pc @xmath50 9.24 @xmath61 cm and m@xmath0=50 m@xmath23 , respectively . its average density is given by @xmath623.0 @xmath63 g @xmath53 . the radius and mass of the second clump are given by r=1 pc @xmath50 3.08 @xmath64 cm and m@xmath0=400 m@xmath23 , respectively . its average density is @xmath656.5 @xmath66 g @xmath53 and its corresponding free fall time is @xmath67 s @xmath56826 247 yr . we emphasize that the radius and mass of the clumps have been changed with respect to those of the cores . in spite of this , it is still possible to make a comparison of the collapse results , as the focus of this paper is to study the effects of changing the values of @xmath3 and @xmath2 on the collapse of different parent gas structures . besides , if we had kept unchanged some physical property of the clump structures , so that we had used the core radius , while the total parent mass is different to that of the cores , then the resulting clump structure would be un - physical , in the statistical sense defined by @xcite . for every sph particle i of a simulation with total number of particles n@xmath68 , that is , i=1 .. n@xmath68 , the particle mass is given by @xmath69 ; then we implement a mass perturbation of the type @xmath70 where the perturbation amplitude is set to the values of a=0.1 and a=0.25 ; the mode is fixed to m=2 and @xmath71 is the azimuthal spherical coordinate . this mass perturbation scheme was successfully applied in many papers on collapse ; see for instance @xcite , @xcite , @xcite , and notably @xcite , when the gadget2 code was proven , among other tests , with the calculation of the isothermal collapse , where the unequal mass particles method was validated . @xcite demonstrated the need of fulfill appropriate spatial resolution requirements in order to avoid the occurrence of artificial fragmentation in a collapse simulation . following @xcite , the smallest mass that a @xmath72 calculation can resolve , @xmath73 , is given by @xmath74 where @xmath75 is the number of neighboring particles included in the @xmath72 kernel and @xmath76 is the spherical jeans mass @xmath76 , which is defined by @xmath77 where @xmath78 is jeans wavelength and @xmath79 is the peak density reached in a simulation . therefore , the smallest mass particle @xmath80 in our simulations must at least be such that @xmath81 . let us assume that @xmath82 , @xmath79=1.0 @xmath83 g @xmath53 and that the sound speed @xmath84 varies within the values 10 369 and 17 868 cm s@xmath85 , which correspond to runs with @xmath3=0.1 and 0.3 of the 1 m@xmath1 model , respectively . in this case , the @xmath73 is within 3.0 @xmath86 and 1.5 @xmath87 g. in the simulations of a 1 m@xmath1 core , we used n@xmath68=2 000 000 of sph particles , so that @xmath80 is @xmath88 g , and then we have that the ratio of masses is @xmath89 0.06 and 0.3 . therefore the number of particles is high enough to fulfill the resolution requirements described by @xcite . for the 5 m@xmath1 core , the particle mass is @xmath80=4.9 @xmath86 g while the sound speed varies within 9500.0 and 12400.0 cm s@xmath85 , so that by using again n@xmath68=2 000 000 of sph particles , the ratios @xmath89 are appropriate up to peak densities smaller than @xmath79=5.0 @xmath90 . for the 50 m@xmath1 clump , we used 12 000 000 of sph particles , so the particle mass is @xmath91 g. in this case , the sound speed varies within 17 497 and 29 576 cm s@xmath85 , and for a peak density of 1.0 @xmath92 , @xmath73 are 0.5 and 0.1 , respectively . for the 400 m@xmath1 clump , a very large number of particles is needed for the simulations to fulfill the resolution requirements . because computational limitations , we could not achieve the desired resolution and despite of this , we evolved these simulations using only 2 million of sph particles ; for this lack of reliability , we consider them useful only for comparison with the lower mass simulations . in order to ensure the change in the thermodynamic regime from isothermal to adiabatic , we here implement the barotropic equation of state proposed by @xcite : @xmath93 \label{beos}\ ] ] where @xmath94=5/3 and c@xmath95 is the sound speed . the critical density @xmath96 has only been given here the value @xmath97 g @xmath53 . as we will show in the following sections , in the simulations considered in this paper , the average peak density increases up to 3 orders of magnitude within the adiabatic regime . to follow the gravitational collapse of our models , in this paper we use the fully parallelized particle - based code gadget2 , which is based on the tree - pm method for computing the gravitational forces and on the standard sph method for solving the euler equations of hydrodynamics ; see @xcite and also @xcite . the gadget2 code has implemented a monaghan - balsara form for the artificial viscosity ; see @xcite and @xcite . the strength of the viscosity is regulated by setting the parameter @xmath98 and @xmath99 ; see equations 11 and 14 in @xcite . we have fixed the courant factor to @xmath100 . herein we used a radial mesh with spherical geometry , such that a set of concentric shells was created and populated with sph particles by means of a monte carlo scheme to set the initial particle configuration . thus , all the sph particles of each simulation , were located randomly in all the available surfaces of each spherical shell . the total mass in each shell is kept constant , so that the global density of the gas structure is also constant . in order to have a constant density distribution in a local sense , we next applied a radial perturbation to all the particles of a given shell such that any particle could be randomly displaced radially outward or inward , but preventing a perturbed particle from reaching another shell ; see section 2.1 of @xcite . in table [ tab : modelos ] we summarize the models considered in this paper and the main configurations obtained . column 1 shows the number of the model ; column 3 shows the total mass of the parent structure ; column 4 shows the value of the mass perturbation amplitude , defined by eq.[masspert ] ; column 5 and 6 show the values of the dimensionless ratios @xmath3 and @xmath2 , respectively ; column 7 shows the type of configuration obtained and column 8 shows the number of the figure and panel , in which the solution configuration is shown . we now emphasize five types of solution configurations that appear frequently , which may be termed as follows : ( i ) a face - to - face configuration , characterized by the presence of two fragments located one in front of the other , so that the connecting filament has been entirely accreted ; ( ii ) a connecting filament configuration , in which the filament still plays a relevant role as a gas bridge between the two fragments ; ( iii ) a central primary configuration , characterized by a rotationally supported solution surrounded by spiral arms ; ( iv ) a filament configuration , characterized by a thin and dense longitudinal collapsed structure without fragments present ; ( v ) a binary configuration , characterized by two fragments , so that they are orbiting one about the other . all of these configurations have already been observed before in the literature ; in the next section we describe how these configurations are transformed one into another because the change in parameters of the parent gas structure . the main results of this paper are shown in color iso - density plots for a slice of particles around the equatorial plane of the spherical gas structure . as we want to compare several collapse models in this paper , we use only one panel per model , which usually will correspond to the last available snapshot , and we use three panels to compose a mosaic in order to illustrate the resulting configurations obtained from a parent gas structure of total mass m@xmath0 . it should be mentioned that the vertical and horizontal axes of the iso - density plots indicate length in terms of the sphere radius , which varies from zero to one in all the plots for the initial snapshot , irrespective of the model ; therefore the cartesian axes x and y vary initially from -1 to 1 . in order to facilitate the comparison of the results , in each mosaic at least , we use the same length scale for all the panels . a comparison between models is still possible even at slightly different output times because most of the configurations have already entered a stable stage . otherwise , the time evolution should be done with the sink technique introduced by @xcite , so that the system can be evolved further in time . the panels of a mosaic have the same value of the virial parameter @xmath7 ( equivalently , with the same @xmath2 values of 0.1,0.3 and 0.48 ) for a given value of @xmath3 . we observed that all the models with the highest @xmath3 and @xmath2 , given by 0.3 and 0.48 , respectively , do not collapse but get dispersed . as we want to include only collapsing configurations , these models have been skipped and for this reason some mosaics are formed only by two panels . the last collapsing configurations have been grouped in other figures , because they are more likely to make better sense in mathematical terms , as their @xmath2 values may be too high to keep physical significance in the theory of star formation . finally , for each model with m@xmath0 , @xmath3 and @xmath2 given , we build two mosaics , each corresponding to a perturbation amplitude @xmath101 with values of @xmath1020.1 and @xmath1020.25 , respectively ; see eq . [ masspert ] ; see table [ tab : modelos ] . the figures [ cprueba60p1 ] , [ cprueba60p2 ] , and [ cprueba60p3 ] show the resulting systems obtained from the gravitational collapse of a m@xmath0= 1 m@xmath1 core , similar to `` the standard test case '' . in fig.[cprueba60p1 ] all the panels show a well - defined pair of mass condensations connected by a filament . when @xmath2 increases , the filament disappears so that the last collapsing configuration shown in panel 1 of fig.[lastcolconfm1 ] is a well defined face - to - face configuration . by observing the axes of the panels it can be seen that the separation of the resulting fragments has significantly been increased for this last collapsing configuration . when @xmath3 is increased from 0.1 to 0.2 , the pair of fragments can move one towards the other , so that an orbiting binary system is formed for lowest value of @xmath2 , and for higher values of @xmath2 , the configurations end with a central mass condensation surrounded by short spiral arms , as can be seen in the left and right panels of fig.[cprueba60p2 ] and also in the panel 2 of fig.[lastcolconfm1 ] . in fig.[cprueba60p3 ] , one can see that even for the lowest value of @xmath2 , the resulting configuration is a central primary , which is rotationally supported , such that the gravitational collapse slows down . it is interesting to observe the change in the outcome of models with the same @xmath2 , but in which the @xmath3 has been increased : from a face - to - face configuration type to a central primary type . this behavior was also observed by @xcite . when the mass perturbation is increased from @xmath1020.1 to @xmath1020.25 , we can see in fig.[cprueba70p1 ] and in panel 4 of fig.[lastcolconfm1 ] that a similar behavior is obtained to that which was seen in fig . [ cprueba60p1 ] , but the binary separations are quite larger than that seen in the corresponding panels of fig.[cprueba60p1 ] . we also noted that the number of orbiting binary systems is greatly increased , as can be seen in figs.[cprueba70p1 ] and [ cprueba70p2 ] ; only the last collapsing configuration finishes as a central primary , as can be seen in panel 6 of fig.[lastcolconfm1 ] . for all the values of @xmath2 , we see that a pair of mass condensations connected by a very smooth filament are formed , as can be seen in fig.[cprueba30p1 ] . when @xmath3 is increased from 0.1 to 0.2 , the mass condensations are very weak or they even disappear , while the filaments are thinner and more pronounced ; see fig . [ cprueba30p2 ] . for the highest value of @xmath1030.3 considered here , we see in fig.[cprueba30p3 ] that the corresponding filaments become thinner and narrower and again a rotationally supported central primary surrounded by long spiral arms is seen as the last collapsing configuration for the highest value of @xmath2 . when the perturbation scale takes the new value of @xmath104 , the filaments almost disappear and instead a pair of mass condensations are seen located face - to - face and more pronounced than before , as one can see in fig.[cprueba40p1 ] . later , when @xmath3 increases to 0.2 , the mass condensations are weaker but still face - to - face , while no dense filaments are seen in fig.[cprueba40p2 ] . we observe again that the binary separation is significantly increased in these runs , shown in fig.[cprueba40p1 ] and fig.[cprueba40p2 ] , as compared with those of fig.[cprueba30p1 ] and fig.[cprueba30p2 ] . for the highest @xmath1030.3 value considered here , we see in fig.[cprueba40p3 ] the appearance of weaker filaments and again the transition to a central primary . in any run of the models with m@xmath0= 5 m@xmath1 we have not directly observed the formation of an orbiting binary system . we first notice that there is almost no difference between the configurations obtained for the clump of m@xmath0= 50 m@xmath1 with those obtained for the m@xmath0= 400 m@xmath1 , as can be seen by comparing figs.[cprueba80p1],[cprueba80p2],[cprueba80p3 ] and [ cprueba90p1],[cprueba90p2 ] , [ cprueba90p3 ] with the corresponding figures of the largest mass clump . we take advantage of this similarity by using only the last collapsing configurations of the m@xmath0= 400 m@xmath1 clump to compare with the lower mass models of previous sections . in fig . [ cprueba80p1 ] one can see that configurations of the face - to - face type connected by a weak filament are mainly formed for the two lower values of @xmath2 . these filaments are fainter than those observed in the m@xmath0= 1 m@xmath1 model shown in fig.[cprueba60p1 ] while they are very similar to those obtained for the m@xmath0= 5 m@xmath1 model shown in fig.[cprueba30p1 ] . in the right panel of fig . [ cprueba0p1 ] and panel 1 of fig.[lastcolconfm400 ] , which corresponds to a @xmath1050.3 and @xmath1050.7 , respectively , still show the face - to - face configuration , as was the case for the two models of smaller total mass considered ; see the panels 1 of fig.[lastcolconfm1 ] and fig.[lastcolconfm5 ] . when @xmath3 takes the value of 0.2 , the filament formed for a large @xmath2 is fragmenting as can be seen in the right panel of fig . [ cprueba80p2 ] , while the opposite case was seen for the m@xmath0= 5 m@xmath1 model in fig.[cprueba30p2 ] ; that is , the filament becomes denser and thinner while in the m@xmath0= 5 m@xmath1 model shown in fig.[cprueba60p2 ] there is no filament for comparison . when @xmath2 takes its highest possible value , still in the collapsing regime , the filament of the m@xmath0= 400 m@xmath1 model is replaced by a central primary configuration surrounded by spiral arms ; see the panel 2 and 3 of fig . [ lastcolconfm400 ] and the bottom left panel of fig . [ cprueba0p3 ] . this behavior was also observed in the m@xmath0= 5 m@xmath1 models in fig.[cprueba30p2 ] and fig.[cprueba30p3 ] but was not observed in the m@xmath0= 1 m@xmath1 models with @xmath1030.2 in fig.[cprueba60p2 ] nor with @xmath1030.1 in fig.[cprueba60p1 ] , although it occurred for this model with @xmath1030.3 , as can be seen in fig.[cprueba60p3 ] . when the mass perturbation is increased to @xmath1020.25 , more configurations of the face - to - face type appear connected by a tenuous bridge of gas ; see fig . [ cprueba50p1 ] . this behavior was also observed in the m@xmath0= 5 m@xmath1 model shown in fig.[cprueba40p1 ] , but it must be mentioned that the binary separation is much larger in the former model than in the latter . something similar was also observed for the m@xmath0= 1 m@xmath1 model , but there the filaments are more pronounced , as can be seen in fig.[cprueba70p1 ] . the runs with @xmath1030.3 are similar for the m@xmath0= 50 m@xmath1 models and m@xmath0= 5 m@xmath1 shown in fig . [ cprueba50p2 ] and fig.[cprueba40p1 ] , respectively , in which we see configurations of the face - to - face type . it should be noted that in the m@xmath0=1 m@xmath1 model , the configurations formed are not at all similar to the ones seen for the corresponding two models with larger parent mass , as an orbiting binary configuration is only seen in fig.[cprueba70p2 ] . for the @xmath106 400 @xmath107 models with @xmath1030.3 , we see a well - defined filament that is transformed to a central primary configuration when @xmath2 takes its highest value ; see fig . [ cprueba50p3 ] ; this was also observed in fig.[cprueba40p3 ] for the m@xmath0= 5 m@xmath1 model . but again , there is no possible comparison with the configuration obtained for the m@xmath0= 1 m@xmath1 model shown in fig.[cprueba70p3 ] , in which orbiting binaries were indeed observed even for this higher value of @xmath3 . for all the models of m@xmath0= 50 m@xmath1 , we did not observe the formation of any orbiting binary system . we measure the uppermost @xmath2 values for all models except for the clump of m@xmath0= 50 m@xmath1 , so that the parent structure is still within the collapsing regime to the left of these curves , while to the right of each curve , there is no collapse . the configurations found are shown in figs . [ lastcolconfm1 ] , [ lastcolconfm5 ] and [ lastcolconfm400 ] . in fig . [ lcurve ] we observe that the curve for the models with a=0.25 are always located to the right of the corresponding curve for a=0.1 ; this fact is easy to understand , as the mass perturbation weighs more , then the resulting fragments are more massive , so they need more rotational energy to reach the level in which the equilibrium between the centrifugal force and the gravitational force is to be overcome . when this happens , the fragments are separated indefinitely and the gas structure does not collapse . we measure the most important integral properties of the resulting collapse configurations : for binaries , we consider the fragment masses and the binary separations ; for the central primaries , we consider the central mass and radius ; for all the resulting fragments or primaries , we also calculate the values of the energy ratios @xmath108 and @xmath109 . in order to calculate these properties , we proceeded as follows : we locate the highest density particle in the fragment s region ; this particle is considered the center of the fragment . the binary separation is simply calculated as the distance between the centers associated with each fragment . we next find all the particles which have density above or equal to some minimum density value , given in advance as @xmath110 for all the models . we finally check that these particles are also located within a given maximum radius r@xmath111 from the fragment s center . this set of particles allowed us to calculate the fragment integral properties , as the mass , including the @xmath108 and @xmath109 ; see @xcite . the results obtained for the face - to - face configurations are shown in figs.[sepfrags ] , [ massfrags ] and [ alphavsbetafrags ] . for the central primary configurations , we report the mass of the central condensations results and their integral properties , which are shown in figs.[masscenp ] and [ alphavsbetacenp ] . all the fragment properties of the face - to - face and of central primary configurations have also been reported in tables [ tab : propfrags ] and [ tab : propfrags2 ] , respectively . the entries of these tables are as follows . the first column shows the number of the model ; the total parent mass m@xmath0 is shown in the second column ; column third shows r@xmath111 in terms of the initial sphere radius r@xmath95 ; columns 4 , 5 and 6 of table [ tab : propfrags ] show the mass of the fragments given in terms of m@xmath0 and the binary separation in astronomical units ; in the case of table [ tab : propfrags2 ] , the last column shows the mass of the central primary region in terms of m@xmath0 . the last columns give the values of the @xmath108 and @xmath109 . the mass seeds , implemented by means of the mass perturbation of eq . [ masspert ] , accrete mass slowly during the early evolution stage of all the simulations until they become a well - defined pair of mass condensations that move through a gas of particles . shortly thereafter , when the mass condensations are massive enough , their translational motion can be slowed down or even stopped as a consequence of the dynamical friction of the surrounding gas . then , the mass condensations appear to be face - to - face . at this time , a competition begins between their gravitational force , which favors their approach , and the centrifugal force due to the parent structure rotation , which favors their separation . also , the mass condensations may or may not be connected by a filament , and as we have seen , these connecting filaments can be of very different types depending on the total mass of the parent structure and on the values of @xmath3 and @xmath2 . for the m@xmath0= 1 m@xmath1 core models , a good combination of these forces occurs such that the assembled mass condensations are sufficiently massive yet the centrifugal force is strong enough for their gravitational attraction to make them approach each other , swing past each other , and finally separate to form an orbiting binary system . we have observed that the occurrence of these events is significantly increased when the mass perturbations weigh more , as the mass condensations are naturally more massive to overcome their centrifugal force more easily , while the @xmath2 must still be high enough to make them swing past one another . by contrast , when the mass condensations are too massive , they approach each other , make contact , and finally merge to form a single central primary condensation . in this case , the gravitational force between the mass condensations has too easily overcome the rotational centrifugal force . for the m@xmath0= 5 m@xmath1 and m@xmath0= 50 m@xmath1 models , we observed that the mass condensations remain face - to - face in most of the runs , such that if @xmath2 increases further , then the separation of the face - to - face configuration gets larger up to the point that the parent structure no longer collapses . however , for higher values of @xmath2 , these face - to - face configurations give rise to a central primary configuration , such that for a slightly higher @xmath2 , the parent structure does not collapse but instead expands outwards but with spherical symmetry . we see in fig.[sepfrags ] that the binary separation is correlated with the mass of the parent gas structure . it seems that no correlation exists in the case of the fragment mass , as can be see in fig.[massfrags ] . this missing correlation is expected on physical grounds , in fact , for the taurus dark cloud , @xcite reported a correlation between the mass of the newly formed stars and the mass of the associated dense proto - stellar cores . more recently , large proto - stellar masses were observed by @xcite . a potentially interesting issue is the fact that some binary systems have a barycentre that is not at the cartesian coordinate origin ; see for instance the panels of fig.[cprueba30p1 ] . a possible explanation for this fact is that the masses of the fragments of some binary systems are different ; see fig.[massfrags ] and table[tab : propfrags ] . while it is true that the perturbation mass implemented in eq.[masspert ] has axial symmetry , it should be remembered that each spherical radial shell was randomly populated by means of a monte carlo method ; see [ subs : setup ] . we think that the origin of the unequal masses of the fragments and the displacement of the barycentre can be explained by the development of an asymmetrical random mass seed , that grow over time and are manifested mainly later , in the more evolved binary systems . the fact that some binary systems show upward displacement while others downward , gives support to this explanation . we notice that the mass fraction of the central primary configurations is similar for all the models reported in table [ tab : propfrags2 ] ; this is so because the number of sph particles entering into the central region is similar for all the models ; therefore , the mass of the central primary configuration does not scale with the mass of the parent structure;see fig.[masscenp ] . it is clear that the values calculated for the energy ratios @xmath108 and @xmath109 , unfortunately depend on the values chosen for @xmath112 and @xmath113 , so that there inevitably is certain ambiguity in defining the fragment s boundaries . despite of this , when we calculated the @xmath108 and @xmath109 values we observe that some binary fragments do show a clear tendency to virialize , as it can be appreciated in fig.[alphavsbetafrags ] . we emphasize that a similar conclusion can be drawn from the calculation of @xcite , where plots of the @xmath108 and @xmath109 time evolution were also presented . on the contrary , for the central primaries , we do not observe any trend to approach the virial line ; see fig.[alphavsbetacenp ] . in this paper , we have considered the gravitational collapse of several rotating spherical gas structures of very different size and mass , such that the collapse was triggered initially by means of the same azimuthal mass perturbation , whose amplitude has been allowed to take the values @xmath1020.1 and @xmath1020.25 . for the sake of comparison between our models , we have used the same set of values of the @xmath3 and @xmath2 ratios . the range of @xmath3 considered here , 0.1 - 0.3 , was similar to that of @xcite ; this range reflects the observational fact that the temperatures of the gas structures are mostly in the range of 9 - 12 k ; see @xcite . the range of @xmath2 considered here , 0.1,0.3 and 0.48 , was initially motivated by recent observations of the virial parameters of clouds , so that the critical virial parameter is within the chosen range , as we mentioned in sect.[intro ] . however , we have extended this range of @xmath2 in order to look for the last collapsing configurations , for which purpose the new range has been increased to values of 0.9 for the lowest @xmath3 and around 0.4 for the highest @xmath3 . this range of @xmath2 values is wider than those considered by other authors ; see for instance , @xcite , who used a maximum @xmath2 of 0.3 . we have thus revisited many binary configurations already seen in other works , but here the configurations change depending on the mass of the parent structure . based on these results , we present the following assessment of the effect of parent mass on the collapse . the resulting configurations for panels with @xmath1030.1 have a well - defined pair of mass condensations connected by a dense filament for the m@xmath0= 1 m@xmath1 model . the connecting filaments become weaker and almost disappear when the total mass increases to m@xmath0= 5 m@xmath1 and m@xmath0= 50 m@xmath1 , respectively . the last collapsing configurations are of the same kind ; that is , of the face - to - face type ; see sect.[sec : resultados ] . when @xmath1030.2 , we have observed the formation of many orbiting binary configurations for the m@xmath0= 1 m@xmath1 model ; this tendency becomes more pronounced when the mass perturbation increases from @xmath1020.1 to @xmath104 , so that the more massive the mass condensations , the greater the probability of an orbiting binary configuration . for the m@xmath0= 5 m@xmath1 model , what has been observed primarily are configurations where thin and dense filaments connect two very small mass condensations . when the mass of the parent gas structure is m@xmath0= 50 m@xmath1 , mainly face - to - face configurations are formed , in which the mass condensations are predominant and the filaments become thicker and less dense than those seen in the previous m@xmath0=5 m@xmath1 model . the last collapsing configurations are of three kinds ; i.e. , some runs finish with a central primary ( see figs.[cprueba60p1 ] and [ cprueba0p1 ] ) ; other runs finish with a binary ( see figs.[cprueba70p2 ] , figs.[cprueba40p2 ] and [ cprueba50p2 ] ) , and finally , only one run finishes with a thin and dense filament ( see fig.[cprueba30p2 ] ) . when @xmath1030.3 , we have seen the formation of most of the types of binary configurations discussed in this paper ; i.e. , some runs show a central primary , and other runs show a well - defined pair of mass condensations connected by a filament that becomes weaker as the total parent mass increases . however , the last collapsing configurations are of the type of a central primary for all the models . it must be emphasized that we have seen more differences between the outcomes of the m@xmath0=1 m@xmath1 model and those obtained from the m@xmath0=5 m@xmath1 , m@xmath0=50 m@xmath1 and m@xmath0=400 m@xmath1 models . in addition , we have seen many similarities in the outcomes of these last two models . these observations indicate that there is a critical total mass not much larger than one solar mass , for which the mass condensations to be formed will be too massive and the centrifugal force will not be able to make them start to orbit around each other . as we saw in sect.[sec : dis ] , the binary systems formed from massive gas structures can be the origin of wide binaries , in which a very large separation between the mass components can be reached for large values of @xmath2 . it is interesting to mention that a high frequency of wide binaries has recently been observed in regions having very different physical properties ; see @xcite . lastly , we mentioned that the mass perturbation shown in eq . [ masspert ] was invented to favor fragmentation of the so - called `` standard isothermal test case '' of m@xmath0=1 m@xmath1 . for the results outlined in this paper and those reported in other papers , we conclude that this mass perturbation scheme works quite well for the low - mass gas structure , but needs some improvement for the higher mass case . in order to calculate the integral properties of the some configurations , we took particles by applying selection criteria based on two parameters , a minimum density and a maximum radius , whose values were fixed in advance . thus , one would expect slight differences in the reported results as they are definition - dependent . nevertheless , we find that there is a clear correlation between total mass of the parent gas structure and the separation of binary fragments ; we do not observe a similar correlation for the mass of the fragments . it must be emphasized that the initial grid of spherical concentric shells , where sph particles were randomly located on all the available radial surface of each shell , appears to be a well - suited setup for numerical simulations of binary formation as we have obtained in a natural way , that the mass of the fragments in binary configurations are not always equal one to the other . the author thankfully acknowledge the computer resources , technical expertise and support provided by the laboratorio nacional de supercmputo del sureste de m ' exico through the grant number o-2016/047 ; i thank acarus - unison for the use of their computing facilities in the development of this manuscript . bodenheimer , p. , tohline , j. e. and black , d. c. , 1980 , international astronomical union , colloquium on stellar hydrodynamics , 58th , los alamos , n. mex . , 12 - 15 , 1980 . space science reviews , vol . 27 , nov .- dec . 1980 , p. 247 - 252 ; discussion , p. 253 . palau , a. , ballesteros - paredes , j. , vazquez - semanedi , e. , sanchez - monge , a. , estallela , r. , fall , s.m . , zapata , l.a . , camacho , v. gomez , l. naranjo - romero , r. , busquet , g. and fontani , f. , 2015 , mnras , * 453 * , issue 4 , pp.3785 - 3797 . lcccccc 1 & 1 & 0.1 & 0.1 & 0.1 & connecting filament & [ cprueba60p1 ] l + 2 & 1 & 0.1 & 0.1 & 0.3 & connecting filament & [ cprueba60p1 ] m + 3 & 1 & 0.1 & 0.1 & 0.48 & connecting filament & [ cprueba60p1 ] r + 4 & 1 & 0.1 & 0.1 & 0.758 & face - to - face & [ lastcolconfm1 ] 1 + 5 & 1 & 0.1 & 0.2 & 0.1 & binary & [ cprueba60p2 ] l + 6 & 1 & 0.1 & 0.2 & 0.3 & connecting filament & [ cprueba60p2 ] m + 7 & 1 & 0.1 & 0.2 & 0.48 & central primary with arms & [ cprueba60p2 ] r + 8 & 1 & 0.1 & 0.2 & 0.5344 & central primary & [ lastcolconfm1 ] 2 + 9 & 1 & 0.1 & 0.3 & 0.1 & central primary with disk & [ cprueba60p3 ] l + 10 & 1 & 0.1 & 0.3 & 0.3 & central primary with arms & [ cprueba60p3 ] m + 11 & 1 & 0.1 & 0.3 & 0.3983 & central primary & [ lastcolconfm1 ] 3 + 12 & 1 & 0.25 & 0.1 & 0.1 & connecting thin filament & [ cprueba70p1 ] l + 13 & 1 & 0.25 & 0.1 & 0.3 & connecting filament & [ cprueba70p1 ] m + 14 & 1 & 0.25 & 0.1 & 0.48 & connecting filament & [ cprueba70p1 ] r + 15 & 1 & 0.25 & 0.1 & 0.8871 & face - to - face & [ lastcolconfm1 ] 4 + 16 & 1 & 0.25 & 0.2 & 0.1 & binary & [ cprueba70p2 ] l + 17 & 1 & 0.25 & 0.2 & 0.3 & binary & [ cprueba70p2 ] m + 18 & 1 & 0.25 & 0.2 & 0.48 & binary & [ cprueba70p2 ] r + 19 & 1 & 0.25 & 0.2 & 0.5411 & binary & [ lastcolconfm1 ] 5 + 20 & 1 & 0.25 & 0.3 & 0.1 & binary & [ cprueba70p3 ] l + 21 & 1 & 0.25 & 0.3 & 0.3 & binary & [ cprueba70p3 ] m + 22 & 1 & 0.25 & 0.3 & 0.3983 & central primary & [ lastcolconfm1 ] 6 + 23 & 5 & 0.1 & 0.1 & 0.1 & connecting filament with disk & [ cprueba30p1 ] l + 24 & 5 & 0.1 & 0.1 & 0.3 & connecting filament & [ cprueba30p1 ] m + 25 & 5 & 0.1 & 0.1 & 0.48 & face - to - face & [ cprueba30p1 ] r + 26 & 5 & 0.1 & 0.1 & 0.6118 & face - to - face & [ lastcolconfm5 ] 1 + 27 & 5 & 0.1 & 0.2 & 0.1 & thin filament & [ cprueba30p2 ] l + 28 & 5 & 0.1 & 0.2 & 0.3 & thin filament & [ cprueba30p2 ] m + 29 & 5 & 0.1 & 0.2 & 0.48 & thin filament & [ cprueba30p2 ] r + 30 & 5 & 0.1 & 0.3 & 0.1 & connecting filament & [ cprueba30p3 ] l + 31 & 5 & 0.1 & 0.3 & 0.3 & thin filament & [ cprueba30p3 ] m + 32 & 5 & 0.1 & 0.3 & 0.3594 & central primary with arms & [ lastcolconfm5 ] 3 + 33 & 5 & 0.25 & 0.1 & 0.1 & face - to - face & [ cprueba40p1 ] l + 34 & 5 & 0.25 & 0.1 & 0.3 & face - to - face & [ cprueba40p1 ] m + 35 & 5 & 0.25 & 0.1 & 0.48 & face - to - face & [ cprueba40p1 ] r + 36 & 5 & 0.25 & 0.1 & 0.6989 & face - to - face & [ lastcolconfm5 ] 4 + 37 & 5 & 0.25 & 0.2 & 0.1 & connecting filament & [ cprueba40p2 ] l + 38 & 5 & 0.25 & 0.2 & 0.3 & face - to - face & [ cprueba40p2 ] m + 39 & 5 & 0.25 & 0.2 & 0.48 & face - to - face & [ cprueba40p2 ] r + 40 & 5 & 0.25 & 0.2 & 0.7467 & face - to - face & [ lastcolconfm5 ] 5 + 41 & 5 & 0.25 & 0.3 & 0.1 & connecting thin filament & [ cprueba40p3 ] l + 42 & 5 & 0.25 & 0.3 & 0.3 & connecting filament & [ cprueba40p3 ] m + 43 & 5 & 0.25 & 0.3 & 0.3918 & central primary with arms & [ lastcolconfm5 ] 6 + 44 & 50 & 0.1 & 0.1 & 0.1 & connecting filament & [ cprueba80p1 ] l + 45 & 50 & 0.1 & 0.1 & 0.3 & face - to - face & [ cprueba80p1 ] m + 46 & 50 & 0.1 & 0.1 & 0.48 & face - to - face & [ cprueba80p1 ] r + 47 & 50 & 0.1 & 0.2 & 0.1 & connecting thin filament & [ cprueba80p2 ] l + 48 & 50 & 0.1 & 0.2 & 0.3 & connecting filament & [ cprueba80p2 ] m + 49 & 50 & 0.1 & 0.2 & 0.48 & thin filament & [ cprueba80p2 ] r + 50 & 50 & 0.1 & 0.3 & 0.1 & connecting filament & [ cprueba80p3 ] l + 51 & 50 & 0.1 & 0.3 & 0.3 & thin filament & [ cprueba80p3 ] m + 52 & 50 & 0.25 & 0.1 & 0.1 & face - to - face & [ cprueba90p1 ] l + 53 & 50 & 0.25 & 0.1 & 0.3 & face - to - face & [ cprueba90p1 ] m + 54 & 50 & 0.25 & 0.1 & 0.48 & face - to - face & [ cprueba90p1]r + 55 & 50 & 0.25 & 0.2 & 0.1 & connecting filament & [ cprueba90p2 ] l + 56 & 50 & 0.25 & 0.2 & 0.3 & face - to - face & [ cprueba90p2 ] m + 57 & 50 & 0.25 & 0.2 & 0.48 & face - to - face & [ cprueba90p2 ] r + 58 & 50 & 0.25 & 0.3 & 0.1 & connecting filament & [ cprueba90p3 ] l + 59 & 50 & 0.25 & 0.3 & 0.3 & face - to - face & [ cprueba90p3 ] m + 60 & 400 & 0.1 & 0.1 & 0.1 & connecting filament with disk & [ cprueba0p1 ] l + 61 & 400 & 0.1 & 0.1 & 0.3 & connecting filament & [ cprueba0p1 ] m + 62 & 400 & 0.1 & 0.1 & 0.48 & connecting filament & [ cprueba0p1 ] r + 63 & 400 & 0.1 & 0.1 & 0.7 & face - to - face & [ lastcolconfm400 ] 1 + 64 & 400 & 0.1 & 0.2 & 0.1 & filament & [ cprueba0p2 ] l + 65 & 400 & 0.1 & 0.2 & 0.3 & connecting filament & [ cprueba0p2 ] m + 66 & 400 & 0.1 & 0.2 & 0.48 & filament & [ cprueba0p2 ] r + 67 & 400 & 0.1 & 0.2 & 0.55 & central primary with arms & [ lastcolconfm400 ] 2 + 68 & 400 & 0.1 & 0.3 & 0.1 & connecting thin filament & [ cprueba0p3 ] l + 69 & 400 & 0.1 & 0.3 & 0.3 & filament & [ cprueba0p3 ] m + 70 & 400 & 0.1 & 0.3 & 0.38 & central primary with arms & [ lastcolconfm400 ] 3 + 71 & 400 & 0.25 & 0.1 & 0.1 & connecting filament & [ cprueba50p1 ] l + 72 & 400 & 0.25 & 0.1 & 0.3 & connecting filament with disk & [ cprueba50p1 ] m + 73 & 400 & 0.25 & 0.1 & 0.48 & face - to - face & [ cprueba50p1 ] r + 74 & 400 & 0.25 & 0.1 & 0.81 & face - to - face & [ lastcolconfm400 ] 4 + 75 & 400 & 0.25 & 0.2 & 0.1 & connecting filament with disk & [ cprueba50p2 ] l + 76 & 400 & 0.25 & 0.2 & 0.3 & connecting weak filament & [ cprueba50p2 ] m + 77 & 400 & 0.25 & 0.2 & 0.48 & face - to - face & [ cprueba50p2 ] r + 78 & 400 & 0.25 & 0.2 & 0.55 & face - to - face & [ lastcolconfm400 ] 5 + 79 & 400 & 0.25 & 0.3 & 0.1 & connecting filament with disk & [ cprueba50p3 ] l + 80 & 400 & 0.25 & 0.3 & 0.3 & connecting weak filament & [ cprueba50p3 ] m + 81 & 400 & 0.25 & 0.3 & 0.385 & central primary with arms & [ lastcolconfm400 ] 6 + lccccccccc 4 & 1 & 0.225 & 5.5e-02 & 4.9e-02 & 6877.1 & 0.11 & 0.28 & 0.12 & 0.22 + 5 & 1 & 0.02 & 1.4e-01 & 1.32e-01 & 133.5 & 0.15 & 0.32 & 0.15 & 0.34 + 6 & 1 & 0.04 & 8.3e-02 & 1.46e-02 & 1062.3 & 0.13 & 0.30 & 0.07 & 0.72 + 15 & 1 & 1.0 & 3.8e-02 & 3.4e-02 & 19004.2 & 0.15 & 0.24 & 0.19 & 0.13 + 17 & 1 & 0.04 & 1.3e-01 & 5.6e-03 & 1856.3 & 0.16 & 0.31 & 0.09 & 0.86 + 18 & 1 & 0.1 & 1.0e-01 & 1.0e-01 & 2720.3 & 0.12 & 0.31 & 0.12 & 0.31 + 19 & 1 & 0.02 & 5.4e-02 & 5.1e-02 & 324.5 & 0.17 & 0.30 & 0.16 & 0.30 + 20 & 1 & 0.015 & 1.3e-01 & 1.2e-01 & 333.6 & 0.19 & 0.26 & 0.19 & 0.26 + 21 & 1 & 0.025 & 9.8e-02 & 9.8e-02 & 333.8 & 0.17 & 0.29 & 0.17 & 0.29 + 25 & 5 & 0.1 & 5.0e-02 & 4.99e-02 & 3206.4 & 7.7 & 0.23 & 6.40 & 0.20 + 26 & 5 & 0.15 & 5.3e-02 & 5.25e-02 & 55284.9 & 0.28 & 0.14 & 0.27 & 0.14 + 33 & 5 & 0.15 & 2.0e-01 & 1.6e-01 & 2002.2 & 2.24 & 0.50 & 0.50 & 0.58 + 34 & 5 & 0.15 & 1.2e-01 & 1.2e-01 & 2848.5 & 5.18 & 0.30 & 5.10 & 0.31 + 35 & 5 & 0.15 & 8.4e-02 & 2.2e-02 & 5005.9 & 5.90 & 0.26 & 0.08 & 0.58 + 36 & 5 & 0.15 & 5.36e-02 & 5.29e-02 & 86903.3 & 0.14 & 0.23 & 0.07 & 0.22 + 38 & 5 & 0.1 & 8.3e-02 & 4.96e-02 & 2502.3 & 5.6 & 0.24 & 0.12 & 0.47 + 39 & 5 & 0.1 & 5.7e-02 & 1.04e-02 & 5005.6 & 14.4 & 0.24 & 0.16 & 0.6 + 40 & 5 & 0.15 & 4.8e-02 & 4.8e-02 & 99934.4 & 0.40 & 0.20 & 0.17 & 0.22 + 45 & 50 & 0.062 & 4.77e-02 & 5.09e-02 & 37642.66 & 0.05 & 0.29 & 0.06 & 0.36 + 46 & 50 & 0.125 & 6.74e-02 & 5.35e-02 & 55592.61 & 0.10 & 0.21 & 0.06 & 0.30 + 52 & 50 & 0.0625 & 8.19e-02 & 7.87e-02 & 33227.5 & 0.03 & 0.45 & 0.02 & 0.45 + 53 & 50 & 0.1 & 7.69e-02 & 7.41e-02 & 53690.4 & 0.05 & 0.31 & 0.04 & 0.28 + 54 & 50 & 0.125 & 6.91e-02 & 6.66e-02 & 77490.5 & 0.08 & 0.28 & 0.05 & 0.23 + 56 & 50 & 0.05 & 4.63e-02 & 4.38e-02 & 40819.7 & 0.17 & 0.26 & 0.07 & 0.26 + 57 & 50 & 0.070 & 4.35e-02 & 4.05e-02 & 61478.6 & 0.14 & 0.28 & 0.11 & 0.21 + 59 & 50 & 0.065 & 6.46e-02 & 6.30e-02 & 21099.1 & 0.14 & 0.26 & 0.12 & 0.24 + 63 & 400 & 0.25 & 1.28e-02 & 7.17e-05 & 308823.5 & 0.13 & 0.04 & 0.15 & 0.35 + 73 & 400 & 0.4 & 2.05e-01 & 2.06e-01 & 259714.3 & 0.2 & 0.2 & 0.15 & 0.23 + 74 & 400 & 0.5 & 6.8e-02 & 6.8e-02 & 627298.3 & 0.45 & 0.068 & 0.29 & 0.18 + 77 & 400 & 0.1 & 8.78e-02 & 8.65e-02 & 205888.0 & 0.15 & 0.23 & 0.14 & 0.22 + 78 & 400 & 0.085 & 1.2e-02 & 4.1e-03 & 124480.4 & 0.17 & 0.03 & 0.23 & 0.23 + lccccc 8 & 1 & 0.03 & 4.6e-02 & 0.18 & 0.20 + 9 & 1 & 0.01 & 1.0e-01 & 0.22 & 0.18 + 10 & 1 & 0.02 & 7.8e-02 & 0.17 & 0.22 + 11 & 1 & 0.035 & 5.37e-02 & 0.18 & 0.21 + 22 & 1 & 0.015 & 1.14e-01 & 0.16 & 0.31 + 32 & 5 & 0.015 & 2.15e-02 & 0.15 & 0.26 + 43 & 5 & 0.02 & 2.13e-02 & 0.19 & 0.24 + 67 & 400 & 0.008 & 9.77e-03 & 0.13 & 0.27 + 70 & 400 & 0.005 & 1.20e-02 & 0.18 & 0.23 + 81 & 400 & 0.01 & 1.82e-02 & 0.18 & 0.27 +

gj 1132 b is a nearby earth - sized exoplanet transiting an m dwarf , and is amongst the most highly characterizable small exoplanets currently known . in this paper we study the interaction of a magma ocean with a water - rich atmosphere on gj 1132b and determine that it must have begun with more than 5 wt% initial water in order to still retain a water - based atmosphere . we also determine the amount of that can build up in the atmosphere as a result of hydrogen dissociation and loss . we find that the magma ocean absorbs at most 10% of the produced , whereas more than 90% is lost to space through hydrodynamic drag . the most common outcome for gj 1132 b from our simulations is a tenuous atmosphere dominated by , although for very large initial water abundances atmospheres with several thousands of bars of are possible . a substantial steam envelope would indicate either the existence of an earlier envelope or low xuv flux over the system s lifetime . a steam atmosphere would also imply the continued existence of a magma ocean on gj 1132 b. further modeling is needed to study the evolution of or -rich atmospheres on gj 1132 b.

1607.03906

with the success of the kepler and k-2 missions and ground - based follow - up efforts of the brightest targets , significant strides have been made in understanding the size and density distribution of planets around other stars ( e.g. * ? ? ? * ; * ? ? ? planets with radii less than 1.5 to 1.6 earth radii and masses less than about 7 earth masses are universally consistent with a rocky , earth - like composition @xcite . however , most of these likely rocky planets have been found at very close orbital periods and are therefore significantly hotter than the earth . some of these planets receive orders of magnitude more stellar insolation than the earth , and their atmospheres will be sculpted and altered by interactions with the stellar insolation , particularly the high energy extreme ultra - violet ( xuv , 1 - 120 nm ) radiation . therefore models of atmospheric loss and evolution for close - in planets are timely . there has been substantial work done on atmospheric loss from planets in the solar system , particularly venus ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? several recent studies extend this type of modeling to atmospheric loss on habitable zone exoplanets with -rich atmospheres @xcite . @xcite have modeled water loss from the recently discovered trappist-1 system of planets around an ultracool dwarf star . others have also studied whether or not close - in rocky exoplanets could be the residual core remnants of gas giant planets stripped of massive atmospheres ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? many of the solar system studies have noted that preferential loss of h from steam atmospheres may lead to build up of in a planet s atmosphere ( e.g. , * ? ? ? * and references therein ) . this is particularly a problem for venus , where minimal is observed , despite an assumed massive early loss of atmospheric water . @xcite applied this type of model to rocky exoplanets in the habitable zones of m and k dwarf stars , where may be a biosignature mimic . in the present paper , we also study atmospheric loss and oxygen build up , but we extend previous models by including an interior model that allows for uptake of by the planet s mantle . our interior model includes both a magma ocean stage , as well as parameterized solid state convection with passive outgassing following solidification . this model is based on magma ocean thermal evolution models long used to study the solar system terrestrial objects ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in comparison , few exoplanet models consider the solid body except as a lower boundary condition for the atmosphere . the present model is an improvement on these treatments and is the first fully coupled model of atmosphere - interior exchange of oxygen . we focus on gj 1132b , a planet only slightly larger than the earth ( m@xmath1 = 1.62 m@xmath2 , r@xmath3 = 1.16 r@xmath2 ) , which was recently discovered by the mearth ground - based transiting planet survey @xcite . gj 1132 is a nearby m3.5 dwarf ( 0.181 m@xmath4 ) located only 12 parsecs away . the planet gj 1132 b has an orbit of 1.6 days and at 0.0153 au , receives @xmath0 19 times more stellar insolation than the earth and 10 times more than venus . with a large relative transit depth , gj 1132 b will be amenable to near - term follow - up both from large ground based telescopes , as well as orbiting observatories like hst and jwst . it is our goal to determine if the planet could have sustained a water or rich atmosphere over its lifetime . we focus on o and h in order to be able to thoroughly explore the parameter space in a timely manner . future models may wish to include a more detailed chemistry incorporating carbon and nitrogen - bearing species . the magma ocean stage on close - in rocky exoplanets may be extremely long - lived . observations of these objects may present a means to test magma ocean models which are also used to study processes occuring during solar system accretion . as such , observations of gj 1132b and other planets like it may help us improve models for our own solar system , in particular , models for water and loss on venus . this paper is organized as follows . section [ sec : escape ] discusses our atmospheric escape model and line - by - line climate model . section [ sec : model ] describes the planetary interior model and the coupling to the atmospheric model . section [ sec : results ] presents results from the coupled model , including the amount of water lost from the planet , the final abundance in the atmosphere , and the mantle compositon . in section [ sec : discussion ] we discuss some of the limitations of the model . finally , in section [ sec : predictions ] , we give predictions for the atmospheric composition of gj 1132 b. as in the solar system , atmospheric erosion from young planets around m dwarfs will be driven by a combination of xuv - driven hydrodynamic escape , erosion by coronal mass ejection events ( cmes ) , blowoff by giant impacts , and a host of more complex processes involving non - thermal effects , ion - pickup and magnetic fields @xcite . the early xuv emission from most m dwarfs is high for an extended period , making xuv - driven hydrodynamic escape one of the most critical effects to model . as it is also more straightforward to calculate escape rates in this case than for many other processes , we focus on it here . for a planet undergoing xuv - driven hydrodynamic escape , the atmospheric escape flux ( kg / m@xmath5/s ) is approximately given by @xcite @xmath6 where @xmath7 is the stellar flux in the xuv wavelength range ( 1 - 120 nm ) and @xmath8 is the gravitational potential at the base of the escaping region . here we take @xmath9 with @xmath10 the gravitational constant , and @xmath11 and @xmath12 the estimated planetary radius and mass of gj 1132 b , respectively ( see table [ tab : params ] ) . @xmath13 is an empirical factor that accounts for radiative losses and 3d effects and typically varies between 0.15 and 0.3 @xcite . equation [ eq : elim2 ] neglects the expansion of the heated upper atmosphere away from the planet s surface , which typically results in a correction factor of up to a few tens of percent . equation [ eq : elim ] also assumes that re - emission of absorbed xuv radiation at infrared wavelengths is not effective . this is a reasonable assumption for a hydrogen - dominated upper atmosphere , but not if the upper atmosphere is dominated by a gas with strong vibrational and rotational modes such as . we also neglect tidal enhancement of the escape flux @xcite , which is likely a much smaller effect than the uncertainty in the xuv flux . the total mass of atmosphere lost as a fraction of the final planet mass is @xmath14 here we are assuming @xmath15 , so that @xmath12 and @xmath11 can be treated as approximately independent of time in eqn ( [ eq : mlost ] ) . the present - day xuv flux from gj 1132 has not yet been measured . however , the star gj 1214 ( 0.15 m@xmath16 ) is similar in mass to gj 1132 ( 0.181 m@xmath16 ) and has a similar activity level @xcite . as such , we use the semi - empirical high - energy spectrum of gj 1214 constructed by @xcite as a proxy for that of gj 1132 ( see fig . [ fig : stellar_flux ] ) . the nuv - to - fuv portion of this spectrum was directly observed with hubble cos and stis @xcite , the euv was estimated from a model - dependent scaling from the lyman @xmath17 emission line @xcite , and the x - ray from a plasma model matched to an earlier xmm detection of a flare from gj 1214 @xcite . in this spectrum , the xuv flux ( 1 - 120 nm ) represents about 3 @xmath18 of the bolometric flux , with an additional 3 @xmath18 of the bolometric flux contributed by the lyman @xmath17 line alone ( 120 - 130 nm ) . based on scaling from gj 1214 , we estimate that gj 1132 b currently receives about 0.8 w / m@xmath19 in the xuv . the xuv flux could be at least @xmath20 above or below this value , due both to uncertainties in reconstructing gj 1214 s intrinsic spectrum ( see * ? ? ? * ) and to the unknown extent to which gj 1132 s high energy behavior tracks that of gj 1214 . the time evolution of xuv from m dwarfs similar in mass to gj 1132 is poorly constrained . for main - sequence stars including m dwarfs , observations indicate that time - averaged xuv from the stellar corona for young , active stars saturates at @xmath21 @xcite . m dwarfs may stay in this active phase for roughly a gigayear @xcite and then fade to lower xuv flux ratios , although the lower limit for quiescent xuv from old , inactive mid - m dwarfs is just starting to be probed @xcite . here , we take two approaches to bracket the range of uncertainty for xuv - driven atmospheric loss . the xuv flux models are shown in figure [ fig : stellar_flux ] . xuv flux model a assumes that xuv emissions are @xmath22 times the evolving stellar luminosity @xcite and declining as a power law after 1 gyr with @xmath23 . xuv model b assumes that throughout its youth , gj1132 s xuv flux is @xmath22 times the present - day stellar luminosity and zero after 1 gyr with @xmath24 . this brackets the likely present - day xuv flux at 5 gyr . from eqn . [ eq : mlost ] , we find @xmath25 for model a. alternatively , model b yields @xmath26 . hence a very large amount of hydrogen ( 2% to 14% of the total mass ) could have been lost from gj1132b since its formation . having demonstrated that even a substantial primordial hydrogen envelope could have been lost from gj 1132 b , we now assess the possibility that the planet has retained an atmosphere of heavier gases . the first thing we need to calculate is the rate at which hydrogen escape would drag away heavier species . the flux received by gj 1132 b places it well within the kombayashi - ingersoll limit for the runaway greenhouse ( @xcite ; fig . [ fig : lbl_goldblatt_intercomp ] ) a planetary albedo of 0.955 is required for stable surface water on gj1132b , which is implausibly high for a planet with an atmosphere . enceladus has an albedo of 0.99 @xcite , but it is airless with a surface composition dominated by fresh water ice . ] @xcite . if it formed with some water it would hence initially have had an -rich upper atmosphere . given an intense early xuv - driven escape regime , the oxygen in this , along with other heavy elements such as or , would have been dragged along with the escaping hydrogen . the loss rate of a heavier species in the neutral hydrodynamic escape regime depends on how effectively the hydrogen drags that species with it . specifically , the number flux @xmath27 of a heavy species 2 [ in molecules / m@xmath5/s ] is given by @xcite @xmath28 where @xmath29 and @xmath30 and @xmath31 and @xmath32 are , respectively , the molar concentrations [ mol / mol ] and molecular masses [ amu ] of species 1 and 2 . the crossover mass @xmath33 is defined as @xmath34 here @xmath35 is boltzmann s constant , @xmath36 is the temperature of the escaping gas , @xmath37 is gravitational acceleration at the escape radius , @xmath38 is the proton mass and @xmath39 is the binary diffusion coefficient for species 1 and 2 . for o atoms dragged by h , @xmath40 m@xmath41 s@xmath41 @xcite . we also define a reference flux @xmath42 for species 1 in the absence of heavy species @xcite . note that when @xmath43 , @xmath44 in general . otherwise , @xmath45 , so that the loss rate depends on the relative abundances of species 1 and 2 . in our coupled model , we set @xmath46 equal to the diffusion - limited loss rate and @xmath47 equal to zero once the abundance of exceeds that of . following @xcite , we use the composition - dependent loss rates for h and o , rather than the stoichiometric loss rates of @xcite . we discuss the possibility of oxygen - dominated escape from gj 1132 b in a later section . equations ( [ eq : crossmass ] ) and ( [ eq : ref_flux ] ) can be used to define the critical xuv flux required for drag to occur . setting @xmath48 , we can write @xmath49 x_1\ ] ] with @xmath50 the potential energy of one molecule of species 1 . for gj1132b , given o drag by h with @xmath23 and @xmath51 k , @xmath52 w / m@xmath5 . as can be seen from figure [ fig : stellar_flux ] , this is smaller than the estimated xuv flux received by gj1132b for the first 10 gy of its lifetime in model a , implying that oxygen will continually be dragged along with escaping hydrogen if an -rich atmosphere is present . however , the planet will still oxidize overall as the escape rate of o is less rapid . whether this oxidation could lead to a detectable atmospheric oxygen signal depends on atmosphere - interior exchange rates , which we address in the section [ sec : model ] . the net buildup rate of o on the planet in the hydrodynamic drag escape regime can be approximated as @xmath53 where the @xmath54 and @xmath55 terms are the molecular number density and individual scale heights of species 1 and 2 , respectively . for o diffusing through h following photolysis @xcite @xmath56 equation ( [ eq : odiff ] ) can be simply physically interpreted as the diffusion rate of o atoms out of the escaping region back to the lower atmosphere . the rate at which a planet exchanges volatiles between the atmosphere and interior is a strong function of temperature . in particular , once the surface is hot enough to be in a magma ocean state , the atmosphere and interior will equilibrate on geologically short timescales . for this reason , climate calculations are necessary to assess the increase in surface temperature due to the atmosphere s greenhouse effect . to calculate surface temperature , we first calculate the outgoing longwave radiation ( olr ) from a pure atmosphere using a line - by - line radiative transfer calculation . we integrate the monochromatic equation for upwelling radiative flux per unit wavenumber ( w / m@xmath5/cm@xmath41 ) @xmath58 where @xmath59 is surface temperature , @xmath60 is the mean path optical depth at a given wavenumber @xmath61 and pressure @xmath62 , @xmath63 is the total optical depth , and @xmath64 is the planck spectral irradiance . mean path optical depth is defined as @xmath65 where @xmath66 is surface pressure , @xmath37 is surface gravity and @xmath67 is the mass absorption coefficient ( m@xmath5/kg ) . in addition , @xmath68 is the mean emission angle cosine , which we take to be a constant 0.5 here . the layer optical depth weighting approach of @xcite is used to ensure accurate model behaviour in high absorption regions of the spectrum . line absorption coefficients for are calculated from the 2010 hitemp line list @xcite , with the voigt function used to describe lineshapes and temperature scaling for the line strengths following standard methods @xcite . the calculation is performed over 30 layers up to a minimum atmospheric pressure of 1 pa . spectral calculations were performed from 1 cm@xmath41 to 5 times the wien peak wavenumber of the planck function at the given surface temperature . we used 5000 points in wavenumber ; sensitivity tests indicated that further increases in spectral resolution had an insignificant effect on the integrated olr . the temperature profile was assumed to be a dry adiabat from the surface to the tropopause , after which a stratospheric temperature equal to the skin temperature for gj1132b given a planetary albedo of 0.75 was assumed ( 344.2 k ) . ideal gas behaviour was assumed when calculating the dry adiabat , which is a reasonable approximation for the range of temperatures and pressures studied @xcite . we accounted for the variation in the specific heat capacity of water vapour as a function of temperature using data from @xcite . continuum opacity due to far - wing absorption of strong lines and other effects was taken into account using the mt - ckd parametrization @xcite . outside of the mt - ckd temperature range of validity , continuum absorption was simply set to its value at the maximum temperature given . spectral lines were truncated at 25 cm@xmath41 to avoid double - counting of the continuum absorption . to render the line - by - line calculation more manageable , we also preprocessed the hitemp-2010 dataset by removing weak lines , which we defined as lines with a reference strength below @xmath69 cm@xmath41 / cm@xmath5 molecule@xmath41 at 1000 k. this approximation means that we slightly underestimate the atmospheric opacity at the highest temperatures and pressures studied . as the planet s surface is already in a magma ocean state under these conditions , however , this has little effect on atmospheric evolution . to validate the code , we first ensured that it reproduced semi - analytic textbook results ( figure 4.5 in @xcite ) . next , we compared the code output with runaway greenhouse calculations for earth @xcite . figure [ fig : lbl_goldblatt_intercomp ] shows the results of this intercomparison . as can be seen , our model agrees closely with published results except in a small region around 1200 cm@xmath41 , most likely due to slightly differing assumptions for the continuum ( c. goldblatt , personal communication ) . given the large uncertainties in other parameters for gj1132b , we decided this agreement was more than sufficient for our purposes . we calculated the olr over a range of surface temperatures from 400 to 4000 k and a range of surface pressures from 1 pa to 1000 bar . at high surface temperatures , the uncertainty in water vapour opacity becomes large due to uncertainty in the scaling of the continuum . however , at these temperatures the planet s surface is in a magma ocean state that permits rapid exchange of oxygen between the atmosphere and interior . hence this uncertainty should have little effect on our key conclusions . we calculate the atmospheric heat flux as a balance of the outgoing longwave radiation ( olr ) and the absorbed shortwave radiation ( asr ) . the asr is given by @xmath71 , where @xmath72 is the planetary albedo and @xmath73 is the bolometric stellar flux received by the planet , which we derived by interpolating data from @xcite to the mass of gj 1132 . the planetary albedo of gj 1132 b is currently unconstrained , although observations indicate low albedos for planets orbiting m dwarfs in general . @xcite did a statistical study of kepler s close - in super - earths and found a median total albedo of @xmath0 0.3 , although values ranged up to 0.92 . note that the sample of planets studied all had equilibrium temperatures significantly larger than gj 1132 b , and they may be more representative of bare rocky planets than those with dense atmosphere . here we take the albedo of venus , as a representative planet with a thick , hot atmosphere , as our nominal constant value @xmath74 , but discuss the effect of lower albedos in the results section . we address the atmosphere - interior rates of exchange by coupling our atmospheric model with a magma ocean model , which includes thermal evolution and the exchange of and o with the atmosphere . the thermal parameterization combines elements of the work of @xcite , @xcite , and @xcite . as in these papers , we assume that solidification of the magma ocean proceeds from the bottom up , due to the fact that mantle adiabats are steeper than the solidus and liquidus curves of silicates . the thermal evolution is governed by two temperatures : the mantle potential temperature , which dictates the degree of melting and convection within the mantle , and the surface temperature , which is governed by heat flux out of the mantle and heat loss from the top of the atmosphere . for most of the duration of the magma ocean phase , these temperatures are the same . however , as the solidification front ( the depth at which the mantle adiabat intersects the mantle solidus ) moves towards the surface , a thermal boundary layer can develop at the surface , which insulates the mantle from additional heat loss . following formation of the thermal boundary layer , the model switches to whole mantle solid - state convection as parameterized in @xcite . the atmosphere is assumed to be composed of , h and o gases . is the only source of atmospheric opacity and the climate is calculated as discussed above . the composition and thickness of the atmosphere depends on mass exchange with the magma ocean and loss of volatiles due to both atmospheric escape and crystallization into the solid mantle . following magma ocean solidification , only passive outgassing of and atmospheric loss occur . we will discuss each of these aspects in more detail below . the thermal evolution of the magma ocean potential temperature is given by : @xmath75 where @xmath76 is the mantle bulk density , @xmath77 is the silicate heat capacity ( @xmath78 ) , @xmath79 is the planetary radius , @xmath80 is the core radius , @xmath81 is the radius of solidification , @xmath82 is the heat of fusion of silicates ( @xmath83 ) , @xmath84 is the mantle heat flux , @xmath85 is the heat generated by radioactive decay . we begin our calculations at @xmath86 = 4000 k , which is hot enough for the magma ocean to extend from the surface to the core - mantle boundary . the heat generated by radioactive decay is limited to the long - lived isotopes , , and . abundances of these elements are assumed to be the same as for the earth s mantle , and the parameterization for @xmath85 is the same as that given by @xcite equation ( 4 ) . although we expect gj 1132 b to have different abundances of the radioactive elements , the results of the magma ocean model are relatively insensitive to them , given the typically short lifetimes of the magma oceans . after solidification , the first term on the rhs disappears and the thermal evolution proceeds as for @xcite . the mantle heat flux is parameterized by the mantle rayleigh number : @xmath87 where @xmath88 is the thermal conductivity ( 4.2 w m@xmath41 k@xmath41 ) , the critical rayleigh number ( @xmath89 ) and the exponent @xmath90 ( 0.33 ) are determined from numerical mantle convection simulations , @xmath17 is the thermal expansion coefficient ( @xmath91 k@xmath41 ) , @xmath92 is the thermal conductivity ( @xmath93 m@xmath19 s@xmath41 ) , and @xmath94 is the kinematic viscosity ( m s@xmath95 ) . the dynamic viscosity @xmath96 for a silicate liquid is very small , of order 0.01 pa s. we therefore assume that the liquid portion of the magma ocean is instantaneously well - mixed . we only consider convection within the magma ocean , not the solid mantle , until solidification of the magma ocean has occurred . as partial crystallization proceeds , the viscosity of the magma ocean increases dramatically . the viscosity depends on the melt fraction , which we do not use here to avoid confusion with the energy - limited escape flux , see eqn ( [ eq : elim2 ] ) ] @xmath97 , which is given by @xmath98 . we use the same viscosity parameterizations as @xcite . below a critical melt fraction ( @xmath99 0.4 ) , the viscosity becomes solid - like , where our solid viscosity is given by @xmath100 , where @xmath101 , @xmath102 kj mole@xmath41 and @xmath103 is the ideal gas constant . the radius of solidification is given by the intersection of the mantle adiabat with the mantle solidus . we derive an equation for @xmath104 by approximating the adiabat as the first taylor expansion , and the solidus as a straight line in two sections , from 0 - 100 km , and from 100 km to the core - mantle boundary . the coefficients for the high pressure region are taken from @xcite ( @xmath105 k gpa@xmath41 , @xmath106 = 1825 k ) , and a linear fit is done to the low pressure dry peridotite solidus from that paper ( @xmath107 k gpa@xmath41 , @xmath106 = 1420k ) . the liquidus is assumed to be larger than the solidus by 600 k. the linear parameterization for the solidus leads to a simple and straightforward analytic expression for the radius of solidification , which yields our second differential equation : @xmath108 = a g \rho_m ( r_p - r_s ) + b\\ \frac{dr_s}{dt } = \frac{c_p(b \alpha - a \rho_m c_p)}{g(a \rho_m c_p - \alpha t_p)^2 } \frac{dt_p}{dt}\end{aligned}\ ] ] the surface temperature of the planet is calculated from the heat loss equation for the surface environment , where we make the simplification that the atmosphere and thermal boundary layer are governed by a single average temperature ( @xmath59 ) : @xmath109 where @xmath110 is the heat flux from the atmosphere , calculated from olr - asr ( see sec . [ sec : linebyline ] ) , and @xmath111 is the thickness of the thermal boundary layer , which is given by : @xmath112 . the boundary layer develops once the melt fraction at the surface of the magma ocean reaches the critical value , causing the viscosity of the magma ocean to increase dramatically . this is the `` mush '' stage of @xcite . a sample run of the thermal model is shown in the top panel of figure [ fig : modelrun ] . the potential temperature and surface temperature are nearly identical until the `` mush '' stage is reached and the boundary layer begins to grow . when the surface temperature reaches the solidus temperature ( 1420 k ) , the magma ocean phase has concluded . when applied to an earth - like planet , our thermal model reproduces the cooling times and heat fluxes found in @xcite and @xcite very well . we deviate at later stages due to the fact that we do not include condensible atmospheric water vapor , which will not be present on gj 1132b . however , the comparison gives us confidence that our thermal model produces reasonable results . water is very soluble in silicate melts , so the pressure at the surface of the planet during the magma ocean stage is set by its solubility in the magma ocean . we use a fit to the solubility data of @xcite : @xmath113 where @xmath114 is the mass fraction of water in the liquid silicate melt . note that the solubility of water within silicates at low pressures is effectively temperature independent . mass balance for water within the magma ocean system is given by : @xmath115 where @xmath116 is the partition coefficient for water between melt and solid ( 0.01 ) , @xmath117 is the mass of water in the magma ocean + atmosphere system on the current time step , and the mass of crystals ( @xmath118 ) within the magma ocean is found from the melt fraction @xmath97 calculated along the adiabatic profile in the magma ocean . the total mass of the magma ocean ( @xmath119)is determined by difference with the radius of solidification . the mass of water in the magma ocean + atmosphere system ( @xmath117 ) and the mass of water in the solid mantle at a given time are determined with the differential equations : @xmath120 where @xmath121 is the xuv - driven atmospheric mass loss rate of h ( in kg m@xmath95 s@xmath41 ) , discussed in section [ sec : primordial ] . we find that our calculations conserve mass when atmospheric loss is turned off . a sample model run is shown in the bottom panel of fig . [ fig : modelrun ] . the figure shows the distribution of water between the three planetary reservoirs : magma ocean , atmosphere and solid mantle . when the magma ocean has cooled , most of the water remains in the solid mantle . atmospheric escape is included in the calculation shown . we assume that is produced in the atmosphere by loss of h from . is then also lost from the atmosphere at a slightly slower rate due to hydrodynamic drag , as discussed in section [ sec : drag ] . the o@xmath122 produced in the atmosphere is also in contact with the feo in the silicate melt . we allow the magma ocean to take up by oxidation of feo to . the equilibrium oxygen fugacity for the magma ocean is given by @xcite : @xmath123 \\ - 7.01 \times 10^{-7}\frac{p(pa)}{t } - 1.54 \times 10^{-10 } \frac{(t-1673)p(pa)}{t } \\ + 3.85 \times 10^{-17 } \frac{p^2(pa)}{t}\end{gathered}\ ] ] where @xmath124 is the molar concentration of the oxides in the silicate melt , and we use the bulk silicate earth ( bse ) as our nominal composition @xcite . this empirical relationship was derived for a wide range of natural silicate melt compositions equilibrated at oxygen fugacities ranging from metal - silicate equilibrium ( iron - wustite buffer ) up to air at temperatures between 1473 and 1900 k. rocky exoplanets are expected to have relatively similar rocky elemental abundances to the earth based on analysis of the observed mass - radius measurements and stellar elemental abundances ( * ? ? ? * and references therein ) . gj 1132 b itself falls very close to an earth - composition track on the mass - radius diagram @xcite . the empirical oxygen fugacity relationship therefore is likely to cover the relevant compositional range for rocky exoplanets . the strongest mantle influence on the oxygen fugacity will likely be the total abundance of feo ( feo * ) in the silicate , so we explore a range of feo abundances in our calculations . note that we assume a metal - free magma ocean for these calculations . mass balance between the atmosphere and the magma ocean is calculated for in the same way as for . oxygen is sequestered into the solid mantle as . we assume no fractionation between liquid and solid , either for feo or , although fe is known to fractionate from mg in the melt ( i.e. , minerals that condense early should be less fe - rich than those that condense later ) . we consider that this will have only a small effect on our oxygen mass balance , but we discuss implications in a later section . the magma ocean + atmosphere system loses oxygen to the solid mantle and to atmospheric escape , while atmospheric escape of h from produces o. this gives us two more differential equations for the abundance of free o in the solid and in the magma ocean + atmosphere system : @xmath125 where @xmath126 is the xuv - driven atmospheric mass loss rate of o ( in kg m@xmath95 s@xmath41 ) and @xmath127 is the mass fraction of in the mantle . following magma ocean solidification , direct exchange of oxygen between the mantle and the atmosphere halts ( @xmath128 ) . oxygen no longer exchanges with the mantle following solidification , but is continuously created by h loss and lost by hydrodynamic drag . although similar models for venus have allowed continued o loss due to oxidation of the crustal layer @xcite , we consider this effect to be small given that the upper mantle will already be significantly oxidized by exchange with the atmosphere during the magma ocean phase . water outgassing continues but there is no return of water to the mantle after magma ocean solidification . outgassing is parameterized similar to @xcite : @xmath129 where @xmath130 is the volume - averaged mass fraction of water in the melt , @xmath131 is the volume - averaged melt fraction of the mantle , @xmath132 is the mantle convection velocity , and @xmath133 is the degassing efficiency , which can vary from 0 ( no degassing ) to 1 ( completely efficient degassing ) . we will explore the effect of the degassing efficiency in our discussion of the results . the mass and radius for gj 1132b are taken from the discovery paper @xcite . the core mass and radius assuming a two - component model ( silicate + metal , no water ) are determined with the online tool of @xcite . values for the planet properties are given in table 1 . the mass of the planet is currently only known to 3@xmath134 , although continued doppler monitoring will shrink the mass uncertainty and enable more detailed compositional models . we note that there is additionally a well - known degeneracy in determining the planet s composition from the density . using the online tool of @xcite , we find that the nominal mass of the planet allows for up to about 20 wt% of the planet to be water . note that this extreme value results in a nearly zero silicate mass fraction , which is highly unphysical , as giant impact simulations show that mantle stripping can produce planets with at most 70% core mass fraction @xcite . however , we will test loss models here for total planetary water abundances up to 20 wt% as a limiting case , while holding the core and silicate mass fractions fixed at the value determined assuming the present measured mass and a two - component silicate - metal model . .parameters for gj1132b used in the modeling . [ cols="<,<",options="header " , ] [ tab : params ] we explore model results for the two xuv flux models . for both models , we vary the initial planetary water inventory and the mantle feo abundance . we explore water inventories ranging from 0.1 up to 1000 earth oceans ( eo = @xmath135 ) of water , which is about 20 ppm in the mantle up to about 20 wt% . we note that while there are measurements of the solubility of water in silicate melts up to this value , the data beyond 10 wt% is sparse and fairly poorly constrained . for mantle feo , we consider abundances ranging from 0.1 to 20 wt% . abundances of feo in the silicate mantles of the terrestrial planets in the solar system span this range . estimates for mercury s mantle are 2 - 3 wt% , earth and venus have about 7 - 8 wt% of feo , whereas mars has a mantle feo abundance of about 18 wt% , and vesta 20 wt% @xcite . the abundance of feo in the mantle is a result of the composition of the protoplanets out of which a planet is made , the conditions under which core formation occurs and any subsequent reducing or oxidizing processes . the abundance of ( or ) in the earth s mantle is fairly small ( /fe@xmath136 , @xcite ) . we examine values from 0 up to 0.03 , and find only a minor difference on the final results of the model . our nominal results use /fe@xmath137 = 0 , which gives the mantle maximum oxygen uptake potential . we also consider the effect of efficient ( @xmath138 ) versus inefficient ( @xmath139 ) degassing after the magma ocean stage has solidified . this parameter has an effect on the final water and abundances , as we discuss below . magma ocean solidification times depend strongly on the initial water abundance of the planet , as well as the xuv flux . figure [ fig : solidification ] shows the solidification times for the two xuv models as a function of initial water abundance . models were run for a total integration time of 5 gyr , consistent with the estimated age of gj 1132 @xcite . the xuv flux model b results in magma oceans that persist roughly an order of magnitude longer than for xuv flux model a. the longer duration is due to the slower loss of water vapor from the atmosphere , which causes the planet to remain hotter for longer . total planetary water loss also depends strongly on both the initial water abundance and the xuv flux , as well as the degassing efficiency in the post - magma ocean state . figure [ fig : h2olost ] shows the fraction of intial water lost for both flux models , as well as different degassing efficiencies . for xuv flux model a , water loss is more than 95% complete for all intial water abundances , except the largest . for xuv flux model b , the figure shows that the amount of water lost for low initial water abundances depends strongly on the degassing efficiency . for efficient degassing , all water is lost except for initial water abundances @xmath140 10 wt% of the planet . for no degassing post - magma ocean , the fraction of the initial water lost increases with increasing water abundance , but only up to 10wt% , above which water loss decreases . this indicates that for low water abundances , most of the remaining water is stored in the mantle and is lost after the magma ocean phase during passive outgassing of the interior . if degassing is inefficient , the water can be permanently trapped in the mantle . a large melting event , possibly caused by late impacts , could induce further outgassing . the amount of remaining water is not sufficient to affect the planet s density except at the very highest water abundances where most of it remains in the atmosphere . for xuv model b and no degassing , the remaining water abundance is @xmath010 wt% of the planet s mass . this is technically consistent with the present mass and radius measurement , but requires un - earth - like silicate(0.36 ) and core(0.54 ) mass fractions @xcite . the majority of the water in this scenario would be locked in solid phases in the mantle . ( i.e. no outgassing after magma ocean solidification ) , whereas thin lines with open points ( a : square , b : circle ) are for @xmath138 ( i.e. perfectly efficient outgassing after magma ocean soldification ) . the difference between the thin and thick lines for xuv model b indicates that most of the planet s water is lost after magma ocean soldification . , title="fig : " ] for models with intial @xmath141 , all free oxygen is produced by destruction of water and loss of hydrogen . the amount of oxygen produced is therefore directly proportional to the amount of water lost , shown in fig . [ fig : h2olost ] . xuv model a therefore typically produces more total oxygen than model b , especially at high initial water abundances and at lower abundances when degassing is inefficient . a fraction of the oxygen that is produced is lost both to atmospheric escape and to the mantle . for both xuv models , between 90 to 99 % of the total oxygen produced by photolysis of water vapor is lost to space , with higher o loss amounts occurring at low water abundances . at most 10% of the total oxygen produced is sequestered into the mantle ( discussed below ) . the remainder of the oxygen resides in the atmosphere , as shown in figure [ fig : finalo2_1 ] . both xuv models can result in residual remaining in the atmosphere at a few bar level , with minimal dependence on feo content . several hundred to several thousand bars can build up only for initial water abundances greater than about 5 wt% of initial water . for xuv model a , the final atmospheric abundance is negligible for water abundances below @xmath05 - 10 wt% initial water , whereas for xuv model b , the abundance in the atmosphere is slightly dependent on the feo content of the mantle , with more atmospheric build up for smaller feo abundances . this is because there is a smaller sink for in the magma ocean with lower feo abundances . degassing efficiency affects the final abundance in the atmosphere for xuv model b with abundances less than 10 wt% as shown in figure [ fig : finalo2_2 ] . at higher water abundances , persistent magma oceans mean that the model never enters the passive degassing state . for water abundances less than 10 wt% , the atmosphere has about 10 times more oxygen than for inefficient degassing . this is due to creation of additional oxygen by dissociation of water degassed in the post - magma ocean time frame . we find that for both xuv models , is more abundant than water vapor in the atmosphere for nearly all of our parameter space , but the atmosphere is likely to be fairly tenuous ( @xmath142 ) . steam dominates the atmosphere only for xuv model b at the highest water abundances , with about a factor of 10 more water vapor than . therefore our models indicate that the atmosphere of gj 1132b may be tenuous and dominated by . if abundant atmospheric water is observed , it is indicative of both a low xuv flux history and high initial abundance . for mantles with initial /fe@xmath137 of 0.02 - 0.03 , we find that atmospheric is relatively unaffected . at low water abundances , is the same as for our nominal calculations . at large water abundances ( @xmath143 wt% for xuv b ) , atmospheric is the same for low initial feo , but is slightly larger than for our nominal model as feo increases . we find a maximum increase at 20% initial feo of about 50% in the atmospheric pressure . planetary albedo has a slightly larger effect on our results . for a lower planetary albedo of 0.3 , we find for xuv model b that the final atmospheric pressure is 60 - 90% of our nominal results . for xuv model a , results are the same ( i.e. , p @xmath144 1)at water abundances below 5 wt% , and are about 75 - 90% of the nominal results at higher water abundances . in both cases we find that the fraction of the nominal abundance increases with increasing water abundance . that is , albedo has a larger effect on models with lower initial water abundances . however , the effect of albedo is small enough that it does not alter our primary conclusions . while more extensive destruction of for xuv model a implies greater production of oxygen , most of the oxygen is directly lost to space and therefore a smaller fraction can be absorbed into the mantle than for xuv model b. for xuv model a , this is at most 8% of the total oxygen produced , whereas slightly more ( 10% ) can be absorbed for model b. for xuv model a extensive absorption of o by the mantle only occurs at high water ( @xmath145 5 wt%)and feo abundances ( @xmath1455 wt% ) . in contrast , extensive oxygen absorption occurs across wide ranges of water ( @xmath1450.05 wt% ) and feo ( @xmath145 0.5 wt% ) for xuv model b , although the highest relative absorption still occurs at the largest feo and water abundances . while high feo abundances lead to more oxygen absorption , the conversion ( or oxidation ) of feo to is more extensive at low feo abundances as shown in figure [ fig : fe2o3 ] . note that we include here the remaining in the magma ocean for those models which do not fully solidify . for xuv model a , the peak oxidation occurs above 15 wt% , at feo abundances less than 8 wt% . for xuv model b , the peak is at 5 wt% for feo abundances less than about 5 wt% . less oxidation occurs for xuv model b at higher abundances because these models lose less h from the atmosphere and therefore produce less free o. figure [ fig : fe2o3profile ] shows the profile of abundance with depth in the solidifying magma ocean at the end of the integration period of 5 gyr . the remaining liquid at larger radii has the same abundance as the last layer of solidified mantle . the outer radius of the magma ocean is smaller than the planetary radius because of the formation of a thermal boundary layer , which insulates the upper mantle . the maximum abundance of is limited by the total feo content , which is fixed at 1 wt% in this figure . as can be seen , the mantle becomes progressively more oxidized as the magma ocean solidifies , and the degree of oxidation is strongly dependent on the total water abundance . less stratification occurs for non - zero /fe@xmath137 starting abundances . the progressive oxidation of the mantle may effect later mantle convection . the density of silicates enriched in will be slightly lower than more reduced silicates , which results in a stably stratified mantle . this may delay the onset of solid state convection after magma ocean solidification . we will discuss this possibility in the next section . to the base of the thermal boundary layer . this is why none of the curves extend to a full planetary radius . for non - zero initial abundances , less stratification in mantle composition is observed , with most of the lower mantle solidifying with the initial abundance . ] in our nominal models , we assume that atmospheric loss is energy - limited , where the loss rates are dependent on the o and h molar concentrations . we assume that energy - limited escape driven by hydrodynamic loss of h occurs until the and total atmospheric pressures are equal . after this cross - over point , we assume that h must diffuse through the o background gas , at which point the hydrodynamic loss halts and o no longer escapes . however , the transition composition is uncertain because h should diffuse more readily into the upper atmosphere than o. we explore the sensitivity of our final results to this transition point in figure [ fig : crossover ] . here we show results for both xuv flux models with @xmath146 for a constant feo abundance of 8 wt% as a function of initial water abundance for different transition points ( @xmath147 = 0.4(nominal ) , 0.1 , and 0.001 ) . for xuv model a , the final pressure is insensitive to the transition point up to @xmath01 wt% of . at higher water abundances , the final pressure is reduced by several orders of magnitude as the transition point drops , except at the very highest water abundance where the magma ocean persists . for xuv model b , the transition point has a strong effect on the abundance for initial water abundances less than @xmath010 wt% . reducing the transition abundance results in more tenuous atmospheres , since more of the o can escape . ) . blue lines are for xuv model a , pink lines are xuv model b. ] it is possible that gj 1132b began with an envelope dominated by gas , rather than . as discussed in section [ sec : primordial ] , a significant mass of can be lost from the planet , up to 15% of the planet s mass over 10 gyr . interaction of an atmosphere with mantle feo might result in reduction of mantle feo to fe metal through a reaction such as : the forward reaction is thermodynamically unfavorable and has been shown to go nearly to completion in the reverse direction ( oxidation of metal ) at all temperatures and pressures on the present and early earth @xcite . in fact , these experimental studies of the iron - water reaction at high pressure have shown that hydrogen liberated from water can be sequestered into a metallic phase via a reaction such as : . however we expect that relative to the total duration of the magma oceans , the presence of metal within the magma ocean was relatively short - lived . we therefore consider that the primary effect of an initial -envelope would be to prolong the magma ocean lifetimes and reduce the loss of water and from those calculated here . is a common atmospheric component that is often included in magma ocean models @xcite , due to both its large abundance and its contribution to greenhouse warming . we do not consider it here in order to minimize the complexity of the model , but we will qualitatively discuss its possible effect on the evolution of gj 1132b . the solubility of in silicate melt is much lower than that of , but it is much more soluble in metal alloy . therefore , numerous papers on earth - based magma ocean models have noted that will be concentrated in either the atmosphere or the core . @xcite argues based on alloy / melt partition coefficients that a magma ocean that equilibrates with only 1 wt% of alloy would lose at least 60% of its total carbon to the core . however , we noted above that the presence of metal within the magma ocean was likely relatively short - lived , so unless carbon is removed during core formation , it seems likely that there should be substantial carbon remaining in the magma ocean and atmosphere . solubility of in the mantle depends on the temperature , pressure , melt composition , and oxygen fugacity : as gj 1132 b becomes more oxidized , should become more soluble in the melt . however , solubility relationships indicate that it is unlikely that more than about 20 - 30% of the could be dissolved in the magma ocean @xcite . @xcite also suggests the possibility of diamond precipitation in the mid to lower mantle or a magma ocean carbon pump to lower atmospheric abundances . this would sequester carbon in the mantle where it would be available for later outgassing during a post - magma ocean state , much like water in our efficient degassing scenarios . however , while this is a possibility , it would require detailed additional modeling to evaluate . in the atmosphere will prolong the magma ocean lifetimes by additional greenhouse warming , which may enhance atmospheric loss of both water vapor and . @xcite showed that in highly irradiated super - earth atmospheres dissociation of can lead to both carbon and oxygen loss , with carbon escaping more rapidly due to its lower atomic weight . @xcite showed that water loss from -rich atmospheres can still be substantial , especially for planets that receive more insolation than the present day earth , such as gj 1132 b. while is effective at cooling the upper atmosphere , which can hinder loss in more temperate planets , a back - of - the - envelope calculation suggests that the degree of cooling from the 15 @xmath148 m non - lte emission would still be far lower than the xuv flux received by gj 1132 b , at least for the first gyr . therefore , cooling of the upper atmosphere would likely be insufficient to hinder the escape of and . therefore , tenuous atmospheres are the most likely scenarios for gj 1132 b after loss of an envelope . non - thermal effects provide additional loss avenues as discussed below . considering non - thermal mechanisms for atmospheric escape from gj1132b , it is safe to assume no planetary magnetic field , in analogy to venus and as a conservative choice . also , while gj1132b is closer than venus to its star in terms of bolometric irradiation and stellar wind flux , it is significantly more massive . as a result , charge exchange ( e.g. , @xmath149 ) and ion escape could increase the loss of hydrogen , and dissociative recombination ( e.g. , @xmath150 ) could increase the loss of oxygen . the latter mechanism releases only @xmath151 j per atom , which is not enough to permit escape given the high mass of gj1132b . the hydrogen non - thermal escape might be insignificant , by analogy to venus ( e.g. pierrehumbert 2010 ) , but a dedicated study is warranted . similarly , the stellar wind of m dwarfs like gj1132 is expected to be too tenuous to lead to significant atmospheric erosion , but no firm conclusion is possible without detailed modeling @xcite . in order to have efficient degassing during the post - magma ocean state , the mantle of gj 1132 b must continue to convect . however , progressive oxidation of the mantle by liberated o should lead to lower density materials at the top of the mantle , as shown in figure [ fig : fe2o3profile ] . this may prevent overturn of mantle and delay the onset of solid state mantle convection , which would lead to reduced outgassing efficiencies . however , @xcite calculated the mineralogy of a solidifying magma ocean ( without atmospheric oxidation ) , and found that the cumulate pile of the solidified magma ocean is unstable due to the partitioning of feo into later ( near surface ) crystal phases . although oxidation of feo to may change the exact mineral condensation sequence , the additional oxygen should not be sufficient to counteract the density effect of enhanced feo abundance in the upper mantle . for low feo abundances , the density instability may be insufficient to cause mantle overturn , in which case gj 1132 b may become stuck in a stagnant lid regime . this would mimic the low degassing efficiency model , which we have shown is only important in the case of the low xuv model b. inefficient degassing reduces the final abundance by about an order of magnitude in pressure . our model suggests that gj 1132 b would require more than @xmath05 wt% by planet mass of initial water in order to retain a substantial steam envelope . substantial oxygen atmospheric abundances ( a few bars up to several thousand bars ) without significant steam ( @xmath144 1 bar ) would imply a relatively high xuv flux and initial water abundances greater than @xmath05 wt% . substantial oxygen atmospheric abundances ( @xmath145 500 bars ) with significant steam ( @xmath145 500 bars ) would imply either low xuv flux over the system s lifetime , a large initial water abundance of more than 250 eo , or the presence of an earlier -rich envelope . the presence of a steam atmosphere implies the continued existence of a magma ocean at gj 1132 b s surface . however , most of our starting conditions result in tenuous atmospheres with at most a few bars of and little to no steam remaining . further constraints on the initial planet composition will require more stringent mass bounds and xuv flux measurements . future observations of gj1132b s atmosphere will allow us to probe these scenarios . the planet s transmission and emission spectra are sensitive to the relative abundances of o@xmath152 , h@xmath152o , and other species @xcite , and these spectra will be measurable with deep observations from the ground @xcite or from space ( * ? ? ? * ; * ? ? ? * and references therein ) . complementary jwst observations of gj1132b s thermal phase curve could reveal its total atmospheric mass , and therefore determine whether the present - day atmosphere is thick or tenuous @xcite . detection of - collisionally - induced absorption features may also be used to constrain the presence and total pressure of a massive atmosphere @xcite . our model is applicable to a wide range of exoplanets inwards of their habitable zones . for instance , water loss from the recently discovered trappist-1 system @xcite was modeled by @xcite . as the host star is an ultra - cool dwarf and the planets therefore receive less total xuv flux , they may retain both massive water vapor and atmospheres , although conclusions must await both planet mass determinations , as well as detailed application of our model . application to planets such as trappist-1d , which is potentially within the habitable zone , will require altering our atmospheric model to allow for condensible water vapor . in our own solar system , venus may have experienced the loss of a similar steam - rich atmosphere as posited here for gj 1132 b , but with nearly 10 times less stellar insolation , the escape rates from venus should have been much lower and magma ocean cooling should have been much faster . future application of this model to venus may help confirm whether an early magma ocean could have taken up the produced by atmospheric loss as suggested by @xcite and others . the authors thank colin goldblatt for providing the runaway greenhouse olr data used to perform the intercomparison shown in figure [ fig : lbl_goldblatt_intercomp ] , and an anonymous referee for a positive and helpful review . the line - by - line opacity computations in this paper were run on the odyssey cluster supported by the fas division of science , research computing group at harvard university . ls and ds acknowledge support from the simons foundation . zkbt gratefully acknowledges support from the mit torres fellowship for exoplanet research . abe , y. , & matsui , t. 1985 , lunar and planetary science conference proceedings , 15 , c545 baraffe , i. , homeier , d. , allard , f. , chabrier , g. 2015 . a&a , 577 , a42 . barstow , j. k. , aigrain , s. , irwin , p. g. j. , kendrew , s. , & fletcher , l. n. 2016 , , 458 , 2657 benneke , b. , & seager , s. 2012 , , 753 , 100 berta - thompson , z. k. , irwin , j. , charbonneau , d. , et al . 2015 , , 527 , 204 . bolmont , e. , selsis , f. , owen , j. e. , et al . 2016 , arxiv:1605.00616 burke , c. j. , christiansen , j. l. , mullally , f. , et al . 2015 , , 809 , 8 chassefire , e. 1996 . jgr , 101 , 26039 . clough , s. , kneizys , f. , & davies , r. 1989 , atmospheric research , 23 , 229 clough , s. a. , iacono , m. j. , moncet , j. l. 1992 . jgr , 97 , 15 . cohen , o. , drake , j. , glocer , a. , et al . 2014 , arxiv preprint arxiv:1405.7707 cowan , n. b. , greene , t. , angerhausen , d. , et al . 2015 , , 127 , 311 demory , b .- o . 2014 , , 789 , l20 dressing , c. d. , & charbonneau , d. 2015 , , 807 , 45 dressing , c. d. , charbonneau , d. , dumusque , x. , et al . 2015 , , 800 , 135 elkins - tanton , l. t. , parmentier , e. m. , hess , p. c. 2003 . met . planet . , 38 , 1753 . elkins - tanton , l.t . e&psl , 271 , 181 . erkaev , n. v. , kulikov , y. n. , lammer , h. , et al . 2007 , , 472 , 329 erkaev , n. v. , lammer , h. , odert , p. , et al . 2013 , astrobiology , 13 , 1011 france , k. , parke loyd , r. o. , youngblood , a. , et al . 2016 , , 820 , 89 frost , d. , & mccammon , c. a. 2008 . earth planet . , 36 , 389 . fukai , y. 1984 . nature , 308 , 174 . gillmann , c. , chassefire , e. & lognonn , p. 2009 . earth planet . lett . , 286 , 503 . gillon , m. , jehin , e. , lederer , s. m. , et al . 2016 , , 533 , 221 goldblatt , c. , robinson , t. d. , zahnle , k. j. , & crisp , d. 2013 , nature geoscience , 6 , 661 hamano , k. , abe , y. , genda , h. 2013 . nature , 497 , 607 . hawley , s. l. , gizis , j. e. , & reid , i. n. 1996 , , 112 , 2799 hirschmann , m. m. 2000 . geosys . 1 , 2000gc000070 hirschmann , m.m . earth planet . 341 - 344 , 48 . holloway , j. r. 1998 . hunten , d. m. , pepin , r. o. , & walker , j. c. g. 1987 , , 69 , 532 ingersoll , a. p. 1969 , journal of atmospheric sciences , 26 , 1191 , j. f. 1988 , icarus , 74 , 472 kasting , j.f . , & pollack , j.b . 1983 . icarus , 53 , 479 . kasting , j.f . space sci . khodachenko , m. l. , ribas , i. , lammer , h. , et al . 2007 , astrobiology , 7 , 167 kislyakova , k. g. , lammer , h. , holmstrm , m. , et al . 2013 , astrobiology , 13 , 1030 koll , d. d. b. , & abbot , d. s. 2015 , , 802 , 21 kombayashi , m. 1967 . j. meteor . japan , 45 , 137 . kress , v. c. & carmichael , i. s. e. 1991 . comp , 108 , 82 . , y. n. , lammer , h. , lichtenegger , h. i. m. , et al . 2006 , planetary and space science , 54 , 1425 kuramoto , k. , & matsui , t. 1996 . jgr 101 , 14909 . lalitha , s. , poppenhaeger , k. , singh , k. p. , czesla , s. , & schmitt , j. h. m. m. 2014 , , 790 , l11 , h. , lichtenegger , h. i. m. , kulikov , y. n. , et al . 2007 , astrobiology , 7 , 185 lammer , h. and 17 others . 2009 , astron . astrophys . 506 , 399 . lebrun , t. , massol , h. , chassefire , e. , davaille , a. , marcq , e. , sarda , p. , leblanc , f. , brandeis , g. 2013 , jgr , 118 , 1155 . lopez , e.d . & fortney , j. j. 2013 . apj , 776 , 2 . lichtenegger , h. i. m. , lammer , h. , griemeier , j .- , et al . 2010 , icarus , 210 , 1 lide , d. p. , ed . 2000 , crc handbook of chemistry and physics , 81st edn . ( crc press ) linsky , j. l. , fontenla , j. , & france , k. 2014 , , 780 , 61 luger , r. , & barnes , r. 2015 , astrobiology , 15 , 119 luger , r. , barnes , r. , lopez , e. , et al . 2015 , astrobiology , 15 , 57 marcus , r. a. , sasselov , d. , hernquist , l. , & stewart , s. t. 2010 , , 712 , l73 newton , e. r. , irwin , j. , charbonneau , d. , berta - thompson , z. k. , & dittmann , j. a. 2016 , , 821 , l19 oneill , h. st . c. & palme , h. 1998 , in the earth s mantle - composition , structure and evolution , ed . i. jackson ( cambridge , uk : cambridge univ . press ) owen , j. e. , & mohanty , s. 2016 , , 459 , 4088 papale , p. 1997 . comp , 126 , 237 . parke loyd , r. o. , france , k. , youngblood , a. , et al . 2016 , arxiv:1604.04776 pierrehumbert , r. t. 2011 . principles of planetary climate . cambridge univ . press . pizzolato , n. , maggio , a. , micela , g. , sciortino , s. , & ventura , p. 2003 , , 397 , 147 ramirez , r. m. , & kaltenegger , l. 2014 , the astrophysical journal letters , 797 , l25 robinson , m.s . , & taylor , g. j. 2001 . meteoritics and planetary science , 36 , 841 . rodler , f. , & lpez - morales , m. 2014 , , 781 , 54 rogers , l. 2015 . apj , 801 , 41 . rothman , l. , gordon , i. , babikov , y. , et al . 2013 , journal of quantitative spectroscopy and radiative transfer , 130 , 4 , l. s. , rinsland , c. p. , goldman , a. , et al . 1998 , journal of quantitative spectroscopy and radiative transfer , 60 , 665 sandu , c. , & kiefer , w. s. 2012 . lett . , 39 , l03201 . schaefer , l & sasselov , d. 2015 . apj , 801:40 . schwieterman , e. w. , meadows , v. s. , domagal - goldman , s. d. , et al . 2016 , , 819 , l13 selsis , f. , wordsworth , r. d. , & forget , f. 2011 , , 532 , a1 shkolnik , e.l . , & barman , t.s . 2014 , astron . j. , 148 , 64 . snellen , i. a. g. , de kok , r. j. , le poole , r. , brogi , m. , & birkby , j. 2013 , , 764 , 182 tian , f. 2009 , , 703 , 905 tian , f. 2015 , earth and planetary science letters , 432 , 126 . tian , f. , & ida , s. 2015 , nature geoscience verbiscer , a. , french , r. , showalter , m. , & helfenstein , p. 2007 , science , 315 , 815 vidotto , a. a. , jardine , m. , morin , j. , et al . 2013 , astronomy & astrophysics , 557 , a67 , j. c. g. , hays , p. b. , & kasting , j. f. 1981 , journal of geophysical research , 86 , 9776 watson , a.j . , donahue , t.m . , walker , j.c . 1981 , icarus , 48 , 150 . weiss , l. m. , & marcy , g. w. 2014 , , 783 , l6 wordsworth , r.d . , pierrehumbert , r.t . apj , 778 , 154 . wordsworth , r. , forget , f. , millour , e. , et al . 2013 , icarus , 222 , 1 wordsworth , r. , & pierrehumbert , r. 2014 , the astrophysical journal letters , 785 , l20 wright , n. j. , drake , j. j. , mamajek , e. e. , & henry , g. w. 2011 , , 743 , 48 youngblood , a. , france , k. , parke loyd , r. o. , et al . 2016 , arxiv:1604.01032 zahnle , k. j. 1986 , journal of geophysical research : atmospheres ( 19842012 ) , 91 , 2819 zahnle , k. , kasting , j. f. , & pollack , j. b. 1990 , icarus , 84 , 502 zendejas , j. , segura , a. , & raga , a. c. 2010 , icarus , 210 , 539 zeng , l. , sasselov , d. d. , & jacobsen , s. b. 2016 , , 819 , 127

blind deconvolution is an ubiquitous non - linear inverse problem in applications like wireless communications and image processing . this problem is generally ill - posed , and there have been efforts to use sparse models for regularizing blind deconvolution to promote signal identifiability . i of this two - part paper characterizes the ambiguity space of blind deconvolution and shows unidentifiability of this inverse problem for _ almost every _ pair of unconstrained input signals . the approach involves lifting the deconvolution problem to a rank one matrix recovery problem and analyzing the rank two null space of the resultant linear operator . a measure theoretically tight ( parametric and recursive ) representation of the key rank two null space is stated and proved . this representation is a novel foundational result for signal and code design strategies promoting identifiability under convolutive observation models . ii of this paper analyzes the identifiability of _ sparsity _ constrained blind deconvolution and establishes surprisingly strong negative results on scaling laws for the sparsity - ambiguity trade - off . = 1 identifiability , rank one matrix recovery , blind deconvolution , parametric representation , rank two null space

1411.3810

deconvolution is a challenging inverse problem that is ubiquitous in applications of signal processing , control theory and wireless communications like blind image restoration @xcite , blind system identification @xcite and , blind channel estimation and equalization @xcite . in the absence of additional constraints , blind deconvolution is known to be ill - posed and each application mentioned above imposes some form of prior knowledge on the underlying signal structures to render the inverse problem better behaved . _ sparsity _ based models have been used extensively in the past decade to capture hidden signal structures in many applications of interest . prominent examples of the exploitation of sparsity include natural images admitting sparse wavelet domain representations @xcite , ultra wide band communication channels exhibiting sparsity in doppler - delay domain representations @xcite , and user preferences and topic models displaying low - rank structures @xcite ( sparsity in eigenvalue domain ) . while there have been a few attempts at exploiting sparsity priors for blind deconvolution and related problems @xcite , there does not appear to be a general characterization of _ identifiability _ of such sparse models , except with very restrictive constraints . for example , @xcite assumes a _ single random subspace _ signal model as opposed to the more commonly used _ union of subspaces _ signal model in compressed sensing @xcite . in this two - part paper , we prove some surprising negative results for signal identifiability in the non - sparse and sparsity constrained blind deconvolution problems , even for the single subspace model . for a principled approach to studying these negative results , the first part of the paper develops a measure theoretically tight characterization of the ambiguity space of blind deconvolution . because of the simplicity of the linear convolution map , its ( non - linear ) kernel can be partially parametrized to yield an analytically useful representation that is instrumental in the derivation of our results . to the best of our knowledge , this is a novel parameterization that has not appeared elsewhere in literature . the second part @xcite of this paper shows that the unfavorable sparsity - ambiguity trade - off results on identifiability are a consequence of the kernel of the linear convolution map being of large dimension ( of the order of the signal dimension ) . we present some constructive numerical examples illustrating the nature and source of unidentifiability for blind linear deconvolution . the discrete - time blind linear deconvolution problem can be stated as the task of recovering ( up to scalar multiplicative ambiguities ) the unknown signal pair @xmath0 from the observation of their noise free linear convolution @xmath1 . it is intuitive to see that if there exists another pair @xmath2 such that @xmath3 but @xmath4 and @xmath5 are not collinear then @xmath6 is _ unidentifiable _ ( information theoretically impossible to recover uniquely ) . below , we illustrate what can go wrong with blind deconvolution for @xmath7 ( beside scalar multiplicative ambiguities ) . consider the vectors @xmath8 and @xmath9 assigned as @xmath10 clearly , @xmath11 and @xmath12 are linearly independent and a simple calculation reveals that @xmath13 implying unidentifiability of both @xmath14 and @xmath15 within @xmath7 . a more non - trivial numerical example can be constructed using the vectors @xmath16 , @xmath17 , @xmath18 and @xmath19 ( which are functions of @xmath11 , @xmath12 , @xmath20 and @xmath21 above ) , defined as @twocolumn [ eqn : numerical example cos sin ] @xmath22 [ eqn : numerical example cos sin ] @xmath23 clearly , @xmath16 and @xmath17 are linearly independent and we observe that @twocolumn @xmath24 @xmath25 implying unidentifiability of both pairs @xmath26 and @xmath27 within @xmath7 . and @xmath28 in the figure , @xmath29 implies @xmath30 for any @xmath31 . ] the construction used in illustrates a form of _ rotational ambiguity _ that generalizes to produce a family of unidentifiable pairs in @xmath32 . [ fig : rotation ambiguity ] represents a pictorial depiction of the following construction . let @xmath33 denote the common convolved output in . consider the parameterized vectors @xmath34 and @xmath35 defined by [ eqn : rotational transform ] @xmath36 where @xmath31 are the parameters , and @xmath37 acts as the set of seed vectors driving the above transformation . clearly , linear independence of @xmath11 and @xmath12 and @xmath31 imply that @xmath38 and @xmath39 are linearly independent . simple algebraic manipulations reveal that @twocolumn [ eqn : parametrized example cos sin ] @xmath40 [ eqn : parametrized example cos sin ] @xmath41 since and have the same , both @xmath42 and @xmath43 are unidentifiable within @xmath7 . since @xmath44 describes a two dimensional parameter space , implies that the unidentifiable subset of @xmath32 is at least two dimensional . notice that the aforementioned construction for unidentifiable signal pairs ( the rotational ambiguity ) remains true , even if the domain of the vectors @xmath38 and @xmath39 is restricted to an annulus @xmath45 for some @xmath46 . prior research on blind system and channel identification @xcite has mainly focused on single - in - multiple - out ( simo ) and multiple - in - multiple - out ( mimo ) systems , also known as the _ blind multi - channel finite - impulse - response ( fir ) estimation _ problem . these systems have multiple output channels that need to be estimated from the observed outputs when the channels are driven by either a single source ( simo system ) or multiple sources ( mimo system ) . a key property necessary for successful identifiability and recovery of the multiple channel vectors is that the channels should display sufficient diversity or richness either stochastically ( cyclostationary second order statistics ) @xcite or deterministically ( no common zero across all channels ) @xcite . as pointed out in @xcite , such diversity is generally unavailable in the single - in - single - out ( siso ) systems thus making them extremely challenging . however , siso systems are our primary concern due to their equivalence to blind deconvolution problems . in this paper , we shall _ not _ be concerned with stochastic formulations where multiple realizations of the channel are observed to infer empirical statistics . instead , we shall consider non - sparse ( here , in i ) and sparse ( in ii ) channel instances and characterize their instance identifiabilities . when blind deconvolution is treated as a channel estimation problem , @xcite shows that guard intervals of sufficient length between blocks of transmitted symbols enable successful blind channel identification , even for the deterministic siso system type . more specifically , for an @xmath47 length channel impulse response ( cir ) a guard interval of length @xmath48 is needed between consecutive blocks of transmitted source symbols , resulting in an absence of any inter - block or inter - symbol interference ( isi ) in the convolved output of the channel . we consider scenarios involving isi and show unidentifiability a wide variety of domain restrictions . although our definition of identifiability is similar to that in @xcite , our results are substantially different in nature . firstly , we deal with real fields ( not algebraically closed ) and with sparse subsets of real vector spaces ( induced by measures that are not absolutely continuous the lebesgue measure ) as opposed to the algebraically closed complex fields and vector spaces considered in @xcite . this requires different proof techniques . secondly , we focus on unidentifiability results and characterize the dimension of the unidentifiable subsets in addition to showing existence results . furthermore , our proofs constructively show unidentifiable subsets as opposed to non - constructive existence proofs . to the best of our knowledge , blind deconvolution was cast as a rank one matrix recovery problem first in @xcite ; we adopted this framework in our earlier works @xcite on characterization of identifiability in general bilinear inverse problems . herein , we specifically consider the blind deconvolution problem and characterize its inherent _ unidentifiability _ under some application - motivated sparsity and subspace constraints . further , @xcite focused on developing heuristic algorithms for the deconvolution problem and did not explicitly address identifiability . the subsequent paper @xcite , by the authors of @xcite , does ( implicitly ) address identifiability ( through a study of recoverability by convex programming ) but assumes the knowledge of the support of the sparse signal . although blind deconvolution is an instance of a bilinear inverse problem , the present work differs significantly from our earlier work @xcite in two ways . firstly , @xcite focused on identifiability results ( analogous to achievability results in information theory ) whereas the present paper derives unidentifiability results ( akin to converse / impossibility results in information theory ) . secondly , ii of this paper @xcite explores more explicit sparsity priors motivated by applications like cooperative communication @xcite , channel estimation for wide - band communication @xcite and musical source separation @xcite , whereas @xcite developed the theory of regularized bilinear inverse problems more abstractly , making assumptions no more specific than non - convex cone constrained priors . a consequence of focusing specifically on blind deconvolution is that we are able to develop tight converse bounds that match our achievability bounds in @xcite . the tractability of this analysis hinges on the structural simplicity of the convolution operator after lifting . to rigorously justify the dimension related results in both parts of the paper , we call upon the theoretical framework of the _ hausdorff dimension _ of sets in metric spaces . a moderately detailed introduction to the topic can be found in @xcite . for the sake of readability , we provide a short review of the concept and some of the useful properties in ii , ii - c of the paper . in this work , we quantify unidentifiability for certain families of noiseless sparse and non - sparse blind deconvolution problems under a _ non - asymptotic _ and _ non - statistical _ setup . specifically , given model orders @xmath49 , i of the paper develops a parameterization of the ambiguity kernel for blind linear deconvolution and shows that almost every input @xmath50 is unidentifiable under blind deconvolution in the absence of further constraints . furthermore , the characterization developed herein is used in ii of the paper to investigate some application driven choices of the separable domain restriction @xmath51 and establish that these restrictions are insufficient to guarantee identifiability of the vectors @xmath5 and @xmath52 from their linearly convolved resultant vector @xmath53 . our approach leads to the following novelties for i ( refer to @xcite for contributions of ii of the paper ) . 1 . we are able to analyze unidentifiability in blind deconvolution by studying the rank constrained null space of a linear operator on matrices . we explicitly demonstrate the almost everywhere unidentifiable nature of unconstrained blind deconvolution by constructing families of adversarial signal pairs for even model orders @xmath54 and @xmath47 . this is a much stronger unidentifiability result than any counterparts in the literature . we state and prove a measure theoretically tight , non - linear , partially parametric and partially recursive characterization of the rank two null space of the lifted linear convolution map . to the best of our knowledge , this is a new result . characterizing the rank two null space is useful for a variety of applications , including the study of scaling laws in bilinear inverse problems @xcite . our proofs are constructive and demonstrate a rotational ambiguity phenomenon in general bilinear inverse problems which is the cause of existence of a large dimensional set of unidentifiable input pairs in blind deconvolution . the rest of the paper is organized as follows . [ sec : model ] describes the system model , sets up the notion of identifiability up to a suitably defined equivalence class , and presents the lifted reformulation of the blind deconvolution problem as a rank one matrix recovery problem . [ sec : parametrization ] develops a novel , measure theoretically tight characterization of the rank two null space of the lifted linear convolution map , exploiting both parametric as well as recursive definitions . [ sec : results without coding ] presents the key unidentifiability result for non - sparse blind deconvolution , exploiting the parametric representation of the rank two null space of the lifted linear convolution map developed in [ sec : parametrization ] . [ sec : conclusion ] concludes the paper . detailed proofs of all the results in the paper appear in the [ sec : nullspace converse proof]-[sec : ae unidentifiability proof ] . all vectors are assumed to be column vectors unless stated otherwise . we shall use lowercase boldface alphabets to denote column vectors ( @xmath55 ) and uppercase boldface alphabets to denote matrices ( @xmath56 ) . the matlab^^ indexing rules will be used to denote parts of a vector / matrix ( @xmath57 denotes the sub - matrix of @xmath56 formed by the rows @xmath58 and columns @xmath59 ) . the all zero vector / matrix shall be denoted by @xmath60 and its dimension would be clear from the usage context . for matrices , @xmath61 and @xmath62 respectively return the transpose and rank of their argument . special sets are denoted by uppercase blackboard bold font ( @xmath63 for real numbers ) . other sets are denoted by uppercase calligraphic font ( @xmath64 ) . linear operators on matrices are denoted by uppercase script font ( @xmath65 ) . for any matrix @xmath66 , we denote its column space by @xmath67 . the standard euclidean inner product on a vector space will be denoted by @xmath68 and the underlying vector space will be clear from the usage context . when we refer to the ` dimension ' of a set , we shall mean the hausdorff dimension of the set . to avoid unnecessarily heavy notation , we shall adopt the following convention : the use of both vector variables ( like @xmath4 , @xmath69 , @xmath70 , @xmath71 , ) as well as matrix variables ( like @xmath72 , @xmath73 , ) is allowed to differ across theorems and proofs ( and even across different subparts of the same proof ) with the understanding that their meanings are restricted to the scope of the individual theorems ( and proofs ) . this facilitates the reuse of the same variable names across different theorems with distinct meanings without confusion . we shall consider the noiseless linear convolution system model @xmath74 where @xmath75 denotes the pair of unknown signals with a given domain restriction @xmath76 , @xmath77 denotes the linear convolution map , and @xmath78 is the vector of observations given by @xmath79 we are interested in solving for the vectors @xmath4 and @xmath69 from the noiseless observation @xmath80 in . the blind linear deconvolution problem corresponding to is represented by the feasibility problem we are interested in whether the pair @xmath75 can be uniquely identified in a meaningful sense . we assume that the lengths ( or model orders ) @xmath54 and @xmath47 , respectively , of vectors @xmath4 and @xmath69 are fixed and known _ a priori_. notice that the deconvolution problem has an inherent scaling ambiguity due to the identity @xmath81 stemming from the bi - linearity of the convolution operator . thus , any meaningful definition of identifiability for blind deconvolution must disregard this type of scaling ambiguity . this leads us to the following definition of identifiability . [ defn : identifiability ] a vector pair @xmath76 is identifiable with respect to the linear convolution map @xmath82 , if @xmath83 satisfying @xmath84 , @xmath85 such that @xmath86 . this definition is in the same spirit as the notion of noise free identifiability described in @xcite , but restricted to a set @xmath87 . it is easy to see that [ defn : identifiability ] induces an equivalence structure on the set of identifiable pairs in @xmath32 , and identifiability refers to the identification of the equivalence class in @xmath32 that generated the observation @xmath80 in . while is an accurate representation of a blind deconvolution problem , it is not easily amenable to an identifiability analysis in the sense of [ defn : identifiability ] . we use the _ lifting _ technique from optimization @xcite to rewrite as a rank minimization problem subject to linear equality constraints @xcite ; a form that is better suited for an identifiability analysis , where @xmath88 is _ any _ set satisfying @xmath89 and @xmath90 is a linear operator which can be deterministically constructed from the linear convolution map . we shall refer to @xmath65 as the _ lifted linear convolution map_. specifically , @xmath91 is the unique linear operator that satisfies @xmath92 by construction , the optimal solution to is a rank one matrix @xmath93 and its singular value decomposition @xmath94 yields a solution @xmath95 to . the exact steps involved in the lifting technique and a proof of equivalence between the lifted and the original problems , in the much broader context of bilinear inverse problems , are available in @xcite . our unidentifiability results in [ sec : results without coding ] will be based on an analysis of . [ rem : projections ] it is well known from functional analysis @xcite that any finite dimensional linear operation can be decomposed into a set of inner product operations that collectively define the linear operation . the lifted linear convolution map @xmath91 can be decomposed into a functionally equivalent set , comprising of @xmath96 matrices , using coordinate projections . let @xmath97 denote the @xmath98 matrix in the decomposition and @xmath99 denote the @xmath98 coordinate projection operator of @xmath96 dimensional vectors to scalars , if @xmath78 then @xmath100 , for @xmath101 . for all @xmath2 , we have the relation @twocolumn @xmath102 @xmath103 where @xmath68 denotes the trace inner product in the space of matrices @xmath104 . by way of exposition , we compute @xmath105 for @xmath106 . let @xmath107 . invoking the expansion in , we have @twocolumn @xmath108 @xmath109 the matrix @xmath110 is given by @xmath111 for @xmath112 and @xmath113 . in general , it is not hard to see that the matrices @xmath114 , @xmath101 , are hankel matrices in @xmath115 specified as @xmath116 for @xmath117 , and @xmath118 . [ fig : lifting example ] illustrates these matrices forming the decomposition corresponding to the lifted linear convolution operator @xmath91 for the case of @xmath119 . for linear convolution map with @xmath120 , @xmath121 and @xmath122 . ] we illustrate an interpretation of the lifted linear convolution operator that forms a key step in the proofs of our results in [ sec : parametrization ] . let @xmath78 be the result of applying the lifted convolution map @xmath91 to the matrix @xmath123 , @xmath124 and consider the case @xmath119 in [ fig : anti - diagonal sum ] as an example . it is clear that @xmath125 is formed by adding all elements of @xmath126 that lie on the @xmath127 arrow ( the @xmath127 anti - diagonal ) in [ fig : anti - diagonal sum ] . in other words , _ the linear operator @xmath91 sums elements of its argument along the anti - diagonals to generate the output_. from the definition of linear convolution in , it is easy to see that this interpretation of the lifted convolution operator holds regardless of the numerical values of the dimensions @xmath54 and @xmath47 . for @xmath106 satisfying @xmath124 . the shorthand @xmath128 has been used with @xmath129 and @xmath130 . ] as an immediate consequence , we make the following observation ( also an important part of the proofs to the results in [ sec : parametrization ] ) . let @xmath131 be any arbitrary matrix and we consider the matrices @xmath132 constructed as @xmath133 clearly , the elements of @xmath134 along the @xmath127 anti - diagonal are a circularly shifted copy of the elements of @xmath135 along the @xmath127 anti - diagonal , @xmath134 is formed by shifting down the elements of @xmath135 by one unit along the anti - diagonals . in particular , this implies that the sum of elements along the @xmath127 anti - diagonal is the same for both @xmath135 and @xmath134 . using the _ sum along anti - diagonals _ interpretation of the lifted linear convolution map @xmath91 , we have @xmath136 where @xmath137 denotes the lifted linear convolution map in one lower dimension than @xmath91 both rows and columns . thus , @xmath135 and @xmath134 are indistinguishable under the action of the linear operator @xmath91 . suppose that @xmath91 is the lifted linear operator corresponding to the linear convolution map . we denote the rank two null space of @xmath91 by @xmath138 which is defined as @xcite @xmath139 in this section , we establish a partially parametric and partially dimension recursive characterization of @xmath138 that is critical to our subsequent results in [ sec : results without coding ] . we first establish [ lem : rank-2 nullspace ] that describes a subset of @xmath138 . [ prop : rank-2 nullspace converse ] shows existence of subsets of @xmath138 not covered by [ lem : rank-2 nullspace ] . [ thm : rank-2 nullspace full ] then provides an almost everywhere characterization of @xmath138 . finally , [ prop : rank-2 nullspace exception set ] shows existence of subsets of @xmath138 that are not covered by [ thm : rank-2 nullspace full ] . we shall make use of the fact that the rank one null space of the linear convolution operator is trivial , @xmath140 which follows from interpreting convolution as polynomial multiplication , since the product of two real polynomials is identically zero if and only if at least one of them is identically zero . [ lem : rank-2 nullspace ] enables the analysis of the rank two null space of @xmath91 , which is equivalent to the non - linear ambiguity space of blind deconvolution in a transformed domain . [ lem : rank-2 nullspace ] let @xmath141 and @xmath142 admit a factorization of the form @xmath143 for some @xmath144 and @xmath145 . then @xmath146 . in [ lem : rank-2 nullspace ] above , we have used the shorthand representations @xmath147 and @xmath148 for brevity , we shall continue to use analogous shorthand representations for matrices in the rest of the paper , with the understanding that these symbolic shorthands are dimension - wise consistent upon element - wise expansion . let @xmath149 admit a factorization as in . then , @xmath150 clearly , @xmath151 in is obtained by shifting down the elements of @xmath152 by one unit along the anti - diagonals and then flipping the sign of each element . since the convolution operator @xmath91 sums elements along the anti - diagonals ( see [ fig : anti - diagonal sum ] for illustration and [ sec : anti - diagonal sum interpretation ] for details ) , the representation of @xmath149 as in immediately implies that @xmath153 . since implies that @xmath154 so we have @xmath146 . we notice that @xmath91 maps @xmath104 to @xmath155 . an @xmath156 dimensional rank two matrix has @xmath157 degrees of freedom ( dof ) , so that @xmath138 has at most @xmath158 dof . since the representation on the of also has @xmath159 dof and @xmath160 , our parametrization is tight up to dof . however , the converse of [ lem : rank-2 nullspace ] is false in general , as we show in [ prop : rank-2 nullspace converse ] below . this implies that while unidentifiability results for blind deconvolution can be shown by proving existence of matrices satisfying , proof of identifiability results for deterministic input signals , on the other hand , requires substantially more mathematical effort and careful analysis on the subset of @xmath138 that is not element - wise representable in the form of . [ prop : rank-2 nullspace converse ] for @xmath161 , there exists a @xmath159 dimensional set @xmath162 such that @xmath163 and for every @xmath164 , @xmath66 is not representable in the form of . [ sec : nullspace converse proof ] . [ prop : rank-2 nullspace converse ] shows that [ lem : rank-2 nullspace ] only provides a partial characterization of @xmath138 . in particular , the elements of @xmath138 that are not representable by form an equally large set . we are ready to state an almost everywhere description of @xmath138 for the lifted linear convolution map @xmath165 . our description will involve a recursion over dimension . to avoid any confusion , we shall explicitly augment the dimension to the symbolic representation of the lifted operator . thus , the lifted linear convolution map @xmath165 will be denoted by @xmath166 and the associated rank two null space will be denoted by @xmath167 . for @xmath141 , we define the following sets @twocolumn @xmath168 @xmath169 @xmath170 @xmath171 [ thm : rank-2 nullspace full ] the following relationships hold . 1 . [ itm : m1 n1 ] @xmath172 and @xmath173 for all positive integers @xmath174 . [ itm : m2 n2 ] @xmath175 and @xmath176 for integers @xmath141 . 3 . [ itm : m n subset ] @xmath177 for integers @xmath161 . m n restricted ] @xmath178{\mathcal{n}_{0}\bb{m , n } \bigcup \mathcal{n}_{2}\bb{m , n } } \setminus \mathcal{m}(m , n)$ ] for integers @xmath161 . [ sec : full null space proof ] . [ thm : rank-2 nullspace full ] provides an _ almost _ full characterization of the ambiguity set for blind deconvolution , beside the constructions used in [ lem : rank-2 nullspace ] and [ prop : rank-2 nullspace converse ] , this ambiguity space contains ( measure theoretically ) very few elements . in fact , for @xmath179 or @xmath180 , the characterization provided by [ lem : rank-2 nullspace ] is complete . another interpretation of [ thm : rank-2 nullspace full ] is that it provides a converse characterization of @xmath138 while [ lem : rank-2 nullspace ] and [ prop : rank-2 nullspace converse ] provide achievability / existence characterizations for @xmath138 . , @xmath181 , @xmath182 and @xmath183 as indicated by [ thm : rank-2 nullspace full ] for @xmath161 . ] the sets @xmath176 and @xmath184 are explicitly parametrized @xmath185 . in principle , to construct elements from @xmath186 for arbitrary @xmath187 with @xmath188 , we can start from the explicitly parameterized set @xmath189 and apply recursively @xmath190 times . [ fig : nullspace - venn - diagram ] shows a venn diagram displaying relationships among the different sets in [ thm : rank-2 nullspace full ] . we note that [ thm : rank-2 nullspace full ] proves equality of the sets @xmath167 and @xmath191 when restricted to @xmath192 . in [ prop : rank-2 nullspace exception set ] below , we show that this restriction is necessary when @xmath193 . for @xmath161 and @xmath194 , it is presently unclear whether @xmath183 is a non - empty set . we leave this as an open question . thus , [ prop : rank-2 nullspace exception set ] characterizes the gap left by the converse characterization in [ thm : rank-2 nullspace full ] for @xmath193 . [ prop : rank-2 nullspace exception set ] for @xmath195 , @xmath196{\mathcal{n}_{0}\bb{n , n } \bigcup \mathcal{n}_{2}\bb{n , n}}$ ] is a non - empty set of dimension at least @xmath197 . [ sec : nullspace exception set proof ] . while we shall only use [ lem : rank-2 nullspace ] in the sequel , we briefly emphasize the importance of the other results in this section . [ prop : rank-2 nullspace converse ] and [ thm : rank-2 nullspace full ] supply the intuition that the rank two null space of the linear convolution map is a substantially complex geometrical object . roughly speaking , the rank two null space @xmath167 can be divided into exactly two measure theoretically significant ( dimension - wise significant ) disjoint parts , @xmath181 and @xmath182 , and a measure theoretically insignificant ( not necessarily disjoint from @xmath191 ) part @xmath183 . both these parts are equally significant since they are of the same dimension , and whatever part of @xmath167 is outside @xmath191 , is contained within @xmath183 ( see [ fig : nullspace - venn - diagram ] for illustration ) and is therefore measure theoretically insignificant . hence , we claim that [ thm : rank-2 nullspace full ] provides an almost everywhere characterization of @xmath138 . of the two measure theoretically significant parts , [ lem : rank-2 nullspace ] gives an analytically simple parametrization for one part , namely @xmath181 . the other part , namely @xmath182 , admits a dimension recursive definition and despite our best efforts , we could not construct an analytically simple parametrization of @xmath182 akin to [ lem : rank-2 nullspace ] . these results hint that development of provably correct _ non - randomized _ coding strategies that promote signal identifiability under blind deconvolution would need to be quite sophisticated . in particular , the codes must disallow ambiguities arising from the recursive definition of @xmath182 . for provably correct _ simple randomized _ linear coding strategies like @xcite , [ prop : rank-2 nullspace converse ] hints that a coding redundancy of @xmath198 is intuitively necessary to prevent bad realizations of random codes with high probability . [ prop : rank-2 nullspace exception set ] lower bounds the dimension of the insignificant set @xmath183 in the special case of @xmath199 . this might be interpreted as the additional ambiguity arising from having to distinguish between equi - dimensional vectors @xmath200 under a bilinear operator like linear convolution ; since @xmath75 being a solution to the unconstrained blind deconvolution problem also implies @xmath201 as a solution . this ambiguity does not arise when @xmath202 , @xmath203 and @xmath194 . we shall use _ identifiability _ in the sense of [ defn : identifiability ] . after disregarding some pathological scenarios in [ sec : pathological examples ] we show a strong unidentifiability result ( [ thm : ae unident ] ) for _ non - sparse _ blind deconvolution in [ sec : nonsparse deconv ] . the proof strategy yields valuable insight for the sparsity constrained blind deconvolution identifiability results that we present in ii of the paper @xcite . throughout this section , we assume that @xmath32 represents the ( not necessarily convex ) feasible _ cone _ in , @xmath204 one has @xmath205 for every @xmath206 . to prove [ thm : ae unident ] , [ lem : rank-2 nullspace ] suffices and we do not need the full power of [ thm : rank-2 nullspace full ] . [ lem : finite quotient set ] states a non - linear re - parameterization cum decomposition result to serve as a building block for constructing adversarial instances of input signals for which deconvolution fails the identifiability test . this is used in the proof of [ thm : ae unident ] below and also to prove the results in ii of the paper , and is motivated by the nonlinear parameterization of the blind deconvolution ambiguity space in [ lem : rank-2 nullspace ] and the rotational ambiguity phenomenon discussed in [ sec : examples ] of the introduction . the astute reader would notice the symbolic connection to the transformation in and the representation in . [ lem : finite quotient set ] let @xmath207 be an arbitrary integer and @xmath208 be an arbitrary vector . the quotient set @xmath209 defined as @twocolumn @xmath210 @xmath211 is finite ( possibly empty ) with cardinality at most @xmath212 . if @xmath213 is an even integer then @xmath209 is non - empty . [ sec : finite quotient set proof ] . let @xmath214 , @xmath215 , and @xmath216 . if @xmath217 , then @xmath218 and @xmath219 for @xmath220 and @xmath221 . [ lem : rank-2 nullspace ] and the representation in imply that under the linear convolution map , @xmath6 is indistinguishable from @xmath222 where @xmath223 and @xmath224 , since @xmath225 is in @xmath138 . as @xmath5 and @xmath226 are linearly independent , @xmath6 is unidentifiable by [ defn : identifiability ] . thus , for identifiability of @xmath6 it is necessary that at least one of @xmath227 or @xmath228 is non - zero . similarly , if @xmath229 then invoking [ lem : rank-2 nullspace ] and representation implies that @xmath6 is indistinguishable ( under the convolution map ) from @xmath230 where @xmath231 and @xmath232 . by linear independence of @xmath5 and @xmath226 , @xmath6 is unidentifiable . thus , for identifiability it is necessary that at least one of @xmath233 or @xmath234 is non - zero . the examples above highlight that model orders @xmath54 and @xmath47 play a critical role for identifiability ( and unidentifiability ) for siso systems , overestimating the model orders is fatal for blind deconvolution . this is a well known fact for simo systems @xcite . let @xmath235 be an arbitrary vector pair in the feasible set @xmath32 . to examine non - pathological examples ( see [ sec : pathological examples ] ) , we will assume that @xmath236 ; the treatment of the other case , @xmath237 , is ideologically identical . [ thm : ae unident ] let @xmath238 be even integers and @xmath202 be an arbitrary vector with @xmath236 . let the feasible set @xmath32 be defined as @xmath239 then @xmath240 is unidentifiable almost everywhere any measure over @xmath69 that is absolutely continuous the @xmath47 dimensional lebesgue measure . [ sec : ae unidentifiability proof ] . [ thm : ae unident ] shows that unidentifiability is the norm rather than the exception . in fact , the number of identifiable pairs is insignificant over continuous distributions . [ prop : rank-2 nullspace converse ] ( and the discussion leading to it ) intuitively suggests that @xmath241 additional constraints on @xmath4 in are almost necessary for identifiability @xmath32 ( defined by ) , from a heuristic dof argument . if these additional constraints on @xmath202 are allowed , then it would turn into a semi - blind deconvolution problem . [ thm : ae unident ] provides somewhat more concrete evidence towards the same intuition . our prior work @xcite discusses an example for multi - hop channel estimation , where @xmath242 additional subspace constraints are imposed on @xmath202 by system design , and this not only leads to identifiability but also to efficient and provably correct recovery algorithms . a philosophically similar situation is discussed in @xcite using a guard interval based system for blind channel identification . we note that [ thm : ae unident ] states a much stronger ( almost everywhere lebesgue measure ) unidentifiability result than any counterparts in the literature ( @xcite ) which only assert existence of some unidentifiable input . the requirement of the model orders @xmath54 and @xmath47 being even positive integers is because @xmath63 is _ not algebraically closed_. a slightly weaker version of [ thm : ae unident ] , asserting the existence of a @xmath243 dimensional set of unidentifiable signal pairs in @xmath32 ( given by ) , follows readily from 2 in ii of the paper ( a result on unidentifiability of blind deconvolution under canonical sparsity constraints ) , and indeed _ does not _ require @xmath54 or @xmath47 to be even integers . this observation agrees with the absence of any conditions on the model orders in @xcite . for completeness , we state an adapted version of ii , 2 below ( which follows from setting the sparse index subsets therein to empty sets ) while deferring the proof to ii of the paper . [ thm : sparse unident moderate ] let @xmath244 be arbitrary integers and @xmath32 be as defined in . then there exists a set @xmath245 of dimension @xmath243 such that every @xmath246 is unidentifiable . we mention in passing that the statement of [ thm : ae unident ] is asymmetric @xmath4 and @xmath69 since it applies to every @xmath4 but not to every @xmath69 ( only almost every @xmath69 ) . this is slightly stronger than the measure theoretically symmetric version of [ thm : ae unident ] that asserts unidentifiability almost everywhere the @xmath243 dimensional lebesgue product measure over the pair @xmath75 . we further note that [ thm : ae unident ] _ can not _ be strengthened to assert unidentifiability for every pair @xmath240 as a result of a simple thought experiment . if such a statement were true , then reinterpreting convolution as polynomial multiplication implies that there can not exist any polynomials over the real field that admit exactly two polynomial factors over the reals ; which is clearly a false statement . through the unidentifiability results in this section , we have attempted to _ quantify _ the ill - posedness of non - sparse blind deconvolution . the main message is that if an application exhibits a bilinear observation model of linear convolution and no additional application specific structure can be imposed on the unknown variables then it is necessary to drastically revise the system design specifications . such revision may incorporate some form of randomized precoding of the unknowns , so that the _ effective _ bilinear operator governing the observation model looks substantially different from the convolution operator ( the gaussian random precoding used in @xcite ) . alternatively , sparsity in non - canonical bases could also be helpful ( rademacher random vector signal model used in @xcite ) , as canonical - sparsity constraints are shown to be insufficient for identifiability in ii of the paper . blind deconvolution is an important non - linear inverse problem routinely encountered in signal processing applications . natively , blind deconvolution is ill - posed from the viewpoint of signal identifiability and it is assumed that application specific additional constraints ( like sparsity ) would suffice to guarantee identifiability for this inverse problem . in the current work , we proved lebesgue almost everywhere unidentifiability in non - sparse blind deconvolution for even model orders ; a much stronger impossibility result than any counterparts in existing literature . our approach built on the _ lifting _ technique from optimization to reformulate blind deconvolution into a rank one matrix recovery problem , and analyzed the rank two null space of the resultant linear operator . while this approach is philosophically applicable to other bilinear inverse problems ( like dictionary learning ) , it is the simplicity of the convolution operator in the lifted domain that made our analysis tractable . more specifically , we explicitly demonstrated a form of rotational ambiguity in bilinear functions that generates a dimension - wise large set of unidentifiable inputs . to establish our results , we developed a _ measure theoretically tight , non - linear , partially parametric and partially recursive _ characterization of the rank two null space of the linear convolution map . our proofs are constructive , dimensions of key sets are explicitly computed , and results hold over the real field which is not algebraically closed . this result is a precursor to non - randomized code design strategies for guaranteeing signal identifiability under the bilinear observation model of linear convolution . the design of such codes is a topic of ongoing research . finally , an unanswered question that ensues from our analysis is whether the dimension - wise insignificant part @xmath247 , of the ambiguity kernel of blind deconvolution @xmath167 , is empty for @xmath161 and @xmath194 . the proof involves contradiction by an explicit dimension recursive construction . let @xmath161 be arbitrary integers and let @xmath137 denote the lifted linear convolution map in one lower dimension than @xmath91 both rows and columns . further , let @xmath248 , @xmath249 be arbitrary vectors and define @xmath250 as @xmath251 since @xmath252 and @xmath253 , using [ lem : rank-2 nullspace ] on @xmath72 implies @xmath254 . for any invertible matrix @xmath255 , let us designate @xmath256 for some @xmath257 and @xmath258 . clearly , @xmath259 . furthermore , the vectors @xmath260 are linearly independent , owing to the invertibility of @xmath56 . consider the matrix @xmath261 constructed as @xmath262 from and in [ sec : anti - diagonal sum interpretation ] , @xmath263 does not change if elements of @xmath264 are shifted down by one unit along the anti - diagonals . this implies @xmath265 for @xmath266 and thus , @xmath267 using and @xmath268 . therefore , @xmath269 since implies @xmath270 , but @xmath73 as represented in is not convertible to the form in because of linear independence of @xmath271 and @xmath272 . to finish the proof , we need to show that the parametric family of matrices @xmath73 as defined by is of dimension @xmath159 . we denote this set by @xmath162 . consider @xmath273 so that @xmath274 is @xmath275 dimensional and @xmath276 is @xmath277 dimensional . @xmath56 represents two independently chosen directions of linear combination to generate @xmath271 and @xmath272 , so @xmath56 is two dimensional . finally , given our calculations , we need to add one dimension for a scalar multiplicative parameter to account for all matrices @xmath278 . thus , @xmath162 is @xmath279 dimensional . throughout this proof , we shall let @xmath280 denote the @xmath127 coordinate projection operator , @xmath281 . we argue that @xmath172 . the proof for @xmath173 is similar . since linear convolution with a scalar is equivalent to multiplication by the same scalar , by definition of @xmath282 , the lifted operator @xmath283 is an identity operator on @xmath284 . this immediately implies @xmath172 since the identity operator has a trivial null space . let @xmath285 be an arbitrary integer . we show that @xmath176 . the proof of @xmath175 for @xmath286 is similar . setting @xmath287 in [ lem : rank-2 nullspace ] , we have @xmath288 . to complete the proof , we need to show @xmath289 . let @xmath290 . we have @xmath291 and by definition of @xmath292 , we have @xmath293 ( see [ fig : anti - diagonal sum ] for illustration ) , implying that @xmath294 . by a similar argument , @xmath295 . by definition of @xmath292 we have , @xmath296 implying that @xmath297 . letting @xmath298 we get @xmath299 since @xmath300 is arbitrary , we have @xmath289 . let @xmath161 be arbitrary integers . we ll show that @xmath301 . from [ lem : rank-2 nullspace ] , we have @xmath302 . it remains to show that @xmath303 . to this end , we borrow constructions from the proof of [ prop : rank-2 nullspace converse ] . let @xmath304 for some @xmath257 and @xmath258 . we construct @xmath305 as in and following the chain of arguments @xmath306 using the anti - diagonal sum interpretation of [ sec : anti - diagonal sum interpretation ] , we have @twocolumn @xmath307 @xmath308 since @xmath309 . further , implies that @xmath270 and therefore @xmath310 . since @xmath311 is arbitrary , we have @xmath303 . let @xmath161 be arbitrary integers . from the preceding part , we have @xmath301 implying that @xmath312{\mathcal{n}_{0}\bb{m , n } \bigcup \mathcal{n}_{2}\bb{m , n } } \setminus \mathcal{m}(m , n ) \subseteq \mathcal{n}\bb{\mathscr{s}_{\bb{m , n } } , 2 } \setminus \mathcal{m}(m , n)$ ] . it remains to show that @xmath313{\mathcal{n}_{0}\bb{m , n } \bigcup \mathcal{n}_{2}\bb{m , n } } \setminus \mathcal{m}(m , n)$ ] . let @xmath314 . we have @xmath315 and by definition of @xmath316 , we have @xmath317 ( see [ fig : anti - diagonal sum ] for illustration ) , implying that @xmath294 . by a similar argument , @xmath318 . _ our proof strategy will be to show that @xmath72 always admits a factorization like and then deduce that this necessarily implies our result . _ since @xmath319 , either @xmath320 or @xmath321 is true . let us first assume that @xmath320 implying that @xmath322 . we have two further possibilities , @xmath323 and @xmath324 . suppose that @xmath323 . we set @xmath325 so that @xmath326 . since @xmath327 , we have @xmath328 and since @xmath322 we conclude that @xmath329 such that @xmath330 and @xmath331 are linearly independent . we choose @xmath332 as follows . if @xmath333 ( @xmath334 for @xmath335 ) , then we set @xmath336 , else we set @xmath337 we claim that the vectors @xmath338 and @xmath339 are linearly independent and span @xmath340 , which would imply that @xmath72 admits a factorization of the form @xmath341 for some vectors @xmath342 . indeed , if @xmath333 then @xmath343 and @xmath344 are linearly independent and span @xmath340 since @xmath328 . in case @xmath345 , setting @xmath346 we observe that @xmath347 since @xmath320 and @xmath348 , @xmath349 and imply that @xmath338 and @xmath339 are linearly independent and span @xmath340 owing to the linear independence of @xmath350 and @xmath351 and @xmath352 being equal to two . thus , holds implying that @xmath353 and @xmath354 . we have @xmath355 by construction so @xmath356 . further , if @xmath357 were true then would imply @xmath324 contradicting our assumption of @xmath323 . thus , @xmath358 which implies @xmath359 . letting @xmath360 and @xmath361 , implies @xmath362 following the same chain of arguments as @xmath306 with the anti - diagonal sum interpretation , implies @twocolumn @xmath363 @xmath364 since @xmath365 . therefore , @xmath366 . if @xmath367 then and imply that @xmath368 . instead , if @xmath369 then @xmath370 and implies @twocolumn @xmath371 @xmath372 and thus @xmath373 by . since @xmath319 by assumption , we get @xmath374{\mathcal{n}_{0}\bb{m , n } \bigcup \mathcal{n}_{2}\bb{m , n } } \setminus \mathcal{m}(m , n)$ ] . now suppose that @xmath324 so that @xmath375 for some matrix @xmath376 . we will use mathematical induction on @xmath54 to settle this case . _ induction step : _ let @xmath377{\mathcal{n}_{0}\bb{m ' , n } \bigcup \mathcal{n}_{2}\bb{m ' , n } } \setminus \mathcal{m}(m',n)$ ] be true for @xmath378 . since @xmath316 sums elements along the anti - diagonals ( see [ fig : anti - diagonal sum ] for illustration ) and @xmath365 , we have @xmath379 implying that @xmath380 . further , @xmath320 by assumption and @xmath381 by implying that @xmath382 by definition . thus , @xmath383 and by the induction hypothesis either @xmath384 or @xmath385 . we separate the analysis for each of these cases below suppose that @xmath384 . then there exist vectors @xmath386 and @xmath258 such that @xmath387 and @xmath388 . now using we get @xmath389 where we have set @xmath390 and @xmath391 . clearly , @xmath392 . to conclude that @xmath393 we need to show that @xmath394 we have , @xmath395 and therefore @twocolumn @xmath396 \mathscr{s}_{\bb{m-1,n-1}}\bb{\vec{u}_{3}\tpose{\vec{v}_{1 } } + \vec{u}_{4}\tpose{\vec{v}_{2 } } } \\ & = \begin{bmatrix } 0 \\ \mathscr{s}_{\bb{m-2,n-1}}\bb{\vec{u}_{1}\tpose{\vec{v}_{1 } } + \vec{u}_{2}\tpose{\vec{v}_{2 } } } \end{bmatrix } = \vec{0 } \end{split } \label{eqn : null space verification}\ ] ] @xmath397 where the first equality is because @xmath398 sums elements along the anti - diagonals and the second equality follows from the fact that @xmath388 and @xmath399 . since @xmath367 , implies that @xmath400 . further , @xmath401 and thus , implies that @xmath394 and hence @xmath393 . by assumption , @xmath319 and therefore @xmath402 . 2 . assume that @xmath385 . then there exist vectors @xmath403 and @xmath144 such that @xmath404 and gives @xmath405 where we have set @xmath406 . clearly , @xmath407 and implies @xmath408 by definition . by assumption , @xmath319 and therefore @xmath409 . _ induction basis : _ we need to show that @xmath410{\mathcal{n}_{0}\bb{3,n } \bigcup \mathcal{n}_{2}\bb{3,n } } \setminus \mathcal{m}\bb{3,n}$ ] . let @xmath411 . if @xmath412 then it follows from [ itm : not zero ] of the proof that @xmath413{\mathcal{n}_{0}\bb{3,n } \bigcup \mathcal{n}_{2}\bb{3,n } } \setminus \mathcal{m}\bb{3,n}$ ] . instead , if @xmath414 then we have @xmath415{\mat{z}\bb{2:3 , : } } \end{bmatrix } = \mathscr{s}_{\bb{2,n}}\bb{\mat{z } } = \vec{0}\ ] ] where the first equality is due to @xmath292 summing elements along the anti - diagonals and the second equality follows from @xmath416 . therefore , @xmath417 implying that @xmath418 from [ sec : full null space proof part 2 ] of the proof . thus , there exist @xmath419 and @xmath144 such that @xmath420 implying that @twocolumn @xmath421 @xmath422 where @xmath423 . thus , implies that @xmath424 . since @xmath425 we have @xmath426 . we have proved that if @xmath427 and @xmath365 then @xmath428 . to finish the proof we need to consider the other scenario for @xmath319 , show that @xmath429 and @xmath365 imply @xmath428 . the arguments for this case are ideologically similar to our approach for @xmath427 and hence we omit it . we explicitly define an @xmath48 dimensional parameterized set of matrices @xmath430 as follows , @twocolumn @xmath431 @xmath432 and show that @xmath433{\mathcal{n}_{0}\bb{n , n } \bigcup \mathcal{n}_{2}\bb{n , n}}$ ] . let @xmath72 be an arbitrary element of @xmath434 and let @xmath435 denote some representative parameters certifying that @xmath436 according to . it can be easily verified that @xmath72 is skew - symmetric . since @xmath437 sums elements along anti - diagonals ( see [ fig : anti - diagonal sum ] for illustration ) , skew - symmetry of @xmath72 implies that @xmath72 is in the null space of @xmath437 . further , @xmath72 admits the factorization @xmath438 implying that @xmath439 and therefore @xmath440 . since @xmath441 we can also conclude that @xmath442 . to finish the proof , we need to show that @xmath443 . we proceed for a proof by contradiction . first , assume that @xmath444 . then there exist vectors @xmath445 such that @xmath446 since @xmath447 and @xmath448 , implies @xmath323 and @xmath449 . now implies that @xmath450 and @xmath451 . therefore , @xmath358 and @xmath452 . using the two different representations for @xmath340 in and we must have @xmath453 for some @xmath454 . however , implies @xmath455 and @xmath456 and thus , @xmath457 since @xmath358 and @xmath452 . this leads to a contradiction and therefore @xmath458 . next , we assume @xmath459 . all of our arguments from the preceding case of @xmath444 are still valid if we set @xmath460 . therefore , once again we reach a contradiction and @xmath461 . so , we have @xmath443 and the proof is complete . if @xmath462 then its cardinality is trivially upper bounded by @xmath463 . so , for the purpose of proving the cardinality upper bound , we assume that @xmath209 is non - empty . let @xmath464 be arbitrarily selected . since @xmath465 by assumption , we have @xmath466 by definition of @xmath209 , and therefore division by any element of the set @xmath467 is well defined . we provide a construction for @xmath468 as follows . we have , @xmath469 and for every @xmath470 , @xmath471 we define the vectors @xmath472 and @xmath473 . for @xmath474 , we have @twocolumn [ eqn : recursive defn ] @xmath475 [ eqn : recursive defn ] @xmath476 where follows from , follows from using basic trigonometric relationships , and follows from the definitions for vectors @xmath477 and @xmath478 . equation represents a set of constraints on the vector variable @xmath478 with @xmath477 known . since @xmath479 by definition , can be solved recursively for @xmath480 ( in that order ) as polynomial expressions in the variable @xmath481 . specifically , for @xmath482 , leads to an expression of @xmath483 as a @xmath484 degree polynomial of the variable @xmath481 and therefore , represents a consistency equation that must be satisfied by every solution to ( by every element of the set @xmath209 ) . furthermore , dividing the consistency equation by @xmath483 on both and results in a polynomial equation of degree @xmath485 in the variable @xmath481 and therefore @xmath483 can admit at most @xmath485 distinct values . if @xmath64 denotes the set of admissible values of @xmath483 then @xmath486 . for a given value of @xmath487 , uniquely determines @xmath478 since it gives a recepie to compute @xmath488 for every @xmath489 and we have @xmath479 by definition . from , we get @xmath490 implying that given a value of @xmath487 , @xmath491 is uniquely determined , and @xmath492 and @xmath493 each admits at most two distinct values . therefore , the pair of scalars @xmath494 admits at most @xmath495 distinct values . using and the definition of @xmath478 , we have @xmath496 implying that @xmath497 is uniquely determined by @xmath494 . therefore , @xmath468 admits at most @xmath498 distinct values . since @xmath468 represents an arbitrary element of @xmath209 , we conclude that @xmath209 is finite with cardinality at most @xmath212 . as a boundary case , notice that when @xmath499 , and are vacuous and suffices to prove the result since it implies the existence of exactly two elements in @xmath500 , @xmath468 and @xmath501 . to prove that @xmath209 is non - empty when @xmath213 is an even integer , we proceed as follows . given @xmath502 with @xmath465 , @xmath472 is well defined , admits @xmath503 and is completely known . let @xmath504 be a vector variable such that it satisfies for @xmath474 and admits @xmath479 . if such a vector @xmath478 could be found with @xmath505 , then using to choose @xmath493 and setting @xmath506 yields and element @xmath464 , since 1 . follows from being able to reverse the sequence of implications in the preceding part of the proof , @xmath507 for every @xmath470 , 2 . follows directly from , @xmath506 and @xmath479 , and 3 . @xmath465 is satisfied since @xmath508 and @xmath509 . for the boundary case of @xmath499 , is vacuous and @xmath510 satisfies for @xmath511 , thus yielding an element in @xmath500 and proving the result . for @xmath512 , we follow the recursive solution process for to arrive at a consistency equation for @xmath483 as previously described , we recursively express @xmath480 ( in that order ) as polynomials in the variable @xmath481 leading to a polynomial expression for @xmath483 of degree @xmath484 in the variable @xmath481 . this further leads to a polynomial equation of degree @xmath485 in the variable @xmath481 with a non - zero constant term and hence admits at least one non - zero real root whenever @xmath513 is an odd integer ( equivalently , if @xmath213 is even ) . using this value of @xmath481 , uniquely determines the vector @xmath478 since it provides a recepie to recursively compute @xmath488 for every @xmath489 and we have @xmath479 by definition . this demonstrates the existence of the vector @xmath504 with the desired characteristics and completes the proof . as in the statement of the theorem , let @xmath240 be an arbitrary vector pair with @xmath32 defined by . since @xmath238 are even integers , invoking [ lem : finite quotient set ] once each for @xmath4 and @xmath69 , we can construct vectors @xmath145 , @xmath144 and scalars @xmath514 such that @xmath515 equation implies that the dependence of @xmath69 is uniformly continuous on the @xmath47 real numbers @xmath516 ( with a non - singular jacobian matrix ) and these @xmath47 real numbers completely parametrize @xmath69 . further , since the measure over @xmath69 is absolutely continuous the @xmath47 dimensional lebesgue measure , it is possible to choose a measure over @xmath517 that is absolutely continuous the one dimensional lebesgue measure , and therefore @xmath518 is true almost everywhere the measure over @xmath517 . finally , since @xmath4 and @xmath69 have disjoint parameterizations , @xmath519 is true almost everywhere the measure over @xmath517 ( owing to absolute continuity one dimensional lebesgue measure ) . thus , @xmath520 is true almost everywhere the measure over @xmath517 and therefore also the measure over @xmath69 . once we have vectors @xmath70 and @xmath71 , we consider the decomposition @xmath521 for @xmath522 , as in . setting @xmath523 and observing and we conclude that @xmath524 and therefore the pairs @xmath75 and @xmath230 produce the same convolved output . since @xmath236 , implies that @xmath525 and therefore implies @xmath526 for @xmath518 . so , assuming @xmath518 implies that @xmath527 . further , @xmath4 and @xmath226 are linearly independent if @xmath519 , implying that @xmath75 is unidentifiable by [ defn : identifiability ] with @xmath230 as the certificate . the proof is complete since @xmath520 is true almost everywhere the measure over @xmath69 .

mayer cluster expansion is an important tool in statistical physics to evaluate grand canonical partition functions . it has recently been applied to the nekrasov instanton partition function of 4d gauge theories . the associated canonical model involves coupled integrations that take the form of a generalized matrix model . it can be studied with the standard techniques of matrix models , in particular collective field theory and loop equations . in the first part of these notes , we explain how the results of collective field theory can be derived from the cluster expansion . the equalities between free energies at first orders is explained by the discrete laplace transform relating canonical and grand canonical models . in a second part , we study the canonical loop equations and associate them to similar relations on the grand canonical side . it leads to relate the multi - point densities , fundamental objects of the matrix model , to the generating functions of multi - rooted clusters . finally , a method is proposed to derive loop equations directly on the grand canonical model . apctp pre2013 - 017 notes on mayer expansions and matrix models jean - emile bourgine + + _ pohang , gyeongbuk 790 - 784 , republic of korea _

1310.3566

the agt correspondence @xcite implies a relation between the canonical partition function of a and the grand canonical partition function of a generalized matrix model . the former represents a correlator of liouville theory , according to the proposal of dijkgraaf and vafa @xcite , further investigated in @xcite . the latter describes the instanton partition function of a 4d @xmath0 supersymmetric gauge theory in the @xmath1-background , as derived using localization techniques in @xcite . here the term generalized matrix model do not pertain to a matrix origin for the model , but instead refers to a set of models that can be studied using techniques initially developed in the realm of matrix models . among these techniques , the _ topological recursion _ @xcite exploits the invariance of the integration measure to derive a tower of nested equations satisfied by the correlators of the model . these equations , referred as _ loop equations _ , are solved employing methods from algebraic geometry . this technique has recently been extended to a wide spectrum of coupled integrals models in @xcite . in a suitable limit of the , agt - equivalent to the nekrasov - shatashvili ( ns ) limit of the @xmath1-background @xcite , loop equations are no longer algebraic but first order linear differential equations . in this context , the is a natural quantization of the hermitian matrix model , to which it reduces at @xmath2 . the first element of this tower of differential equations has been mapped to the tq relation derived in @xcite that describes the dual susy gauge theory in the ns limit @xcite . it is then natural to ask about the existence of a structure similar to loop equations on the gauge side of the correspondence . but so far , the loop equation technique has not been applied to grand canonical matrix models . on the other hand , the cluster expansion of mayer and montroll @xcite has been successfully employed to derived an effective action relevant to the ns limit @xcite . can we relate this cluster expansion to the topological expansion of a generalized matrix model ? is there an equivalent of the loop equations technique on the grand canonical side ? and more generally , how do canonical and grand canonical coupled integrals relate to each other ? these are the issues we propose to address in these notes . for this purpose , we consider the following grand canonical generalized matrix model , @xmath3 in analogy with the nekrasov partition function , integrals are understood as contour integrals over the real line . the potential @xmath4 and the kernel @xmath5 are free of singularities over the real axis . we propose to study the expansion of @xmath6 when the kernel is close to one . more precisely , we assume the form @xmath7 with @xmath8 an even function , non - vanishing at @xmath9 . although the results of these notes are very general , what we have in mind for the function @xmath8 is typically @xmath10 it is crucial for our considerations that @xmath8 is independent of @xmath11 . in this way , we exclude a class of models more relevant to the study of nekrasov partition functions . for instance , setting @xmath12 , one recovers the model proposed by j. hoppe in @xcite . this model is a one - parameter version of the nekrasov partition function that depends on two @xmath13-background equivariant deformation parameters @xmath14 and @xmath15 @xcite . as @xmath16 , it exhibits a phenomenon referred as _ instanton clustering _ in the context of susy gauge theories @xcite . it corresponds to poles coming from the kernel and pinching the integration contour . such poles should be avoided by a deformation of the contour , picking up the corresponding residues . as a result , terms of the @xmath11-expansions we are considering are reshuffled and the results presented here are no longer valid . these notes are organized as follows . in the second section , we compare the mayer cluster expansion of the grand canonical model with the collective field theory describing the large @xmath17 limit of the canonical model . taking the coupled limit @xmath16 and @xmath18 with @xmath19 fixed , we derive relations between the free energies at first orders . these relations are a consequence of the fact that the grand canonical partition function is the discrete laplace transform of the canonical one . we go on with the study of the canonical loop equations . we show that they relate to graphical identities between generating functions of rooted clusters . such generating functions show up in the mayer expansion and are identified with the multi - point densities . finally , we present a technique to derive directly the grand canonical loop equations . the main results are summarized in the concluding section . the cluster expansion was introduced by mayer and montroll as a way to compute the free energy knowing the form of the interaction between particles @xcite ( see also the book @xcite and the excellent review by andersen @xcite ) . it allows to derive the equation of state for various types of fluids . to do so , the kernel is expanded in @xmath11 , which corresponds to strength of molecular interactions in the case of non - ideal gases . the terms of the series consist of coupled integrals with the kernel @xmath8 instead of @xmath20 , and their expression is encoded into clusters . here , a cluster is a set of vertices connected by at most one link . the partition function is a sum over disconnected clusters , but after taking the logarithm the summation is restricted to connected ones . we denote by @xmath21 a generic connected cluster with @xmath22 vertices , @xmath23 the set of its links ( or edges ) and @xmath24 the set of vertices . to each vertex @xmath25 of a cluster is associated an integration over the particle of coordinate @xmath26 with measure @xmath27 . the edge @xmath28 between particles @xmath25 and @xmath29 represents the kernel @xmath30 . thus , the logarithm of the partition function writes @xmath31 where the symmetry factor @xmath32 is the cardinal of the group of automorphism for the cluster , i.e. the number of permutations of vertices that leave @xmath21 invariant . the first terms of the expansion and their symmetry coefficients are given in figure . the mayer expansion [ sum_clusters ] is an expansion at small ( bare ) fugacity @xmath33 . we would like to reformulate it as a @xmath33-exact expansion in the parameter @xmath11 . we will also renormalize the fugacity , keeping @xmath34 fixed . by analogy , @xmath33 would encode the gauge coupling constant of the nekrasov partition function , and the mayer cluster expansion is an expansion upon the number of instantons . more precisely , @xmath33 would correspond to @xmath35 and should be renormalized by a factor @xmath15 in the ns limit @xmath36 . in this context , the @xmath11-expansion we study corresponds to an expansion in the @xmath1-background parameter @xmath15 , or in the agt dual , to the semi - classical expansion of liouville correlators . since each link brings a factor of @xmath11 , at first order only the clusters with a minimal number of links contribute . these clusters , denoted @xmath37 , have a tree structure , with @xmath38 links for @xmath22 vertices . thus , at first order in @xmath11 the free energy is given by the following sum over trees , @xmath39 where we used the shortcut notation @xmath40 . note that we have renormalized the free energy by a factor of @xmath11 , which is reminiscent of the volume of the @xmath1-background @xmath41 by which the prepotential should be multiplied in order to be finite in the @xmath42 limit @xmath43 . the first terms of this expansion are given in the figure . to evaluate @xmath44 , it is convenient to consider the generating function of rooted trees @xmath45 , defined as @xmath46 where with a slight abuse of notations we denoted the root and its coordinate by the same letter @xmath47 . the first order terms of this expansion are given in figure . this function is interpreted as a tree - level dressed vertex . we should also emphasize that rooting a tree , or marking a vertex , reduces the symmetry factor @xmath48 since automorphisms are now constraint to leave the root , or the marked vertex , invariant . .,width=340 ] the function @xmath49 obeys an integral equation that can be obtained as follows . let us assume that the root @xmath47 is directly connected to @xmath50 vertices , and sum over the possible numbers @xmath50 . each of these @xmath50 vertices is the root of a new tree , and we deduce the relation , @xmath51 graphically represented on figure . the symmetry factor @xmath52 takes into account the possibility of permuting the @xmath50 vertices . performing the summation , and taking the logarithm , we obtain the integral equation satisfied by @xmath53 , @xmath54 .,width=188 ] it remains to relate the free energy to the generating function @xmath53 . this is done using the following formula due to b. basso , a. sever and p. vieira @xcite , @xmath55 it is easy to see that both terms in the rhs will produce a sum over clusters weighted by the same integrals as in [ sum_trees ] , but with different symmetry factors . a combinatorial proof of this formula is given in appendix . it is useful to reformulate the previous expression [ basso ] of the free energy at first order as the value of an effective action @xmath56 $ ] at its extremum @xmath57 , @xmath58,\quad \text{such that}\quad \left.\dfrac{\d \sgc}{\d y_0}\right|_{y_0=y_0^\ast}=0,\ ] ] this action is obtained after introducing the integral equation [ int_equ ] into [ basso ] , @xmath59=\hf\int_{\mathbb{r}^2}{\dfrac{dx}{2i\pi}\dfrac{dy}{2i\pi}y_0(x)y_0(y)f(x - y)}-\int_\mathbb{r}{\dfrac{dx}{2i\pi}y_0(x)\left[\log y_0(x)-1\right]}+\int_\mathbb{r}{\dfrac{dx}{2i\pi}y_0(x)\log\left(q q(x)\right)}.\ ] ] it is remarkable that the saddle point equation derived from this action is nothing else than the integral equation [ int_equ ] . it is also worth noticing that when instanton clustering phenomenon is taken into account , one arrive at a similar expression , with logarithms replaced by dilogarithms in the second term . for instance , the effective action derived by nekrasov and shatashvili to describe @xmath0 sym reads @xmath60=\hf\int{\rho(x)\rho(y)g(x - y)}+\int{\left[\text{li}_2(1-e^{-\rho(x)})-\rho(x)\log(1-e^{-\rho(x)})\right]dx}+\int{\rho(x)\log\left(qq(x)\right)dx},\ ] ] where we used the notations of @xcite . expanding the middle term in @xmath61 , we recover at the order @xmath62 the action [ eff_action ] obtained previously . this type of cut - off term for the action has been studied in @xcite . [ [ subleading - order ] ] subleading order + + + + + + + + + + + + + + + + we now focus on @xmath63 , the second order term in the @xmath11-expansion of the free energy at fixed @xmath64 , @xmath65 at this order , clusters that contribute have @xmath22 links for @xmath22 vertices , which means that they have exactly one cycle . such clusters will be denoted @xmath66 . the relevant terms of the mayer expansion for the free energy are @xmath67 the @xmath64-expansion starts here at @xmath68 since at least three vertices are needed to form a cycle . as we go to higher orders in @xmath11 , more vertices will be needed to form the cycles , leading to a higher first order term in @xmath64 . thus , @xmath44 fully determines the @xmath64-expansion of the free energy up to order @xmath69 , and @xmath70 up to @xmath71 . to evaluate the summation over clusters @xmath66 , we first consider the clusters depicted on figure , and for which all vertices belong to the cycle . such clusters have a symmetry factor of @xmath72 due to the invariance under @xmath22 rotations , and a reflexion symmetry . their contribution writes @xmath73 where indices are taken modulo @xmath22 . all the clusters of type @xmath66 may be obtained by dressing the vertices of a pure cycle cluster by appropriate trees . summing over the dressing possibilities boils down to replace @xmath74 in the formula [ contrib_cycle ] by the tree - level dressed vertex @xmath75 . the expression for the free energy correction follows , @xmath76 this is actually the expansion of the logarithm of a fredholm determinant where the first two terms are missing . taking the exponential , we find @xmath77}}\exp\left(-\dfrac14\int{\dfrac{dx}{2i\pi}\dfrac{dy}{2i\pi}y_0(x)y_0(y)f(x - y)^2}\right).\ ] ] the two missing terms correspond to a tadpole ( a vertex with a link looping back to it ) and two vertices doubly connected . the action [ eff_action ] obtained above describes a dyson gas of particles with the non - singular interaction @xmath8 at @xmath78 @xcite . it is also the effective action of a collective field theory for a generalized matrix model at large @xmath17 @xcite . we will show here that the corresponding matrix model is simply the canonical model @xmath79 defined in the introduction ( up to minor corrections ) . the fact that grand canonical and canonical models share the same effective action further leads to relate the rooted vertex generating function at tree level @xmath53 with the collective field at large @xmath17 . at first order in @xmath11 , the canonical partition function is equivalent to @xmath80 the collective field is by definition a generating function of invariants under the permutation of eigenvalues . it is convenient to use the eigenvalue density , @xmath81 that has been normalized to one . it is usual for matrix models to assume that in the large @xmath17 limit , eigenvalues condense into a finite union of connected sets , typically a union of intervals for hermitian matrices . this set @xmath82 is the support of a continuous eigenvalue density @xmath83 obtained as the large @xmath17 limit of the finite densities defined in [ def_rho0 ] . depending on the explicit form of potential and interaction , this assumption might not be valid . we will nonetheless work in this framework , the results derived following this approach being consistent with those obtained on the grand canonical model . in the collective field theory approach , the canonical free energy is given at first order by the extrema of an effective action @xmath84 , @xmath85},\quad \p_x\left.\dfrac{\d \sc}{\d\rho_0(x)}\right|_{\rho_0=\rho_0^\ast}=0.\ ] ] the factor @xmath86 has been introduced here to facilitate later comparison with the previous subsection . the canonical action is a sum of three terms , @xmath87=\hf\e n\int{f(x - y)\rho_0(x)\rho_0(y)dxdy}+\int{\rho_0(x)\log\left(\dfrac{q(x)}{2i\pi}\right)dx}-\int{\rho_0(x)\left[\log\rho_0(x)-1\right]dx}.\ ] ] the derivation of the first two terms is rather straightforward since it is sufficient to write down the integrand of [ zc_order1 ] in an exponential form , and replace the sum over eigenvalues by integrals of the density . the third term corresponds to the entropic term introduced by dyson in @xcite . it is a gibbs factor , coming from the fact that the coulomb gas charges are indistinguishable . following @xcite , it is re - derived in the appendix as a jacobian in the change of measure from the discrete set of variables @xmath88 to the functional integral over @xmath89 $ ] . in the case of hermitian matrix models , such entropic factors cancel with the energetic term coming from the regularization of the kernel at coinciding eigenvalues . however , here @xmath90 is finite at @xmath9 and cancellation does not occur . comparing [ def_smm ] and [ eff_action ] , we deduce that the effective actions are equivalent , upon the identification of @xmath49 with the density @xmath91 , and provided we set @xmath92 . however , by definition the density @xmath83 is normalized to one , and this identification would require @xmath53 to have also a unite norm . to resolve this issue , we introduce the norm @xmath93 of @xmath53 and identify as follows , @xmath94 this identification requires to set @xmath95 in the limit @xmath18 and @xmath16 . this relation signifies that the summation [ def_cz ] defining the grand canonical model is dominated at @xmath16 by the term with @xmath96 variables . similarly , the nekrasov partition function expressed as a sum over young tableaux is dominated in the seiberg - witten limit @xmath43 by a partition with @xmath97 boxes @xcite . it also justifies the approach of @xcite to the study of the ns limit . under the previous identification between the dressed vertex @xmath53 and the density @xmath83 , canonical and grand canonical actions are related through @xmath98=\dfrac1\a\sgc[y_0]+\log(\a / q).\ ] ] the term proportional to @xmath99 is missing from the action [ def_smm ] , but it can be introduced by hand , exploiting the fact that the density @xmath83 is normalized to one . in this case , @xmath99 plays the role of a lagrange multiplier imposing the unit norm . the two actions @xmath84 and @xmath100 produce equivalent equations of motions , and the free energies satisfy at first order @xmath101 since @xmath102 depends on @xmath17 but not on @xmath64 , and the opposite for @xmath103 , this relation only holds for a specific value of @xmath104 or @xmath105 . more comments on this will follow in the next subsection where this relation is re - derived by exploiting the fact that @xmath106 is the discrete laplace transform of @xmath79 . [ [ one - loop - determinant ] ] one loop determinant + + + + + + + + + + + + + + + + + + + + the subleading , or genus one , correction to the free energy can also be computed in the framework of the collective field theory . there are two types of corrections . the first one corresponds to amend the canonical action by a subleading term @xmath107 , and the second type to the gaussian fluctuations around the saddle point . the modification of the action is due to an earlier kernel approximation that should now be refined . indeed , at the second order in @xmath11 the approximation [ zc_order1 ] of @xmath79 is no longer acceptable and must be replaced by @xmath108 the correction to the kernel is responsible for an additional contribution to the canonical action , @xmath109 which reproduces the term in the second exponential of the expression [ f1 ] for @xmath110 , provided we set again @xmath111 . the first exponential corresponds to the factor in front of the integrals in [ def_zn_ii ] , and comes from the diagonal part of the kernel . it is well known that the integration of gaussian fluctuations around the classical solution produces the inverse square root of ( minus ) the hessian matrix determinant , @xmath112=e^{-\int{\log \rho_0(x)dx}}\det\left[\d(x - y)-\e n\rho_0(x)f(x - y)\right].\ ] ] the prefactor involving the integral of @xmath113 cancels with the sub - leading order of the entropic term computed in appendix , formula [ sub_lead ] . the remaining determinant reproduces the one which appears in [ f1 ] , upon the identification @xmath114 and @xmath111 . gathering all contributions , we find @xmath115}}.\ ] ] comparing with [ f1 ] , we conclude that the sub - leading contributions to the free energy of both models are equal . again , this equality holds only for a specific value of @xmath105 or @xmath104 . the observed relations between free energies at first orders originate in the discrete laplace transform , also called @xmath116-transform , performed in [ def_cz ] to define the grand canonical model . this transformation can be inverted by considering a contour integral over @xmath117 circling the origin , @xmath118 in the large @xmath17 limit , it is possible to evaluate the integral using a saddle point technique @xcite , and the relation between grand canonical and canonical free energies is a simple legendre transform , @xmath119 on the lhs , the additional terms are due to the factor @xmath120 and can be absorbed in the definition of @xmath121 . under this transformation , the number of particles @xmath17 and the chemical potential @xmath122 are conjugate variables . they are related through the saddle point equation , @xmath123 this equation can be solved in terms of @xmath104 and [ legendre_fe ] provides the grand canonical free energy knowing the canonical one . inverting the legendre transform , @xmath102 can be derived from @xmath103 with @xmath124 . in the previous considerations , @xmath11 was a simple spectator . the novelty in these notes is to tune the parameter @xmath11 toward zero as the number of particles is sent to infinity , keeping @xmath111 fixed . the legendre transformation [ legendre_fe ] survives this limit and produces the relation [ legendre ] between first orders free energies . in this limit , the conjugate variables are @xmath93 and @xmath125 . it is also interesting to note that the saddle point equation [ saddle ] gives the normalization condition for the dressed vertex @xmath126 , @xmath127 the dressed vertex is the generating function of connected rooted clusters . its expression is given by [ def_y0 ] after replacing the summation over rooted tree by a general summation over rooted clusters @xmath128 with appropriate @xmath11 factors . the first equality in [ qdq ] is shown in the appendix using the mayer expansion [ sum_clusters ] of the free energy . it is the equivalent of the matone relation for susy gauge theories @xcite . the normalization condition expands in @xmath11 , providing refined approximations for the saddle point @xmath129 . in order to investigate the subleading orders , we need to introduce some notation for the large @xmath17 expansion of the canonical model at @xmath130 with @xmath93 fixed , @xmath131 let us emphasize that this expansion is different from the standard topological expansion at fixed @xmath11 . it is the reason why the one - loop term in [ zc_1 ] is not only given by the determinant but also contains corrective terms to the action . the inverse discrete laplace transform [ z_trans ] with the constraint @xmath111 specializes to @xmath132}}.\ ] ] at sub - leading order , this integral is approximately equal to @xmath133}+o(1/n),\ ] ] with @xmath134 solution of the normalization condition [ qdq ] , and @xmath135\right|_{q = q^\ast}.\ ] ] this quantity @xmath136 can be expressed in terms of the norm @xmath137 of the two points grand canonical density @xmath138 , defined in [ def_gc_dens ] , using a formula derived in appendix , @xmath139 at the saddle point , we have @xmath140 . expanding [ ztrans_2nd_order ] in @xmath11 , we obtain at second order the following relation between free energies , @xmath141 to retrieve the equality previously observed among the free energies at subleading order , we have to assume that the norm @xmath142 of @xmath143 is equal to @xmath93 at the saddle point . it implies that the tree - level propagator @xmath144 , which is the generating function of bi - rooted trees , has a vanishing norm ( see [ brho_y_2pts ] below ) . it is however possible that we missed a factor in our treatment of the canonical partition function , in particular when we discarded the zero - mode in appendix . this is why we will remain cautious and keep the critical value of @xmath142 arbitrary in the following . [ [ density - and - dressed - vertex ] ] density and dressed vertex + + + + + + + + + + + + + + + + + + + + + + + + + + the comparison of the effective actions led us to propose an identification between the tree - level dressed vertex of the grand canonical model and the large @xmath17 eigenvalue density associated to the canonical model . this identification can also be derived by general considerations involving the discrete laplace transform . it will be done here in two steps . first we have to relate the dressed vertex @xmath126 to a grand canonical density @xmath145 . then , we will exploit the inverse laplace transformation to deduce an equality between canonical and grand canonical densities at first order . to complete our program , we need to define the grand canonical vev of an operator @xmath146 , @xmath147 it is expressed in terms of the canonical vevs , @xmath148 where the operator depends on @xmath17 fields in a permutation invariant manner . we focus on the density operator , and consider the sourced partition function @xmath149=\zgc\la\exp\left(\int{dx j(x)\cd(x)}\right)\ra,\quad \cd(x)=\sum_i\d(x-\phi_i).\ ] ] the grand canonical density is defined as @xmath150}{\d j(x)}\right|_{j=0}.\ ] ] introducing the source term @xmath151 in the partition function corresponds to replace the potential by @xmath152 , as can be seen from @xmath149=\zgc\la\prod_ie^{j(\phi_i)}\ra.\ ] ] thus , the mayer expansion also applies to the sourced quantity , leading to [ sum_clusters ] with @xmath153 inserted into the product over vertices . from this expression , we compute the derivative @xmath154}{\d j(x)}\right|_{j=0}=\sum_{l=0}^\infty{\bq^l\sum_{c_l}\dfrac{l}{\s(c_l)}\dfrac{q(x)}{2i\pi}\int{\prod_{i\in v(c_l)\setminus\{x\}}q(\phi_i)\dfrac{d\phi_i}{2i\pi}\prod_{<ij>\in e(c_l)}\e f(\phi_{ij})}}.\ ] ] the identity [ sum_sym ] demonstrated in appendix allows to replace the clusters summation by a summation over rooted clusters . in doing so , we obtain exactly the dressed vertex @xmath154}{\d j(x)}\right|_{j=0}=\dfrac1{2i\pi\e } y(x),\ ] ] and we deduce from [ def_brho_ii ] the identity @xmath155 at the level of one marked point . it is easy to check that this identification is compatible with the property [ qdq ] by computing the norm of the density @xmath145 from the definition [ def_brho_ii ] . it remains to take the laplace transform . the canonical density is by definition @xmath156 it reduces to the collective field at large @xmath17 , @xmath157 . hence , @xmath158 is related to @xmath159 by a discrete transformation similar to [ def_cz ] . using a saddle point technique , we find at subleading order the relation @xmath160 , in agreement with the proposed identification between @xmath53 and @xmath83 . it is also possible to derive this relation considering the inverse laplace transform of the sourced partition function . it implies a legendre relation of the type [ legendre_fe ] among sourced free energies . taking the functional derivative with respect to the source @xmath151 , we recover the relation between one - point density . one has to be careful because the saddle point depends on the source . but , contrary to the case of 2-points densities treated below , the dependence vanishes here . [ [ higher - point - densities - and - cluster - generating - functions ] ] higher point densities and cluster generating functions + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the previous argument generalizes to a higher number of marked vertices and multi - points densities . the two points grand canonical density defined as the connected correlator as @xmath161 , we have the factorization property @xcite , @xmath162 it ensures that @xmath163 . ] @xmath164}{\d j(x)\d j(y)}\right|_{j=0}=\e\la\cd(x)\cd(y)\ra_c,\ ] ] relates to the full propagator @xmath165 , generating function of bi - rooted trees , as @xmath166 this identity is obtained by taking the second derivative of the sourced free energy @xmath167 $ ] . the additional term in the rhs of this relation corresponds to coinciding eigenvalues in the decomposition @xmath168 the second term in the rhs produces a delta function of @xmath169 times the one - point density , and @xmath165 corresponds to the non - diagonal terms . the second part of the argument exploits the fact that the sourced partition functions are also related through a discrete laplace transform . but now the saddle point @xmath170 depends on the source @xmath151 , for instance @xmath171 at leading order , the sourced free energies satisfy the equation [ legendre ] . taking twice the derivative with respect to the source , we obtain the relation between two points densities at first order , @xmath172 where @xmath173 is the leading order of the canonical two points connected density @xmath174}{\d j(x)\d j(y)}\right|_{j=0}=\dfrac1n\lan\cd(x)\cd(y)\ran_c.\ ] ] the second term in [ rel_2pts ] is due to the dependence of the saddle point in the source . this expression is compatible with the requirement of vanishing norm for the connected correlator @xmath173 . at higher points , we expect relations similar to [ brho_y_2pts ] to hold between multi - rooted clusters generating functions and grand canonical densities . they can be derived by performing higher derivations of the free energy with respect to the source . on the other hand , the relation between canonical and grand canonical densities becomes increasingly complicated and can not be worked out easily using this method , even at the planar order . in the previous section , we have compared canonical and grand canonical models at the level of free energies . we have shown how to recover the collective field theory description of the canonical model from the cluster expansion of the grand canonical partition function . this comparison was restricted to the two first orders in large @xmath17 and small @xmath11 . on the canonical side , it is possible compute higher order terms by employing the recursive technique of loop equations . this technique , originally developed for matrix models , has recently been extended to a large class of models to which @xmath79 belongs @xcite . and @xmath175 in their notations . for this value of @xmath176 , and depending on the explicit expression for the kernel , some of the assumptions considered in their paper may not be satisfied . this would have to be checked case by case . ] our goal in this section is to map these loop equations to similar relations among objects pertaining to the cluster expansion . these objects are the @xmath137-points @xmath177-functions , generating functions of @xmath137-rooted clusters . we have already encountered the cases @xmath178 and @xmath179 , corresponding respectively to the dressed vertex @xmath126 and the propagator @xmath165 . loop equations for the canonical densities are obtained in the following manner . first , the invariance of the measure allows to write a set of linear relations among ( non - connected ) correlators . these correlators are decomposed into connected parts . the connected correlators involved are resolvents , i.e. multiple cauchy transforms of the densities . as such , they have a branch cut along the support @xmath82 of the densities in each of their variables . taking the discontinuities of the previous equations , we are able to derive a set of coupled integral equations among densities . these equations can be expanded in large @xmath17 , and solved recursively . the recursion involves both the genus , that is the order in @xmath180 and the number of points . at this level , loop equations also depends on the derivative of densities . they can be integrated with a little bit of algebra . the resulting primitive equations no longer contain the densities derivative . in the process , a constant of integration appears . it is fixed by imposing a vanishing norm to the @xmath137-points densities with @xmath181 . grand - canonical densities also obey the canonical loop equations . indeed , those equations are linear in the ( non - connected ) canonical correlators , and valid for any @xmath17 . they can be summed over @xmath17 with appropriate coefficients to produce equations among grand canonical correlators . next , these correlators are decomposed into connected parts . we must emphasize that the connected grand canonical correlators are no - longer the discrete laplace transform of canonical ones . these connected correlators are also the resolvents associated to the multi - points grand canonical densities . for @xmath11 infinitesimal , these densities are assumed to be continuous on a connected support , just like the canonical ones . the discontinuity process still works , leading to the same derivative loop equations . densities are then expanded in @xmath11 , which plays a role equivalent to the the large @xmath17 topological expansion for the canonical model . after integration , we recover the same integral equations , but with different constant of integrations since grand canonical and canonical densities have a different norm . in this section , the strategy is as follows . we first provide the derivation of the canonical loop equation , and re - write them in the integrated form . then , we compare this equation with a relation among @xmath177-functions derived using the mayer expansion . we deduce from the relation between @xmath177 and @xmath182 that this density obey the integrated canonical loop equation . we conclude that the @xmath177-function relations are the equivalent of loop equations . finally , a technique to derive the loop equation for grand canonical densities is presented in subsection . the simplest loop equation is derived from the identity @xmath183}.\ ] ] it produces an equation satisfied by the resolvent @xmath184 which is the cauchy transform of the density @xmath185 , @xmath186 we will also need to introduce an auxiliary quantity @xmath187 defined as @xmath188 with @xmath189 the standard matrix model potential . then , the first loop equation takes the form @xmath190 with the shortcut notation @xmath191 for the logarithmic derivative of the kernel . we have assumed that the eigenvalues condense on the support @xmath82 of @xmath61 in the large @xmath17 limit . it implies that @xmath184 has a branch cut on @xmath82 , with a discontinuity given by @xmath192 . on the other hand , by construction @xmath187 is not singular over @xmath82 . thus , taking the discontinuity of the loop equation [ 1st_leq ] over @xmath82 allows to eliminate @xmath187 and write an equation involving only densities , @xmath193 the next step is to expand the density at large @xmath17 , we denote @xmath194 the term of order @xmath195 . at the first order , the dependence in the two points density drops , and we find an integral equation for @xmath83 which is precisely the equation of motion [ eom_sc ] derived from the canonical action @xmath84 . integrating once , we recover the integral equation [ int_equ ] obeyed by @xmath53 , @xmath196 provided we choose the integration constant @xmath197 to be @xmath99 . a priori , the unit norm constraint over the density should fix this integration constant . however , it is very non - trivial to impose this condition in practice due to the complicated form of the integral equation . at the saddle point , @xmath198 and @xmath199 . the subleading order of the first loop equation contains the two points density at first order @xmath173 . to compute this quantity , we need a second loop equation , derived from the identity @xmath200}.\ ] ] it provides an equation satisfied by the two points resolvent @xmath201 and involving an auxiliary quantity @xmath202 , @xmath203 this equation also involves the three - points density , but this dependence drops at leading order . the first loop equation [ 1st_leq ] can be used to simplify the result , which gives @xmath204dx_1dx_2dx_3}. \end{split}\end{aligned}\ ] ] just like @xmath187 , @xmath202 has not branch on @xmath82 for its variable @xmath205 , and will be eliminated by taking the discontinuity of the equation . in this process , the difference of resolvents must be regularized at coincident values as follows , regular on the branch cut @xmath82 , and consider @xmath206 since @xmath47 belongs to the integration contour . ] @xmath207 taking the discontinuity over the variable @xmath205 and @xmath208 , extracting the first order and integrating once , we get @xmath209 with the integration constant @xmath210 that may depend on @xmath211 . this degree of freedom is fixed by imposing that @xmath173 is a symmetric function of its parameters , and has zero norm since it is a connected density . we would like to recover the loop equation [ equ_r0_xy ] using the mayer expansion . according to our previous discussion in subsection , this equation should be obeyed by the tree - level propagator @xmath144 , generating function of bi - rooted tree . let us recall its definition , @xmath212 where @xmath213 is the generating function of bi - rooted trees such that the roots are connected through a chain of @xmath22 intermediate vertices ( and @xmath214 links ) . we should also supply the definition of the first member of this set of functions , @xmath215 , obtained when the roots are directly connected . if we merge the two roots in order to build a cycle , @xmath216 ] contrary to the functions @xmath217 , @xmath144 is a symmetric function of @xmath47 and @xmath211 . the functions @xmath218 obey an obvious recursion relation that is interpreted as attaching to the vertex @xmath47 a new rooted vertex @xmath205 , @xmath219 in this process , @xmath47 is still marked in the sense that it is determined uniquely being the first vertex attached to @xmath205 on the path to @xmath211 , but we will not consider it as a root anymore , preferring the endpoint @xmath205 . summing over @xmath22 , we deduce the integral equation obeyed by @xmath144 , @xmath220 making use of the relation [ brho_y_2pts ] between the grand canonical two points density @xmath138 and the propagator @xmath165 , we deduce that at first order in @xmath11 , @xmath221 obey the integrated loop equation [ equ_r0_xy ] with a vanishing integration constant @xmath222 , @xmath223 the relation [ rel_2pts ] between @xmath173 and @xmath143 is compatible with the loop equations [ equ_r0_xy ] and [ equ_br0_xy ] . this can be shown by taking the @xmath64-derivative of the equation [ int_equ ] satisfied by @xmath224 , @xmath225 the second equality is a consequence of [ b7 ] and [ brho_y_2pts ] . we deduce the expression of @xmath210 at the saddle point , @xmath226 to obtain the equation satisfied by the genus one correction to the 1-point density @xmath227 , we examine the first loop equation [ leq1 ] at subleading order . again , the result can be simplified using the first order result [ leq1_0 ] , and integrated , leading to @xmath228 where @xmath229 contains the contribution of the two points density , @xmath230 the equation [ equ_r0_xy ] obtained upon the 2-points density can be used to simplify the expression [ s_d ] and integrate it , leading to the following loop equation for @xmath227 , has originally be obtained in the form @xmath231 integrating this expression multiplied by @xmath232 over @xmath211 , and then using the primitive relation [ equ_r0_xy ] to simplify the result , we obtain the identity @xmath233dy},\ ] ] since @xmath234 . this identity is then plugged into the expression [ s_d ] of @xmath235 . ] @xmath236dy}+\g_1,\ ] ] where the integration constant @xmath237 is fixed by a normalization condition . the integrated loop equation [ leq1_iii ] we have obtained for @xmath227 should be compared to the equation satisfied by @xmath238 , the generating function of rooted clusters with exactly one cycle , @xmath239 in this expression , represented graphically on figure , the first term corresponds to the case where @xmath47 does not belong to the cycle . hence , there is a vertex @xmath211 , directly linked to @xmath47 such that if we remove this link , the cycle is present in the cluster rooted by @xmath211 . in the second term , @xmath47 directly belongs to the cycle . in this case , we choose a vertex @xmath211 from the cycle and directly connected to @xmath47 . cutting the link @xmath169 , we obtain a bi - rooted tree . in the process , we gain a symmetry factor @xmath240 due to the choice of @xmath211 . finally , the third term correspond to trees of @xmath144 for which @xmath47 and @xmath211 are directly related . for those clusters , @xmath47 and @xmath211 can not get an extra link , and their contribution must be withdrawn from the previous term . comparing [ leq1_iii ] and [ equ_y1 ] , we deduce that @xmath241 satisfies the loop equation [ leq1_iii ] with vanishing integration constants @xmath242 . another possibility to establish grand canonical loop equations is to start from the definition of the density @xmath145 and make use of the @xmath243-function to fix one of the integration variables . the canonical correlator of @xmath17 variables reduces to a correlator of @xmath244 variables , and after summation over @xmath17 we obtain an equation satisfied by the grand canonical density , @xmath245 at first order in @xmath11 , it is possible to use the factorization property @xmath246 the sum over @xmath26 can be replaced by an integral over density @xmath185 . expanding [ gc_leq1 ] in @xmath11 , and keeping only the first order , we deduce that @xmath247 satisfies the integral equation [ int_equ ] with @xmath248 . at the subleading order in @xmath11 , the factorization property becomes @xmath249.\ ] ] in the rhs bracket , the first term comes from the expansion of @xmath250 , the last term is the first order correction to the factorization property . replacing sum over variables @xmath26 with densities , and expanding in @xmath11 , we get at the second order an equation satisfied by the subleading correction to @xmath145 , @xmath251 where we have used the first order equation to simplify the result . using the equation [ equ_br0_xy ] to deal with the last term , this loop equation reproduces the equation [ equ_y1 ] obtained from the cluster expansion upon the identification @xmath155 and [ brho_y_2pts ] of the densities . the same argument can be repeated for the two points density . fixing an integration variable using the operator @xmath252 in the two - point correlator , we find @xmath253 where the result has been simplified using the first loop equation [ gc_leq1 ] . from the factorization property @xmath254 we recover at subleading order in @xmath11 the equation [ equ_br0_xy ] satisfied by the two points density @xmath143 . in these notes we compared the cluster expansion of a grand canonical model with the standard matrix model treatment of its canonical partition function . at tree level , the grand canonical free energy is given by the minimum of an effective action which is identical to the one provided by the collective field theory approach applied to the canonical model . the correspondence extends to the level of one - loop corrections , where the sum over one - cycle clusters reproduces the expansion of the fredholm determinant computed from the integration over gaussian fluctuations in the collective field theory . the matching of free energies can be explained by the discrete laplace transform relating canonical and grand canonical models . introducing a source term , we were also able to find the relation satisfied by the one - point and two points densities . we continued with the study of canonical loop equations , and realized that a similar set of equations can be derived from the cluster expansions . instead of @xmath137-point connected densities , these equations involve the generating functions of @xmath137-rooted clusters , denoted @xmath177 . using these equations , and the relation between @xmath177-functions and densities derived earlier , we verified that grand canonical densities also obey the canonical loop equations . it implies that the whole loop equation structure is present in the cluster expansion , and takes the form of graphical relations among clusters . finally , we proposed a method to derive directly this set of loop equations within the grand canonical model . our study is restricted to the first two orders in the large @xmath17 and small @xmath11 expansion . the generalization to higher orders still needs to be done , and the general form of loop equations to be worked out . once the full set of equations identified , it may be possible to apply the topological recursion to the grand canonical model . another important point that remains is the description of instantons clustering relevant to the instanton partition function of @xmath0 susy gauge theories . the dual description grand - canonical / canonical may allow a better understanding of this phenomenon . a possible application for this work could be the derivation of the subleading correction in @xmath15 to the partition function , and the investigation of its presumed integrable properties . in this scope , it is tempting to assume that the instanton clustering in sym is entirely described by the effective action [ action_ns ] , and conjecture that the subleading order is given by the associated determinant , @xmath255}}e^{\frac12\int{dx\log\left(e^{\rho(x)}-1\right)}-\frac12\int{dx\log\rho(x)}}\\ & \times\exp\left(\hf g(0)\int{dx(1-e^{\rho(x)})}-\dfrac14\int{dxdy(1-e^{\rho(x)})(1-e^{\rho(y)})g(x - y)^2}\right ) . \end{split}\end{aligned}\ ] ] this proposal is very naive , but it could be tested using the agt correspondence with the representation of liouville correlators . the grand canonical model we studied has a very specific form of interaction and may not be relevant to statistical systems . it would be interesting to consider more physical models . one may also wonder if the topological reduction employed in this context have a matrix model analogue . nevertheless , the results presented here are very general and could be relevant for a large spectrum of problems . they have deep connections with integrable models and the tba equation @xcite . they play a role in the computation of light - like wilson loops at strong coupling in @xmath256 sym @xcite . they may also be applied to the study of 3-points function of scalar operators in this theory @xcite . i would like to thank dima volin , yutaka matsuo and ivan kostov for valuable discussions , and in particular benjamin basso for sharing his unpublished results . it is also a pleasure to acknowledge the hospitality of ewha university at the occasion of the workshop `` solving ads / cft '' , and of cquest ( sogang u. ) and tokyo university where parts of this work has been done . i acknowledge the korea ministry of education , science and technology ( mest ) for the support of the young scientist training program at the asia pacific center for theoretical physics ( apctp ) . in this appendix , we give a demonstration for the formula [ basso ] where the sum over tree clusters @xmath37 is given by the difference of the two terms , @xmath257 it is easy to see that both terms expand as a sum of tree clusters @xmath37 , weighted as in [ sum_trees ] , but with different symmetry factors . the second term @xmath258 contains trees with at least one link , associated to the function @xmath8 present in the integral [ def_ab ] , which we call the main link . such clusters are formed by gluing two trees along this main link . to all the terms in the expansion of @xmath259 and @xmath258 correspond a term of the summation [ sum_trees ] . on the other hand , a single term in [ sum_trees ] corresponds to many terms of @xmath259 and @xmath258 series , since a tree @xmath37 can be rooted from any of its vertices , leading to @xmath22 terms in @xmath259 , and has @xmath38 links that can be associated to the main link of a @xmath258-term . the fact that the formula [ basso ] holds has to do with the property that a tree @xmath37 has exactly @xmath22 links for @xmath38 vertices . the strategy we follow is to consider the terms in @xmath259 and @xmath258 cluster expansions that can be identified with a given cluster @xmath37 of the summation [ sum_trees ] . the corresponding terms in @xmath259 are obtained by rooting the vertices of the cluster @xmath37 . in the same way , terms from the @xmath258-expansion are derived from edging the links of the tree @xmath37 . we then show that cancellation occurs between terms of @xmath259 and @xmath258 expansions due to the coincidence of symmetry factors . the remaining term provides the contribution of @xmath37 to the free energy with the correct symmetry factor . it is better to understand what is going on over a few examples . here we focus on linear trees , i.e. chains of vertices , denoted @xmath260 . to facilitate the argument , vertices will be numbered according to their order in the chain , from left to right . we further call @xmath25th link the edge linking the vertices @xmath25 and @xmath261 . these chains appear in the free energy expression [ sum_trees ] with a factor @xmath262 corresponding to a reflexion symmetry . we start with the case of a chain with odd length , @xmath263 . in the @xmath259-expansion , @xmath260 is associated to @xmath264 terms , corresponding to rooting the vertex @xmath25 ( or @xmath265 ) for @xmath266 and the vertex @xmath264 . this procedure is depicted in figure ( left ) . the index @xmath25 runs from one to @xmath267 since rooting the vertex @xmath25 or @xmath265 leads to the same tree . the first @xmath267th rooted trees obtained in this way have a symmetry factor of one since they are composed of two branches of length @xmath268 and @xmath265 joining at the root . the remaining rooted tree , associated to the central vertex , has the symmetry factor @xmath269 since now both branches have the same length and can thus be exchanged . vertices.,width=453 ] now , let us turn to the @xmath258-expansion . we consider the @xmath258 term having the @xmath25th link of @xmath260 as its main link . this @xmath258-term corresponds to a rooted chain of @xmath25 vertices glued to another rooted chain of @xmath265 vertices through the main link . it is represented on figure ( right ) . its symmetry factor is one since the two trees on both sides of the main link have symmetry factor one and different length . we also have to take into account a chirality factor of two counting the possibility of exchanging the two trees . this factor two is canceled by the factor @xmath240 in the definition of @xmath258 . here again we choose @xmath25 running only from one to @xmath267 since we get the same term after exchanging @xmath270 . since the first @xmath267th @xmath259-terms are canceled by the @xmath258-terms , it remains only the contribution from the central vertex . as already mentioned , this contribution is weighted by @xmath240 , thus providing the correct symmetry factor for the free energy cluster . there is a lesson to learn from this example . as we will see later , it is possible to associate uniquely a link to each vertex but one by a recursive procedure . the corresponding terms in the @xmath259- and @xmath258-expansions cancel , and only the last vertex contributes . the rooted tree of this vertex has the same symmetry factor than the original cluster . however a subtlety may appear . it is illustrated by our second example , the case of a chain with even length @xmath271 . rooting the tree @xmath260 from the vertex @xmath25 ( or @xmath272 ) with @xmath266 , we obtain a tree with two branches of length @xmath268 and @xmath273 . these rooted trees have symmetry factor equal to one . then , we consider the @xmath25-th edges with @xmath274 . they give @xmath258-terms consisting of two rooted chain , of length @xmath25 and @xmath272 , glued through the main link . the symmetry factor is one , and there is a chirality factor of two , again eliminated by the factor @xmath240 . on the other hand , the central link , numbered @xmath267 , is associated to two rooted trees of the same length @xmath267 , and in such a case there is no chirality since right and left side of the @xmath258-terms are the same . we note that the @xmath275th first @xmath259-terms are eliminated by the chiral @xmath258-terms , and the last @xmath259-term gets subtracted by a half of its value , which corresponds to the @xmath258-term having the central edge as its main link . the difference @xmath276 thus reproduces the symmetry factor @xmath269 needed for the free energy . we now consider an arbitrary tree @xmath37 and associate recursively to each vertex a unique link , to which it connects directly , using the following procedure . first , the leaves , i.e. the vertices connected to only one other vertex , are naturally associated to the only link that end on them . then , we remove those leaves in order to obtain a strictly smaller tree to which we repeat the procedure . at the end of the recursion , only two configurations may arise . in the first case , as for the odd chain , only a single vertex remains , all others are uniquely associated to a link . in the second case , typically for the even chain , a set of two vertices connected by a link remains . in this configuration , there is an ambiguity in the choice of the vertex to associate to the remaining link . as a second step , we argue that the @xmath259-term of a rooted vertex and the @xmath258-term of the cluster edged from the associated link cancel . to do so , we have to show that the rooted tree have the same symmetry factor than the tree with associated edge as main link . let us take the @xmath258-term associated to an edge of the cluster @xmath37 which is not the final stage of the previous procedure . it consists of two rooted tree @xmath277 and @xmath278 , with symmetry factors @xmath279 and @xmath280 , such that the total factor is @xmath281 . we will see that these two composing trees are always different , so there is no symmetry enhancement . this object is chiral , which cancel the factor @xmath240 in front of the @xmath258 integral . it is associated to a rooted tree in the @xmath259-series which is displayed in figure . by construction , the rooted vertex @xmath47 is an ending point of the main edge ( in yellow on the figure ) . it is attached to the tree with the smallest deepness , i.e the smallest maximal distance between the leaves and the root . corresponding to the step of the recurrence at which vertex and link have been associated . then , in the tree rooted at @xmath47 , the root is attached to the main edge , and to a tree of deepness @xmath282 . on the other side of the main edge lies a tree with strictly larger deepness . ] the trees @xmath283 and @xmath284 can not have the same deepness , otherwise the procedure of associating the vertex to an edge would not be unique . the main edge still increases the deepness of the deep tree on the right by one , and no symmetry enhancement can occur . thus , the symmetry factor for the @xmath259-term is also @xmath281 . this shows the cancellation between @xmath259 and @xmath258 terms . , and its associated link ( highlighted ) . it consists of two rooted subtrees , @xmath283 by @xmath47 and @xmath284 by @xmath211 , linked by the @xmath258-term main edge.,width=226 ] it remains to study the final stage of the procedure . the simplest case is when only one vertex remains . then the symmetry factor of the rooted tree is equal to the one of the cluster @xmath37 we started from . otherwise there would exist an automorphism exchanging the final vertex of @xmath37 with another one . but this is not possible as the root has been determined uniquely through the procedure described above . finally , we consider the case where one edge and two vertices remain at the end of the recursion . this edge connects two trees of a @xmath258-term with the same deepness , leaving to possibility of the two tree to be identical . if they are not identical , we can choose any of the two vertices to be the one associated with the remaining edge . then , we can repeat the previous argument to show that the corresponding @xmath259 and @xmath258 terms cancel . no symmetry enhancement happens due to the fact that a node is added to one of the trees , thus incrementing its deepness . the rooted tree associated to the last vertex has again the same symmetry factor than the initial cluster . otherwise it could be exchanged with the vertex we removed previously , meaning that the two trees where actually the same . the case where the two trees are identical has already been encountered in the example of the even chain . the @xmath258-term have symmetry factor @xmath285 where @xmath286 is the symmetry factor of the composing trees , and we included the @xmath240 prefactor which is no longer canceled by the chirality . the two vertices related to this edge leads to the same rooted tree , which has the symmetry factor @xmath287 . taking the difference @xmath276 , we get @xmath288 which is exactly the symmetry factor of the initial cluster @xmath37 , the factor of two taking into account the possibility of reflexion with respect to the final edge . in this appendix , we give a proof of the formulas [ qdq ] and [ qdq2 ] . first , examine the action of @xmath289 on the grand canonical free energy expressed as a sum over clusters as in [ sum_clusters ] , @xmath290 the only effect of this operation is to multiply the cluster integrals by the number of vertices @xmath22 . this expression should be compared to the cluster expansion of the integral of @xmath126 . the integral of rooted clusters reproduces the cluster contributions of the free energy expansion , and we should only be concerned about the symmetry factor . to a given cluster @xmath21 of the free energy expansion corresponds @xmath22 rooted clusters @xmath128 obtained by rooting the vertices @xmath291 . however some of these rooted clusters are identical . to avoid over - counting these terms , we separate the set of vertices into the sets @xmath292 of vertices producing equivalent rooted clusters , @xmath293 as already mentioned , the integral contributions of integrated rooted clusters producing the same cluster @xmath21 are equal and can be factorized . thus , in order to prove [ qdq ] we just need to establish @xmath294 now , let us discuss the group of automorphisms of the cluster @xmath21 , denoted @xmath295 . consider one vertex @xmath296 and the group of automorphism for the rooted cluster @xmath128 . it is obvious that this group @xmath297 is a subgroup of @xmath295 consisting of the automorphism of @xmath21 that leave the vertex @xmath47 invariant . those groups are subgroups of the group of permutations for the set vertices @xmath298 and can be decomposed into a product of transpositions . we also need a formal definition of @xmath292 . two vertices @xmath47 and @xmath211 produce an identical rooted cluster if and only if there exists an automorphism of @xmath21 mapping one into the other . suppose we take an element @xmath299 , then this set is the orbit of @xmath300 under the group of automorphisms , @xmath301 note also that the groups of automorphisms for two vertices from @xmath292 are isomorphic , @xmath302 , although they have a different representation on the cluster @xmath21 . let @xmath303 and @xmath291 . there is a unique @xmath267 such that @xmath304 . from the definition of @xmath292 , @xmath305 also belongs to @xmath292 and we denote this element @xmath300 . by construction @xmath306 leaves the vertex @xmath47 invariant since @xmath307 is the transposition that exchanges @xmath47 and @xmath300 . therefore it is an element of @xmath297 that we denote @xmath308 , and we have @xmath309 . it means that given a vertex @xmath47 , any automorphism @xmath310 can be decomposed uniquely into its action on @xmath47 , given by @xmath307 and another automorphism that leaves @xmath47 invariant , namely @xmath308 . with @xmath311 and @xmath312 such that @xmath313 . it implies that @xmath314 in contradiction with the fact that @xmath315 . ] we deduce @xmath316 which implies [ sum_sym ] . the demonstration remains valid if we replace the cluster @xmath21 by a rooted cluster @xmath128 and the rooted clusters @xmath317 with bi - rooted ones @xmath318 , and [ sum_sym ] becomes @xmath319 since we now have only @xmath38 unmarked vertices . we deduce the following relation between rooted and bi - rooted generating functions , @xmath320 which implies [ qdq2 ] . similar formulas can be obtained for a higher number of roots , @xmath321 by recursion , it implies @xmath322 the jacobian from the change of variable @xmath88 to @xmath323 $ ] can be obtained using the faddeev - poppov approach . we consider @xmath324\prod_x\d\left(\rho_0(x)-\dfrac1n\sum_{i=1}^n\d(x-\phi_i)\right)},\ ] ] the product of delta functions can be represented with the help of a ghost field @xmath325 and gives @xmath326e^{i\int{dx\l(x)\rho_0(x)}}\prod_{i=1}^n e^{-\frac{i}{n}\l(\phi_i)}},\ ] ] where @xmath327-factors were included in the measure @xmath328 $ ] . it implies that @xmath329e^{i\int{dx\l(x)\rho_0(x)}}\left(\int{e^{-\frac{i}{n}\l(\phi)}d\phi}\right)^n}.\ ] ] or , introducing an effective action , @xmath329e^{s[\rho_0,\l]}},\quad s[\rho_0,\l]=i\int{dx\l(x)\rho_0(x)}+n\log\left(\int{e^{-\frac{i}{n}\l(x)}dx}\right).\ ] ] at leading order , the integral over the ghost field can be evaluated as a saddle point . the equation of motion implies @xmath330 after replacing @xmath325 in the effective action , we notice that the @xmath331-dependence drops since @xmath83 is normalized to one . we end up with the entropic term @xmath329e^{-n\int{dx\rho_0(x)\log\rho_0(x)}}}.\ ] ] the subleading contribution is equal to the inverse of the square root of ( minus ) the hessian determinant . the hessian matrix evaluated at the saddle point gives @xmath332.\ ] ] the factor of @xmath333 may be absorbed in the integration measure and will be discarded . the remaining determinant is a fredholm determinant that can be computed exactly , @xmath334=e^{\int\log\rho_0(x)dx},\ ] ] in doing so , we removed a zero - mode associated to the unite norm of the density , @xmath335=1-\int{\rho_0(y)dy}.\ ] ] thus , at the second order we have @xmath336e^{-n\int{dx\rho_0(x)\log\rho_0(x)}-\frac12\int{dx\log\rho_0(x)}}},\ ] ] with factors @xmath337 absorbed in the measure of integration ( the factors @xmath338 cancel with the ones coming from the saddle point integration in [ zc_1 ] ) .

these lecture notes provide an introduction to mm / submm extragalactic astronomy , focused on agn studies , with the final goal of preparing students to their future exploitation of the alma capabilities . i first provide an overview of the current results obtained through mm / submm observations of galaxies and agns , both local and at high redshift . then i summarize the main mm / submm facilities that are currently available . alma is then presented with a general description and by providing some details on its observing capabilities . finally , i discuss some of the scientific goals that will be achievable with alma in extragalactic astronomy , and for agn studies in particular . galaxies : active , evolution , formation , high - redshift , nuclei , quasars , seyfert , starburst , millimeter , submillimeter , instrumentation : high angular resolution , interferometers 95.85.fm , 95.85.bh , 98.54.-h , 98.54.cm , 98.54.ep , 98.54.kt

0806.0695

the atacama large millimeter array ( alma ) is one of the largest ground - based astronomy projects of the next decade , which will revolutionize several fields of astronomy . a large community of scientists is expected to use alma to tackle several outstanding questions in astrophysics . however , mm / submm astronomy is often considered a field restricted to experts . in the case of students and young scientists in particular , the limited familiarity with mm / submm facilities and observations may prevent them to fully exploit the alma capabilities in the future . these lecture notes are aimed at providing students and young researches some background on mm / submm extragalactic astronomy , with a focus on the investigation of agns . i will first provide a quick overview of the current results obtained through extragalactic mm / submm observations , by focusing on agns ( [ sec_mm_astronomy ] ) . i will then summarize the currently available ( and forthcoming ) mm - submm facilities ( [ sec_current_facilities ] ) . then i will shortly describe alma and summarize its observing capabilities ( [ sec_alma ] ) . finally , i will discuss some of the alma prospects for extragalactic studies , and in particular for agns , both in the local universe and at cosmological distances ( [ sec_alma_prospects ] ) . these lecture notes are far from being exhaustive ; several scientific cases will not be discussed at all ; the main goal of these notes is only to provide an introduction to mm / submm extragalactic astronomy and to highlight some scientific cases that alma will be able to tackle . this branch of astronomy includes observations at wavelengths between @xmath010 mm and @xmath0300 @xmath1 m . longer wavelengths are traditionally identified as radio - astronomy domain . shorter wavelengths , out to mid - ir wavelengths , are unobservable from ground because of the nearly complete atmospheric absorption ( although some sites , under exceptional conditions , allow observations out to @xmath2 m . ) . even within the mm - submm range not all wavelengths are equally easy to observe , since the sky transparency on average decreases rapidly at shorter wavelengths . at @xmath3 only a few atmospheric windows are accessible , and only under optimal weather conditions . this issue is clearly illustrated in fig . [ fig_atm_transmission ] , which shows the atmospheric transmission at the alma site . the main source of opacity at these wavelengths is the water vapor . this is the reason for locating mm - submm observatories at dry and high altitude sites , where the amount of water vapor is much reduced . however , even at these optimal sites there are strong variations of the the water vapor , which make the atmospheric transmission change strongly ( fig . [ fig_atm_transmission ] ) both on long ( seasonal ) and short ( day / night ) time scales . given the difficulties of observing at these wavelengths one may wonder why international agencies are investing so much effort to develop facilities with enhanced observing capabilities in these bands . the mm - submm band contains a wealth of information that can not be inferred from any other band . most of the @xmath0150 molecules known so far in the _ cold _ interstellar medium ( see http://astrochemistry.net for an updated list ) emit their rotational transitions in the mm - submm bands , with a density of about 70 lines / ghz . all of these transitions are important diagnostics of the chemistry , of the physics and of the dynamics of the inter stellar medium ( ism ) from which stars form . some of these lines are so strong ( e.g. the co transitions ) to be powerful tools to trace the dynamics and the gas physics even in distant galaxies . furthermore , some of the strongest lines emitted by the ism of any galaxy , such as the [ cii]158@xmath1 m and the [ oi]63@xmath1 m fine structure lines ( the two main coolants of the ism ) , are redshifted into the mm - submm bands at z@xmath424 . within the context of the continuum emission , the mm - submm bands encompass the rayleigh - jeans region of the warm dust thermal emission ( which traces star formation and the dust mass ) , the high frequency tail of the synchrotron emission ( dominating the radio emission in most galaxies ) and of the free - free emission ( tracing hii regions ) . at high redshift the prominent ir dust thermal bump ( which dominates the spectral energy distribution sed in starburst galaxies ) is shifted into the submm band , therefore making this one of the best spectral regions to search and characterize high - z star forming galaxies . this was just a very quick glance at the scientific motivations behind the development of mm - submm facilities , and mostly limited to the extragalactic field . young stellar objects , protostars and proto - planetary systems are , for instance , additional fields where the mm - submm range is crucial for a thorough investigation . the importance of the mm - submm band within the extragalactic context will become more obvious in the following sections , where i will provide some ( shallow ) background on what we currently know of external galaxies based on mm - submm observations , and where some extragalactic alma science cases will be discussed . on the technical side , it is important to mention that the ( sub)mm is currently the shortest wavelength where sensitive , many - elements coherent detection interferometers are feasible from the ground . these can simultaneously provide high angular resolution , sensitivity , and image reconstruction fidelity . direct detection interferometers at shorter wavelengths ( e.g. mid / near - ir ) can achieve similar angular resolution , but are more severely constrained in terms of sensitivity and image fidelity . m ( _ right _ ) . note that most of the far - ir emission comes from a region that is heavily obscured at optical wavelengths . credit of the space telescope science institute ( optical hst image ) and of c. dowell ( submm cso image ) . ] the warm dust emitting at far ir wavelengths is mostly heated by the uv radiation field of young massive stars in star forming regions . as a consequence , the far infrared luminosity @xmath5 and its submm rayleigh - jeans part are considered good tracers of star formation in galaxies . in particular , these bands are useful to trace obscured star formation , since they are virtually unaffected by dust extinction . this is evident in fig . [ fig_antennae ] , where the 350@xmath1 m map of the interacting galaxies `` antennae '' ( obtained at the cso telescope , c. dowell , priv . is compared with the optical hst 20@xmath6 ) . both these issues will no longer be a problem with alma , which will have sensitivities orders of magnitude better and an angular resolution similar to hst . as already mentioned , also most of the _ gas _ phase of the cold ism emits in the mm - submm range . more specifically , it is in this band that most of the molecular gas transitions are observed . however , _ cold _ molecular hydrogen h@xmath7 ( by far the most abundant molecule in the cold gas phase ) can not be detected directly , since it has no electric dipole moment ( therefore rotational transitions with @xmath8 are not allowed ) . carbon monoxide co is the second most abundant molecule : its rotational levels are excited by collisions with h@xmath7 , producing the brightest molecular lines in the spectrum of any galaxy . the luminosity of the co rotational transitions , and in particular the fundamental one j(1@xmath90 ) , are widely used as tracers of the molecular gas mass through a linear relation : @xmath10 . however , the conversion factor @xmath11 is found to depend on the gas metallicity as well as on its physical conditions ( temperature and density ) . the critical density of the co transitions is relatively low , at least for the low rotational levels . for instance , the critical density of co ( 10 ) ( at 115 ghz ) is about @xmath12 . as a consequence , co is generally little effective in tracing the high density regions of molecular clouds . moreover , the large optical thickness the co lines ( at least for the low rational levels ) prevents us to penetrate dense molecular regions . dense regions are better traced by other species , such as hcn , hco@xmath13 and cs , which are characterized by higher critical densities for transitions in the same frequency bands as the co ones . for instance hcn ( 10 ) , hco@xmath13 ( 10 ) and cs ( 21 ) ( observable in the 3 mm band ) , have critical densities of @xmath14 , @xmath15 and @xmath16 , respectively . however , these lines are typically one order of magnitude fainter than co. , indicated by the solid line . the right - hand axis translates the far - ir luminosity into star formation rate . _ right : _ @xmath17 as a function of the far ir luminosity . the right - hand axis gives the gas exhaustion timescale , where @xmath11 is the @xmath18to h@xmath7 conversion factor ( @xmath19 ) . both figures are from @xcite.,title="fig : " ] , indicated by the solid line . the right - hand axis translates the far - ir luminosity into star formation rate . _ right : _ @xmath17 as a function of the far ir luminosity . the right - hand axis gives the gas exhaustion timescale , where @xmath11 is the @xmath18to h@xmath7 conversion factor ( @xmath19 ) . both figures are from @xcite.,title="fig : " ] there is a relationship between infrared luminosity and co luminosity , which is by itself not surprising ( as most luminosity - luminosity relations ) . more interesting is the fact that the relation is not linear , being @xmath20 ( fig . [ fig_fir_co ] , from @xcite ) . since @xmath5 is proportional to the star formation rate sfr , while @xmath21 is proportional to the molecular gas content m(h@xmath7 ) ( which is the fuel for star formation ) , the non linear relation implies that the star formation efficiency , defined as @xmath22 increases with the luminosity of the system . the inverse of the star formation efficiency is the gas exhaustion timescale , @xmath23 , i.e. the time required by the starburst to totally consume the available molecular gas , if the star formation proceeds at the rate currently observed . the non linear relation between @xmath5 and @xmath21 implies that the gas exhaustion timescale decreases with luminosity , reaching values as low as @xmath24 , as illustrated in fig . [ fig_fir_co ] . it is clear that the most powerful starbursts are also short lived . however , it should be mentioned that this picture may be biased , especially for what concerns high redshift sources , due to the incomplete census of galaxy populations . indeed , @xcite have recently identified some _ powerful _ starburst galaxies characterized by _ low _ sfe at z@xmath02 . the far - ir luminosity appears to be _ linearly _ correlated with the luminosity of dense gas tracers , such as hcn . the relation is linear over a very broad luminosity range spanning from galactic star forming regions to powerful starburst galaxies @xcite . this result suggests that star formation always occurs in dense molecular clouds , and that star formation in starburst systems may simply be a scaled up version of galactic star formation . however , also the linearity between @xmath5 and @xmath25 has been recently questioned through indications that the ratio between these two quantities increases with the luminosity @xcite . moreover , the presence of an agn may also affect the intensity of the hcn emission , as discussed in [ sec_past_agn_local ] . interferometers allowed astronomers to map the molecular gas in large samples of galaxies , generally by exploiting the co lines , but more recently also through transitions of various other species . the co emission in galaxies show a variety of morphologies : spiral patterns , bars , rings , nuclear concentrations and irregular distributions @xcite . the co maps also reveal important kinematic information . in the case of disk or ring - like morphologies the gas kinematics is generally dominated by ( nearly ) regular rotation . in the case of barred galaxies , the molecular gas kinematics is often characterized by prominent streaming motions along the stellar bar ( e.g. * ? ? ? * ; * ? ? ? this is regarded as direct evidence that the non - axisymmetric potential of a bar is effective in funneling gas into the central region ( which may eventually produce a central starburst , or be further driven into the nuclear region to fuel an agn ) . agns generally heat their circumnuclear dust to temperatures much higher than starburst galaxies . indeed , active nuclei are generally characterized by strong mid - ir and near - ir emission , indicating dust temperatures of several hundred degrees and reaching the dust sublimation temperature ( 15002000 k ) . at far - ir and submillimeter wavelengths the relative contribution of agn and star formation to the dust heating is debated . spectral decomposition techniques , as well as the correlation between far - ir emission and other tracers of star formation , suggest that the far - ir and submm emission is dominated by star formation in the host galaxies , even in powerful qsos @xcite . disentangling _ spatially _ the far - ir / submm emission due to the agn from the host galaxy emission is very difficult , if not impossible , with current facilities due to the lack of angular resolution at these wavelengths . disentangling the two components is important not only to determine the contribution of agns to the far - ir / submm radiation , but also to constrain models of the obscuring dusty torus , as we shall see in [ sec_alma_local ] . identifying the main mechanism responsible for fuelling agns has been one of the hottest topics in the last decade . non - axisymmetric potentials introduced by stellar bars and galaxy interactions were considered as promising mechanisms , but various studies failed in finding any excess of these morphologies in agns . co observations offer the possibility of directly witnessing the gas fuelling towards the nucleus . intensive campaigns have been performed with millimetric interferometers to investigate this issue by mapping the co emission in galaxies with different types of nuclear activity ( e.g. the nuga project , * ? ? ? * ; * ? ? ? * ; * ? ? ? the main result is that there is no evidence for systematic differences , in terms of molecular gas distribution and kinematics , between galaxies hosting agns and quiescent ones . in particular , seyfert galaxies appear characterized by a wide variety of molecular gas distributions : streaming motions along stellar bars , rings , nuclear concentrations , nuclear voids and irregular distributions . the lack of any relationship between the presence of an agn and the co morphology / dynamics , indicates that there is no ubiquitous evidence for current fuelling of the agn . one possibility , is that the large scale fuelling phase and the agn phase may not be simultaneous . another consideration is that local agns have low luminosities and do not require large fuelling rates from the host galaxies . more specifically , most local seyfert nuclei are characterized by black hole accretion rates of about @xmath26 ; at these rates even a single molecular cloud of @xmath27 can keep the nucleus active for about 1 gyr . the fuelling problem is more serious for powerful qsos , where the accretion rates may exceed @xmath28 . in qsos some mechanism capable of funneling molecular gas from the host galaxy into the nuclear region is actually required . however , qsos are much more distant than seyfert galaxies . the limited sensitivity and angular resolution of current mm interferometers hampers our capability of mapping the molecular gas distribution in qso hosts , which will instead be easy with alma . ngc 1068 is one of the best studied agns also at mm - submm wavelengths . at the highest resolution currently achievable the nuclear co distribution shows a broken ring - like structure , about 2@xmath6 in size ( corresponding to 140 pc ) , as illustrated in fig . [ fig_n1068 ] @xcite . at the specific location of the agn ( cross in fig . [ fig_n1068 ] ) little co emission is observed . does this mean that little or no molecular gas is present in the vicinity of the agn ? this is somewhat contrary to expectations , given the need of gas in the vicinity of the agn both to feed it and to provide the heavy obscuration characterizing ngc 1068 @xcite . actually , we know that in the vicinity of the agn , on scales of 110 pc , dense molecular gas and dust are present , based on radio and mid - ir interferometric observations @xcite . possibly most of the molecular gas in the nuclear region is dense and warm , and therefore not properly sampled by the co lines ( at least not by the low rotational transitions ) . maps of other lines tracing dense gas , such as hcn , have lower angular resolution @xcite , but do show a much higher concentration on the nucleus with respect to co. @xcite detected several molecular species in the nucleus of ngc 1068 tracing not only dense gas , but also a very complex chemistry . these findings indicate that the agn has created a giant `` x - ray dominated region '' ( xdr ) . hard x - ray photons emitted by the agn can penetrate deep into the circumnuclear molecular clouds and keep the temperature high over an extended region . the high gas temperature in xdr s favors the formation of various molecular species such as hcn @xcite . in contrast , photo dissociation regions ( pdr ) , which are generated by the uv photons of star forming regions , are characterized by a much narrower region with enhanced temperature , making the production of various molecular species much less efficient than in xdr . these fundamental differences between xdr s and pdr s suggest that xdr - enhanced species can be used to unveil the presence of heavily obscured agns that escaped detection at other wavelengths . within this context @xcite and @xcite developed a diagnostic diagram involving the line ratios hcn / hco@xmath13 versus hcn / co , where pure seyfert nuclei are clearly separated from starburst nuclei , in the sense that the former show enhanced hcn emission ( but see also caveats discussed in * ? ? ? * ) . this and other complementary diagnostic diagrams will be usable with alma to identify obscured agns even in distant galaxies . ( scaled from the sed of m82 ) at different redshifts ( z@xmath292 , 5 , 7 and 10 , as labelled ) . note that the observed flux at @xmath30 changes very little . the thick blue and red lines show the alma and jwst sensitivities for continuum detection at 5@xmath31 with an integration of @xmath32 . ] as the spectrum of a starburst galaxy moves in redshift , the global flux is reduced according to the 1/d@xmath33 cosmological dimming , but the intrinsic luminosity observed at a fixed mm - submm wavelength increases as a consequence of the steeply rising submm continuum . at z@xmath41 such strong negative k correction counteracts completely the cosmological dimming , so that detecting a source at z@xmath2910 is as easy as detecting a source at z@xmath291 ( for a given intrinsic luminosity ) . this effect is illustrated in fig . [ fig_sed_z ] , which shows the observed flux distribution of a star forming galaxy at different redshifts . the flux observed at @xmath34 mm is essentially unchanged within the redshift range @xmath35 . [ fig_kneg ] shows the same effect by plotting , for a given starburst template , the observed flux at various wavelengths ( and specifically in the alma bands ) as a function of redshift . the observed flux remains nearly constant in the redshift range 1@xmath36z@xmath3610 at wavelengths longer than about 800@xmath1 m . at wavelengths shorter than @xmath0500@xmath1 correction is not strong enough to compensate for the cosmological dimming . at wavelengths longer than @xmath02 mm the k - correction is strong , but the observed flux is more than one order of magnitude fainter than observed at 1 mm ; moreover , at @xmath37 there is an increasing `` risk '' of contamination by non - thermal sources . ( by scaling the m82 template ) observed in the different mm - submm alma bands . note the nearly flat trend at 1@xmath36z@xmath3610 at wavelengths close to @xmath01 mm . the predicted flux density in the optical ( @xmath38 ) is also shown for comparison . ] an interesting consequence of the strongly negative k correction at mm - submm wavelengths is that in any deep field one expects to see many more galaxies at z@xmath41.5 than at lower redshifts . this is just the opposite of what observed in the optical , where the distribution of observed fluxes is dominated by the cosmological dimming ( fig . [ fig_kneg ] ) , which makes any optical deep field dominated by galaxies at z@xmath361.5 , and only a small fraction of galaxies at higher redshifts . the negative k correction allowed the discovery of a large number ( a few 100 ) of starburst galaxies at high - z , thanks to extensive surveys exploiting array of bolometers available on single dish telescopes ( see e.g. * ? ? ? * ; * ? ? ? * for a review ) . dusty starburst galaxies at high - z discovered through the detection of their submm continuum are often dubbed as sub - millimeter galaxies ( smgs ) . although , high - z dusty starbursts have been also discovered through observations in the mm band , for sake of simplicity we will refer to the whole population as `` smgs '' . much effort has been invested for several years in the identification of the optical ( or near - ir ) counterparts of smgs , mostly with the goal of determining their redshift through spectroscopic followup . however , as we will discuss in [ sec_current_facilities ] , the angular resolution of single dish telescopes is so low ( 11@xmath618@xmath6 ) that several optical / near - ir candidate counterparts are found within the telescope beam . the optimal way to identify the true counterpart would be to obtain mm - submm observations at higher angular resolution with a mm - submm interferometer . however , as discussed in [ sec_current_facilities ] , the sensitivity to continuum of current mm - submm interferometers is low , making this approach very expensive in terms of observing time , especially if a statistically meaningful sample is needed . @xcite employed the alternative strategy of exploiting deep radio vla observations . given the tight correlation between far - ir thermal emission and radio synchrotron emission observed in galaxies , most smgs should be associated with a detectable radio source . the higher angular resolution achievable with the vla allowed @xcite to locate the position of the smgs with an accuracy high enough to position the slit for the spectroscopic identification . note also that the large vla primary beam , resulting into a large field of view , allowed the simultaneous radio detection of many smgs in each single pointing . the identification of a prominent ly@xmath11 in many of the deep spectra provided the redshift for several tens smgs . the smgs redshift distribution resulting from these surveys is shown in fig . [ fig_smg_z ] ( left ) , and it is characterized by a median redshift @xmath39 . the redshift distribution is very similar to that of qsos , suggesting a link between the two populations of objects . the sample of smgs for which a redshift has been measured is limited to a submm flux of a few mjy . the inferred redshifts imply far - ir luminosities of several times @xmath40 , i.e. extreme ultra - luminous infrared galaxies ( more luminous , on average , than local ulirg samples ) . therefore , currently identified smgs represent only the most luminous population of high - z dusty starbursts ( the tip of the iceberg ) . in fig . [ fig_smg_z ] ( right ) the inferred contribution of smgs to the density of star formation ( sfrd ) is compared with those inferred by uv / optical surveys and radio / ir tracers @xcite . smgs are indicated with large squares , while small squares show an attempt to account for the incompleteness due to the fact that current surveys only sample the most luminous sources , as discussed above . clearly , once incompleteness is accounted for , smg contribute significantly to the history of star formation , especially at redshifts around two . again , their evolution appears very similar to that of qsos , suggesting a close link between smgs and the evolution of massive halos which hosts qsos . the redshift determination of smgs allowed also their millimetric spectroscopic followup with interferometers aimed at the detection of molecular transitions ( which requires an accurate knowledge of the redshift due to the limited bandwidth ) . intensive campaigns , especially with the iram pdb interferometer yielded the detection of co rotational transitions for several smgs @xcite , indicating that these galaxies also host huge amounts of molecular gas . the profile of the co lines is often double - peaked , indicative of a rotating disk or of a merging of two galaxies . followup interferometric observations at higher angular resolution have found that both the continuum and line emission are very compact in most smgs , a few kpc or even less @xcite . the compactness of the smgs provides constraints on the inferred central densities , which result much higher than in any other population of starburst galaxies , but comparable to those of compact passive galaxies found in the same redshift range . these findings , along with the huge inferred star formation rates , suggest that smgs are progenitors of local massive ellipticals , experiencing their main episode of star formation , but which must undergo further structural evolution in order to reach the sizes observed in local ellipticals . these interferometric data also provide constraints on the global dynamical mass , typically yielding masses of a few times @xmath41 ( e.g. * ? ? ? * ; * ? ? ? interestingly , the inferred high density of massive galaxies ( @xmath42 ) at z@xmath02 is difficult to reconcile with classical hierarchical models of galaxy evolution . millimetric and submillimetric observations of distant agns are currently limited mostly to powerful qsos ( although we will later discuss the case of lower luminosity agns discovered in smgs ) . bolometric observations of large qso samples have detected mm - submm continuum emission at the level of a few mjy in about 60 optically selected qsos and about 20 radio galaxies at z@xmath41 , out to z=6.4 @xcite . generally the detection rate achieved with current facilities is about 30% among bright qsos . detected qsos have far - ir luminosities of about @xmath43 , but these clearly represent the tip of the iceberg due to our current sensitivity limits . m luminosity ( which is a tracer of star formation ) and far - ir luminosity ( inferred from the mm / submm emission in high - z qsos ) . high - z mm - bright qsos follow the same relation of local starburst ulirgs , indicating that the far - ir emission of the former is mostly due to star formation ( from * ? ? ? ] for a fraction of high - z qso the inferred far - ir luminosity follows the same correlation with the radio luminosity as local star forming galaxies , suggesting that the far - ir luminosity is powered mostly by star formation @xcite . this result is supported by the spitzer mid - ir detection , in a number of mm - bright qsos , of strong pah emissionm . these species are mostly destroyed in the strong radiation field produced by agns . ] , which is considered a good tracer of star formation @xcite . moreover , the pah luminosity is found to correlate with the far - ir luminosity and following the same relation of starburst galaxies ( fig.[fig_pah ] ) , further supporting the starburst origin of the far - ir emission . the star formation rates inferred from the far - ir luminosity ( as well as from the pah luminosity ) are as high as a few times @xmath44 . this suggests that in mm - bright qso we are witnessing the simultaneous growth of black holes ( as traced by the optical and x - ray agn emission ) and of the stellar mass in their host galaxy ( as traced by the far - ir ) , which will probably evolve into massive ellipticals @xcite . such a co - coeval growth is expected in models of bh - galaxy evolution aimed at explaining the local relation between spheroids and bh mass @xcite . however , the relation between star formation and black hole accretion ( i.e. the relation between @xmath5 and @xmath45 ) seems to saturate at high luminosities @xcite , suggesting that at high - z the bh growth proceeds more rapidly than the host growth . ) already present in this object close to the epoch of re - ionization . adapted from @xcite . ] the high far - ir luminosities inferred from the mm - submm observations also imply large amount of dust in the host of high - z qsos . accurate measurements of the dust mass are however difficult to obtain due to the unknown dust temperature and emissivity . multiple band observations greatly help to remove the degeneracy between these quantities @xcite . fig.[fig_j1148_dust ] shows the rest - frame infrared sed of one of the most distant qsos ( z=6.4 ) known so far , for which the far - ir thermal bump is relatively well sampled thanks to various submm and mm observations . the inferred dust masses are as high as several times @xmath46 . the discovery of such huge masses of dust in the most distant qsos is puzzling . indeed , in the local universe the main source of dust are the atmospheres of evolved stars ( mostly agb ) . however , at z@xmath46 the age of the universe is less than 1 gyr , which is the minimum time required for agb stars to evolve in large numbers and to significantly enrich the ism with dust . this suggests that in the early universe other mechanisms dominate the dust production . core - collapse sne are candidate sources of dust on short timescales . observations have indeed found dust formed in sn ejecta @xcite , as also expected on theoretical grounds @xcite . the extinction curve inferred in high - z qsos and galaxies appears in agreement with that expected for dust produced in sn ejecta @xcite , supporting the idea that sne may indeed be the major source of dust in the early universe . however , even with the highest dust yield observed so far , the huge mass of dust inferred from mm - submm observations of the most distant qsos remains difficult to account for @xcite . extensive reviews on the co emission in high - z objects , and in agns in particular , are given in @xcite and in @xcite . more than half of the @xmath040 co detections obtained at high redshift ( z@xmath41 ) are in agns ( qsos or radio galaxies ) . in particular , all of the detections at z@xmath43.5 are in agns . the higher detection rate in high - z qsos is mostly due to their huge far - ir luminosities and to the co - fir correlation ( [ sec_past_normal_local ] ) . the molecular gas masses inferred from the co detections are of the order of a few times @xmath47 . except for the huge amounts of molecular gas , the general properties of the co emission in high - z qsos and radio - galaxies do not differ strongly from local and lower redshift powerful starbursts . the observation of multiple co transitions suggests that the co excitation temperature is higher in some qsos ( e.g. * ? ? ? * ) , but the statistics are still very low . the co emission follows the same trend with the fir luminosity observed in fig . [ fig_fir_co ] , indicating that qso host galaxies are experiencing a strong starburst event in the process of rapidly exhausting the available molecular gas , on a time scale of only @xmath48 . it is interesting to note that the relation between co luminosity ( tracing the amount of molecular gas ) and optical luminosity ( tracing the black hole accretion rate in qsos ) is not linear , as shown in fig . [ fig_co_opt ] @xcite . this result indicates that , while the black hole can accrete at very high rates ( limited only by its eddington luminosity ) , the host galaxy has only a limited amount of molecular gas available for star formation ( given by the galaxy evolutionary mechanism ) ; hence the two formation processes probably occur on different time scales ( as already discussed above ) . deep observations have allowed the detection of transitions from other molecular and atomic species , such as hcn , hco@xmath13 , hnc , cn an ci , which are tracers of high density gas and of the gas chemistry and excitation ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? however , these detections are limited to very few bright sources , due to the limited sensitivity of current facilities . within the limited statistics available , the intensity of these lines relative to co and to @xmath5 do not differ strongly from lower redshift and local starburst galaxies ( although there are indications that the intensity of hcn decreases at high luminosities ; * ? ? ? * ; * ? ? ? 158@xmath1 m fine structure line in one of the most distant quasars known , j1148 + 5152 , at z=6.4 @xcite . _ lower panel : _ the co(65 ) transition observed in the same objects is shown for comparison @xcite . note the different flux scales on the two panels . ] recently , deep observations achieved the first detection of the [ cii ] fine structure line at 158@xmath1 m in two high - z qsos ( fig . [ fig_cii ] ) @xcite . this line is emitted by photo - dissociation regions ( pdrs ) in star forming galaxies , and it is the strongest line in the spectrum of nearly any galaxy . the ratio @xmath49}/l_{fir}$ ] observed in these two high - z qsos is about @xmath50 , i.e. similar to the value observed in local ulirgs . the detection of [ cii ] with a luminosity relative to @xmath5 similar to local sources suggests that these high - z systems have already been enriched with heavy elements , and with carbon in particular @xcite . generally the angular resolution and sensitivity are not good enough to trace co rotational curves and therefore to obtain accurate constraints on the dynamical mass . however , in some high - z bright agn the kinematics is resolved , especially in lensed cases , where the lensing shear helps to resolve the galaxy structures ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? in other unresolved , or marginally resolved qsos the width of the co line , along with measurements ( or limits ) on the size of the co emitting region provide some constraints on the dynamical mass ( e.g * ? ? ? typical dynamical masses inferred for qso hosts and radio galaxies are of the order of a few times @xmath41 . however , we should recall that current observations sample only the `` tip of the iceberg '' of the agn population at high redshift . recently , @xcite attempted to calibrate the width of co lines in galaxies as a proxy of their dynamical masses . once this calibration is applied to the host of qsos , the resulting galaxy mass can be compared with the black hole mass inferred from their optical - uv emission lines . an interesting result is that high - z qsos appear to deviate from the local bh - galaxy mass relation , in the sense that their hosts are less massive than expected . this result is in agreement with other independent indications that the bh - galaxy mass relation evolves at high redshift @xcite . however , one should be careful when using the co line width as a tracer of the dynamical mass in ( unobscured ) qsos . indeed , selection effects make qsos hosts to be viewed preferentially face - on , since edge - on views tend to be obscured . as a consequence , the observed width of the co line in qsos is likely reduced due to projection effects @xcite , and may not be representative of the intrinsic rotation of the co disk @xcite . corresponds to about 800 pc . ] an interesting particular result is the co(32 ) map of one of the most distant qsos , j1148 + 5152 , at z=6.4 ( fig . [ fig_j1148_map ] ) @xcite . the map shows resolved emission extending over about 5 kpc and two central peaks separated by 1.7 kpc . from the velocity structure @xcite infer a dynamical mass of about @xmath51 . this mass is much lower than the mass of a few times @xmath52 that would be expected from the black hole mass derived for the same object ( a few times @xmath53 ) if applying the local bh - bulge mass relation . this result would provide further support to the scenario where at high - z black holes grow faster then their host galaxies , while the latter reach the local bh - bulge relation only at lower redshifts . however , observations at higher angular resolution and with higher sensitivity are certainly required to better determine the dynamical mass , and remove the degeneracy with the inclination angle of the putative disk . above we have discussed mm - submm observations aimed at specifically investigating objects known to host powerful agns ( qsos or radio - galaxies ) . we can however consider the orthogonal approach : do galaxies discovered at mm - submm wavelengths ( smgs ) host agns ? this issue has been investigated with the aid of deep x - ray observations . @xcite found that most smgs do host a x - ray source whose luminosity and spectral shape can only be ascribed to the presence of an obscured agn . however , even when corrected for absorption , the inferred agn luminosity can not account for far - ir luminosity observed in smgs , hence the main source of far - ir emission in smgs is due to star formation . similar results were obtained more recently thanks to spitzer mid - ir spectroscopy @xcite . the detection of strong pahs in the spectra of most smgs supports the scenario where most of the luminosity is due to star formation . the detection of a weak mid - ir continuum due to hot dust does trace the presence of an obscured agn in some smgs , but with a bolometric power much lower than the starburst component . summarizing , smgs represent a critical phase in the formation of massive galaxies ; black hole accretion is accompanying this major star formation event , but at a rate much lower than observed in qsos . millimetric and submillimetric facilities can be divided in two main classes : single dish telescopes and interferometers . these two classes of facilities are often complementary , especially in extragalactic astronomy , one generally being characterized by wider field of view and higher sensitivity to continuum , the other providing higher angular resolution and higher sensitivity for line detection . this section provides a short overview of the current facilities in both classes , by also discussing their main relative advantages and limitations . a list of the main currently available single dish mm - submm telescopes is given in tab . [ tab_sing_dish ] . current single dish telescopes have a diffraction limited angular resolution of only about 1020@xmath6 . however , one of the main advantages of single dish telescopes relative to interferometers is their wide field of view , which makes single dish telescopes optimally suited for wide area surveys . all single dish telescopes are equipped with various heterodyne receivers for spectroscopy . however , an additional advantage of single dish telescopes is the possibility of using bolometers , which ( thanks to their wider band width ) are more sensitive to the continuum emission than heterodyne receivers . to fully exploit the large field of view for the detection of mm - submm sources , observatories have developed cameras hosting large bolometer arrays that allow the simultaneous observation of wide areas of the sky . [ tab_sing_dish ] lists the main bolometer cameras currently available , with a short summary of their characteristics ( number of detectors , wavelength of operation , field of view ) . most of the high redshift submillimeter galaxies known so far have been discovered thanks to such bolometer arrays . some of these single dish observatories are dedicating an increasing fraction of their total observing time to extensive surveys by exploiting these bolometer cameras . probably this will be one of the main working modes for some single dish telescopes when alma will be in operation : single dish telescopes will provide continuum targets through wide area surveys to be then observed in depth with alma ( to obtain spectroscopic information , multi - band data , and angular resolution information ) . .main mm - submm single dish telescopes currently available [ cols= " < , < , < , < , < , < , < , < " , ] + notes : + @xmath54 angular resolution in arcsec in the compact ( 200 m ) and in the most extended configuration ( 18 km ) . + @xmath55 field of view in arcsec estimated as the full width half maximum of the primary beam . + @xmath56 continuum sensitivity : rms ( in mjy ) in one minute of integration , derived by using the eso etc . the weather conditions were automatically selected by the etc depending on the band . + @xmath57 emission line sensitivity : rms ( in mjy ) in one minute of integration , derived by using the eso etc and by assuming a line width of 300 km / s . + @xmath58 initially only 6 antennas will be equipped with band 5 ( which explains the much lower sensitivity in this band ) . + @xmath59 funding of band 10 is still pending , awaiting results from the technical development . the sensitivity and frequency range quoted for this band are just indicative , and may be subject to significant variations depending on technical constraints . alma consists of at least 54@xmath6012 m antennas and 12@xmath607 m antennas for a total collecting area of at least 6500 m@xmath33 . the antenna configurations will cover baselines ranging from 200 m out to 18 km , yielding a maximum angular resolution better than 0.1@xmath6 at 3 mm , and better than 0.01@xmath6 at submm wavelengths ( tab . [ tab_alma ] ) . the field of view is limited by the primary beam of individual antennas and it ranges from 1@xmath61 at 3 mm to about 10@xmath6 at submm wavelengths ( tab . [ tab_alma ] ) . [ fig_alma_art ] shows an artistic view of the alma array in extended and compact configurations . the 7 m antennas and 4 of the 12 m antennas are clustered into the atacama compact array , aca ( fig . [ fig_alma_art ] ) , which will have a compact fixed configuration aimed at covering the _ uv _ plane on short spacings . aca will be crucial to recover diffuse emission extending on scales larger than a few arcsec ( which is resolved out by the alma array ) . for more than 50% of the time , allowing very sensitive observations at submm wavelengths . courtesy of the european southern observatory . ] alma will be located on chajnantor plateau , in the atacama desert ( chile ) , at an altitude of 5000 m. this is an exceptionally dry site providing a good transparency even in the 350@xmath1 m and 450@xmath1 m windows for a significant fraction of the time . in particular , the precipitable water vapor ( the main source of opacity ) is @xmath62 for more than 50% of the time , at least during the winter months ( fig . [ fig_pwv ] ) . the vast flat topography of chajnantor plateau is optimal to accommodate the various array configurations , providing some constraints on the location of the antennas only in the most extended configuration . the antennas will be moved from one station to the other by means of huge trucks at a rate of 2 antennas / day . as listed in tab . [ tab_alma ] , receivers for 6 frequency bands are currently fully funded ( bands 3 , 4 and 69 ) . band 5 ( 163211 ghz ) is currently planned only for 6 antennas . the funding and the specifications of band 10 are pending , since the receivers development is still in progress . the distribution of the alma bands relative to the atmospheric windows is illustrated in fig . [ fig_atm_transmission ] . the maximum simultaneous frequency coverage allowed by the huge correlator is 8 ghz , which allows alma to reach high sensitivities even for the continuum detection [ tab_alma ] summarizes the alma sensitivities ( both for continuum and line detection ) . for instance , in band 7 ( 870@xmath1 m ) alma reaches the extraordinary sensitivity of @xmath00.1 mjy ( rms ) in one minute of integration , to be compared with the scuba sensitivity of about 1 mjy ( rms ) in 1 hour of integration . constructions at the alma site started already a few years ago and are now nearly completed . at the time of writing the first eight antennas have arrived for integration and testing at the operations support facilities ( osf , located at about 30 km from the alma site ) . in early 2009 the first three antennas will be moved to the alma operations site ( aos ) for commissioning . in 2010 a call for proposal is planned for early science with an initial set of at least 16 antennas . the full array is scheduled to come in full operation in 2012 . an important feature of the alma organization is the plan of making it an observing facility easily accessible also by non - experts in the field . both the software to prepare the observations and the pipeline for the data reduction are designed to be easy to use also by people that are not necessarily experts in interferometry nor necessarily acquainted with mm / submm observing techniques . this is an important characteristic that will make alma totally open to a broad astronomical community . it is obvious that the alma capabilities will allow astronomers to tackle numerous outstanding issues , spanning most of the research fields in astronomy . a summary of some of the main scientific goals can be found in the design reference science plan ( drsp ) , available at http://www.strw.leidenuniv.nl/alma/drsp11.shtml . it is beyond the scope of these lecture notes to provide an overview of the several outstanding scientific cases for alma . in this section i will only focus on some of the alma prospects for agn studies and some related topics , but even in this field the discussion is not meant to be exhaustive . one of the most hotly debated issues on agns is the physics , structure and dynamics of the circumnuclear molecular medium , which is both responsible for the obscuration of agns and presumably related to their feeding . some models assume a uniform gas distribution in a toroidal geometry , whose thickness is supported by ir radiation pressure ( e.g. * ? ? ? other models propose that the circumnuclear obscuring medium is clumpy and originated by the outflow of the accretion disk @xcite . yet other models assume a clumpy distribution due to the effects of supernova explosions or stellar winds , which are also responsible for making the medium `` thick '' @xcite . the high angular resolution , along with its spectroscopic capabilities , will allow alma to clearly distinguish between these models , both through morphological analyses and by investigating the kinematics of the nuclear and circumnuclear medium . [ fig_wada ] shows the expected distribution of circumnuclear co emission according to the torus model in @xcite , and convolved with the angular resolution of alma projected at the distance of ngc 1068 ( panel _ c _ ) : the clumpy nature of the `` torus '' is clearly resolved . actually , the structure shown in fig . [ fig_wada ] describes the extended component of the torus ( radius @xmath032 pc ) , which contains a large amount of gas , but accounts only for a small fraction of the nuclear obscuration . most of the obscuration is due to dense gas on the parsec scale @xcite , as also inferred by mid - ir interferometric observations @xcite . in clumpy models the dust within each individual cloud spans a wide range of temperatures and therefore emits strongly at all wavelengths , from mid - ir to submm . as a consequence , an interesting prediction of clumpy models is that the nuclear `` torus '' , made of out of several clouds , should show the same morphology at all infrared wavelengths . in contrast , uniform `` torus '' models expect a strong radial temperature gradient and , therefore , the innermost warm dust should emit much more at mid - ir wavelengths , while submm radiation from cold dust should be observed mostly in the outer regions . at high frequencies alma has an angular resolution matching vlti ( i.e. sub - pc scale at the distance of famous agns such as ngc 1068 and circinus ) , therefore it will be possible to directly compare the mid - ir morphology with the submm morphology and directly test the predictions of clumpy and uniform torus models . the gas kinematics inferred from alma observations of the nuclear molecular emission lines will provide important information on the dynamical stability and origin of the circumnuclear medium . in the case of a simple uniform structure , supported by radiation pressure , we expect gas motions to be dominated mostly by rotation . in the case of a medium originated by the accretion disk wind @xcite we expect the molecular gas kinematics to also show a clear outflow component . in the case of a strong contribution by sn and stellar winds we also expect a strong turbulent component , hence a large width of the emission lines even within resolved regions . obviously investigating the molecular gas in the innermost regions will require the observation of species typically tracing dense and highly excited gas ( e.g. hcn , hco@xmath13 , ... ) , which are much fainter than co , but easy to detect with alma . a detailed modelling of the excitation and emissivity distribution of these high density tracers in agn torii has been recently performed by @xcite . according to these models the flux ratios between these molecular lines is expected to be spatially highly inhomogeneous , reflecting the inhomogeneous structure of the molecular torus . only the excellent angular resolution of alma will allow us to reveal such complex structures for the first time . the horizontal blue dashed lines indicate the angular resolution at @xmath63 of various current or planned optical / near - ir facilities ( jwst , 8 m telescope , 30 m telescope , vlti ) . the red shaded area gives the range of angular resolution achievable with alma at different frequencies . courtesy of a. marconi . ] by resolving the molecular gas dynamics on such small scales it will be possible to measure the mass of the nuclear black hole . for instance , in ngc 1068 the @xmath64 black hole has a sphere of influence with a radius of about 4 pc , requiring a beam projected on the sky smaller than 0.05@xmath6 to be resolved , which is achievable with alma at most frequencies . in ngc 1068 , and a few additional nearby sy2s , the bh mass has been already measured thanks to vlbi observations of a maser h@xmath7o disk @xcite ; however , the latter technique can only be exploited in a few cases where the maser disk is observed nearly edge - on ( within 15@xmath65 of the line of sight ) . alma observations of the nuclear molecular gas will allow us to extend the measurement of black hole masses in agns to large samples , without stringent constraints on the geometry of the nuclear gas distribution . [ fig_bh_rinf ] shows the radius of influence , as projected on the sky , of black holes of different masses in the nuclei of galaxies , as a function of their distance ( the central stellar gravitational potential is assumed to match the @xmath66 relation ) . the red shaded area indicates the range of angular resolution achievable with alma at various frequencies . clearly alma is in principle able to measure even very small black holes ( @xmath67 ) in nearby systems ( @xmath68 ) ; therefore alma will allow us to extend the investigation of the @xmath66 relation to very low masses . alma will also allow us to measure massive black holes ( @xmath69 ) in distant systems , @xmath70 ; at these distances we can find the closest cases of merging systems ( most of which are powerful ulirgs ) . in these cases alma will measure the black hole mass within each of the two merging systems , which will allow us to place the two individual black holes on the @xmath66 relation and to compare their location with the evolutionary path expected by models of hierarchical black hole growth ( * ? ? ? * ; * ? ? ? * e.g. ) . at mm - submm wavelengths with the angular resolution achievable with a mm - submm vlbi . the left panels illustrate the result of the model , showing a clear `` shadow '' due to the light bending by the gravitational field of the supermassive black hole . the other panels show the image obtained after convolution with the beam expected for the mm - submm vlbi . the panels on the top are for a rotating black hole , while the panels on the bottom are for a non - rotating black hole . from @xcite . ] alma will also be a powerful machine to discover heavily obscured agns . it is now clear that a population of obscured agn are missed by optical spectroscopic surveys aimed at identifying agns through their characteristic narrow emission lines @xcite . some of these obscured agns are missed because the narrow line region is also obscured by dust in the host galaxy @xcite . in other cases the nlr is not formed at all ( as inferred by the lack of nlr lines even in the mid - ir , * ? ? ? * ) , probably because the ionizing radiation is absorbed in all directions ( i.e. 4@xmath71 obscuration rather than a torus - like geometry ) . however , the x - rays emitted by the agns create extended xdrs , which are characterized by enhanced temperatures favoring the formation of several molecular species characteristics of these regions , as discussed in [ sec_past_normal_local ] . the molecular transitions and diagnostics typical of these regions ( e.g. enhanced hcn / hco@xmath13 * ? ? ? * ; * ? ? ? * ) do not suffer any dust absorption and will be easily detected with alma even in distant systems . therefore , alma will be able to provide an unbiased census of the agn population , by including also those heavily obscured agns that are not detectable at other wavelengths . i conclude this section by discussing the potentialities for the investigation of our galactic center . the nuclear radio source sgra@xmath72 is also a mm / submm source . if the emitting region is uniformly distributed around the supermassive black hole , then @xcite showed that photons are expected to be deviated by the strong gravitational field within a few schwarzschild radii , therefore creating a `` shadow '' in a putative mm / submm high resolution image ( fig . [ fig_falcke ] , left ) . there is a plan of combining alma with other mm / submm observatories distributed world wide to create ( sub-)millimetric vlbi network , reaching an angular resolution of a few 10@xmath1arcsec , which would allow us to resolve the black hole `` shadow '' ( fig . [ fig_falcke ] , right ) . this would really be a major result , essentially the first `` picture '' of a black hole . the shape and the contrast of the `` shadow '' would also allow us to determine whether the black hole is rotating ( upper panels in fig . [ fig_falcke ] ) or not ( lower panels ) . we have seen in [ sec_past_starburst_highz ] that current mm / submm observations of high - z objects are limited to very luminous systems ( @xmath43 ) , not representative of the bulk of the galaxy population at high redshift . alma will be able to detect galaxies at least two orders of magnitude fainter , therefore providing an unbiased view of the evolution of galaxies from the mm / submm perspective . [ fig_sed_z ] , compares the continuum sensitivity of alma ( 24 hours of integration ) with the continuum flux expected by a galaxy with @xmath73 ( i.e. @xmath74 ) at various redshifts . it is clear that with the same integration time alma will be able to detect the thermal dust continuum of galaxies that are three times fainter . it is also interesting to note that for a sed typical of starburst galaxies the alma sensitivity nicely matches the mid / near - ir sensitivity of jwst , another major facility which will be launched in 2013 , i.e. when alma will be already fully operating . these two facilities will be fully complementary for the investigation of faint distant galaxies that are out of reach for current observatories . even in its compact configuration alma will have a resolution high enough ( @xmath75 ) to make the confusion noise negligible , which is instead one of the main limitations of current surveys with single dish telescopes . the excellent alma resolution will allow astronomers to identify the optical / near - ir counterpart of mm - submm sources without ambiguities . the optical / near - ir counterpart may provide the redshift of the galaxy / agn through optical / near - ir followup spectroscopy . however , alma will directly provide the redshift of the mm / submm sources through the detection of molecular lines or fine structure atomic lines . [ fig_lines_z ] shows the observed frequency of the co rotational transitions as a function of redshift overlaid on the alma frequency bands . at z@xmath43 at least two co transitions are observable within band 3 ( the most sensitive one ) , therefore unambiguously providing the redshift of the source . at z@xmath363 the redshift confirmation requires the observation an additional co line in a higher frequency band . [ fig_alma_lines ] illustrates the alma sensitivity for the detection of the co(65 ) line ( which in smgs is close to the peak of the co lines intensities , * ? ? ? * ) for a galaxy with @xmath73 . alma will clearly be able to detect this co transition in luminous infrared galaxies ( lirgs ) out to z@xmath05 . note that to scale the diagram in fig . [ fig_alma_lines ] to other fir luminosities one has to keep in mind the non - linear relation between @xmath21 and @xmath5 ( fig . [ fig_fir_co ] ) . 158@xmath1 m line fluxes expected for a galaxy with @xmath73 as a function of redshift ( keep in mind that the luminosity of these lines does not scale linearly with @xmath5 ) . _ dashed lines : _ sensitivity of alma ( 24 hours of integration ) for a 5@xmath31 detection of emission lines ( width of 300 km / s ) in the frequency bands corresponding to the co(65 ) and [ cii ] redshifted lines . here i have only considered the frequencies around the center of the bands and far from deep atmospheric absorption features ( tab . [ tab_alma ] ) , and i have neglected band 5 ( which has currently much lower sensitivity because available only for 6 antennas ) . the [ cii ] luminosity was inferred by fitting to the @xmath49}/l_{fir}$ ] relation reported in @xcite . the co luminosity was inferred by assuming the @xmath76 relation in fig.[fig_fir_co ] and by assuming the relative luminosities of the co transitions observed in smm j16359 + 6612 ( z=2.5 ) by @xcite . ] at z@xmath47 only co transitions higher than ( 65 ) will be observable within the alma bands . however , such high transitions are generally little excited in most galaxies @xcite and therefore more difficult to observe relative to the lower transitions . yet , already at z@xmath41 the strong [ cii]158@xmath1 m line will be observable in the alma bands , and it will be one of the main tools to identify the redshift of distant sources , and in particular at z@xmath47 ( fig . [ fig_lines_z ] ) . fig . [ fig_alma_lines ] shows that in lirgs ( @xmath77 ) alma will be able to detect [ cii ] out to z@xmath010 . note that also in this case the diagram can not be scaled linearly to other fir luminosities , since the relation between @xmath49}$ ] and @xmath5 is not linear @xcite . the second brightest line is generally [ oi]63@xmath1 m ( in some objects this line is even stronger than [ cii ] , but in some exceptional objects it may be self - absorbed ) . [ oi]63@xmath1 m will be observable at z@xmath44 in the alma bands , and will help to obtain the redshift confirmation along with the [ cii ] detection . the [ oiii]88@xmath1 m line is also generally relatively strong , especially if an agn is present ( @xmath78}\sim 0.5 - 0.3 ~f_{[cii]}$ ] ) , and can be detected with alma at high redshifts . however , since [ oi ] and [ oiii ] will be observable at high - z mostly in the high frequency bands , which are the least sensitive , the detection of these lines will probably be limited to ulirgs ( unless chemical evolutionary effects favor the emission of these oxygen lines , as discussed below ) . the far - ir emission inferred from the mm / submm thermal continuum sed , along with the redshift inferred from the co , [ cii ] and [ oi ] lines , will allow alma to provide a self - consistent , unbiased view of the evolution of the star formation rate through the cosmic epochs . for what concerns high - z agns , alma will allow us to directly trace the coevolution of black hole growth ( through the x - ray and optical emission ) and of the formation of stellar mass in the host galaxies ( through the far - ir emission ) . as discussed in [ sec_past_agn_highz ] , current mm / submm studies have found tentative indications of this co - evolution for very luminous systems ( @xmath79 ) ; alma will extend this investigation to much more quiescent systems ( @xmath80 ) , more representative of the bulk of the galaxy population at high redshift . the luminosity of the co emission will provide the molecular gas mass for most of these galaxies and , when compared with the sfr , will provide a direct measure of the star formation efficiency ( i.e. the star formation per unit gas mass ) as a function of redshift . in high redshift galaxies ( z@xmath41 ) the angular resolution of alma will allow us to resolve morphologies on sub - kpc scales . besides providing the extension of the star formation activity and the distribution of the molecular gas , the alma maps will deliver precious information on the gas dynamics . by resolving rotation curves it will be possible to constrain the dynamical mass of individual galaxies , which can be compared with the expectations by hierarchical models of galaxy evolution . in the case of qsos and type 1 agns it will be possible to directly compare the galaxy dynamical mass to the black hole mass ( inferred from the broad optical / uv emission lines ) , therefore tracing any putative evolution of the @xmath81 relation . the observed evolution of the latter relation as a function of redshift will be directly comparable with the predictions of models dealing with the galaxy - bh co - evolution @xcite , and will therefore provide tight constraints on the same models . the dynamics and morphology of the molecular gas will also provide information on the agn fuelling mechanism . in [ sec_past_agn_local ] we saw that local agns do not show obvious systematic signatures of nuclear fuelling from the host galaxy , but also that such low luminosity agns do not actually require large fuelling from the host galaxy to maintain their low accretion rates . the fuelling demand from the host galaxy is however much larger in powerful qsos , whose accretion rates may exceed a few @xmath82 . therefore , it is important to trace the dynamics of the molecular gas in the host of these powerful qsos , and investigate for instance whether in these systems gas funneled by bars or driven by galaxy merging is common . however , such luminous qsos are not found in the local universe , while are abundant at high redshift . current facilities are in general unable to resolve the molecular gas dynamics in such distant systems , with very few exceptions . alma will allow us to trace in detail the gas dynamics in a large number of qso hosts at high redshift and to statistically investigate the occurrence of various possible fuelling mechanisms . detailed simulations are already being performed to assess the detectability and capability of resolving molecular gas morphologies in high - z agns . @xcite showed that not only alma will be able to resolve and detect the molecular gas in the host galaxy , but also the 100 pc gaseous torii expected to surround most of the supermassive black holes at z@xmath02 . 63@xmath1 m to [ cii]158@xmath1 m intensity ratio as a function of galaxy age , by assuming the chemical evolutionary model in @xcite . the model also assumes log(u)=-2.5 , @xmath83 and that the excitation conditions do not change over time . courtesy of m. kaufman . ] the alma detection of multiple molecular and atomic lines in high - z galaxies and qsos will also provide precious information on the chemical evolutionary status of the galaxy . for instance , the ratio of [ oi]63@xmath1 m and [ cii]158@xmath1 m is sensitive to the relative abundance of oxygen and carbon c / o . since carbon is subject to a delayed enrichment with respect to oxygen ( by several 100 myr ) , the c / o abundance ratio is a sensitive tracer of the evolutionary stage of a galaxy . fig . [ fig_kauf ] shows the expected @xmath8463\mu m}/l_{[cii]158\mu m}$ ] ratio as a function of the age of a galaxy , by exploiting the chemical models in @xcite ( the model assumes the physical conditions derived for the qso at z=6.4 , as discussed in @xcite , and that the excitation conditions do not change over time ) . therefore , by measuring both these lines , alma will allow us to constrain the evolutionary stage of distant galaxies . in particular , it will be extremely interesting to investigate the c / o abundance ratio in agns and galaxies at z@xmath07 , close to the re - ionization epoch , where the age of the universe is close to the minimum enrichment timescale for carbon ; at these redshifts the c / o ratio is expected to be very low , and therefore the [ oi]/[cii ] ratio is expected to be very high . in very luminous high - z sources , such as the qsos already detected with current facilities , alma will be able to easily detect several other molecular ( e.g. hcn , hco@xmath13 , ... ) and atomic ( e.g. [ nii]122@xmath1 m ) transitions , which will allow us to perform an accurate modelling of the physics of the ism and to infer accurate chemical abundances ( hence the evolutionary stage of the system ) . as an example , fig . [ fig_wotten ] shows the simulated alma spectrum ( 24 hours of integration ) of a quasar with the same redshift ( z=6.4 ) and luminosity of the qso j1148 + 5251 , which is currently the most distant qso with a co detection . beside the co(65 ) transition , the 8 ghz alma band is expected to reveal several other molecular transitions with excellent signal - to - noise . finally , alma will allow us to investigate in detail the evolution of dust production in the early universe . the dust mass in galaxies can only be inferred by measuring the ( rest - frame ) infrared to submm sed . currently dust masses have been measured only for very powerful systems at high - z ( qsos and smgs ) , providing an incomplete and very biased view , not representative of the global evolution of dust through the cosmic epochs . alma will measure the dust mass in large samples of high - z systems , even in relatively quiescent ones ( sfr of a few @xmath82 ) . therefore , it will be possible to trace the evolution of dust mass as a function of redshift and for different classes of galaxies , which can be directly compared with models of dust evolution @xcite . in particular , it will be important to clarify whether dust produced by sne can actually be the main source of dust in the early universe , or whether other mechanisms of dust production are required @xcite . i am grateful to the organizers of the school for their kind invitation . i thank m. walmsley , p. caselli , k. wada and l. testi for many useful comments on the manuscript . i am grateful to c. dowell , a. beelen , a. marconi , m. kaufman and a. wotten for providing some of the figures in these lecture notes . some of the images were reproduced with kind permission of the space telescope science institute and of the european southern observatory . finally , i acknowledge financial support by the national institute for astrophysics ( inaf ) and by the italian space agency ( asi ) .

we present next - to - leading order qcd predictions for the total cross section and for a comprehensive set of transverse - momentum distributions in -jet production at the large hadron collider . we neglect the small contributions from subleading - color virtual terms , top quarks and some terms containing four quark pairs . we also present ratios of total cross sections , and use them to obtain an extrapolation formula to an even larger number of jets . we include the decay of the boson into leptons . this is the first such computation with six final - state vector bosons or jets . we use blackhat together with sherpa to carry out the computation .

1304.1253

reliable theoretical predictions for standard - model processes at the large hadron collider ( lhc ) are important to ongoing searches for new physics . they are also important to the increasingly precise studies of the newly discovered higgs - like boson @xcite , of the top quark , and of vector boson self - interactions . new - physics signals very typically lie beneath standard - model backgrounds in a broad range of search strategies . ferreting out the signals requires a good quantitative understanding of the backgrounds and their uncertainties . with the increasing jet multiplicities used in cutting - edge search strategies , this becomes more and more challenging . some of the uncertainty surrounding predictions of standard - model background rates can be alleviated through use of data - driven estimates , but this technique also requires theoretical input to predict the ratios of background processes in signal regions to those for control processes or in control regions . predictions for background rates at the lhc rely on perturbative qcd , which enters all aspects of short - distance collisions at a hadron collider . leading - order ( lo ) predictions in qcd suffer from a strong dependence on the unphysical renormalization and factorization scales . this dependence gets stronger with increasing jet multiplicity . next - to - leading ( nlo ) results generally reduce this dependence dramatically , typically to a 10 - 15% residual sensitivity . thus they offer the first quantitatively reliable order in perturbation theory . the production of a @xmath0 boson in association with jets has played a special role in collider physics . it was the dominant background to top - quark pair production at the tevatron . at the lhc it remains an important background for precision studies , including those of top quarks . it is important to many new physics searches involving missing energy , including those for supersymmetry . recent searches have made use of samples with high jet multiplicity , and proposed searches aim to push to higher multiplicities yet . precise quantitative control over the theoretical predictions leads to improved sensitivity to new phenomena . measurements of @xmath0 boson production in association with multiple jets have been made by the cdf @xcite and d0 @xcite collaborations at the tevatron , and by the atlas @xcite and cms @xcite collaborations at the lhc . such measurements also permit stringent tests of the predictions of the standard model . theoretical predictions for the production of vector bosons with a lower multiplicity of jets ( one or two jets ) have been available at nlo in qcd for many years @xcite . in recent years , the advent of new on - shell techniques @xcite for computing one - loop amplitudes at larger multiplicity has led to nlo results for three @xcite and four @xcite associated jets . other new results include those for the production of vector - boson pairs @xcite or top anti - top pairs @xcite in association with two jets . another recent approach @xcite has been demonstrated in the production of up to seven jets in @xmath1 collisions , and shows promise for lhc physics as well . there have also been important advances with more traditional methods , especially for the case of heavy quarks @xcite . in the present article , we take another step forward in multiplicity , presenting nlo results for inclusive -jet production at the lhc . these are the first nlo qcd results at a hadron collider with six or more electroweak bosons or jets in the final state . we incorporate the decay of the @xmath0 boson into leptons , so that there are seven final - state objects to track . in the present paper we use on - shell methods as implemented in numerical form in the blackhat software library @xcite . this library , together with the sherpa package @xcite , has previously been used to make nlo predictions for -jet production @xcite , for -jet production @xcite , and for four - jet production @xcite . it has also been used in investigations of high-@xmath2 @xmath0 polarization @xcite , and to compute -jet to -jet ratios for assessing theoretical uncertainties @xcite in the cms searches @xcite for supersymmetric particles . the atlas collaboration has also used results from blackhat computations with sherpa for standard - model studies of electroweak vector - boson production in association with three or more jets @xcite . other programs that use on - shell methods are described in refs . @xcite . sherpa is used to manage the numerous partonic subprocesses entering the calculation , to integrate over phase space , to construct physical distributions , and to output @xcite @xmath3-tuples . in contrast to earlier computations , we use the comix package @xcite to compute born and real - emission matrix elements , along with the catani seymour @xcite dipole subtraction terms . rather than repeating the entire computation for each scale and for each parton distribution function ( pdf ) set , we store intermediate results in @xmath3-tuple format , recording momenta for all partons in an event , along with the coefficients of various scale- or pdf - dependent functions in the event weight . the @xmath3-tuple storage makes it possible to evaluate cross sections and distributions for different scales and pdf error sets . we perform the basic calculation with loose cuts , also making it possible to choose different ( tighter ) cuts without recomputing the time - consuming matrix elements . in this paper , we compute the total cross sections at nlo for inclusive -jet and -jet production with @xmath4 and describe @xmath5 ratios and -jet/-jet ratios . such ratios can be sensitive probes of new physics . we also study two types of distributions : the differential cross section in the total hadronic transverse energy @xmath6 , and the complete set of differential cross sections in the jet transverse momenta . for four and five jets we make use of a leading - color approximation for the virtual contributions . this approximation has been shown to have subleading - color corrections of under 3% for processes with four or fewer associated jets @xcite . this paper is organized as follows . in we summarize the basic setup of the computation . in we present our results for cross sections , ratios and distributions . we give our summary and conclusions in . , @xmath7 and @xmath8 , followed by the decay of the @xmath0 boson to leptons . ] and @xmath9 , followed by the decay of the @xmath0 boson to leptons . ] in this paper we compute the -jet processes in nlo qcd , followed by leptonic @xmath0-boson decay , @xmath10 these processes receive contributions from several partonic subprocesses . at leading order , and in the virtual nlo contributions , the @xmath11 subprocesses are all obtained from @xmath12 by crossing five of the partons into the final state . similarly , we obtain the subprocesses for the @xmath13 case from the various crossings of the subprocesses @xmath14 the @xmath0 boson changes the quark flavor and couples to the @xmath15-@xmath16 line . both the labels @xmath15 and @xmath17 denote light quarks . amplitudes with multiple identical quark flavors are obtained by appropriate symmetrization . sample feynman diagrams illustrating virtual contributions with 1 , 2 and 3 external quark pairs are shown in ( although our calculation is not based on feynman diagrams ) . all contributions to the virtual corrections are included in a leading - color approximation described below . besides the virtual contributions , nlo qcd requires also real - emission contributions with an additional parton in the final state . here we keep the full color dependence . however , we drop the finite contributions from tree amplitudes with four external quark pairs ; they contribute well under 1% to the cross section . sample real - emission diagrams are displayed in . the decay of the vector boson ( @xmath18 ) into a charged lepton and neutrino is included at the amplitude level ; no on - shell approximation is made for the @xmath0 boson . the lepton - pair invariant mass follows a relativistic breit - wigner distribution with width given by @xmath19 gev and mass @xmath20 gev . ( the other electroweak parameters are also chosen as in ref . we take the leptonic decay products to be massless . in this approximation , of course , the results for muon final states are identical to those for electrons . the five light quarks , @xmath21 , are all treated as massless . we do not include contributions to the amplitudes from a real or virtual top quark ; its omission should have a percent - level effect on the overall result @xcite . we also approximate the cabibbo - kobayashi - maskawa matrix by the unit matrix . as previously determined for the three - jet case , this approximation causes a change of under 1% in total cross sections for the cuts we impose @xcite , and should also be completely negligible in our study . we use standard kinematic variables , whose definitions may be found in appendix a of ref . the renormalization and factorization scales in ref . @xcite were chosen as multiples of a total partonic transverse energy @xmath22 . we will use a modified version of it here , @xmath23 where the sum runs over all final - state partons @xmath24 and @xmath25 . all partons @xmath24 are included in @xmath26 , whether or not they are inside jets that pass the cuts . this quantity is not directly measurable ; however , it is very similar to the more usual jet - based total transverse energy , and it is more practical for use as a dynamical scale choice . both @xmath22 and the modified version @xmath26 are independent of the experimental cuts . thus , modifying the cuts will not affect the value of the matrix element at a point in phase space . this makes it suitable as a choice of renormalization or factorization scale , avoiding unwanted dependence on experimental cuts . later we will compute the distribution in the jet - based observable @xmath6 . this variable is similar to the partonic version , @xmath26 , except that the @xmath0 boson @xmath27 is omitted , and it is based on jets passing all cuts . we define jets using the anti-@xmath28 algorithm @xcite with parameter @xmath29 . the jets are ordered in @xmath2 , and are labeled @xmath30 in order of decreasing transverse momentum @xmath31 , with jet @xmath32 being the leading ( hardest ) jet . the transverse mass of the @xmath0 boson is computed from the transverse momenta of its leptonic decay products , @xmath33 . in our study , we consider the inclusive process @xmath34 jets at an lhc center - of - mass energy of @xmath35 tev with the following set of cuts : @xmath36 , to equal the neutrino transverse energy , @xmath37 . in carrying out the computation we imposed a set of looser cuts and generated @xcite format @xmath3-tuples . as mentioned above and described further below , the @xmath3-tuples store intermediate results such as parton momenta and coefficients associated with the event weights for the events passing the looser cuts , whose only restriction is that the minimum jet transverse momentum is @xmath38 gev . the @xmath3-tuples are also valid for anti-@xmath28 , @xmath28 and siscone algorithms @xcite for @xmath39 , as implemented in the fastjet package @xcite . in the siscone case the merging parameter @xmath40 is chosen to be @xmath41 . this allows the @xmath3-tuples to be used for studying the effects of varying the jet algorithm , along with variations due to parton distributions , scale choices , and experimental cuts . in our study , we use the mstw2008 lo and nlo pdfs @xcite at the respective orders . we use the five - flavor running @xmath42 and the value of @xmath43 supplied with the parton distribution functions . our predictions are at parton level . we do not apply corrections due to non - perturbative effects such as those induced by the underlying event or hadronization . for comparisons to experiment it is important to incorporate these effects , although for most cross - section ratios we do not expect them to be large . parton - shower event generators such as powheg and mc@nlo @xcite , and further refinements of these methods @xcite , have been developed that consistently include a parton shower and maintain nlo accuracy for events with a specified jet multiplicity . more recently , advances have been made in maintaining the nlo accuracy across different jet multiplicities in a single sample @xcite . these advances mark an important step in significantly reducing theoretical uncertainties associated with hadron - level predictions of many types of lhc events . we look forward to applying them in the future to the production of @xmath0 bosons with up to five additional jets . the new techniques we use for obtaining virtual contributions are collectively known as on - shell methods , and are reviewed in refs . these methods rely on underlying properties of amplitudes factorization and unitarity in order to express them in terms of simpler , on - shell amplitudes of lower multiplicity . while amplitudes necessarily contain off - shell states inside loops or trees , avoiding direct use of these states allows the method to avoid the gauge dependence they induce . eliminating the gauge dependence greatly reduces the enormous cancellations of intermediate terms that would plague a textbook feynman - diagram calculation . the first application of the unitarity method @xcite to collider physics was to obtain the analytic matrix elements for @xmath44 and @xmath45 ( @xmath46 or @xmath47 ) @xcite , used in the nlo program mcfm @xcite . more recently , on - shell methods been implemented in a more flexible numerical form , breaking the long - standing bottleneck to nlo computations for higher - multiplicity final states posed by the one - loop ( virtual ) corrections . these methods scale well as the number of external legs increases @xcite . there have also been important advances in computing virtual corrections with more traditional methods @xcite . one - loop amplitudes in qcd with massless quarks may be expressed as a sum over three different types of feynman integrals ( boxes , triangles , and bubbles ) with additional so - called rational terms . the integrals are universal and well - tabulated , so the aim of the calculation is to compute their coefficients , along with the rational terms . in an on - shell approach , the integral coefficients may be computed using four - dimensional generalized unitarity @xcite , while the rational terms may be computed either by a loop - level version @xcite of on - shell recursion @xcite or using @xmath48-dimensional unitarity @xcite . we use a numerical version @xcite of forde s method @xcite for the integral coefficients , and subtract box and triangle integrands similar to the ossola pittau procedure @xcite , improving the numerical stability . to compute the rational terms , we use a numerical implementation of badger s massive continuation method @xcite , which is related to @xmath48-dimensional unitarity . these algorithms are implemented in an enhanced version of the blackhat code @xcite . blackhat organizes the computation of the amplitudes in terms of elementary gauge - invariant `` primitive amplitude '' building blocks @xcite . many primitive amplitudes can be associated with feynman diagrams in which all external partons touch the loop ( i.e. there are no nontrivial trees attached to the loop ) . representative feynman diagrams for the leading - color primitive amplitudes used in the present calculation are shown in . the primitive amplitudes are then assembled into partial amplitudes , which are the kinematic coefficients of the different color tensors that can appear in the amplitude . the complete virtual cross section is obtained by interfering the one - loop partial amplitudes with the tree - level amplitude and summing over spins and color indices . the color factors arising from the color sum in the assembly of primitive amplitudes into partial amplitudes become highly nontrivial as the number of quark lines increases ; we use a general solution given in ref . an important feature is that each primitive building block has a relatively simple analytic structure with only a limited number of spurious singularities present . a given primitive amplitude can appear in multiple partial amplitudes and does not have to be recomputed for each one . this approach also allows for a straightforward separation of leading- and subleading - color contributions . this separation can be exploited to significantly enhance the efficiency of the monte carlo integration @xcite : the subleading - color contributions are much smaller , yet more computationally costly ; separating them out allows them to be evaluated at far fewer phase - space points than the leading - color contributions , in order to obtain similar absolute uncertainties . in the -jet calculations we drop altogether the small but time - consuming subleading - color contributions to the virtual corrections . as explicitly verified for three- @xcite and four - jet production @xcite , the omitted subleading - color contributions to the virtual corrections are typically 10% of the leading - color virtual terms , and under 3% of the total cross section . we expect the dropped subleading - color contributions to be similarly small for -jet production . the precise version of the leading - color approximation for the virtual terms used here is the one of ref . it retains full - color dependence in all contributions multiplying poles in the dimensional regularization parameter @xmath49 in the virtual corrections . in the finite parts of the virtual corrections , it drops certain contributions that are subleading in the number of colors @xmath50 in the formal limit @xmath51 , with @xmath52 held fixed . in particular , we drop those finite parts of the leading - color partial amplitudes that are suppressed by explicit powers of @xmath53 , as well as all finite parts of the subleading - color partial amplitudes . in forming the color - summed interference of the surviving parts of the one - loop amplitudes with the tree amplitudes , we do not drop any further terms . the blackhat code uses the four - dimensional helicity ( fdh ) scheme @xcite internally but automatically shifts the result before output to the t hooft veltman scheme @xcite using the full - color dependence in the shift . the approximation differs in these last two aspects from the one used in the earlier blackhat calculation of -jet production @xcite . this causes very slight shifts , under a percent , in our reported cross sections for -jet production for @xmath54 compared to ref . ( a somewhat larger shift arises from the choice of a five - flavor scheme for the running of the coupling instead of the six - flavor scheme used earlier . ) the nlo result also requires real - emission corrections to the lo process , which arise from tree - level amplitudes with one additional parton ; sample contributions are illustrated in . for @xmath0 production with four or five associated jets we use the comix code @xcite , included in the sherpa framework @xcite , to compute these contributions , including the catani - seymour dipole subtraction terms @xcite . the comix code is based on a color - dressed form @xcite of the berends - giele recursion relations @xcite , making it very efficient for processes with high multiplicities . when there are fewer jets , we use the amegic++ package @xcite instead . in order to carry out the monte carlo integration over phase space we use an efficient hierarchical phase - space generator based on qcd antenna structures @xcite , as incorporated into sherpa . the integration is performed using an adaptive algorithm @xcite . running the algorithm in a stable and convergent manner for the real - emission contributions is highly nontrivial , in particular given the intricate structure of their infrared subtractions . in general , the same physical distributions need to be analyzed at different pdf error sets , different renormalization or factorization scales , and for different jet algorithms or experimental cuts . we have organized the computation so the matrix elements do not have to be reevaluated for each choice of parameters @xcite . for each event we generate , we record the momenta for all partons , along with the numerical values of the coefficients of the various scale- or pdf - dependent functions . each term contains a simple function we wish to vary , such as a logarithm of the renormalization scale , multiplied by a numerical coefficient independent of such variation . we store the intermediate information in -format @xmath3-tuple files @xcite . at the end of the main computation we assemble the stored matrix - element coefficients , the pdf and scale choices to obtain cross sections . the availability of these intermediate results makes it straightforward to evaluate cross sections flexibly , for different scales , pdf error sets , experimental cuts or jet - based observables . this format has also been used by the experimental collaborations to compare results from blackhat @xmath55 sherpa to experimental data @xcite . -12 mm , @xmath56 and @xmath57 , reading from left to right . the horizontal axis is the logarithm of the relative error ( [ relerror ] ) between an evaluation by blackhat , running in production mode , and a target expression evaluated at higher precision . the vertical axis shows the number of phase - space points having that relative error . the dark ( black ) line labeled by `` double '' shows the @xmath58 term ; the darkly shaded ( red ) curve labeled by `` single '' , the @xmath59 term ; and the lightly shaded ( green ) curve labeled by `` finite '' , the finite ( @xmath60 ) term . each plot is based on approximately 10,000 points in phase space , distributed as in the actual calculation . , title="fig : " ] -10 mm , @xmath56 and @xmath57 , reading from left to right . the horizontal axis is the logarithm of the relative error ( [ relerror ] ) between an evaluation by blackhat , running in production mode , and a target expression evaluated at higher precision . the vertical axis shows the number of phase - space points having that relative error . the dark ( black ) line labeled by `` double '' shows the @xmath58 term ; the darkly shaded ( red ) curve labeled by `` single '' , the @xmath59 term ; and the lightly shaded ( green ) curve labeled by `` finite '' , the finite ( @xmath60 ) term . each plot is based on approximately 10,000 points in phase space , distributed as in the actual calculation . , title="fig : " ] -10 mm , @xmath56 and @xmath57 , reading from left to right . the horizontal axis is the logarithm of the relative error ( [ relerror ] ) between an evaluation by blackhat , running in production mode , and a target expression evaluated at higher precision . the vertical axis shows the number of phase - space points having that relative error . the dark ( black ) line labeled by `` double '' shows the @xmath58 term ; the darkly shaded ( red ) curve labeled by `` single '' , the @xmath59 term ; and the lightly shaded ( green ) curve labeled by `` finite '' , the finite ( @xmath60 ) term . each plot is based on approximately 10,000 points in phase space , distributed as in the actual calculation . , title="fig : " ] -1.1 cm @xmath61 the different terms in the virtual contributions to matrix elements typically contain poles at unphysical locations in phase space . these poles cancel out when summing over all terms . when different terms are computed independently , one must ensure that the numerical precision of the computations suffices for cancellation of these spurious singularities anywhere in the phase space , so as to avoid unwanted loss of precision in the full matrix element . while the degree of the spurious singularities is in fact typically lower when using on - shell methods than with traditional ones , they are nonetheless present . they may cancel between coefficients of different integrals , each computed numerically . obtaining a numerically stable result for the virtual terms at each point in phase space is accordingly nontrivial . the blackhat code detects instabilities following the criteria described in refs . when faced with an unstable point in phase space , the code switches to higher precision arithmetic , and recomputes only those terms which suffer from the instability . the higher - precision computations are performed in software rather than in hardware , and are accordingly much slower than those at native precision . the recomputation of a limited number of terms ( as opposed to the entire amplitude ) minimizes the additional computer time . we use the qd package @xcite for higher - precision arithmetic . in , we illustrate the stability of the virtual contribution to the differential cross section , @xmath62 , summed over colors and over all helicity configurations for the three subprocesses , @xmath63 and @xmath56 and @xmath64 . in each plot , the horizontal axis represents the logarithmic error , @xmath65 for each of the three components : @xmath66 , @xmath67 and @xmath68 , where @xmath69 is the dimensional regulator . in , @xmath70 is the cross section computed by blackhat as it normally operates . the target value @xmath71 is the cross section computed by blackhat using multiprecision arithmetic with approximately @xmath72 digits , and approximately @xmath73 digits if the point is deemed unstable using the criteria described in refs . the phase - space points are selected in the same way as those used to compute cross sections . we note that an overwhelming majority of events are accurate to better than one part in @xmath74 that is , to the left of the ` @xmath75 ' mark on the horizontal axis . we have explicitly checked that the few points to the right of this mark produce completely negligible errors in the final cross section or distributions , as their cross - section values are not especially large . we now present our nlo results for -jet production at the lhc . we first discuss the renormalization - scale dependence of the total cross section . then we provide the total hadronic energy distribution as an example distribution . finally we present results for the total cross sections for -jet and -jet production and for the @xmath2 distributions of the five jets . . the top four panels give the renormalization - scale dependence at both lo and nlo for -jets through -jets . the bottom panel shows the @xmath76 factors for these cases , with the top curve for -jets and the bottom one for -jets . the factorization scale is held fixed . ] , comparing lo to nlo at the lhc at @xmath77 tev . in the upper panels , the nlo predictions are shown as solid ( black ) lines , while the lo predictions are shown as dashed ( blue ) lines . the thin vertical line in the center of each bin ( where visible ) gives its numerical integration error , corresponding to the fluctuations in the plots . the lower panels show the predictions for the lo distribution and scale - dependence bands , normalized to the nlo prediction at the scale @xmath78 . the lo distribution is the dashed ( blue ) line , and the scale - dependence bands are shaded ( gray ) for nlo and cross - hatched ( brown ) for lo . ] we expect perturbative results to be more stable under variation of the renormalization and factorization scales as the perturbative order is increased . the residual variability has been used as a proxy for the expected uncertainty due to higher - order corrections beyond the calculated order . in previous papers @xcite , we have seen that the variability increases substantially with a growing number of jets at lo , but stabilizes at under 20% at nlo . this trend continues as the number of jets grows beyond the multiplicity considered in our just - cited studies . in , we show the variation of the total cross section for -jet production with the renormalization scale around a central choice of @xmath79 , for @xmath80 at lo and at nlo , along with the so - called `` @xmath76 factor , '' the ratio of the nlo to lo cross sections . we vary the scale down by a factor of four and upwards by a factor of eight . a fixed scale of @xmath81 is appropriate for the total cross section , as it is dominated by total transverse energies of the order of a small multiple of this scale . we hold the factorization scale fixed in order to eliminate changes in the pdfs as we vary the scale . this makes it simpler to see the trends as we change from two to five jets . similar improvements in scale dependence are also observed when we include the variation of the factorization scale . the four upper panels of show that the scale variation at nlo is greatly reduced with respect to that at lo . furthermore , the lo variation grows substantially with an increasing number of jets , while the nlo variation is fairly stable . this increase is expected , because there is an additional power of @xmath82 for every additional jet ; the variation of @xmath82 is uncompensated at lo , but compensated at nlo by the virtual corrections . the relative stability of the nlo prediction in contrast to the lo one is also reflected in the bottom panel . as one increases the number of jets from the smallest ( two ) , the @xmath76-factor curve steepens at first . this steepening slows down as the number of jets reaches five . this reflects a slowing down of the relative stabilization with growing number of jets . for the -jet process at lo , the change in the total cross section is on the order of a factor of @xmath83 if we vary the renormalization scale by a factor of 2 around @xmath84 as in . in contrast , at nlo the dependence is cut to about @xmath85% . the distributions we study have a large dynamic range . accordingly , for physics studies we choose an event - by - event scale to match typical energy scales individually rather than merely on average . following ref . @xcite , we use a central scale equal to half the total partonic final - state transverse energy , @xmath86 where @xmath87 is defined in . as an illustration of the scale dependence in a distribution using this choice , we show the variation of the lo and nlo -jet cross section as a function of the total jet transverse energy @xmath88 in . the bands in the figure show the results from varying the scale up and down by a factor of 2 around the central value ( [ scalechoice ] ) , taking the minimum and maximum of the observable evaluated at five values : @xmath89 . the figure shows the markedly reduced scale dependence at nlo compared to that at lo . it also shows a remarkably flat ratio between the lo and nlo distributions . other authors have suggested alternate choices of dynamical scale @xcite . .4 cm .total cross sections in pb for jet production at the lhc at @xmath77 tev , using the anti-@xmath28 jet algorithm with @xmath90 . the nlo results for -jet production use the leading - color approximation discussed in the text . the numerical integration uncertainty is given in parentheses , and the scale dependence is quoted in superscripts and subscripts . [ cols="^,^,^,^,^",options="header " , ] in , we present the charge - asymmetry ratio , which is the ratio between the production of a @xmath13 boson to a @xmath11 boson , each accompanied by up to five jets . this ratio can be a sensitive probe of new physics @xcite . the table also shows the jet - production ratios @xcite for either sign of the @xmath0 charge , here defined by the ratio of the total cross section for -jet to -jet production . the charge - asymmetry ratios are all significantly greater than unity , and grow with increasing numbers of jets . the jet - production ratios are of order @xmath91 , and decrease with increasing numbers of jets . the nlo corrections to the charge - asymmetry are quite small , and the corrections to the jet - production ratios are modest but noticeable . these values of the charge - asymmetry ratio reflect the excess of up quarks over down quarks in the proton . the @xmath13 bosons are necessarily emitted by up - type quarks , whereas @xmath11 bosons are emitted by down - type quarks . the up - quark excess in the proton then leads to larger @xmath13 cross sections . as the number of jets increases , production of a @xmath0 requires a larger value of the momentum fraction @xmath92 . this alters the mix of subprocesses that contribute to vector - boson production , and also increases the @xmath93 ratio . the case of -jet production is special , because the @xmath94 initial state is absent at lo . in general , the @xmath94 initial state contribution is expected to decrease with increasing @xmath92 ; but for @xmath95 production this contribution actually increases with the number of jets , for production in association with two through four jets ( judged by lo fractions ) ; only for more than four jets does it start to decrease as expected . ( in @xmath13 production , it starts to decrease above three rather than four accompanying jets . ) the @xmath96 initial state decreases , but the @xmath97 initial state increases ; the net effect , along with the increase in the @xmath93 fraction , is an increase in the @xmath5 ratio . the results presented here extend our previous nlo analysis of these ratios with up to four accompanying jets @xcite to the case of five accompanying jets . both the lo and the nlo ratios are quite insensitive to correlated variations of the renormalization and factorization scales in numerator and denominator , so we do not quote the variation here , and only show the uncertainty from the numerical integration . the increase in typical values of @xmath92 with increasing numbers of jets also reduces the values of both the strong coupling @xmath82 and the derivatives of the parton distributions , leading to a decrease in the jet - production ratios . the double ratios the ratios of the lo cross sections in the fifth column of to those in the third column , and of the nlo cross sections in the last column to those in the fourth column are roughly constant , suggesting that the decrease is primarily due to the decrease of @xmath82 with increasing scale . we can use the ratios to extrapolate to larger number of jets for the cuts used in this study . this approach was recently investigated using jet calculus and found to be a good approximation when the jets are required to have the same minimum transverse momenta @xcite . ( substantially different cuts could give rise to different behavior . ) while ratios involving -jet production behave differently from the rest , both because of strong kinematic constraints and because of missing production channels ( at lo ) , the remaining ratios turn out to allow an excellent fit to a straight line . for the charge ratio ( to ) , the results through -jet production would suffice to yield a non - trivial prediction ; the ratio for -jet production confirms this prediction . for the jet - production ratios ( to ) , the - to -jet production ratio is essential to making a non - trivial prediction , as it provides a third point in the fit . while these extrapolations should not be taken to too large a number of jets @xmath3 , we expect them to provide a reasonable prediction for @xmath3 somewhat beyond five . for the lo charge ratio with @xmath3 jets , we obtain the following prediction ( @xmath98 ) , @xmath99 for the nlo ratio , @xmath100 for the @xmath11 jet - production ratio at lo , we find the following prediction ( @xmath101 ) , @xmath102 at nlo , we find , @xmath103 similarly , for the @xmath13 jet - production ratio at lo , we find the following prediction ( @xmath101 ) , @xmath104 and at nlo , we find , @xmath105 the slopes for @xmath13 and @xmath11 differ slightly at lo but are essentially the same at nlo . these predictions are based on fits to the data in . ( more precisely , they are based on an ensemble of 10,000 fits to synthetic data distributed in a gaussian according to the cross sections and statistical error in from which the ratios in were computed . ) from , we obtain the following predictions for the nlo cross sections for production of a @xmath106 in association with six jets , @xmath107 matching the experimental cuts used for . the ratio of these two predictions , @xmath108 , is consistent with an extrapolation using , @xmath109 . in this paper , we presented the first nlo qcd results for inclusive -jet production at the lhc at @xmath77 tev . this process is an important background to many new physics searches involving missing energy , as well as to precise top - quark measurements . in addition to its phenomenological usefulness , it also sets a new bar for the state of the art in perturbative qcd at next - to - leading order , at hadron - collider processes with six final - state objects including jets . we have adopted a number of approximations in this work : the leading - color approximation for the virtual terms ; neglecting top - quark loops ; and neglecting infrared - finite parts of real - emission contributions with four quark pairs . based on studies of -jet production and @xmath0 production in associated with fewer jets @xcite , we expect the leading - color approximation to change cross sections by no more than 3% , and the other approximations to be smaller yet . hence these approximations should have no phenomenological significance , given the other theoretical uncertainties . we find a dramatic reduction in scale dependence in nlo predictions for the total cross section , and for differential distributions as well . the scale dependence of observables shrinks from more than a factor of two variation at lo to a 20% sensitivity at nlo for -jet production . with our dynamical scale choice in , we find @xmath76 factors typically between 0.6 and 1 , with moderate though non - trivial changes in shapes of distributions . we have studied a number of ratios , between @xmath13 and @xmath11 production , and for processes differing by the addition of one jet . the qcd corrections to these ratios are more modest than to total cross sections , and they should also benefit from milder experimental systematical uncertainties . the ratios show interesting trends with increasing number of jets , and with results for -jet production in hand , we can make plausible extrapolations to results for additional jets . these ratios also probe the evolution of different subprocesses with increasing parton fraction @xmath92 . the present study brings an unprecedented level of precision to -jet production . we look forward to comparing the nlo results for this process , and for the extrapolations to yet higher numbers of jets based upon it , with lhc data . we thank joey huston , david saltzberg , maria spiropulu , gerben stavenga , eric takasugi and matthias webber for helpful discussions . this research was supported by the us department of energy under contracts de fg0391er40662 and de ac0276sf00515 . dak s research is supported by the european research council under advanced investigator grant erc adg228301 . dm s work was supported by the research executive agency ( rea ) of the european union under the grant agreement number pitn ga2010264564 ( lhcphenonet ) . the work of kjo and sh was partly supported by a grant from the us lhc theory initiative through nsf contract phy0705682 . this research used resources of academic technology services at ucla , and of the national energy research scientific computing center , which is supported by the office of science of the u.s . department of energy under contract no . de ac0205ch11231 . g. aad _ et al . _ [ atlas collaboration ] , phys . b * 716 * , 1 ( 2012 ) [ arxiv:1207.7214 [ hep - ex ] ] . s. chatrchyan _ et al . _ [ cms collaboration ] , phys . lett . b * 716 * , 30 ( 2012 ) [ arxiv:1207.7235 [ hep - ex ] ] . f. abe _ et al . _ [ cdf collaboration ] , phys . lett . * 70 * , 4042 ( 1993 ) ; phys . rev . lett . * 79 * , 4760 ( 1997 ) [ hep - ex/9709016 ] ; + t. aaltonen _ et al . _ [ cdf collaboration ] , phys . d * 77 * , 011108 ( 2008 ) [ arxiv:0711.4044 [ hep - ex ] ] . v. m. abazov _ et al . _ [ d0 collaboration ] , phys . lett . b * 705 * , 200 ( 2011 ) [ arxiv:1106.1457 [ hep - ex ] ] ; arxiv:1302.6508 [ hep - ex ] . g. aad _ et al . _ [ atlas collaboration ] , phys . b * 698 * , 325 ( 2011 ) [ arxiv:1012.5382 [ hep - ex ] ] ; new j. phys . * 15 * , 033038 ( 2013 ) [ arxiv:1301.6872 [ hep - ex ] ] g. aad _ et al . _ [ atlas collaboration ] , phys . d * 85 * , 032009 ( 2012 ) [ arxiv:1111.2690 [ hep - ex ] ] ; + g. aad _ et al . _ [ atlas collaboration ] , phys . d * 85 * , 092002 ( 2012 ) [ arxiv:1201.1276 [ hep - ex ] ] . s. chatrchyan _ et al . _ [ cms collaboration ] , jhep * 1201 * , 010 ( 2012 ) [ arxiv:1110.3226 [ hep - ex ] ] ; phys . * 109 * , 251801 ( 2012 ) [ arxiv:1208.3477 [ hep - ex ] ] . p. b. arnold and m. h. reno , nucl . b * 319 * , 37 ( 1989 ) [ erratum - ibid . b * 330 * , 284 ( 1990 ) ] ; + p. b. arnold , r. k. ellis and m. h. reno , phys . d * 40 * , 912 ( 1989 ) . j. m. campbell and r. k. ellis , phys . d * 65 * , 113007 ( 2002 ) [ hep - ph/0202176 ] . f. febres cordero , l. reina and d. wackeroth , phys . d * 74 * , 034007 ( 2006 ) [ hep - ph/0606102 ] ; + j. m. campbell , r. k. ellis , f. febres cordero , f. maltoni , l. reina , d. wackeroth and s. willenbrock , phys . d * 79 * , 034023 ( 2009 ) [ arxiv:0809.3003 [ hep - ph ] ] . z. bern , l. j. dixon , d. c. dunbar and d. a. kosower , nucl . b * 425 * , 217 ( 1994 ) [ hep - ph/9403226 ] ; nucl . b * 435 * , 59 ( 1995 ) [ hep - ph/9409265 ] . z. bern and a. g. morgan , nucl . b * 467 * , 479 ( 1996 ) [ hep - ph/9511336 ] ; + z. bern , l. j. dixon , d. c. dunbar and d. a. kosower , phys . b * 394 * ( 1997 ) 105 [ hep - th/9611127 ] ; + c. anastasiou , r. britto , b. feng , z. kunszt and p. mastrolia , phys . b * 645 * ( 2007 ) 213 [ hep - ph/0609191 ] ; + r. britto and b. feng , jhep * 0802 * , 095 ( 2008 ) [ arxiv:0711.4284 [ hep - ph ] ] ; + w. t. giele , z. kunszt and k. melnikov , jhep * 0804 * , 049 ( 2008 ) [ arxiv:0801.2237 [ hep - ph ] ] ; + r. britto , b. feng and p. mastrolia , phys . d * 78 * , 025031 ( 2008 ) [ arxiv:0803.1989 [ hep - ph ] ] ; + r. k. ellis , w. t. giele , z. kunszt and k. melnikov , nucl . phys . b * 822 * ( 2009 ) 270 [ arxiv:0806.3467 [ hep - ph ] ] . z. bern , l. j. dixon and d. a. kosower , nucl . b * 513 * , 3 ( 1998 ) [ hep - ph/9708239 ] . r. britto , f. cachazo and b. feng , nucl . b * 725 * , 275 ( 2005 ) [ hep - th/0412103 ] . r. britto , f. cachazo , b. feng and e. witten , phys . lett . * 94 * , 181602 ( 2005 ) [ hep - th/0501052 ] . c. f. berger , z. bern , l. j. dixon , d. forde and d. a. kosower , phys . d * 74 * , 036009 ( 2006 ) [ hep - ph/0604195 ] . g. ossola , c. g. papadopoulos and r. pittau , nucl . b * 763 * , 147 ( 2007 ) [ hep - ph/0609007 ] . d. forde , phys . d * 75 * , 125019 ( 2007 ) [ arxiv:0704.1835 [ hep - ph ] ] . s. d. badger , jhep * 0901 * , 049 ( 2009 ) [ arxiv:0806.4600 [ hep - ph ] ] . c. f. berger , z. bern , l. j. dixon , f. febres cordero , d. forde , t. gleisberg , h. ita , d. a. kosower and d. matre , phys . * 102 * , 222001 ( 2009 ) [ arxiv:0902.2760 [ hep - ph ] ] . r. k. ellis , k. melnikov and g. zanderighi , jhep * 0904 * , 077 ( 2009 ) [ arxiv:0901.4101 [ hep - ph ] ] ; phys . d * 80 * , 094002 ( 2009 ) [ arxiv:0906.1445 [ hep - ph ] ] . c. f. berger , z. bern , l. j. dixon , f. febres cordero , d. forde , t. gleisberg , h. ita , d. a. kosower and d. matre , phys . d * 80 * , 074036 ( 2009 ) [ arxiv:0907.1984 [ hep - ph ] ] . c. f. berger , z. bern , l. j. dixon , f. febres cordero , d. forde , t. gleisberg , h. ita , d. a. kosower and d. matre , phys . lett . * 106 * , 092001 ( 2011 ) [ arxiv:1009.2338 [ hep - ph ] ] . h. ita , z. bern , l. j. dixon , f. febres cordero , d. a. kosower and d. matre , phys . d * 85 * , 031501 ( 2012 ) [ arxiv:1108.2229 [ hep - ph ] ] . t. melia , k. melnikov , r. rntsch and g. zanderighi , jhep * 1012 * , 053 ( 2010 ) [ arxiv:1007.5313 [ hep - ph ] ] ; phys . rev . d * 83 * , 114043 ( 2011 ) [ arxiv:1104.2327 [ hep - ph ] ] ; + n. greiner , g. heinrich , p. mastrolia , g. ossola , t. reiter andf . tramontano , phys . b * 713 * , 277 ( 2012 ) [ arxiv:1202.6004 [ hep - ph ] ] . a. bredenstein , a. denner , s. dittmaier and s. pozzorini , jhep * 1003 * , 021 ( 2010 ) [ arxiv:1001.4006 [ hep - ph ] ] . g. bevilacqua , m. czakon , c. g. papadopoulos , r. pittau and m. worek , jhep * 0909 * , 109 ( 2009 ) [ arxiv:0907.4723 [ hep - ph ] ] ; + g. bevilacqua , m. czakon , c. g. papadopoulos and m. worek , phys . lett . * 104 * , 162002 ( 2010 ) [ arxiv:1002.4009 [ hep - ph ] ] ; phys . d * 84 * , 114017 ( 2011 ) [ arxiv:1108.2851 [ hep - ph ] ] . s. becker , d. goetz , c. reuschle , c. schwan and s. weinzierl , phys . lett . * 108 * , 032005 ( 2012 ) [ arxiv:1111.1733 [ hep - ph ] ] . a. bredenstein , a. denner , s. dittmaier and s. pozzorini , jhep * 0808 * , 108 ( 2008 ) [ arxiv:0807.1248 [ hep - ph ] ] ; phys . rev . lett . * 103 * , 012002 ( 2009 ) [ arxiv:0905.0110 [ hep - ph ] ] ; + t. binoth , n. greiner , a. guffanti , j. p. guillet , t. reiter and j. reuter , phys . b * 685 * , 293 ( 2010 ) [ arxiv:0910.4379 [ hep - ph ] ] ; + n. greiner , a. guffanti , t. reiter and j. reuter , phys . rev . lett . * 107 * , 102002 ( 2011 ) [ arxiv:1105.3624 [ hep - ph ] ] . f. cascioli , p. maierhofer and s. pozzorini , phys . lett . * 108 * , 111601 ( 2012 ) [ arxiv:1111.5206 [ hep - ph ] ] . c. f. berger , z. bern , l. j. dixon , f. febres cordero , d. forde , h. ita , d. a. kosower and d. matre , phys . d * 78 * , 036003 ( 2008 ) [ arxiv:0803.4180 [ hep - ph ] ] . t. gleisberg , s. hche , f. krauss , a. schlicke , s. schumann and j. c. winter , jhep * 0402 * , 056 ( 2004 ) [ hep - ph/0311263 ] ; + t. gleisberg , s. hche , f. krauss , m. schnherr , s. schumann , f. siegert and j. winter , jhep * 0902 * , 007 ( 2009 ) [ arxiv:0811.4622 [ hep - ph ] ] . z. bern , g. diana , l. j. dixon , f. febres cordero , s. hche , d. a. kosower , h. ita , d. matre and k. ozeren , phys . lett . * 109 * , 042001 ( 2012 ) [ arxiv:1112.3940 [ hep - ph ] ] . z. bern , g. diana , l. j. dixon , f. febres cordero , d. forde , t. gleisberg , s. hche , h. ita , d. a. kosower , d. matre and k. ozeren , phys . d * 84 * , 034008 ( 2011 ) [ arxiv:1103.5445 [ hep - ph ] ] . z. bern , g. diana , l. j. dixon , f. febres cordero , s. hche , h. ita , d. a. kosower , d. matre and k. j. ozeren , phys . d * 84 * , 114002 ( 2011 ) [ arxiv:1106.1423 [ hep - ph ] ] . z. bern , g. diana , l. j. dixon , f. febres cordero , s. hche , h. ita , d. a. kosower , d. matre and k. j. ozeren , phys . d * 87 * , 034026 ( 2013 ) [ arxiv:1206.6064 [ hep - ph ] ] . s. chatrchyan _ et al . _ [ cms collaboration ] , jhep * 1108 * , 155 ( 2011 ) [ arxiv:1106.4503 [ hep - ex ] ] ; + s. chatrchyan _ et al . _ [ cms collaboration ] , phys . lett . * 109 * , 171803 ( 2012 ) [ arxiv:1207.1898 [ hep - ex ] ] . g. ossola , c. g. papadopoulos and r. pittau , jhep * 0803 * , 042 ( 2008 ) [ arxiv:0711.3596 [ hep - ph ] ] ; + j .- c . winter and w. t. giele , arxiv:0902.0094 [ hep - ph ] ; + p. mastrolia , g. ossola , t. reiter and f. tramontano , jhep * 1008 * , 080 ( 2010 ) [ arxiv:1006.0710 [ hep - ph ] ] ; + g. bevilacqua , m. czakon , m. v. garzelli , a. van hameren , a. kardos , c. g. papadopoulos , r. pittau and m. worek , comput . commun . * 184 * , 986 ( 2013 ) [ arxiv:1110.1499 [ hep - ph ] ] ; + g. cullen , n. greiner , g. heinrich , g. luisoni , p. mastrolia , g. ossola , t. reiter and f. tramontano , eur . j. c * 72 * , 1889 ( 2012 ) [ arxiv:1111.2034 [ hep - ph ] ] ; + s. badger , b. biedermann , p. uwer and v. yundin , phys . b * 718 * , 965 ( 2013 ) [ arxiv:1209.0098 [ hep - ph ] ] ; comput . commun . * 184 * , 1981 ( 2013 ) [ arxiv:1209.0100 [ hep - ph ] ] . r. brun and f. rademakers , nucl . instrum . a * 389 * , 81 ( 1997 ) . t. gleisberg and s. hche , jhep * 0812 * , 039 ( 2008 ) [ arxiv:0808.3674 [ hep - ph ] ] . s. catani and m. h. seymour , nucl . b * 485 * , 291 ( 1997 ) [ erratum - ibid . b * 510 * , 503 ( 1998 ) ] [ hep - ph/9605323 ] . h. ita and k. ozeren , jhep * 1202 * , 118 ( 2012 ) [ arxiv:1111.4193 [ hep - ph ] ] . m. cacciari , g. p. salam and g. soyez , jhep * 0804 * , 063 ( 2008 ) [ arxiv:0802.1189 [ hep - ph ] ] . s. catani , y. l. dokshitzer , m. h. seymour and b. r. webber , nucl . b * 406 * , 187 ( 1993 ) ; + g. p. salam and g. soyez , jhep * 0705 * , 086 ( 2007 ) [ arxiv:0704.0292 [ hep - ph ] ] . m. cacciari , g. p. salam and g. soyez , eur . j. c * 72 * , 1896 ( 2012 ) [ arxiv:1111.6097 [ hep - ph ] ] . a. d. martin , w. j. stirling , r. s. thorne and g. watt , eur . j. c * 63 * , 189 ( 2009 ) [ arxiv:0901.0002 [ hep - ph ] ] . s. frixione and b. r. webber , jhep * 0206 * , 029 ( 2002 ) [ hep - ph/0204244 ] ; + p. nason , jhep * 0411 * , 040 ( 2004 ) [ hep - ph/0409146 ] ; + s. frixione , p. nason and c. oleari , jhep * 0711 * , 070 ( 2007 ) [ 0709.2092 [ hep - ph ] ] . s. alioli , p. nason , c. oleari and e. re , jhep * 1006 * , 043 ( 2010 ) [ 1002.2581 [ hep - ph ] ] ; + k. hamilton and p. nason , jhep * 1006 * , 039 ( 2010 ) [ 1004.1764 [ hep - ph ] ] ; + s. hche , f. krauss , m. schnherr and f. siegert , jhep * 1104 * , 024 ( 2011 ) [ arxiv:1008.5399 [ hep - ph ] ] ; jhep * 1108 * , 123 ( 2011 ) [ arxiv:1009.1127 [ hep - ph ] ] ; + r. frederix , s. frixione , v. hirschi , f. maltoni , r. pittau and p. torrielli , phys . lett . b * 701 * , 427 ( 2011 ) [ arxiv:1104.5613 [ hep - ph ] ] ; jhep * 1109 * , 061 ( 2011 ) [ arxiv:1106.6019 [ hep - ph ] ] . n. lavesson and l. lnnblad , jhep * 0812 * , 070 ( 2008 ) [ arxiv:0811.2912 [ hep - ph ] ] ; + s. hche , f. krauss , m. schnherr and f. siegert , jhep * 1304 * , 027 ( 2013 ) [ arxiv:1207.5030 [ hep - ph ] ] ; + t. gehrmann , s. hche , f. krauss , m. schnherr and f. siegert , jhep * 1301 * , 144 ( 2013 ) [ arxiv:1207.5031 [ hep - ph ] ] ; + r. frederix and s. frixione , jhep * 1212 * , 061 ( 2012 ) [ arxiv:1209.6215 [ hep - ph ] ] . l. lnnblad and s. prestel , jhep * 1302 * , 094 ( 2013 ) [ arxiv:1211.4827 [ hep - ph ] ] ; jhep * 1303 * , 166 ( 2013 ) [ arxiv:1211.7278 [ hep - ph ] ] ; + s. alioli , c. w. bauer , c. j. berggren , a. hornig , f. j. tackmann , c. k. vermilion , j. r. walsh and s. zuberi , arxiv:1211.7049 [ hep - ph ] ; + k. hamilton , p. nason , c. oleari and g. zanderighi , jhep * 1305 * , 082 ( 2013 ) [ arxiv:1212.4504 [ hep - ph ] ] . z. bern , l. j. dixon and d. a. kosower , ann . nucl . part . * 46 * ( 1996 ) 109 [ hep - ph/9602280 ] ; + z. bern , l. j. dixon and d. a. kosower , annals phys . * 322 * ( 2007 ) 1587 [ arxiv:0704.2798 [ hep - ph ] ] ; + r. britto , j. phys . a * 44 * ( 2011 ) 454006 [ arxiv:1012.4493 [ hep - th ] ] ; + h. ita , j. phys . a * 44 * ( 2011 ) 454005 [ arxiv:1109.6527 [ hep - th ] ] ; + r. k. ellis , z. kunszt , k. melnikov and g. zanderighi , phys . * 518 * ( 2012 ) 141 [ arxiv:1105.4319 [ hep - ph ] ] . w. t. giele and g. zanderighi , jhep * 0806 * ( 2008 ) 038 [ arxiv:0805.2152 [ hep - ph ] ] . z. bern , l. j. dixon and d. a. kosower , nucl . b * 437 * , 259 ( 1995 ) [ hep - ph/9409393 ] . z. bern and d. a. kosower , nucl . b * 379 * , 451 ( 1992 ) ; + z. bern , a. de freitas , l. j. dixon and h. l. wong , phys . d * 66 * , 085002 ( 2002 ) [ hep - ph/0202271 ] . g. t hooft and m. veltman , nucl . b * 44 * , 189 ( 1972 ) . c. duhr , s. hche and f. maltoni , jhep * 0608 * , 062 ( 2006 ) [ hep - ph/0607057 ] . f. a. berends and w. t. giele , nucl . b * 306 * , 759 ( 1988 ) . f. krauss , r. kuhn and g. soff , jhep * 0202 * , 044 ( 2002 ) [ hep - ph/0109036 ] ; + t. gleisberg and f. krauss , eur . j. c * 53 * , 501 ( 2008 ) [ arxiv:0709.2881 [ hep - ph ] ] . a. van hameren and c. g. papadopoulos , eur . j. c * 25 * , 563 ( 2002 ) [ hep - ph/0204055 ] ; + t. gleisberg , s. hche and f. krauss , arxiv:0808.3672 [ hep - ph ] . g. p. lepage , j. comput . * 27 * , 192 ( 1978 ) ; + t. ohl , comput . commun . * 120 * , 13 ( 1999 ) [ hep - ph/9806432 ] . y. hida , x. s. li and d. h. bailey , http://crd.lbl.gov/~dhbailey/mpdist , report lbnl-46996 . c. w. bauer and b. o. lange , arxiv:0905.4739 [ hep - ph ] ; + k. melnikov and g. zanderighi , phys . d * 81 * , 074025 ( 2010 ) [ arxiv:0910.3671 [ hep - ph ] ] . kom and w. j. stirling , eur . j. c * 69 * , 67 ( 2010 ) [ arxiv:1004.3404 [ hep - ph ] ] ; eur . j. c * 71 * , 1546 ( 2011 ) [ arxiv:1010.2988 [ hep - ph ] ] . s. d. ellis , r. kleiss and w. j. stirling , phys . b * 154 * , 435 ( 1985 ) ; + f. a. berends _ et al . _ , b * 224 * , 237 ( 1989 ) ; + f. a. berends , h. kuijf , b. tausk and w. t. giele , nucl . b * 357 * , 32 ( 1991 ) ; + e. abouzaid and h. j. frisch , phys . d * 68 * , 033014 ( 2003 ) [ hep - ph/0303088 ] . c. englert , t. plehn , p. schichtel and s. schumann , phys . d * 83 * , 095009 ( 2011 ) [ arxiv:1102.4615 [ hep - ph ] ] ; + e. gerwick , t. plehn , s. schumann and p. schichtel , jhep * 1210 * , 162 ( 2012 ) [ arxiv:1208.3676 [ hep - ph ] ] .

we report the discovery of nine metal - poor stars with high levels of-process enhancement (.81 [ eu / fe ] .13 ) , including six subgiants and three stars on the red horizontal branch . we also analyze four previously - known-process - enhanced metal - poor red giants . from this sample of 13 stars , we draw the following conclusions . ( 1 ) high levels of-process enhancement are found in a broad range of stellar evolutionary states , reaffirming that this phenomenon is not associated with a chemical peculiarity of red giant atmospheres . ( 2 ) only 1 of 10 stars observed at multiple epochs shows radial velocity variations , reaffirming that stars with high levels of-process enhancement are not preferentially found among binaries . ( 3 ) only 2 of the 13 stars are highly - enhanced in c and n , indicating that there is no connection between high levels of-process enhancement and high levels of c and n. ( 4 ) the dispersions in [ sr / ba ] and [ sr / eu ] are larger than the dispersions in [ ba / eu ] and [ yb / eu ] , suggesting that the elements below the second-process peak do not always scale with those in the rare earth domain , even within the class of highly--process - enhanced stars . ( 5 ) the light - element ( 12 30 ) abundances of highly--process - enhanced stars are indistinguishable from those with normal levels of-process material at the limit of our data , 3.5 per cent ( 0.015 dex ) on average . the nucleosynthetic sites responsible for the large-process enhancements did not produce any detectable light - element abundance signatures distinct from normal core - collapse supernovae . [ firstpage ] nuclear reactions , nucleosynthesis , abundances stars : abundances stars : atmospheres stars : population ii

1409.5810

elements heavier than the iron group comprise a miniscule fraction of all atoms by number or mass , but they comprise approximately 70 per cent of the stable or long - lived isotopes found naturally on earth . the rapid neutron - capture process , also known as the , is one of the two general nucleosynthesis mechanisms responsible for production of most isotopes heavier than the iron group @xcite . two decades ago , observations of the metal - poor star ( [ fe / h ] @xmath53.1 ) revealed a surprising consistency between its heavy - element abundance pattern and the predicted component of the solar system distribution @xcite . over the next decade , three other stars were found with similar abundance patterns ( , @xcite ; , @xcite ; and , @xcite ) , and since then tens more have been identified ( e.g. , @xcite ) . the ratios of heavy elements to fe ( e.g. , [ eu / fe ] ) in many of these stars are an order of magnitude larger than in the solar system , and this is one of two defining characteristics of the `` @xmath0-ii '' class of stars @xcite . ( the other defining characteristic is that the star has [ ba / eu ] @xmath6 0 . ) the existence of this pattern in metal - poor stars nearly as old as the universe itself was not predicted by theory , and the stellar sites responsible for nucleosynthesis are still debated today . our use of the phrase _ nucleosynthesis _ refers to the generic process described by @xcite:``a large flux of neutrons becomes available in a short time interval for addition to elements of the iron group '' ( p. 587 ) . that which produces the predicted abundance pattern in solar system material is the so - called _ main component _ of the . this is what @xcite were trying to reproduce with their analytic description , even though it would be another 16 years before others would attempt to separate the contribution explicitly from the slow neutron - capture ( ) contribution in the solar system total isotopic distribution @xcite . the @xcite definition is , however , sufficiently flexible to account for the `` incomplete synthesis '' ( @xcite , p. 1305 ) abundance pattern , or so - called _ weak component _ , found in many other metal - poor stars and exemplified by ( e.g. , @xcite ) . the patterns typical of the main and weak components may represent the extremes of a range of nucleosynthesis outcomes that depend on the physical conditions at the time of nucleosynthesis . they may also represent two distinct processes @xcite . abundance patterns intermediate between these two extremes are found , and the full range spans at least @xmath70.5 @xmath2 [ eu / fe ] @xmath31.9 ( e.g. , sneden , cowan , & gallino @xcite ) . this can be illustrated , for example , with the [ sr / ba ] ratio in @xmath8 100 stars as shown in in figure 17 of @xcite . alternatively , the same effect is shown for many elements in 16 stars in figure 11 of @xcite . @xcite and @xcite also find these intermediate values in their samples of metal - poor stars with various levels of enhancement ( see their figures 15 and 28 , respectively ) . phenomenological chemical evolution models ( e.g. , @xcite ) and physically - motivated calculations ( e.g. , @xcite ) can reproduce the observed distributions . these explanations , however , are not uniquely capable of predicting the range of [ sr / ba ] ratios in low metallicity stars ( e.g. , @xcite ) . in the literature , references to _ the _ frequently refer to the event producing the main component pattern . observations indicating that the is a rare event and those indicating the nearly - ubiquitous presence of material in stars are not mutually exclusive . abundance patterns like that found in the star are rare , and so presumably the nucleosynthesis events that produce them are also rare . the greater concentration of stars with [ ba / fe ] and [ ba / sr ] @xmath6 0 and low [ c / fe ] ratios in figure 7 of @xcite indicates that stars with these characteristics , like , are not rare ( cf . thus the events that give rise to these patterns are probably not rare , either . in this paper we present nine new members of the class of stars highly - enhanced in material . until recent years the members of this class included only evolved red giants , as summarized by @xcite . this understandably led to concern that perhaps the consistent pattern observed in some metal - poor stars was an artifact of the analysis or related to physics of the stellar photospheres rather than nucleosynthesis . the one exception is a faint main sequence dwarf identified by @xcite . the nine stars presented here are subgiants ( including stars at the main - sequence turn - off , msto ) and the field - star equivalents of red horizontal branch ( rhb ) stars . these stars have been identified among the metal - poor candidates presented by beers , preston , & shectman ( @xcite ) and analyzed in detail by @xcite . the details of sample selection , observations , and a summary of the analysis techniques are presented in sections [ sample ] , [ r14summary ] , and [ compare ] . we examine these stars for evidence of radial velocity variations in section [ rv ] . section [ patterns ] demonstrates the close matches between the abundance patterns in these stars and four previously - known highly -- enhanced giants , including . sections [ hb ] , [ carbon ] , and [ light ] discuss our results , and section [ summary ] summarizes our findings . throughout this work we adopt the standard definitions of elemental abundances and ratios . for element x , the logarithmic abundance is defined as the number of atoms of x per 10@xmath9 hydrogen atoms , @xmath10(x ) @xmath11 12.0 . for elements x and y , the logarithmic abundance ratio relative to the solar ratio , denoted [ x / y ] , is defined as @xmath12 . we adopt the solar abundances given by @xcite and listed in table 13 of @xcite . abundances or ratios denoted with the ionization state indicate the total elemental abundance as derived from transitions of that particular state after ionization corrections have been applied . abundance ratios for elements x and y compare the total abundances of x and y derived from like ionization states ; i.e. , neutrals with neutrals and ions with ions . we draw our sample from the catalog of 313 metal - poor stars observed and analyzed by @xcite . we adopt a traditional indicator of enrichment to select stars highly - enhanced in material , the [ eu / fe ] ratio . the @xcite study included four red giants previously identified as being highly - enhanced in material , , , , and . prior studies have generally confirmed that [ eu / fe ] @xmath131.0 in each of these stars . as discussed in section [ compare ] , the @xcite [ eu / fe ] ratios are lower than those found by previous studies of these four stars by an average of 0.22 dex ( @xmath14 0.10 ) . therefore , to remain consistent with the @xmath0-ii classification scheme , we consider stars with [ eu / fe ] @xmath150.8 as candidates for inclusion in this class . figure [ euplot ] illustrates the [ eu / fe ] ratios for all stars in the @xcite sample . the red line marks [ eu / fe ] @xmath160.78 . eu ii lines are detected in 15 stars whose [ eu / fe ] ratios exceed this limit . four of these stars are the previously - known giants mentioned above . as we discuss in section [ patterns ] , nine of the remaining eu - rich stars have abundance patterns consistent with that found in previously - identified highly -- enhanced stars and the calculated solar abundance pattern . these stars are marked with large red circles in figure [ euplot ] . these nine stars include six subgiants ( , , , , , ) and three field stars that occupy the same region of the temperature - gravity diagram as the rhb stars in globular clusters ( , , ) . @xcite presented abundances for a limited selection of heavy elements in these three rhb stars . their results are similar to those presented here ( section [ compare ] ) , but that study did not discuss the high levels of enhancement found in these stars . [ eu / fe ] ratios for all stars in the @xcite sample . the red line marks [ eu / fe ] @xmath160.78 , the lower limit for classifying a star as highly - enhanced in material . red circles mark stars examined in our study , and the blue squares mark stars with high levels of [ eu / fe ] resulting from enrichment . the dotted line marks the solar ratio . , width=321 ] portions of the spectra of the 13 highly -- enhanced stars are shown in figures [ rgbspecplot][rhbspecplot ] . the absorption lines are substantially stronger in the cool giants , and there are many lines present . this illustrates why highly -- enhanced stars have been preferentially identified among cool giants . absorption lines that are weak in cool giants would be swamped by the continuous opacity in warmer stars , rendering them undetectable . none of these subgiants or rhb stars were selected for observation based on the strength of their absorption lines . two other stars have [ eu / fe ] ratios in excess of the lower limit established to identify highly -- enhanced stars ( and ) . their abundance patterns are clearly indicative of enrichment by large amounts of material produced by nucleosynthesis . these stars are marked with blue squares in figure [ euplot ] and discussed separately in the appendix . we shall not consider them further here . table [ obstab ] presents a record of observations . observations were made with the magellan inamori kyocera echelle ( mike ) spectrograph @xcite on the 6.5 m walter baade and landon clay telescopes ( magellan i and ii ) at las campanas observatory . these spectra were taken with the 07@xmath1750 slit , yielding a resolving power of @xmath18 41,000 in the blue and @xmath19 35,000 in the red as measured from isolated thar lines in the comparison lamp images . the red and blue arms are split by a dichroic at @xmath8 4950 . this setup achieves complete wavelength coverage from 33509150 . signal to noise ( s / n ) estimates at several reference wavelengths are listed in table [ sntab ] . the footnotes to table [ sntab ] identify stars that were reimaged during the course of the las campanas objective - prism survey @xcite or the hamburg - eso survey @xcite . lcccccc star & telescope/ & exposure & date & ut at & heliocentric & heliocentric + & instrument & length & & mid - exposure & julian date & radial velocity + & & ( s ) & & & & ( ) + + cs 22183031 & magellan - clay / mike & 1400 & 2006 aug 03 & 10:23 & 2453950.935 & @xmath123.1 + cs 22892052 & magellan - baade / mike & 2400 & 2003 jun 26 & 10:12 & 2452816.928 & @xmath112.9 + & magellan - clay / mike & 2400 & 2004 jun 29 & 09:11 & 2453185.886 & @xmath113.0 + cs 22953003 & magellan - clay / mike & 2000 & 2006 aug 03 & 09:46 & 2453950.910 & @xmath1208.3 + cs 31082001 & magellan - clay / mike & 1800 & 2003 jul 19 & 10:27 & 2452839.937 & @xmath1138.9 + + cs 22886012 & magellan - clay / mike & 3600 & 2008 sep 12 & 02:59 & 2454721.630 & @xmath756.8 + & magellan - clay / mike & 1800 & 2009 sep 05 & 03:08 & 2455079.636 & @xmath757.0 + cs 22943132 & magellan - clay / mike & 1400 & 2006 aug 03 & 02:19 & 2453950.602 & @xmath118.9 + cs 22945017 & magellan - clay / mike & 3400 & 2008 sep 11 & 02:58 & 2454720.627 & @xmath1101.8 + cs 22945058 & magellan - clay / mike & 5400 & 2008 sep 11 & 04:28 & 2454720.689 & @xmath123.1 + & magellan - clay / mike & 1550 & 2009 jul 26 & 10:29 & 2455038.940 & @xmath123.7 + cs 22958052 & magellan - clay / mike & 3000 & 2004 sep 26 & 09:12 & 2453274.886 & @xmath188.9 + cs 29529054 & magellan - baade / mike & 3600 & 2003 jan 17 & 03:27 & 2452656.643 & @xmath1113.3 + & magellan - baade / mike & 6000 & 2003 jan 20 & 01:08 & 2452659.547 & @xmath1112.9 + & magellan - clay / mike & 1900 & 2009 sep 05 & 09:59 & 2455079.917 & @xmath1113.4 + + cs 22875029 & magellan - baade / mike & 3600 & 2003 jun 12 & 09:59 & 2452802.918 & @xmath173.1 + & magellan - clay / mike & 3600 & 2003 oct 12 & 04:25 & 2452924.687 & @xmath172.7 + cs 22882001 & magellan - baade / mike & 4800 & 2003 jun 15 & 09:24 & 2452805.892 & @xmath126.9 + & magellan - clay / mike & 6000 & 2003 oct 09 & 06:05 & 2452921.758 & @xmath126.1 + cs 22888047 & magellan - baade / mike & 4400 & 2003 jun 14 & 09:00 & 2452804.876 & @xmath7158.5 + & magellan - clay / mike & 6000 & 2003 oct 13 & 05:15 & 2452925.723 & @xmath7158.1 + lcccccc star & total exp . & no . & s / n & s / n & s / n & s / n + & time ( s ) & obs . & 3950 & 4550 & 5200 & 6750 + + cs 22183031 & 1400 & 1 & 75 & 105 & 75 & 105 + cs 228920521654 ] & 4800 & 2 & 150 & 225 & 175 & 270 + cs 22953003 & 2000 & 1 & 65 & 95 & 75 & 100 + cs 31082001 & 1800 & 1 & 125 & 145 & 230 & 350 + + cs 22886012 & 5400 & 2 & 60 & 85 & 80 & 130 + cs 22943132 & 1400 & 1 & 75 & 105 & 75 & 95 + cs 22945017 & 3400 & 1 & 60 & 75 & 75 & 115 + cs 22945058 & 6950 & 2 & 60 & 85 & 80 & 125 + cs 22958052 & 3000 & 1 & 80 & 105 & 55 & 75 + cs 29529054 & 11500 & 3 & 60 & 100 & 110 & 155 + + cs 22875029 & 7200 & 2 & 110 & 115 & 160 & 220 + cs 22882001 & 10800 & 2 & 85 & 95 & 120 & 170 + cs 22888047 & 10400 & 2 & 95 & 105 & 150 & 205 + @xcite used model atmospheres interpolated from the grid of one - dimensional marcs models @xcite and performed the analysis using a recent version of the spectral line analysis code moog ( @xcite ; see discussion in @xcite ) . effective temperatures ( ) and microturbulent velocities ( ) were derived by requiring that abundances derived from fe i lines showed no trend with the excitation potential and line strength . for stars on the horizontal branch , was derived by requiring that the fe abundance derived from fe i lines matched that derived from fe ii lines . for the other stars , surface gravities ( , in cgs units ) were calculated from the relationship between and given by theoretical isochrones in the y@xmath20 grid @xcite assuming an age of 12 @xmath21 1.5 gyr . the fe abundance derived from fe ii lines was taken to represent the overall metallicity , [ m / h ] . the derived model parameters and their statistical ( internal ) uncertainties are presented in table [ atmtab ] . these values are identical to those presented in table 7 of @xcite and are reproduced here for convenience . @xcite estimated the systematic uncertainties by comparing model parameters with those derived in previous studies . for red giants , subgiants , and stars on the horizontal branch , these comparisons for the full sample yielded standard deviations of 151 , 211 , and 156 k in , 0.40 , 0.34 , and 0.42 in , 0.41 , 0.33 , and 0.26 in , and 0.24 , 0.22 , and 0.16 dex in [ fe ii / h ] . we adopt these as the systematic uncertainties in the model atmosphere parameters . ccccccccc star & & & & [ fe / h ] & [ sr / fe ] & [ ba / fe ] & [ eu / fe ] & [ yb / fe ] + & ( k ) & & ( ) & & & & & + + cs 31082001 & 4650 ( 35 ) & 1.05 ( 0.14 ) & 1.55 ( 0.06 ) & @xmath73.03 ( 0.07 ) & @xmath10.67 ( 0.12 ) & @xmath10.92 ( 0.17 ) & @xmath11.37 ( 0.16 ) & @xmath11.38 ( 0.27 ) + cs 22892052 & 4690 ( 37 ) & 1.15 ( 0.15 ) & 1.50 ( 0.06 ) & @xmath73.16 ( 0.07 ) & @xmath10.68 ( 0.17 ) & @xmath10.94 ( 0.16 ) & @xmath11.35 ( 0.22 ) & @xmath11.39 ( 0.27 ) + cs 22183031 & 4850 ( 36 ) & 1.60 ( 0.14 ) & 1.55 ( 0.06 ) & @xmath73.50 ( 0.07 ) & @xmath70.22 ( 0.23 ) & @xmath10.13 ( 0.15 ) & @xmath10.84 ( 0.16 ) & @xmath10.68 ( 0.18 ) + cs 22953003 & 4860 ( 37 ) & 1.65 ( 0.17 ) & 1.45 ( 0.06 ) & @xmath73.00 ( 0.07 ) & @xmath10.32 ( 0.24 ) & @xmath10.29 ( 0.15 ) & @xmath10.79 ( 0.15 ) & @xmath10.63 ( 0.20 ) + + cs 22945017 & 6080 ( 54 ) & 3.70 ( 0.19 ) & 1.25 ( 0.06 ) & @xmath72.73 ( 0.08 ) & @xmath10.39 ( 0.24 ) & @xmath10.49 ( 0.14 ) & @xmath11.13 ( 0.19 ) & @xmath11.02 ( 0.18 ) + cs 22945058 & 5990 ( 45 ) & 3.65 ( 0.17 ) & 1.55 ( 0.06 ) & @xmath72.71 ( 0.08 ) & @xmath10.34 ( 0.24 ) & @xmath10.28 ( 0.14 ) & @xmath11.13 ( 0.19 ) & @xmath10.83 ( 0.20 ) + cs 22958052 & 6090 ( 46 ) & 3.75 ( 0.18 ) & 1.95 ( 0.06 ) & @xmath72.42 ( 0.07 ) & @xmath10.13 ( 0.24 ) & @xmath10.00 ( 0.15 ) & @xmath11.00 ( 0.18 ) & @xmath10.93 ( 0.17 ) + cs 29529054 & 5710 ( 40 ) & 3.55 ( 0.18 ) & 1.65 ( 0.06 ) & @xmath72.75 ( 0.07 ) & @xmath70.10 ( 0.24 ) & @xmath70.02 ( 0.15 ) & @xmath10.90 ( 0.21 ) & @xmath10.86 ( 0.24 ) + cs 22943132 & 5850 ( 41 ) & 3.60 ( 0.16 ) & 1.40 ( 0.06 ) & @xmath72.67 ( 0.08 ) & @xmath10.27 ( 0.24 ) & @xmath70.05 ( 0.14 ) & @xmath10.86 ( 0.22 ) & @xmath10.37 ( 0.32 ) + cs 22886012 & 5650 ( 42 ) & 3.50 ( 0.15 ) & 1.45 ( 0.06 ) & @xmath72.61 ( 0.07 ) & @xmath10.31 ( 0.23 ) & @xmath10.14 ( 0.14 ) & @xmath10.85 ( 0.17 ) & @xmath10.45 ( 0.22 ) + + cs 22875029 & 5990 ( 44 ) & 1.85 ( 0.27 ) & 2.80 ( 0.06 ) & @xmath72.69 ( 0.06 ) & @xmath10.85 ( 0.22 ) & @xmath10.30 ( 0.15 ) & @xmath10.92 ( 0.16 ) & @xmath10.86 ( 0.19 ) + cs 22888047 & 5950 ( 46 ) & 1.90 ( 0.27 ) & 3.00 ( 0.06 ) & @xmath72.54 ( 0.06 ) & @xmath10.60 ( 0.22 ) & @xmath10.04 ( 0.16 ) & @xmath10.86 ( 0.16 ) & @xmath10.67 ( 0.18 ) + cs 22882001 & 5930 ( 52 ) & 1.90 ( 0.32 ) & 3.00 ( 0.06 ) & @xmath72.62 ( 0.06 ) & @xmath10.16 ( 0.25 ) & @xmath10.06 ( 0.16 ) & @xmath10.81 ( 0.16 ) & @xmath10.56 ( 0.19 ) + + & & & & @xmath72.80 ( 0.08 ) & @xmath10.34 ( 0.09 ) & @xmath10.27 ( 0.09 ) & @xmath10.99 ( 0.06 ) & @xmath10.82 ( 0.09 ) + table 8 of @xcite lists the atomic data for each transition studied . spectrum synthesis was performed for lines broadened by hyperfine splitting or in cases where a significant isotope shift may be present . for unblended lines , @xcite used moog to compute theoretical equivalent widths , which were then forced to match measured equivalent widths by adjusting the abundance . @xcite derived 3@xmath22 upper limits when a line was not detected . table 11 of @xcite lists the abundances derived from each line in each star . that study also adopted corrections to account for departures from local thermodynamic equilibrium ( lte ) in the line formation regions for li i @xcite , o i @xcite , na i @xcite , and k i @xcite . weighted mean abundances and uncertainties were computed using the formalism presented in @xcite . we do not repeat the full set of abundances here , but a few key element ratios are reproduced in table [ atmtab ] for convenience . table [ comparetab ] compares the derived stellar parameters and several abundance ratios with those found by previous investigators for the four red giant and three rhb stars in our sample . most previous studies of red giants calculated using color- relations , frequently leading to warmer and higher metallicity than that found by @xcite . for the four red giant stars listed in table [ comparetab ] , our metallicities are lower than those found by previous studies by 0.19 dex ( @xmath14 0.13 ) . for the rhb stars , our metallicities are lower than those found by previous studies by 0.09 dex ( @xmath14 0.10 ) . no comparisons are available for the subgiants , but @xcite found that the derived metallicities of 40 subgiants were lower than those found by previous studies by only 0.04 dex ( @xmath14 0.18 ) . the [ eu / fe ] ratio is often adopted as a proxy for the overall level of enhancement of a star , and our [ eu / fe ] ratios are lower than found by previous studies by 0.22 dex ( @xmath14 0.10 ) for the red giants and lower by 0.09 dex ( @xmath14 0.08 ) for the stars on the rhb . the reasons for these differences are investigated in detail in section 9.4 of @xcite in summary , this offset can be explained by the combined effects of different model atmosphere parameters , updates to the analysis code , lines available for analysis , and quality of the observed spectra . cccccccccc star & & & & [ fe / h ] & [ sr / fe ] & [ ba / fe ] & [ eu / fe ] & [ yb / fe ] & ref . + & ( k ) & & ( ) & & & & & & + cs 22183031 & 4850 & 1.60 & 1.55 & @xmath73.50 & @xmath70.22 & @xmath10.13 & @xmath10.84 & @xmath10.68 & 1 + ( rg ) & 5270 & 2.80 & 1.20 & @xmath72.93 & @xmath10.10 & @xmath10.38 & @xmath11.16 & & 2 + cs 22892052 & 4690 & 1.15 & 1.50 & @xmath73.16 & @xmath10.68 & @xmath10.94 & @xmath11.35 & @xmath11.39 & 1 + ( rg ) & 4790 & 1.60 & 1.80 & @xmath72.92 & @xmath10.44 & @xmath10.92 & @xmath11.51 & @xmath11.29 & 2 + & 4760 & 1.30 & 2.29 & @xmath73.03 & @xmath10.68 & @xmath10.93 & @xmath11.48 & & 3 , 4 + & 4850 & 1.60 & 1.90 & @xmath73.02 & @xmath10.53 & @xmath11.01 & @xmath11.49 & & 5 , 6 + & 4800 & 1.50 & 1.95 & @xmath73.12 & @xmath10.58 & @xmath10.92 & @xmath11.65 & @xmath11.63 & 7 , 8 + & 4884 & 1.81 & 1.67 & @xmath72.95 & @xmath10.61 & @xmath11.19 & @xmath11.54 & & 9 + & 4725 & 1.00 & 2.00 & @xmath73.16 & @xmath10.52 & @xmath10.89 & @xmath11.56 & & 10 + & 4760 & 1.30 & 2.30 & @xmath73.03 & @xmath10.63 & @xmath10.89 & @xmath11.63 & @xmath11.15 & 11 + cs 22953003 & 4860 & 1.65 & 1.45 & @xmath73.00 & @xmath10.32 & @xmath10.29 & @xmath10.79 & @xmath10.63 & 1 + ( rg ) & 4960 & 1.70 & 1.95 & @xmath72.83 & @xmath10.31 & @xmath10.00 & @xmath10.72 & & 3 , 4 + & 5100 & 2.30 & 1.70 & @xmath72.82 & @xmath10.22 & @xmath10.49 & @xmath11.05 & @xmath11.02 & 5 , 6 + cs 31082001 & 4650 & 1.05 & 1.55 & @xmath73.03 & @xmath10.67 & @xmath10.92 & @xmath11.37 & @xmath11.38 & 1 + ( rg ) & 4825 & 1.50 & 1.80 & @xmath72.92 & @xmath10.65 & @xmath11.17 & @xmath11.63 & & 12 + & 4790 & 1.80 & 1.90 & @xmath72.81 & @xmath10.47 & @xmath11.02 & @xmath11.67 & @xmath11.54 & 2 + & 4825 & 1.50 & 1.80 & @xmath72.92 & @xmath10.73 & @xmath11.16 & @xmath11.69 & @xmath11.66 & 13 + & 4825 & 1.50 & 1.80 & @xmath72.92 & & & @xmath11.68 & @xmath11.57 & 8 + & 4922 & 1.90 & 1.88 & @xmath72.78 & @xmath10.53 & @xmath11.18 & @xmath11.66 & & 9 + & 4925 & 1.51 & 1.40 & @xmath72.81 & @xmath10.66 & @xmath11.43 & @xmath11.53 & & 14 + cs 22875029 & 5990 & 1.85 & 2.80 & @xmath72.69 & @xmath10.85 & @xmath10.30 & @xmath10.92 & @xmath10.86 & 1 + ( rhb ) & 6000 & 2.35 & 3.05 & @xmath72.63 & @xmath10.83 & @xmath10.48 & @xmath11.10 & & 15 + & 6000 & 2.05 & 3.00 & @xmath72.66 & @xmath10.86 & @xmath10.44 & @xmath10.91 & & 16 + cs 22882001 & 5930 & 1.90 & 3.00 & @xmath72.62 & @xmath10.16 & @xmath10.06 & @xmath10.81 & @xmath10.56 & 1 + ( rhb ) & 5950 & 2.50 & 3.20 & @xmath72.45 & @xmath10.28 & @xmath10.20 & @xmath10.92 & & 15 + & 5950 & 2.00 & 3.05 & @xmath72.54 & @xmath10.22 & @xmath10.16 & @xmath10.84 & & 16 + cs 22888047 & 5950 & 1.90 & 3.00 & @xmath72.54 & @xmath10.60 & @xmath10.04 & @xmath10.86 & @xmath10.67 & 1 + ( rhb ) & 6000 & 2.50 & 3.00 & @xmath72.30 & @xmath10.63 & @xmath10.29 & @xmath11.03 & & 15 + & 5850 & 1.70 & 3.20 & @xmath72.57 & @xmath10.31 & @xmath10.23 & @xmath10.93 & & 16 + + mean & @xmath7169 & @xmath70.46 & @xmath70.33 & @xmath70.19 & @xmath10.06 & @xmath70.11 & @xmath70.22 & @xmath70.13 & + standard deviation & 105 & 0.35 & 0.31 & 0.13 & 0.13 & 0.19 & 0.10 & 0.22 & + no . comparisons & 14 & 14 & 14 & 14 & 15 & 15 & 16 & 7 & + + mean & @xmath72 & @xmath70.30 & @xmath70.15 & @xmath70.09 & @xmath10.02 & @xmath70.05 & @xmath70.09 & & + standard deviation & 52 & 0.32 & 0.10 & 0.10 & 0.14 & 0.15 & 0.08 & & + no . comparisons & 6 & 6 & 6 & 6 & 6 & 6 & 6 & & + + studies that adopted model parameters without change from an earlier study have been excluded from computing the average deviations of , , , and [ fe / h ] . + references:(1 ) this study ; ( 2 ) @xcite ; ( 3 ) @xcite ; ( 4 ) @xcite ; ( 5 ) @xcite ; ( 6 ) @xcite ; ( 7 ) @xcite ; ( 8) @xcite ; ( 9 ) @xcite ; ( 10 ) @xcite ; ( 11 ) @xcite ; ( 12 ) @xcite ; ( 13 ) @xcite ; ( 14 ) @xcite ; ( 15 ) @xcite ; ( 16 ) @xcite @xcite measured radial velocities by cross - correlating the spectral order containing the mg i b lines against metal - poor template standards . heliocentric corrections were computed using the iraf _ rvcorrect _ task . typical uncertainties are @xmath8 0.60.8 per observation . table [ obstab ] lists the heliocentric velocity measurements for each observation . repeat observations of , , , , , , and show no evidence of radial velocity variations in our data . @xcite reported a possible detection of binarity for , but that preliminary result was based on a tentative phasing of measurements with semi - amplitude of only 1.0 . subsequent velocity monitoring by our group and others @xcite finds no compelling evidence for velocity variations of . comparisons with prior studies of ( @xcite ) and ( @xcite ) also show no evidence of radial velocity variations . our measured velocity of is 11 different from two measurements made by @xcite . we are aware of only our single - epoch radial velocity measurements of , , and . of the 10 stars in our sample with observations at multiple epochs from all sources , only shows compelling evidence for velocity variations . this frequency , 10 per cent , is in reasonable agreement with the 18 per cent binary frequency ( 3 of 17 stars ) of other highly -- enhanced stars reported by @xcite based on long - term velocity monitoring . our binary frequency may be underestimated since we have velocity measurements at only two or three epochs for most stars , thus long - period or low - amplitude binaries may evade our search . in any case , our data support the conclusions of @xcitethat there is no evidence to suggest that all highly -- enhanced stars are members of binary or multiple star systems . the heavy element abundance patterns are illustrated in figures [ rprorgfig ] , [ rprosgfig ] , and [ rprohbfig ] . three standard abundance templates are shown for comparison . one pattern traces the heavy element abundances in as derived by @xcite and @xcite from higher quality data . it is reassuring that our derived abundance pattern for matches this comparison set so closely . the small deviations for sr and ba can be explained by the different microturbulent velocities derived by us ( 1.50 ) and @xcite ( 1.95 ) , which cause us to derive larger abundances from sr ii and ba ii lines approaching saturation . another pattern traces the heavy element abundances in the metal - poor halo star , which has normal abundances of the lighter elements and a deficiency of the heaviest elements . the pattern found in this star may be considered representative of the so - called weak component of the . the third pattern traces one outcome of nucleosynthesis @xcite . in each figure , these patterns are rescaled to match the stellar eu abundance . + + + + + the heavy element distributions in the three stars on the horizontal branch : , , and . symbols are the same as in figure [ rprorgfig ] . , title="fig:",width=278 ] + the heavy element distributions in the three stars on the horizontal branch : , , and . symbols are the same as in figure [ rprorgfig ] . , title="fig:",width=278 ] + the heavy element distributions in the three stars on the horizontal branch : , , and . symbols are the same as in figure [ rprorgfig ] . , title="fig:",width=278 ] + the largest numbers of elements are detected in the red giants , shown in figure [ rprorgfig ] . nearly all of the rare earth elements ( 57 @xmath4 71 , plus neighboring elements ba and hf ) are detected in each case . ir , a member of the third peak , and the actinide th are also detected in three of these stars . among the lighter elements , sr , y , and zr are always detectable , and mo and ru are detected in two of the stars . fewer elements are detectable in the subgiants ( figure [ rprosgfig ] ) or field rhb stars ( figure [ rprohbfig ] ) . yet in all of these stars the rare earth elements and those beyond match the pattern in established by previous studies . only when comparing the ratios of the lighter elements with the rare earth elements do significant differences emerge . this is revealed in figures [ rprorgfig][rprohbfig ] by the sr , y , and zr abundances that are sometimes enhanced relative to the scaled main component of the ( solid red line ) . numerous studies have pointed out this characteristic in other samples of -enhanced field stars , among them @xcite , @xcite , @xcite , @xcite , and @xcite . curiously , one -enhanced metal - poor star analyzed by @xcite hints that the so - called universal nature may extend to elements at all three peaks ( i.e. , se , te , os , pt ) , regardless of the variations that may occur between the first and second peaks . we quantify the difference between the lighter and heavier elements in our sample by calculating the dispersions in the [ sr / ba ] , [ sr / eu ] , [ ba / eu ] , and [ eu / yb ] ratios . the [ sr / ba ] and [ sr / eu ] ratios have dispersions of 0.29 and 0.27 dex , respectively . the [ ba / eu ] and [ eu / yb ] ratios , however , have dispersions of only 0.19 and 0.16 dex . the latter two values are consistent with the measurement uncertainties listed in table [ atmtab ] . this indicates that a real dispersion exists between the lighter and heavier elements , even among members of the class of highly -- enhanced stars . the small dispersion in the [ ba / yb ] ratios ( 0.18 dex ) also suggests that the truncated @xcite did not affect the nucleosynthesis of the heavy elements found in these stars . in the solar system , approximately 9498 per cent of the eu is predicted to have originated via nucleosynthesis , and the remainder is attributed to nucleosynthesis ( e.g. , @xcite ) . in contrast , only about 1115 per cent of ba in the solar system is predicted to have originated via nucleosynthesis . thus the [ ba / eu ] ratio is commonly used to quantify the relative @xmath0- and contributions to a given star or ensemble of stars . the mean [ ba / eu ] ratio for all 13 stars in our sample is @xmath70.71 @xmath21 0.05 ( @xmath14 0.19 ) . this compares well with other recent estimates by various methods ( e.g. , @xmath70.70 , @xcite ; @xmath70.81 , @xcite ; @xmath70.65 , @xcite ; @xmath70.78 , @xcite ) . highly -- enhanced stars have now been identified across a broad range of evolutionary states that low - mass stars experience . the first of these such stars identified , , was a red giant , and for the following 15 years all other members of the class of highly -- enhanced stars identified were also red giants . @xcite identified the first highly -- enhanced star on the main sequence , . our study has identified six new highly -- enhanced subgiants and three such rhb stars . highly -- enhanced subgiants have been found before by the hamburg / eso r - process enhanced star survey @xcite , but the stars relatively weak lines and the moderate spectral resolution used by @xcite limited their ability to derive abundances of many elements . @xcite also studied the same three rhb stars that we have analyzed , but the focus of that study was not on deriving large numbers of elements to study the abundances patterns in detail . the unease among practitioners in the field that highly -- enhanced stars are found almost exclusively among giants ( e.g. , @xcite ) is put to rest . the bias is observational . our results and those of @xcite demonstrate that this phenomenon is not limited to red giants . enhanced [ c / fe ] and [ n / fe ] ratios were found in the first highly -- enhanced star discovered , ( [ c / fe ] @xmath160.88 , [ n / fe ] @xmath161.01 ; @xcite ) . this raised the question of whether the c- and n - enhancement were related to the enhancement . is included in our sample . is also highly - enhanced in c and n ( [ c / fe ] @xmath161.78 , [ n / fe ] @xmath162.05 ) . exhibits modest c and n enhancement ( [ c / fe ] @xmath160.69 , [ n / fe ] @xmath160.49 ) , and c is also modestly enhanced in ( [ c / fe ] @xmath160.58 ) . all three of these stars are subgiants . detections of the ch , nh , or cn bands in other highly -- enhanced stars indicate that [ c / fe ] and [ n / fe ] are not enhanced . four subgiants and all three rhb stars yield only uninteresting upper limits on [ n / fe ] , but the upper limits derived from the non - detection of ch indicate that [ c / fe ] @xmath231.0 in each of the rhb stars . the fraction of carbon - enhanced metal - poor stars is known to increase with decreasing metallicity , and these stars typically constitute @xmath8 732 per cent of local stellar samples with [ fe / h ] @xmath242.0 @xcite . these estimates range by a factor of several because of differing definitions of c enhancement , metallicity ranges considered , sample sizes and selection techniques , and median distance from the sun . some of these stars are found in binary systems and presumably acquired their c enhancement through mass transfer from a more evolved companion . long - term velocity monitoring and detailed studies of the abundance patterns indicate that this explanation can not be applied to other carbon - enhanced metal - poor stars ( e.g. , @xcite , @xcite , @xcite ) . in our sample of highly -- enhanced stars , finding 2 of 13 stars as carbon - enhanced ( 15 per cent ) is consistent with previous estimates . there is no compelling reason to suspect that the and c - enhancement are related . the large , homogeneous dataset of @xcite affords a unique opportunity to compare abundances in individual stars to abundances in large samples of stars with similar stellar parameters . these comparisons will be minimally affected by uncertainties in the stellar parameters , non - lte effects , poorly - known atomic data , etc . subtle but significant outlying abundance ratios may be identified this way . we use this technique to identify any abundance signatures among the light elements , those with 6 @xmath4 30 , that are peculiar to the highly -- enhanced stars relative to normal metal - poor stars . any peculiar abundance signatures found could be associated with the supernovae or other sites capable of generating large amounts of elements beyond the second peak . we perform this test for each of the 13 stars in our sample . we begin by identifying a comparison sample of stars for each -enhanced star that are in the same evolutionary state , have within @xmath21 200 k , and are within @xmath21 0.3 dex of the metallicity of the -enhanced star . we exclude other highly -- enhanced stars from the comparison samples . we also exclude stars whose heavy elements indicate the presence of a large amount of material , since the present - day composition of these stars may not accurately reflect the composition of their birth clouds . the comparison samples typically include 2030 stars from @xcite , but there as many as 61 stars or as few as 6 stars for comparison in some cases . figures [ comprgfig][comphbfig ] illustrate the [ x / fe ] ratios for each enhanced star . the mean @xmath21 1@xmath22 standard deviation of each [ x / fe ] ratio for the comparison sample is marked by the shaded box . the number of comparison stars is indicated in each figure caption . these figures illustrate that the light - element abundance ratios of highly -- enhanced stars rarely deviate by more than 1@xmath22 from those of other metal - poor stars . like other metal - poor stars , the highly -- enhanced stars show [ @xmath25/fe ] ratios ( where @xmath25 indicates o , mg , si , ca , and ti ) enhanced by factors of a few relative to the solar ratios . the iron - group elements are typically found in solar or sub - solar ratios in both the enhanced stars and the normal metal - poor stars . by construction , the elements sr , y , zr , ba , and eu are significantly enhanced relative to the comparison stars . + + + + + comparison of abundances in the enhanced stars on the horizontal branch with the average abundances of other stars with within @xmath21 200 k and [ fe / h ] within @xmath21 0.3 dex of each . for , , and , the numbers of stars in the comparison samples are eight , seven , and six , respectively . symbols are the same as in figure [ comprgfig ] . , title="fig:",width=278 ] + comparison of abundances in the enhanced stars on the horizontal branch with the average abundances of other stars with within @xmath21 200 k and [ fe / h ] within @xmath21 0.3 dex of each . for , , and , the numbers of stars in the comparison samples are eight , seven , and six , respectively . symbols are the same as in figure [ comprgfig ] . , title="fig:",width=278 ] + comparison of abundances in the enhanced stars on the horizontal branch with the average abundances of other stars with within @xmath21 200 k and [ fe / h ] within @xmath21 0.3 dex of each . for , , and , the numbers of stars in the comparison samples are eight , seven , and six , respectively . symbols are the same as in figure [ comprgfig ] . , title="fig:",width=278 ] + finally , we illustrate the mean deviations of each star from its comparison sample in figure [ meandelplot ] . these values are listed in table [ meandeltab ] . the points in figure [ meandelplot ] represent an unweighted mean of [ x / fe]@xmath26[x / fe]@xmath27 calculated for each element x in each enhanced star . the error bars in figure [ meandelplot ] represent the standard deviation of the mean . when multiple ionization states of an element are detected , each ionization state is considered separately . species are only illustrated in figure [ meandelplot ] when the mean of the differences has been computed from more than three stars . none of the differences illustrated in figure [ meandelplot ] is significant at the 2@xmath22 level or greater , and most light - element ratios in enhanced stars are consistent with the comparison samples at the 1@xmath22 level . the mean of the absolute values of these differences for 15 [ x / fe ] ratios from mg to zn is 0.015 @xmath21 0.008 dex ( @xmath14 0.030 dex ) , or 3.5 per cent . in summary , we find no compelling evidence that any of the light - element ratios are significantly different in the highly -- enhanced stars and metal - poor stars without high levels of enhancement . this complements results derived from a sample of seven moderately -- enhanced stars @xcite . that study found no difference between the [ x / fe ] ratios ( where x represents any of 13 elements between mg and ni ) among the -enhanced and unenhanced groups of stars . mean residuals between each [ x / fe ] ratio in each star and its comparison sample . the uncertainties correspond to the standard deviation of the mean . only comparisons between more than three stars are illustrated . none of the mean residuals is significant at the 2@xmath22 level . , width=321 ] .mean differences between @xmath0-process - enhanced stars and comparison samples [ meandeltab ] [ cols="^,^,^,^,^",options="header " , ] we conclude that the nucleosynthetic sites responsible for producing the large enhancement did not produce any detectable light - element abundance signatures unique from the core - collapse supernovae widely believed to have produced the metals observed in the vast majority of metal - poor stars . this scenario may come about naturally by decoupling the sites of large enhancement from core - collapse supernovae entirely ; e.g. , if mergers of neutron stars are the source of this kind of enhancement . alternatively , if development of the conditions required for this kind of nucleosynthesis is effectively decoupled from the deep regions of supernovae where light - element production occurs , this would also satisfy the observations . a decoupling between light - element and nucleosynthesis has been proposed previously ( e.g. , @xcite ; @xcite ) to satisfy the observed dispersion of [ eu / fe ] ratios at low metallicity compared to , e.g. , the [ mg / fe ] ratios , as illustrated in figure 14 of @xcite . our results reaffirm this situation , extend it to other elements , and constrain the amount of variation allowed in the individual light - element yields . we have examined the abundance patterns found in stars with high levels of enrichment as indicated by their enhanced [ eu / fe ] ratios . our study differs from the survey of @xcite by deriving abundances from higher - resolution spectra that cover a wider wavelength range . thirteen of the 313 metal - poor stars analyzed by @xcite are identified as members of the class of @xmath0-ii stars after correcting for offsets in the derived [ eu / fe ] ratios ( [ eu / fe ] @xmath150.8 ) . four of these stars are red giants whose high levels of enhancement were known previously . we identify six new subgiants ( including the msto phase ) and three new rhb members of this class . @xcite also identified one such star on the main sequence . once limited to the realm of red giants , highly -- enhanced stars are now known all across a broad range of stellar evolutionary states . the first highly -- enhanced star discovered , , was also enhanced in c and n , which signaled a possible connection between these chemical signatures . only 2 of the 13 stars in our sample ( including ) are enhanced in c and n , however , offering no compelling evidence for such a connection . we also find no compelling evidence to suggest that a disproportionately high fraction of highly -- enhanced stars are members of binary or multiple star systems , confirming the conclusions of a dedicated radial velocity survey by @xcite . the dispersion in the [ sr / ba ] and [ sr / eu ] ratios is larger than the dispersion in the [ ba / eu ] or [ yb / eu ] ratios , indicating that the robust pattern does not extend to the elements between the first and second peaks even within the class of highly -- enhanced stars . we compare the light - element ( @xmath28 30 ) abundance pattern in each highly -- enhanced star with the light - element abundance pattern in a comparison set of stars that have similar stellar parameters but lack high levels of enhancement . this test reveals no obvious light - element abundance signatures that are unique to the highly -- enhanced stars . the nucleosynthetic sites responsible for producing the large enhancement apparently did not produce any detectable light - element abundance signatures distinct from normal core - collapse supernovae responsible for much of the early metal production in the universe . we thank the referee , t. beers , for a prompt and helpful report . this research has made use of nasa s astrophysics data system bibliographic services , the arxiv preprint server operated by cornell university , the simbad and vizier databases hosted by the strasbourg astronomical data center , and the atomic spectra database hosted by the national institute of standards and technology . iraf is distributed by the national optical astronomy observatories , which are operated by the association of universities for research in astronomy , inc . , under cooperative agreement with the national science foundation . c. s. is supported by the u.s . national science foundation ( grant ast 12 - 11585 ) aoki w. , norris j. e. , ryan s. g. , beers t. c. , ando h. 2002 , , 567 , 1166 aoki w. et al . 2005 , , 632 , 611 aoki w. , beers t. c. , honda s. , carollo d. 2010 , , 723 , l201 aoki w. , suda t. , boyd r. n. , kajino t. , famiano m. a. 2013 , , 766 , l13 arlandini c. , kppeler f. , wisshak k. , gallino r. , lugaro m. , busso m. , straniero o. 1999 , , 525 , 886 asplund m. , grevesse n. , sauval a. j. , scott p. 2009 , , 47 , 481 barklem p. s. et al . 2005 , , 439 , 129 beers t. c. , preston g. w. , shectman s. a. 1992 , , 103 , 1987 beers t. c. , christlieb n. 2005 , , 43 , 531 bernstein r. , shectman s. a. , gunnels s. m. , mochnacki s. , athey a. e. 2003 , , 4841 , 1694 bisterzo s. , gallino r. , straniero o. , cristallo s. , kppeler f. 2011 , , 418 , 284 bonifacio p. et al . 2009 , , 501 , 519 boyd r. n. , famiano m. a. , meyer b. s. , motizuki y. , kajino t. , roederer i. u. 2012 , , 744 , l14 burbidge e. m. , burbidge g. r. , fowler w. a. , hoyle f. 1957 , rev . mod . phys . , 29 , 547 burris d. l. , pilachowski c. a. , armandroff t. e. , sneden c. , cowan j. j. , roe h. 2000 , , 544 , 302 cameron a. g. w. 1973 , , 15 , 121 carollo d.et al . 2012 , , 744 , 195 carollo d. , freeman k. , beers t. c. , placco v. m. , tumlinson j. , martell s. l.2014 , , 788 , 180 cayrel r.et al . 2004 , , 416 , 1117 cescutti g. , chiappini c. , hirschi r. , meynet g. , frischknecht u. 2013 , , 553 , a51 christlieb n. , schrck t. , frebel a. , beers t. c. , wistozki l. , reimers d.2008 , , 484 , 721 cohen j. g. et al . 2005 , , 633 , l109 cohen j. , christlieb n. , thompson i. , mcwilliam a. , shectman s. , reimers d. , wisotzki l. , kirby e. 2013 , , 778 , 56 cowan j. j. , burris d. l. , sneden c. , mcwilliam a. , preston g. w. 1995 , , 439 , l51 cowan j. j. et al . 2002 , , 572 , 861 cowan j. j. , roederer i. u. , sneden c. , lawler j. e. 2011 , rr lyrae stars , metal - poor stars , and the galaxy , 223 demarque p. , woo j .- h . , kim y .- c . , yi s. k. 2004 , , 155 , 667 fabbian d. , asplund m. , barklem p. s. , carlsson m. , kiselman d.2009 , , 500 , 1221 fields b. d. , truran j. w. , cowan j. j. 2002 , , 575 , 845 for b .- q . , sneden c. 2010 , , 140 , 1694 franois p. 2007 , , 476 , 935 frebel a. et al . 2006 , , 652 , 1585 gallino r. , arlandini c. , busso m. , lugaro m. , travaglio c. , straniero o. , chieffi a. , limongi m. 1998 , , 497 , 388 gustafsson b. , edvardsson b. , eriksson k. , jrgensen u. g. , nordlund . , plez b. 2008 , , 486 , 951 hansen t. , andersen j. , nordstrm b. , buchhave l. a. , beers t. c. 2011 , , 743 , l1 hansen c. j. et al . 2012 , , 545 , a31 hill v. et al . 2002 , , 387 , 560 honda s. et al . 2004a , , 152 , 113 honda s. , aoki w. , kajino t. , ando h. , beers t. c. , izumiura h. , sadakane k. , takada - hidai m. 2004b , , 607 , 474 honda s. , aoki w. , ishimaru y. , wanajo s. , ryan s. g. 2006 , , 643 , 1180 ivans i. i. , sneden c. , gallino r. , cowan j. j. , preston g. w. 2005 , , 627 , l145 johnson j. a. , bolte m. 2002 , , 579 , 616 kelson d. d. 2003 , , 115 , 688 kratz k .- , farouqi k. , pfeiffer b. , truran j. w. , sneden c. , cowan j. j. 2007 , , 662 , 39 kurucz r. l. , bell b. 1995 , kurucz cd - rom , cambridge , ma : smithsonian astrophysical observatory lai d. k. , johnson j. a. , bolte m. , lucatello s. 2007 , , 667 , 1185 lee y. s. , et al . 2013 , , 146 , 132 lind k. , asplund m. , barklem p. s. 2009 , , 503 , 541 lind k. , asplund m. , barklem p. s. , belyaev a. k. 2011 , , 528 , a103 lucatello s. , beers t. c. , christlieb n. , barklem p. s. , rossi s. , marstellar b. , sivarani t. , lee y. s. 2006 , , 652 , l37 mashonkina l.christlieb n. 2014 , , 565 , a123 mcwilliam a. , preston g. w. , sneden c. , shectman s. 1995a , , 109 , 2736 mcwilliam a. , preston g. w. , sneden c. , searle l. 1995b , , 109 , 2757 mcwilliam a. 1998 , , 115 , 1640 montes f. et al.2007 , , 671 , 1685 norris j. e. , ryan s. g. , beers t. c. 1997 , , 488 , 350 placco v. m. et al . 2011 , , 142 , 188 preston g. w. , sneden c. 2000 , , 120 , 1014 preston g. w. , sneden c. 2001 , , 122 , 1545 preston g. w. , sneden c. , thompson i. b. , shectman s. a. , burley g. s. 2006 , , 132 , 85 qian y .- z . , wasserburg g. j. 2008 , , 687 , 272 roederer i. u. , kratz k .- l . , frebel a. , christlieb n. , pfeiffer b. , cowan j. j. , sneden c. 2009 , , 698 , 1963 roederer i. u. , cowan j. j. , karakas a. i. , kratz k .- l . , lugaro m. , simmerer j. , farouqi k. , sneden c. 2010 , , 724 , 975 roederer i. u. et al . 2012 , , 203 , 27 roederer i. u. 2013 , , 145 , 26 roederer i. u. , preston g. w. , thompson i. b. , shectman s. a. , sneden c. , burley g. s. , kelson d.2014a , , 147 , 136 roederer i. u. , preston g. w. , thompson i. b. , shectman s. a. , sneden c.2014b , , 784 , 158 roederer i. u.et al . 2014c , , 791 , 32 rossi s. , beers t. c. , sneden c. , sevastyanenko t. , rhee j. , marstellar b. 2005 , , 130 , 2804 ryan s. g. , aoki w. , norris j. e. , beers t. c. 2005 , , 635 , 349 siqueira mello c. et al . 2013 , , 550 , a122 siqueira - mello c.et al . 2014 , , 565 , a93 sneden c. a. 1973 , ph.d . thesis , univ . of texas at austin sneden c. , parthasarathy m. 1983 , , 267 , 757 sneden c. , preston g. w. , mcwilliam a. , searle l. 1994 , , 431 , l27 sneden c. , mcwilliam a. , preston g. w. , cowan j. j. , burris d. l. , armosky b. j. 1996 , , 467 , 819 sneden c. et al . 2003a , , 591 , 936 sneden c. , preston g. w. , cowan j. j. 2003 , , 592 , 504 sneden c. , cowan j. j. , gallino r. 2008 , , 46 , 241 sneden c. , lawler j. e. , cowan j. j. , ivans i. i. , den hartog e. a. 2009 , , 182 , 80 sobeck j. s. et al . 2011 , , 141 , 175 takeda y. , zhao g. , chen y .- q . , takada - hidai m. 2002 , , 54 , 275 truran j. w. , cowan j. j. , pilachowski c. a. , sneden c. 2002 , , 114 , 1293 wallerstein g. , greenstein j. l. , parker r. , helfer h. l. , aller l. h. 1963 , , 137 , 280 wasserburg g. j. , qian y .- z . 2000 , , 529 , l21 westin j. , sneden c. , gustafsson b. , cowan j. j. 2000 , , 530 , 783 yong d. et al . 2013 , , 762 , 26 has been identified previously using traditional signatures of enrichment from a binary companion that passed through the thermally - pulsing asymptotic giant branch ( tp - agb ) phase of evolution : radial velocity variations , a strong ch g band indicating substantial c enhancement , and an abundance pattern indicative of enhancement @xcite . @xcite derived [ c / fe ] @xmath162.38 , [ ba / fe ] @xmath162.75 , [ eu / fe ] @xmath161.70 , and [ pb / fe ] @xmath163.62 for . the abundance pattern of is illustrated in figure [ sprofig ] . relative to the template pattern shown for comparison , elements at the first peak ( 38 @xmath4 40 ) are underabundant and lead ( @xmath30 82 ) is overabundant . this is explained by the high neutron - to - seed ratio found in low - mass tp - agb stars at low metallicity ( e.g. , @xcite ) . also exhibits similar characteristics , as well as a weak but detectable c@xmath31 band - head near 5165 . @xcite identified as a carbon - enhanced metal - poor star , but @xcite published the first detailed abundance pattern derived from high resolution spectroscopic observations . @xcite also uncovered evidence of radial velocity variations for this star . @xcite derived [ c / fe ] @xmath162.03 , [ ba / fe ] @xmath161.81 , [ eu / fe ] @xmath160.82 , and [ pb / fe ] @xmath162.43 for . the abundance pattern of is also illustrated in figure [ sprofig ] . each of these stars exhibits unmistakable evidence that the high levels of eu enhancement observed can not be mainly attributed to nucleosynthesis .

all lorentzian spacetimes with vanishing invariants constructed from the riemann tensor and its covariant derivatives are determined . a subclass of the kundt spacetimes results and we display the corresponding metrics in local coordinates . some potential applications of these spacetimes are discussed .

gr-qc0209024

a curvature invariant of order @xmath0 is a scalar obtained by contraction from a polynomial in the riemann tensor and its covariant derivatives up to the order @xmath0 . in general there are 14 algebraically independent curvature invariants of zeroth order , the simplest being the ricci scalar . many papers are devoted to studying the properties of the zeroth order curvature invariants ( see @xcite @xcite and references therein ) but higher order curvature invariants remain largely unexplored . recently it was shown that for spacetimes in which the ricci tensor does not possess a null eigenvector , an appropriately chosen set of zeroth order curvature invariants contains all of the information that is present in the riemann tensor @xcite . this is certainly not true for vacuum petrov type n spacetimes with nonzero expansion or twist , all of whose zeroth and first order curvature invariants vanish , but for which there are non - vanishing curvature invariants of the second order @xcite ; and for some non - flat spacetimes in which all curvature invariants of all orders vanish @xcite-@xcite . in this paper we shall determine all lorentzian spacetimes for which all curvature invariants of all orders are zero . indeed , we shall prove the following : [ th : main ] all curvature invariants of all orders vanish if and only if the following two conditions are satisfied : 1 . the spacetime possesses a non - diverging sfr ( shear - free , geodesic null congruence ) . 2 . relative to the above null congruence , all curvature scalars with non - negative boost - weight vanish . the analytic form of the condition ( a ) , expressed relative to any spin basis where @xmath1 is aligned with the null congruence in question , is simply @xmath2 and the analytic form of condition ( b ) is @xmath3 spacetimes that satisfy condition ( a ) belong to kundt s class @xcite ( also , see section [ sect : spacetimes ] and [ ap : kundt ] ) . condition ( b ) implies that the spacetime is of petrov type iii , n , or o ( see eq . ( [ eq : psivanish ] ) ) with the ricci tensor restricted by ( [ eq : phivanish ] ) and ( [ eq : ricvanish ] ) . ( note : throughout this paper we follow the notation of @xcite ; @xmath4 is not the cosmological constant , it is the ricci scalar up to a constant factor . ) the ghp formalism @xcite assigns an integer , called the boost weight , to curvature scalars and certain connection coefficients and operators . this is important for this work , and we shall summarize some of the key details of this notion in the next section . the outline for the rest of the paper is as follows . section [ sect : sufficient ] is devoted to the proof that the above conditions are sufficient for vanishing of curvature invariants . the `` necessary '' part of theorem 1 is proved in section [ sect : necessary ] . the curvature invariants constructed in this section may be also useful for computer - aided classification of spacetimes . kundt s class of spacetimes admits a conveniently specialized system of coordinates , and so it is possible to classify and explicitly describe all spacetimes with vanishing curvature invariants . this is briefly summarized in section [ sect : spacetimes ] , and some of the details are presented in [ ap : kundt ] . we conclude with a discussion . perhaps the best known class of spacetimes with vanishing curvature invariants are the pp - waves ( or plane - fronted gravitational waves with parallel rays ) , which are characterized as ricci - flat ( vacuum ) type n spacetimes that admit a covariantly constant null vector field . the vanishing of curvature invariants in pp - wave spacetimes has been known for a long time @xcite , and the spacetimes obtained here can perhaps be regarded as extensions and generalizations of these important spacetimes . in many applications ( e.g. , in vacuum pp - wave spacetimes ) the resulting exact solutions have a five - dimensional isometry group acting on three - dimensional null orbits ( which includes translations in the transverse direction along the wave front ) and hence the solutions are plane waves . however , the solutions studied here need not be plane waves , and are not necessarily vacuum solutions . in particular , non - vacuum spacetimes with a covariantly constant null vector are often referred to as generalized pp - wave and typically have no further symmetries ( the arbitrary function in the metric is not subject to a further differential equation , namely laplace s equation , when the ricci tensor has the form of null radiation ) . the pp - wave spacetimes have a number of remarkable symmetry properties and have been the subject of much research @xcite . for example , the existence of a homothety in spacetimes with plane wave symmetry and the scaling properties of generally covariant field equations has been used to show that all generally covariant scalars are constant @xcite and that metrics with plane wave symmetry trivially satisfy every system of generally covariant vacuum field equations _ except _ the einstein equations @xcite . in addition to pp - waves , presently there are known to be three classes of metrics with vanishing curvature invariants : the conformally flat pure radiation spacetime given in @xcite ; the vacuum petrov type - n nonexpanding and nontwisting spacetimes @xcite ( this class contains the pp - waves ) ; and vacuum petrov type - iii nonexpanding and nontwisting spacetimes @xcite . naturally all of these spacetimes are subcases of the class studied here . the spacetimes studied in @xcite also belong to our class . there are two important applications of the class of spacetimes obtained in this paper . a knowledge of all lorentzian spacetimes for which all of the curvature invariants constructed from the riemann tensor and its covariant derivatives are zero , which implies that all covariant two - tensors constructed thus are zero except for the ricci tensor , will be of potential relevance in the equivalence problem and the classification of spacetimes , and may be a useful first step toward addressing the important question of when a spacetime can be uniquely characterized by its curvature invariants . more importantly perhaps , the spacetimes obtained in this paper are also of physical interest . for example , pp - wave spacetimes are exact solutions in string theory ( to all perturbative orders in the string tension ) @xcite and they are of importance in quantum gravity @xcite . it is likely that all of the spacetimes for which all of the curvature invariants vanish will have similar applications and it would be worthwhile investigating these metrics further . finally , we note that it is possible to generalize theorem [ th : main ] by including spacetimes with non - vanishing cosmological constant . the assumptions regarding the weyl and traceless ricci tensors remain the same . even in this general case , the invariants constructed from the weyl tensor , the traceless ricci tensor and their arbitrary covariant derivatives vanish . the only non - vanishing curvature invariants are order zero curvature invariants constructed as various polynomials of the cosmological constant . it must be noted , however , that there may exist other types of spacetimes with constant curvature invariants . before tackling the proof of the main theorem , we make some necessary definitions and establish a number of auxiliary results . we shall make use of the newmann - penrose ( np ) and the compacted ( ghp ) formalisms @xcite . throughout we work with a normalized spin basis @xmath5 , i.e. @xmath6 the corresponding null tetrad is given by @xmath7 with the only nonzero scalar products @xmath8 we also recall that o_ao^a = 0= .[oo_ii ] the spinorial form of the riemann tensor @xmath9 is r _ x_abcd + |x _ + _ ab + |_cd , [ rozpr ] where x_abcd = _ abcd + ( + ) and @xmath10 , with @xmath11 the scalar curvature . the weyl spinor @xmath12 is related to the weyl tensor @xmath13 by c _ _ abcd + | _ . [ wtenz ] projections of @xmath14 onto the basis spinors @xmath15 , @xmath16 give five complex scalar quantities @xmath17 . the ricci spinor @xmath18 is connected to the traceless ricci tensor @xmath19 _ ab a -s_ab . [ sbsric ] the projections of @xmath20 onto the basis spinors @xmath15 , @xmath16 are denoted @xmath21 , @xmath22 , @xmath23 , @xmath24 , @xmath25 , and @xmath26 . ( [ eq : psivanish ] ) and ( [ eq : phivanish ] ) imply _ abcd & = & _ 4 o_a o_b o_c o_d - 4 _ 3 o_(a o_b o_c , [ weylspin ] + _ ab & = & _ 22 o_a o_b |o _ |o _ - 2 _ 12 o_b ) |o _ |o _ - 2 _ 21 o_a o_b | _ |o _ .[riccispin ] following the convention established in @xcite , we say that @xmath27 is a weighted quantity ( a scalar , a spinor , a tensor , or an operator ) of type @xmath28 if for every non - vanishing scalar field @xmath29 a transformation of the form @xmath30 representing a boost in @xmath31@xmath32 plane and a spatial rotation in @xmath33@xmath34 plane , transforms @xmath27 according to @xmath35 the boost weight , @xmath36 , of a weighted quantity is defined by @xmath37 directional derivatives are defined by d&= l^ _ = o^a |o^ , & = m^ _ = o^a |^ , + d & = n^ _ = |^ , & = |m^ _ = |o^ and thus ^ = |^ d + o^a |o^ d - |o^ - o^a |^ . [ covder ] in the ghp formalism new derivative operators @xmath38 , @xmath39 , @xmath40 , @xmath41 , which are additive and obey the leibniz rule , are introduced . they act on a scalar , spinor , or tensor @xmath27 of type @xmath28 as follows : let us explicitly write down how the operators @xmath38 , @xmath39 , @xmath40 , @xmath41 act on the basis spinors o^a&=- , |o^ = -||^ , & = -o^a , |^=-||o^ , + o^a&=- , |o^ = -||^ , & = -o^a , |^=-||o^ , + o^a&=- , |o^ = -||^ , & = -o^a , |^=-||o^ , [ derbasis ] + o^a&=- , |o^ = -||^ , & = -o^a , |^=-||o^ . the types and boost - weights of various weighted quantities encountered in the ghp formalism are summarized in table 1 . @xmath42 henceforth we shall assume that conditions ( a ) and ( b ) of theorem [ th : main ] hold , and by implication that equations ( [ eq : kundt ] ) , ( [ eq : psivanish ] ) , ( [ eq : phivanish ] ) , ( [ eq : ricvanish ] ) hold also . without loss of generality we also assume that @xmath1 and @xmath43 are parallely propagated along @xmath44 . analytically , this condition takes the form of the following two additional relations : @xmath45 assumptions ( a ) , ( b ) and conditions ( [ eq : pwlog ] ) greatly simplify the form of the spin - coefficient equations , the bianchi and the commutators identities @xcite . most of these relations assume the form @xmath46 . some of the non - trivial relations are as follows : @xmath47 extending an idea introduced in @xcite , we make the following key definition . [ balancedscalar ] we shall say that a weighted scalar @xmath27 with the boost - weight @xmath36 is _ balanced _ if & ^-b=0 b<0 + & = 0 b0 . we can now prove the following . [ lem : bal2 ] if @xmath27 is a balanced scalar then @xmath48 is also balanced . by definition , a weighted scalar @xmath27 of type @xmath28 is changed by complex conjugation to a weighted scalar @xmath48 of type @xmath49 . the boost weight , however , remains unchanged . let us also recall that @xmath50 and hence that @xmath51 as desired . [ lem : bal1 ] if @xmath27 is a balanced scalar then @xmath52 are all balanced as well . let @xmath36 be the boost - weight of a balanced scalar @xmath27 . from table [ tab : boost ] we see that the scalars listed in ( [ eq : bal1 ] ) have boost - weights @xmath36 , @xmath53 , @xmath53 , @xmath54 , respectively . hence , it suffices to show that the following quantities are all zero : @xmath55 this follows from the leibniz rule and from equations ( [ eq : sct ] ) , ( [ eq : scr ] ) , ( [ eq : scs ] ) , ( [ eq : sck ] ) , ( [ eq : bianchi21 ] ) , and ( [ eq : bianchi3 ] ) . next we show that the scalars in ( [ eq : bal2 ] ) are balanced as well . these scalars have boost weights @xmath56 , @xmath36 , @xmath36 , @xmath53 , respectively . hence , it suffices to show that the following scalars are all zero : @xmath57 the vanishing of the first quantity follows immediately from definition [ balancedscalar ] . using the commutator relation ( [ eq : com2 ] ) we have @xmath58 as desired . vanishing of the quantity involving @xmath41 follows by considering the complex - conjugate of ( [ eq : ethrel ] ) and using the relation @xmath59 and lemma [ lem : bal2 ] . to show that the quantity involving @xmath39 vanishes , we employ ( [ eq : sct ] ) , ( [ eq : com1 ] ) , and ( [ eq : ethrel ] ) to obtain @xmath60 we now proceed inductively and conclude that @xmath61 [ lem : sum ] if @xmath62 , @xmath63 are balanced scalars both of type @xmath64 then @xmath65 is a balanced scalar of type @xmath64 as well . the sum @xmath65 satisfies & & _ 1+_2 ^p |^q ( _ 1+_2 ) , + & & ^-b(_1+_2)=^-b_1+^-b_2=0 and thus it is a balanced scalar of type @xmath64 . [ lem : bal3 ] if @xmath62 , @xmath63 are balanced scalars then @xmath66 is also balanced . let @xmath67 be the respective boost weights . boost - weights are additive and hence the boost - weight of the product is @xmath68 . setting @xmath69 and applying the leibniz rule gives @xmath70 for @xmath71 , the factor @xmath72 vanishes . for @xmath73 , we have @xmath74 and hence the other factor vanishes . therefore the entire sum vanishes . a balanced spinor is a weighted spinor of type @xmath75 whose components are all balanced scalars . [ lem : prodspin ] if @xmath76 , @xmath77 are balanced spinors then @xmath78 is also a balanced spinor . the product @xmath78 is a weighted spinor of type @xmath75 and its components are balanced scalars thanks to lemma [ lem : bal3 ] . [ lem : covder ] a covariant derivative of an arbitrary order of a balanced spinor @xmath79 is again a balanced spinor . applying the covariant derivative ( [ covder ] ) to a balanced spinor @xmath79 , we obtain @xmath80 from table [ tab : boost ] , it follows that @xmath81 is again a weighted spinor of type @xmath75 and its components are balanced scalars due to ( [ derbasis ] ) and lemmas [ lem : bal2 ] , [ lem : bal1 ] , and [ lem : sum ] . [ lem : contrac ] a scalar constructed as a contraction of a balanced spinor is equal to zero . a scalar constructed as a contraction of a balanced spinor also has zero boost - weight , and therefore vanishes by definition [ balancedscalar ] . let us explain more intuitively how this works . a balanced spinor has the form @xmath82 where @xmath83 are balanced scalars and @xmath84 are the basis spinors ( products of @xmath1s , @xmath43s , @xmath85s , and @xmath86s ) . since the boost - weight of each @xmath83 is negative and the boost - weight of each @xmath87 is zero it follows that the boost - weight of each @xmath84 is positive , i.e. there are more @xmath1s and @xmath85s then @xmath43s and @xmath86s in @xmath84 . as a consequence of ( [ oo_ii ] ) a full contraction of each @xmath84 vanishes . in a nutshell : all scalars constructed as a contraction of a balanced spinor vanish because each term contains more @xmath88 s than @xmath89 s . we are now ready to prove that the conditions ( a ) and ( b ) of theorem [ th : main ] are sufficient for vanishing of all curvature invariants . from table [ tab : boost ] and eqs . ( [ eq : bianchi21 ] ) , ( [ eq : bianchi3 ] ) , ( [ eq : bianchi22 ] ) , and ( [ eq : bianchi4 ] ) it follows that the weyl spinor ( [ weylspin ] ) and the ricci spinor ( [ riccispin ] ) and their complex conjugates ( lemma [ lem : bal2 ] ) are balanced spinors . their products and covariant derivatives of arbitrary orders are balanced spinors as well ( lemmas [ lem : prodspin ] , [ lem : covder ] ) . finally , due to lemma [ lem : contrac ] and eqs . ( [ rozpr])([sbsric ] ) all curvature invariants constructed from the riemann tensor and its covariant derivatives of arbitrary order vanish . in this section we consider a spacetime with vanishing curvature invariants and prove that this spacetime satisfies the conditions listed in theorem [ th : main ] . the ricci scalar , being a curvature invariant , must vanish . to prove the other conditions , we consider various zeroth , first , and second order invariants formed from the weyl and the ricci spinors , as well as the newmann - penrose equations and the bianchi identities . in the following we will employ these newmann - penrose equations and the bianchi identities & & _ 3 - _ 2 - _ 21 + _ 20 - 2 = 2 _ 1 - 3 _ 2 + 2 _ 3 - _ 4 + & & - 2 _ 10 + 2 _ 11 + _ 20 - 2 _ 21 + _ 22 , + & & _ 4 - _ 3 + _ 20 - _ 21 = 3 _ 2 - 4 _ 3 + _ 4 + & & - 2 _ 10 + 2 _11 + _ 20 -2_21 + _ 22 , [ rceb_1 - 4 ] + & & _ 22 + _ 11 - _ 21 - _ 12 + 3 = ( + ) _ 22 + 2 ( + ) _ 11 + & & -(+ 2 ) _ 21 - ( 2 + ) _ 12 - _ 10 - _ 01 + _ 02 + _ 20 , + & & _ 2 - _ 3 + _ 22 - _ 21 + 2 = _ 4 - 2 _ 3 + 3 _ 2 - 2 _ 1 + & & + _ 22 - 2 _ 21 - 2 _ 12 + 2 _ 11 + _ 20 . first , we consider the well - known invariants i= , j= . [ invij ] it is generally known that these invariants vanish if and only if the petrov type is iii , n , or 0 . in the following we choose the spinor basis @xmath1 and @xmath43 in such a way that for the petrov types iii and n , @xmath1 is the multiple eigenspinor of the weyl spinor . thus the condition ( [ eq : psivanish ] ) is satisfied . we consider the three petrov types case by case . * petrov type n : + @xmath90 + demanding that the following invariant _ | _ = ( 2_4 |_4 ) ^2 [ inv_1 ] vanishes we obtain = 0 . [ rcekappa ] in further calculations we assume that ( [ rcekappa ] ) is valid . + vanishing of another invariant _ 2= k^f f e e _ m m l l |k^m m l l _ f f e e = ( 24 _ 4 |_4)^2 ( + ) ^4 , [ inv_2 ] where k^f f e e _ m m l l = ^ _ a b c d , [ k ] implies that = = 0 [ rcesigmarho ] and therefore the condition ( [ eq : kundt ] ) , i.e. the condition ( a ) of theorem 1 , holds . + substituting ( [ rcekappa ] ) and ( [ rcesigmarho ] ) into eqs . ( [ rcenp_1,2 ] ) we get _ 00=_01=_10=0 .[phi_0001 ] + and finally from the vanishing of the invariant [ isest ] ^aba _ aba = 4_11 ^ 2 + 2_02_20 + 2_00_22 -4_10_12 - 4_01_21 using ( [ phi_0001 ] ) it follows @xmath91 and thus the condition ( [ eq : phivanish ] ) , i.e. the condition ( b ) of theorem 1 , is also satisfied . * petrov type iii : + providing that the weyl spinor @xmath92 is of petrov type iii , we can construct another spinor @xmath93 _ abcd = _ abef = -2_3 ^ 2 o_a o_b o_c o_d [ psi3 ] which is of petrov type n. now we can construct analogical curvature invariants from @xmath93 as we did from @xmath94 for type n and again conclude that @xmath95 and @xmath96 for metrics with all curvature invariants vanishing . * petrov type 0 : + recall that the totally symmetric pleba ' nski spinor is defined by _ abcd = _ cd ) . [ pleb ] its components are + in analogy with the petrov classification of the weyl tensor , it is possible to define the plebaski - petrov type ( pp - type ) of the plebaski spinor @xcite . thus vanishing of curvature invariants analogous to @xmath97 and @xmath98 ( [ invij ] ) constructed from the plebaski spinor implies that the pp - type is iii , n , or o. + for the pp - types iii and n we can argue as we did for the cases of the petrov types iii and n and conclude that @xmath95 and @xmath96 . substituting these results into ( [ pleb_0 - 4 ] ) we obtain @xmath99 and thus the pp - type iii is excluded . + it remains to consider the pp - type 0 case . + without loss of generality we can choose a null tetrad so that @xmath100 then @xmath101 in ( [ pleb_0 - 4 ] ) implies @xmath102 hence , the vanishing of the invariant ( [ isest ] ) gives @xmath103 demanding @xmath104 in ( [ pleb_0 - 4 ] ) we get @xmath105 thus the only non - vanishing component of the ricci spinor is @xmath106 . the bianchi identities ( [ rceb_1 - 4 ] ) take the form @xmath107 which implies @xmath108 up to now we have proven that spacetimes with vanishing invariants constructed from the weyl and the ricci _ spinors _ and their arbitrary derivatives satisfy the conditions ( a ) and ( b ) of theorem 1 . invariants constructed from the riemann _ tensor _ and its derivatives are combinations of corresponding spinorial invariants . thus one could argue that there might exist a very special class of spacetimes for which all tensorial invariants vanish even though there exist nonzero spinorial invariants . to prove that this does not happen we now construct several tensorial curvature invariants . they may also be useful for computer - aided classification of spacetimes . the curvature invariants shown in ( [ invij ] ) can be given as & & c_^ c_^ - i c^*_^ c_^ , + & & c_^ c_^ c_^ - i c_^ c^*_^ c_^ , where @xmath109 denotes the dual of the weyl tensor . their vanishing implies that the petrov type is iii , n , or o. * petrov type n : + from the vanishing of i_1=c^ ; c _ ; c^ ; c _ ; = 8 i_1 = 2 ( 4 _ 4 |_4 ) ^2 @xmath110 follows . + to obtain @xmath111 we have to construct the tensorial invariants @xmath112 , @xmath113 , @xmath114 : i_2=c^ ; c _ ; c^ ; c _ ; = 4i_2 + 2i_3 |i_3 + 4 ( i_4 + |i_4 ) , where _ 3 = k^f f e e _ f f e e = 6 ( 4 _ 4)^2 , + _ 4 = ^ _ ^ _ + = 18 ( 4 _ 4)^4 , and thus @xmath112 is equal to @xmath115 \ .\ ] ] @xmath113 is defined by i_3=c^ ; c _ ; c^ ; c _ ; = 16 ( 2 i_2 + i_5 + |i_5 ) , where _ 5 = k^f f e e _ m m l l k^m m l l _ f f e e = 2 ^ 10 3 ^ 2 ( _ 4)^4 , and so it takes the form @xmath116 \ .\ ] ] the curvature invariant @xmath114 is a linear combination of @xmath112 and @xmath113 @xmath117 demanding @xmath118 we obtain @xmath119 and then the vanishing of @xmath113 implies @xmath120 + as in the spinorial case , the newmann - penrose equations ( [ rcenp_1,2 ] ) imply @xmath121 , and finally the invariant constructed from the traceless ricci tensor corresponding to ( [ isest ] ) @xmath122 is zero if @xmath123 . thus the conditions ( a ) and ( b ) of theorem 1 are satisfied . * petrov type iii : + in analogy to ( [ psi3 ] ) , the tensor @xmath124 can be defined in terms of the petrov type - iii weyl tensor @xmath125 which is traceless , has the same symmetries as the weyl tensor and is of the petrov type n @xmath126 we can construct curvature invariants from @xmath127 similar to those made from @xmath128 for type n and again show that their vanishing leads to @xmath95 and @xmath96 . * petrov type o : + it is possible to define the traceless pleba ' nski tensor corresponding to ( [ pleb ] ) which is endowed with the same symmetries as the weyl tensor in terms of traceless ricci tensor @xmath129 ( see @xcite ) @xmath130}_{\ \ \d ] } + \d^{[\na}_{\ \ [ \ng } s_{\d ] \nl } s^{\nb ] \nl } - { { { \scriptstyle{\frac{1}{6 } } } } } \d^{[\na}_{\ \ [ \ng } \d^{\nb]}_{\ \ \d ] } s_{\nm \nn } s^{\nm \nn}\ .\ ] ] with the pleba ' nski tensor we can proceed in the same way as in the spinorial case . another way to prove necessity of the conditions ( [ eq : kundt])([eq : ricvanish ] ) for the vanishing of all curvature invariants is to use the result from paper @xcite that the invariants @xmath131 , @xmath132 , and @xmath133 constructed from the ricci spinor are equal to zero only if all four eigenvalues of the ricci tensor are equal to zero . consequently the segre types of the ricci tensor are @xmath134 ( i.e. pp - type n with the only non - vanishing components @xmath135 and @xmath136 @xcite ) , @xmath137 ( i.e. pp - type o with the only non - vanishing component @xmath136 ) , or @xmath138 ( i.e. vacuum ) . in non - vacuum cases , the multiple null eigenvector @xmath139 of the ricci tensor may in general differ from the repeated null vector of the weyl tensor @xmath140 ; however , by demanding vanishing of the mixed invariants @xmath141 , @xmath142 , @xmath143 @xcite constructed from both the weyl and the ricci tensors we arrive at the condition @xmath144 . then the bianchi identities for non - vanishing @xmath145 , @xmath146 , @xmath147 , and @xmath106 imply @xmath148 . and finally the vanishing of the invariant ( [ inv_2 ] ) for p - types iii and n results in @xmath149 . let us describe the metric , written in an adapted coordinate form , of all of the spacetimes with vanishing curvature invariants ( i.e. those satisfying theorem [ th : main ] ) . we recall that spacetimes with vanishing curvature invariants satisfy ( [ eq : kundt ] ) ( i.e. , belong to the kundt class @xcite ) , are of the petrov type iii , n , or o ( i.e. , the weyl spinor @xmath150 is of the form ( [ weylspin ] ) ) , and the ricci spinor @xmath151 has the form ( [ riccispin ] ) that corresponds to the ricci tensor @xmath152 consequently , the pleba ' nski spinor ( [ pleb ] ) has the form _ abcd = -2_21 ^ 2 o_a o_b o_c o_d and the pleba ' nski - petrov type ( pp - type ) is n for @xmath153 or o for @xmath154 . we note that for pp - type n , using a null rotation about @xmath31 we can transform away the ricci component @xmath106 and using further a boost in the @xmath155 plane and a spatial rotation in the @xmath156 plane set @xmath157 . for pp - type o it is possible to set @xmath158 by performing a boost in the @xmath155 plane . the ricci tensor ( [ ricci ] ) has all four eigenvalues equal to zero and its segre type is @xmath134 ( @xmath153 ) , @xmath137 ( @xmath154 and @xmath159 ) , or @xmath138 ( for vacuum @xmath160 ) . the most physically interesting non - vacuum case @xmath137 corresponds to a pure null radiation field @xcite . it can be shown that an electromagnetic field compatible with ( [ ricci ] ) has to be null . other energy - momentum tensors , including a fluid with anisotropic pressure and heat flux , can correspond to a ricci tensor of pp - type o. indeed , it is known that no energy - momentum tensor for a spacetime corresponding to a ricci tensor of segre type @xmath161 ( or its degeneracies ) can satisfy the weak energy conditions ( see @xcite , p 72 ) , and hence spacetimes of pp - type n are not regarded as physical in classical general relativity and hence usually attention is restricted to pp - type o models . however , for mathematical completeness we will discuss all of the models here . in addition , in view of possible applications in high energy physics in which the energy conditions are not necessarily satisfied , these models may have physical applications . the most general form of the kundt metric in adapted coordinates @xmath162 @xcite is s^2=2d u [ h du+d v + w d+|w d| ] -2 p^-2d , [ dskundt ] where the metric functions @xmath163 satisfy the einstein equations ( see @xcite and [ ap : kundt ] ) . for the spacetimes considered here , we may , without loss of generality , put @xmath164 . the following tables summarize the kundt metrics for different subcases in the studied class . .all spacetimes with vanishing invariants with @xmath165 and @xmath159 , i.e. pp - n , are displayed . for details and references see [ ap : kundt ] . [ cols="^,^,^",options="header " , ] it is of interest to find the conditions for which the repeated null eigenvector @xmath31 of the weyl tensor is recurrent for the kundt class . the vector @xmath31 satisfies @xmath166}l^{\na;\nb}=0\ ] ] and its covariant derivative has in general the form @xmath167 performing a boost in the @xmath155 plane @xmath168 with @xmath169 satisfying @xmath170 i.e. putting @xmath171 , @xmath172 ( see e.g. @xcite for transformation properties of np quantities ) we obtain _ ; = ( + |)l _ _ + |(l_m_-l_m _ ) + ( l__-l _ _ ) with @xmath173 satisfying _ lg_=l_;+l_;=2 ( + | ) l _ _ . [ almkill ] this normalization is called `` an almost killing normalization '' in @xcite . as @xmath174 can not be transformed away by any transformation of the tetrad preserving the @xmath140-direction and one can even show that @xmath175 is invariant with respect to all tetrad transformations preserving the @xmath140-direction , the repeated null eigenvector @xmath31 of the weyl tensor is proportional to a recurrent vector @xmath173 if and only if @xmath176 . to summarize : all kundt spacetimes with @xmath176 admit a recurrent null vector . finally , let us present the relation between quantities @xmath177 and @xmath178 given in @xcite and np - quantities when @xmath31 satisfies ( [ almkill ] ) @xmath179 the pp - wave spacetimes have a number of important physical applications , many of which also apply to the other spacetimes obtained in this paper . as mentioned earlier , pp - wave spacetimes are exact vacuum solutions to string theory to all order in @xmath180 , the scale set by the string tension @xcite . horowitz and steif @xcite generalized this result to include the dilaton field and antisymmetric tensor fields which are also massless fields of string theory using a more geometrical approach . they showed that pp - wave metrics satisfy all other field equations that are symmetric rank two tensors covariantly constructed from curvature invariants and polynomials in the curvature and their covariant derivatives , and since the curvature is null all higher order corrections to einstein s equation constructed from higher powers of the riemann tensor automatically vanish . therefore , all higher - order terms in the string equations of motion are automatically zero . many of the spacetimes obtained here will have similar properties . in addition , solutions of classical field equations for which the counter terms required to regularize quantum fluctuations vanish are of particular importance because they offer insights into the behaviour of the full quantum theory . the coefficients of quantum corrections to ricci flat solutions of einstein s theory of gravity in four dimensions have been calculated up to two loops . in particular , a class of ricci flat ( vacuum ) lorentzian 4-metrics , which includes the pp - wave spacetimes and some special petrov type iii or n spacetimes , have vanishing counter terms up to and including two loops . thus these lorentzian metrics suffer no quantum corrections to all loop orders @xcite . in view of the vanishing of all quantum corrections it is possible that all of the metrics summarized in tables 2 4 are of physical import and merit further investigation . string theory in pp - wave backgrounds has been studied by many authors @xcite , partly in a search for a connection between quantum gravity and gauge theory dynamics . such string backgrounds are technically tractable and have direct applications to the four dimensional conformal theories from the point of view of a duality between string and gauge theories . indeed , pp - waves provide exact solutions of string theory @xcite and type - iib superstrings in this background were shown to be exactly solvable even in of the presence of the rr five - form field strength @xcite . as a result the spectrum of the theory can be explicitly obtained . this model is expected to provide some hints for the study of superstrings on more general backgrounds . there is also an interesting connection between pp - wave backgrounds and gauge field theories . it is known that any solution of einstein gravity admits plane - wave backgrounds in the penrose limit @xcite . this was extended to solutions of supergravities in @xcite . it was shown that the super - pp - wave background can be derived by the penrose limit from the @xmath181 backgrounds in @xcite . the penrose limit was recognized to be important in an exploration of the ads / cft correspondence beyond massless string modes in @xcite . maximally supersymmetric pp - wave backgrounds of supergravity theories in eleven- and ten - dimensions have attracted great interests @xcite . recently the idea that our universe is embedded in a higher dimensional world has received a great deal of renewed attention @xcite . due to the importance of branes in understanding the non - perturbative dynamics of string theories , a number of classical solutions of branes in the background of a pp - wave have been studied ; in particular a new brane - world model has been introduced in which the bulk solution consists of outgoing plane waves ( only ) , which avoids the problem that the evolution requires initial data specified in the bulk @xcite . finally , in @xcite an example of non - isometric spacetimes with non - vanishing curvature scalars which can not be distinguished by curvature invariants was presented . this example represents a solution of einstein s equation with a negative cosmological term and a minimally coupled massless scalar field . in this paper we have noted the existence of a class of spacetimes in which all of the curvature invariants are constants ( depending on the cosmological constant ) . these results and their extensions to higher dimensions are consequently also of physical interest . the authors would like to thank r. zalaletdinov for discussions . a.a.c . and r.m were supported , in part , by research grants from n.s.e.r.c . a.p . and v.p . would like to thank dalhousie university for the hospitality while this work was carried out . v.p . was supported by grant gacr-202/00/p030 and a.p . by grant gacr-202/00/p031 . let us present here more details on the spacetimes with all curvature invariants vanishing , which were briefly summarized in section [ sect : spacetimes ] . [ app : ppn ] , [ app : pponull ] , and [ app : ppovac ] correspond to tables [ tab : pp - n ] , [ tab : pp - onull ] , and [ tab : pp - ovacuum ] , respectively . since all of the spacetimes satisfying condition ( a ) of theorem 1 ( i.e. , that satisfy ( [ eq : kundt ] ) ) , belong to the kundt class we start with the metric given by ( [ dskundt ] ) in coordinates @xmath182 @xcite , where the null tetrad is given by l=_v , n=_u-(h+p^2w|w)_v+p^2(|w_+w_| ) , m = p _ .[kundttetrada ] only certain coordinate transformations and tetrad rotations can be performed which preserve the form of the metric ( [ dskundt ] ) and the null tetrad ( [ kundttetrada ] ) ( see @xcite ) @xmath183 in particular , in these coordinates it is not possible , in general , to simultaneously simplify the forms of the ricci spinor components in ( [ ricci ] ) in pp - types n and o by boosts and null and spatial rotations . in most cases it is possible to specialize the solution form by an appropriate choice of coordinates , thereby narrowing the range of allowed coordinate transformations . the remaining coordinate freedom will be described below on a case by case basis . * petrov type iii + the functions @xmath184 , @xmath185 , and @xmath186 have to satisfy equations which follow from the fact that we assume the petrov types iii , n , or o ( @xmath187 ) and have the ricci tensor of the form ( [ ricci ] ) . + for the kundt class , @xmath188 vanishes identically and @xmath189 and thus w,_vv=0 . [ wvv ] then @xmath190 and @xmath191 reduce to & & _ 2=- [ h,_vv+2(p,_p,_|-pp , _ ) + p^2(2w,_v|-|w,_v)]=0 , [ hvv ] + & & w,_v|=|w,_v , [ wvcz ] + & & 2w,_v|=w,_v|w,_v [ wvwv ] and r_=-2(p),_|=0 .[trp ] the gaussian curvature , @xmath192 , of wave surfaces determined uniquely by the spacetime geometry is a spacetime invariant and since it vanishes for the studied class of spacetimes they are characterized by plane wave surfaces @xcite . + from ( [ trp ] ) , using a type i coordinate transformation ( [ eq : coord1 ] ) we can put p=1 .[p ] this restricts the type i transformations to =e^i ( u ) + f(u ) . [ eq : coord1a]then , equations ( [ wvv ] ) , ( [ wvcz ] ) , and ( [ wvwv ] ) , together with r_=- w,_v+w,_v^2=0 , [ wvz ] after another type i coordinate transformation ( [ eq : coord1a ] ) , give without loss of generality @xcite w(u , v,,|)=n+w_0(u,,| ) [ w ] with @xmath193 or @xmath194 . if @xmath195 , the wave surfaces are polarized , and consequently type i transformations are further restricted to @xmath196 + finally , eqs . ( [ hvv ] ) and r_uv =- h,_vv- w,_v |w,_v=0 [ hvvd ] are identical and have the solution @xcite h(u , v,,|)=n+vh_1(u,,| ) + h_0(u,,| ) .[h ] + employing ( [ wvv])([wvwv ] ) , ( [ p ] ) , ( [ wvz ] ) , and ( [ hvvd ] ) , the remaining einstein equations are r_uu= 2(|wh,_v+wh,_v| ) -2h , _ + h,_v(w,_|+|w , _ ) -(h,_,_v + h,_|w,_v ) + -hw,_v|w,_v -(w|w,_uv+|ww,_uv ) + w,_u|+|w,_u -w,_||w , _ + + ( w|w,_v-|ww,_v)(-|w,_+w,_| ) + ( w,_|^2+|w,_^2 + w^2 = -2[_22 - 2(w_21+|w_12)][eeruu ] + r_u= -h,_v+ ( w,_uv -w,_+|w , _ + w,_vw,_|-|w,_vw , _ ) -w,_v ( w |w,_v+|w w,_v ) + = -2_12 .[eeruz ] + the np quantities read & = = = = 0 , + & = -=2=-2=- n , + & = -n-|w_0,_| , + & = -n -(w_0,_|+|w_0 , _ ) , + & = n -vh_1,_|+2n -2n -h_0,_|- ( w_0|w_0),_| , + & = n+h_1 -n + ( w_0,_|-|w_0 , _ ) , + _ 3&=-2h_1,_|+|w_0,_-w_0,_|| + 2n -2n , [ psit ] + _ 4 & = v - h_1,_|| + 2n + 2n + 4n + & + h_1|w_0,_|-h_0,_||+|w_0,_u| + |w_0(|w_0,_-w_0,_|| ) + & + 2n -2n .[psic ] + the remaining coordinate freedom for the case @xmath195 is @xmath197 one could , without loss of generality , take @xmath198 . + the remaining coordinate freedom for the @xmath193 case is @xmath199 one could therefore without loss of generality take @xmath200 . + in general , we can not make any further progress unless we identify a specific source , e.g. , null radiation or null electromagnetic field , which then yields additional field equations through eqs . ( [ eeruu ] ) and ( [ eeruz ] ) ( and , for example , the maxwell equations ) . * petrov type n + in this case @xmath201 and eq . ( [ psit ] ) constitutes an additional differential equation that must be satisfied . this equation can be integrated to obtain a more specialized form of the metric . * petrov type o + in this case @xmath202 , i.e. right hand sides of eqs . ( [ psit ] ) , ( [ psic ] ) must vanish . these equations can be integrated to obtain a fully specified form of the metric . conformally ricci - flat pure radiation metrics , studied in @xcite , all belong to this class . in fact , in @xcite the authors present all pure radiation solutions belonging to kundt s class of petrov types n and o for @xmath203 and of petrov types iii , n , and o for @xmath176 . for pure radiation , one of the remaining einstein equations simply serve to define the radiation energy - density . for specific sources , such as a null electromagnetic field , these equations ( e.g. , eqs . ( [ eeruu ] ) and ( [ eeruz ] ) ) lead to additional differential equations . in the case of vacuum , all solutions can be explicitly written down ( see the next subsection ) . * petrov type iii + for @xmath193 the einstein equation @xmath204 ( [ eeruz ] ) becomes @xmath205,_\zeta=0\ .\ ] ] using a type ii transformation ( [ eq : coord2 ] ) , ( [ eq : coord2n=0 ] ) ( see @xcite for a discussion ) and ( [ w ] ) , ( [ h ] ) , its solution turns out to be + the metric functions are subject to the only remaining einstein equation ( [ eeruu ] ) _ 22 = h_0,_-(w_0w_0,_|| + w_0,_u|+w_0,_|^2 ) .[nuliiit0_f22 ] + the np quantities take the form + this choice of metric form restricts the type i , ii , and iii transformations ( [ eq : coord2n=0 ] ) to four following cases &=&+f(u ) ; + v&=&v+g_1(u)+|g_1(u)|+g_0(u ) ; + u&=&a_1 u+a_0 , v=v / a_1 ; + u&=&h(u),v= v / h,_u - ( h,_uu / h,_u^2 ) | , where @xmath206 , @xmath207 , @xmath208 , and @xmath209 are arbitrary functions of @xmath210 and @xmath211 , @xmath212 are real constants . * n + for type - n spacetimes ( @xmath201 @xmath213 @xmath214 ) , @xmath215 can be transformed away ( [ eq : coord2n=0 ] ) @xcite and thus the metric functions ( [ nuliiit0_wh ] ) are w=0 , h = h_0(u,,|)[nulnt0_wh ] and the np quantities ( [ nuliiit0_np ] ) read + the only remaining einstein equation ( [ nuliiit0_f22 ] ) now becomes _ 22=h_0 , _ .[enulln ] + the remaining coordinate freedom comes from a mixed type i and ii transformation : [ eq : coord.2n ] & = & e^i ( + f(u ) ) , v=v+|f,_u+f,_u| + g(u ) , + & = & h_0 -g,_u+f,_u |f,_u - |f,_uu- f,_uu | , where @xmath216 is a real constant , and @xmath210 is determined up to a affine transformation . + these spacetimes are known as generalized pp - wave solutions . in the case of a null electromagnetic field , energy momentum tensor , eq . ( [ enulln ] ) and maxwell s equations lead to a further differential equation for @xmath217 , whose solution is known @xcite . * petrov type o + all metrics belonging to this class are given in @xcite ( see ( 12 ) therein ) . + the condition @xmath218 from ( [ nulnt0_np ] ) is @xmath219 with the solution h_0=h_02(u ) [ nulot0_h ] after a transformation ( [ eq : coord.2n ] ) . the metric functions are thus given by ( [ nulnt0_wh ] ) with ( [ nulot0_h ] ) and the einstein equation ( [ enulln ] ) becomes @xmath220 . + the coordinates are fixed up to an 8-parameter group of transformations : & = & e^i ( + f(u ) ) , + v & = & v / a_1+|f,_u+f,_u| + ( f|f),_u+g_0 , + u & = & a_1 u + a_0 , where @xmath221 is a complex - valued solution of @xmath222 and @xmath216 , @xmath211 , @xmath212 , @xmath208 are real constants . * petrov type iii + for @xmath195 , the einstein equation @xmath204 ( [ eeruz ] ) is @xmath223_{,\zeta}= -\frac{w_{0},_{\bar{\zeta}}+\overline{w}_{0},_{\zeta}}{\zeta+\bar{\zeta}}\ .\ ] ] + again using a type ii transformations ( [ eq : coord2n=1 ] ) ( as in @xcite ) , we obtain the solution ( [ w ] ) , ( [ h ] ) + the remaining einstein equation ( [ eeruu ] ) then reads _ 22= ( + | ) , _ -w_0,__0,_| .[nuliiit_f22 ] + the np quantities are as follows + the remaining coordinate freedom is * petrov type n + all type - n pure radiation metrics were found in @xcite . + for type - n spacetimes ( @xmath201 @xmath213 @xmath224 ) , @xmath215 can be transformed away again using ( [ eq : coord2n=0 ] ) , ( [ eq : coord.22iii ] ) , and the metric functions ( [ nuliiit_wh ] ) take the form w= , h= + h_0(u,,| ) [ nulnt_wh ] with @xmath217 satisfying ( [ nuliiit_f22])_22 = ( + | ) , _ . [ euunulln ] + the np quantities are as follows + the remaining coordinate freedom is given by where @xmath225 is a real constant , and @xmath226 is an arbitrary real function . * petrov type o + all conformally flat pure radiation metrics ( both with @xmath176 and @xmath203 ) , generalizing the solutions found in @xcite and @xcite , were given in @xcite . the physical interpretation of this class of spacetimes is discussed in @xcite . + the equation @xmath218 in ( [ nulnt_np ] ) has the solution ( see ( 16 ) in @xcite ) h_0=h_00(u)[1+h_01(u)+|h_01(u)| + h_02(u)](+| ) [ nulot_h ] which is to be substituted into the metric functions ( [ nulnt_wh ] ) . + the einstein equation ( [ euunulln ] ) turns to be @xmath227 + the only coordinate freedom is a translation of @xmath228 by an imaginary constant and @xmath210 is determined up to a affine transformation . + einstein - maxwell null fields , massless scalar fields and neutrino fields _ do not _ exist for this class of metrics @xcite . the vacuum petrov types - iii , n , and o kundt metrics are reviewed in @xcite ( chap . the form of the metric , and the remaining coordinate freedom are as in [ app : pponull ] , with the vacuum condition imposing an additional constraint on the metric parameters . * petrov type iii + for vacuum petrov type - iii spacetimes , the metric and the np quantities are given by ( [ nuliiit0_wh ] ) and ( [ nuliiit0_np ] ) , respectively , where @xmath217 satisfies the einstein equation ( [ nuliiit0_f22 ] ) h_0,_-(w_0w_0,_|| + w_0,_u|+w_0,_|^2)=0 . + petrov @xcite found an example belonging to this class ( in different coordinates ) s^2=x(v - e^x)du^2 - 2dudv + e^x(dx^2+e^-2udz^2 ) . * petrov type n pp waves + the metric functions ( [ nulnt0_wh ] ) and the np quantities ( [ nulnt0_np ] ) of vacuum petrov type - n spacetimes satisfy ( [ enulln ] ) h_0,_=0 , h_0=h_00(u,)+|h_00(u,| ) .[vacnt0_h ] + these spacetimes belong to the class of pp - wave spacetimes ( see chap . 21.5 in @xcite ) which admit a covariantly constant null vector that is consequently also a null killing vector . * petrov type o flat spacetime + for flat spacetime , eq . ( [ vacnt0_h ] ) reduces the solution ( [ nulot0_h ] ) to @xmath200 , a flat metric . * petrov type iii + for petrov type - iii vacuum spacetimes with non - vanishing @xmath174 , the remaining einstein equation ( [ nuliiit_f22 ] ) turns out to be ( + | ) , _ = w_0,__0,_| .[eq : e - vaciii ] its solution determines the metric ( [ nuliiit_wh ] ) and the np quantities ( [ nuliiit_np ] ) . + an example from this class , which was originally found by kundt @xcite , with @xmath230 satisfying @xmath231 , is known ( see @xcite ) . * n + for petrov type - n vacuum spacetimes , the einstein equation ( [ eq : e - vaciii ] ) simplifies to @xmath232 with the solution h_0=[h_00(u,)+|h_00(u,|)](+| ) . [ vacnt_h ] the metric and np quantities are then given by ( [ nulnt_wh ] ) and ( [ nulnt_np ] ) with ( [ vacnt_h ] ) . * petrov type o flat spacetime + for the flat spacetime the condition @xmath218 in ( [ nulnt_np ] ) , i.e. @xmath233 has the solution @xmath234(\zeta+{\bar\zeta})\ .\ ] ] ghniau j and debever r 1956 les invariants de courbure de lespace de riemann quatre dimensions _ acad . belg . bull . cl . sci . _ * 42 * 114 sneddon g e 1996 the identities of the algebraic invariants of the four - dimensional riemann tensor 1059 + sneddon g e 1998 the identities of the algebraic invariants of the four - dimensional riemann tensor ii 1659 + sneddon g e 1999 the identities of the algebraic invariants of the four - dimensional riemann tensor iii 5905 zakhary e and carminati j 2001 on the problem of algebraic completeness for the invariants of the riemann tensor i 1474 + carminati j , zakhary e and mclenaghan r g 2002 on the problem of algebraic completeness for the invariants of the riemann tensor ii 492 + carminati j and zakhary e 2002 on the problem of algebraic completeness for the invariants of the riemann tensor iii 4020 joly g c and maccallum m a h 1990 computer - aided classification of the ricci tensor in general relativity 541 seixas w 1991 extension to the computer - aided classification of the ricci tensor 1577 ludwig g 1980 space - times with geodesic , shear free , twist - free , nonexpanding rays 2543 + ludwig g 1981 space - times with geodesic , shear free , twist - free , nonexpanding rays ii 2240 metsaev r r 2002 type iib green - schwarz superstring in plane wave ramond - ramond background _ _ b * 625 * 70 + metsaev r r and tseytlin a a 2002 exactly solvable model of superstring in ramond - ramond plane wave background _ phys . d * 65 * 126004 blau m , figueroa - ofarrill j , hull c and papadopoulos g 2002 a new maximally supersymmetric background of iib superstring theory _ jhep _ * 0201 * 047 + blau m , figueroa - ofarrill j , hull c and papadopoulos g 2001 penrose limits and maximal supersymmetry _ class . quant . * 19 * l87 + blau m , figueroa - ofarrill j and papadopoulos g 2001 penrose limits , supergravity and brane dynamics _ preprint _ hep - th/0202111 + meessen p 2002 a small note on pp - wave vacua in 6 and 5 dimensions d * 65 * 087501 rubakov v and shaposhnikov m 1983 extra space - time dimensions : towards a solution to the cosmological constant problem _ phys . _ b * 125 * 139 + arkani - hamed n , dimopoulos s and dvali g 1998 the hierarchy problem and new dimensions at a millimeter _ phys . _ b * 429 * 263 + randall l and sundrum r 1999 a large mass hierarchy from a small extra dimension 3370 + randall l and sundrum r 1999 an alternative to compactification 4690

we present results from a photometric survey of 22 nuclear rings , aiming to provide insight into their star formation properties , including age distribution , dynamical timescales , star formation rates , and galactic bar influence . we find a clear relationship between the position angles and ellipticities of the rings and those of their host galaxies , which indicates the rings are in the same plane as the disk and circular . we use population synthesis models to estimate ages of each-emitting ( hii ) region , which range from 1 myr to 10 myrs throughout the rings . we find that approximately half of the rings contain azimuthal age gradients that encompass at least 25 of the ring , although there is no apparent relationship between the presence or absence of age gradients and the morphology of the rings or their host galaxies . ngc 1343 , ngc 1530 , and ngc 4321 show clear bipolar age gradients , where the youngest regions are located near the two contact points of the bar and ring . we speculate in these cases that the gradients are related to an increased mass inflow rate and/or an overall higher gas density in the ring , which would allow for massive star formation to occur on short timescales , after which the galactic rotation would transport the regions around the ring as they age . two - thirds of the barred galaxies show correlation between the locations of the youngest region(s ) in the ring and the location of the contact points , which is consistent with predictions from numerical modeling .

0709.2915

nuclear rings are viewed as a type of starburst , consisting of distinct compact areas of young massive stars near the centers of spiral galaxies . the intense bursts , attributed by a star formation rate ( sfr ) that significantly exceeds the average found among normal galaxies , can dominate the overall star formation in early hubble type barred spiral galaxies , and dramatically alter the structure of the host galaxy with respect to morphology , gas and dust content , metallicity , and stellar population ( e.g. , martinet 1995 ; buta & combes 1996 ; kormendy & kennicutt 2004 ) . typically located within the central kiloparsec , nuclear rings contain a mixture of neutral and ionized gas and dust with masses of @xmath2 to @xmath3 ( heller & shlosman 1996 ; rubin , kenney , & young 1997 ) . the great majority of nuclear rings are found in barred spiral galaxies , where the bar provides essential fuel for the formation and evolution of the ring via shock - induced channeling of disk gas to the nuclear regions ( e.g. , schwarz 1981 ; combes & gerin 1985 ; shlosman , begelman , & frank 1990 ; athanassoula 1992 ; knapen et al . 1995b ; heller & shlosman 1994 ; heller & shlosman 1996 ) . the shocks can act as instigators for angular momentum loss , which is crucial for directing the gas towards the central regions and forming nuclear rings . in such a picture massive sf would result and be most intense near where the bar dust lanes intersect the ring , called the contact points . the contact points can lead the bar by as much as 90 with respect to the galactic rotation ( knapen et al . 1995b ; heller & shlosman , 1996 ; regan , vogel , & teuben 1997 ; regan et al . 1999 ; ryder , knapen , & takamiya 2001 ; allard , peletier , & knapen 2005 ) . while many galaxies with nuclear star - forming rings have been identified ( knapen 2005 states that 20% of local spirals host them ) , it is hard to generalize on the physical mechanisms behind the triggering and propagation of massive sf due to the inconsistent and incomplete nature of the existing data . to allow for a more statistical approach to answering some of the mysteries of these rings , we performed an imaging survey of 73 galaxies in @xmath0 and the @xmath4 and @xmath5 bands , many of which host nuclear rings . an overview of this survey is presented in knapen et al . ( 2006 ; hereafter paper i ) . in paper i we classified the morphology of the nuclear and circumnuclear hii regions and confirmed that ( 1 ) most late - type galaxies have a patchy circumnuclear appearance in @xmath0 , ( 2 ) nuclear rings occur primarily in spiral types sa - sbc , and ( 3 ) found that the presence or absence of a close companion does not significantly affect the @xmath0 morphology around the nucleus nor the strength of the nuclear @xmath0 peak . of these 73 galaxies , we identified 22 that host nuclear rings as judged from our @xmath0 images , of which 19 were previously known as such . this paper isolates the individual regions forming the nuclear rings . we present information on the observations and data reduction in section 2 . we derive morphological parameters of the rings using ellipse fitting , and then compare those to the disk characteristics , in section 3 . in section 4 , we identify and analyze the regions forming the rings , compute their equivalent widths ( ews ) , and estimate error budgets . using evolutionary models , we convert the ews of the regions to ages and discuss the assumptions we make with respect to the type of starburst activity occurring in and around the rings in section 5 . we present our age distribution analysis , star formation rate estimations , and a discussion of the stellar bar dynamics with respect to the rings in section 6 , before summarizing our conclusions in section 7 . the ring images presented here are part of a larger william herschel telescope ( wht ) imaging survey conducted to characterize the nuclear and circumnuclear morphology in nearby galaxies ( paper i ) . we selected galaxies with some evidence for @xmath0 structure in their central regions , either from the literature ( e.g. , buta & combes 1996 ; buta & crocker 1993 ) or from our own past work . we adopted the definition of a nuclear ring , nuclear spiral , and psuedo - ring from buta & crocker ( 1993 ) in paper i and follow it here for consistency . since one of the aims of our overall study is the detailed analysis of nuclear rings , our sample was biased towards sbc galaxies known to host such features . we chose rings with enough structure to analyze , and which are considered to be the brightest feature in the circumnuclear region . given the nature of our original survey from which the targets were selected , a distance bias is to some extent unavoidable . we recognize the wide range of distances and ring sizes in the sample , and confirm that rings in farther galaxies are larger than those in nearby galaxies . however , since our selection was with respect to the angular size of the ring , we find that there is no correlation of distance to the number of identifiable regions . the images presented in this paper have been obtained using the auxiliary port ( aux port ) camera on the 4.2 m wht , operated by the isaac newton group in la palma . the camera s circular field of view ( fov ) is @xmath61@xmath78 in diameter , imaged with a 1024 @xmath8 1024 pixel tek ccd . the angular sizes of the nuclear rings in our sample vary between @xmath9 and @xmath10 , which fit well into the fov . we used harris @xmath4 and @xmath5 filters , and one of four narrow - band @xmath0 filters , depending on the systemic velocity of the galaxy . we refer the reader to table 1 of paper i for a listing of the complete sample , along with their global parameters . table 1 of the present paper lists morphological type , ring ellipticity , position angle ( pa ) of the disk and bar , and derived ring size for our subsample of 22 galactic nuclear rings as identified in paper i. we observed spectrophotometric standards to calibrate the @xmath0 images , and photometric standards to calibrate the _ b _ and _ i _ band images . we observed the standards through the same filters as the galaxies , with exposure times ranging from 5 to 20 seconds . the instrumental magnitude of each star in each filter was obtained using the iraf package phot . this photometry package calculates the magnitude , among other parameters , based on user - defined values for the radius and annulus of a sky aperture . to convert from instrumental magnitude to ab magnitude for the spectrophotometric standard stars , an interpolation method was used based on results from oke ( 1990 ) . oke defines values of ab at points where no ambiguity exists for instrument calibration problems due to spectral resolution . therefore , to obtain magnitudes for the filters used in our survey , magnitudes with respect to the wavelengths just above and below the interested filter wavelength were chosen . interpolating between these values yields the absolute magnitude for the five @xmath0 filters . the resulting integrated fluxes of the rings ( on the order of 10@xmath11 @xmath12 ) and luminosities of the regions ( on the order of 10@xmath13 @xmath14 ; see appendix a ) are consistent with published data ( e.g. , pogge 1989 ; storchi - bergmann , wilson , & baldwin 1996 ; regan et al . 1996 ; sheth et al . we fitted the nuclear rings of the 22 galaxies with the iraf task ellipse . we obtained the central position of each ring from the peak in the @xmath5-band image , and fixed the center while iteratively fitting elliptical isophotes to the @xmath0 image . we masked out the nucleus and foreground stars that could affect the fitting process , and provided initial estimates of the major axis length , ellipticity , and pa . the task produced measurements of the azimuthally integrated intensity , observed ellipticity , and pa of the ring major axis as a function of radius . we determined the best - fitting ellipse , which corresponded to the peak isophotal intensity across the radial range of the ring ( see table 1 ) . figure 1 shows the resulting fits of the 22 rings , overlaid on the @xmath0 images . the pixel - based ring diameters produced from ellipse were converted to arcseconds and kiloparsecs ( see table 1 , columns 5 & 6 , respectively ) . the rings vary in major axis from 0.2 kpc to 1.7 kpc , which is typical for nuclear rings ( buta & crocker 1993 ) . in most cases , the galaxy distance was obtained from the nearby galaxies catalog ( tully 1988 ) ; in the cases where there were no catalog values recorded , we computed them by dividing the recessional velocity by the hubble constant ( taken at 75@xmath15 ) see paper i for details . table 1 lists the parameters of the isophote best representing each ring , chosen on the basis of the peak isophotal intensity across the radial range of the ring . we compare the ellipticity and pa of each ring to that of its host disk . disk ellipticities and pas are from the rc3 , with the exception of the values for ngc 1530 , which are based on the kinematical data of regan et al . figure 2 compares the pas and ellipticities of the rings to their host disks . the plots show a clear linear relationship between the ring and its host disk with respect to both pa and ellipticity . this indicates that the rings are in the same plane as the disk and are nearly circular when deprojected . in some cases ( ngc 278 , ngc 4303 , ngc 4314 , ic 1438 , and ngc 7742 ) the disk is nearly circular ; those galaxies are seen face - on and do not appear in figure 2 . the disk morphology in these cases is consistent with that of the rings , which all have very low ellipticities ( less than 0.1 ) . the only exception to this general statement is ngc 7570 whose ring pa and ellipticity differ significantly from those of the disk . the ring appears to be nearly face - on so the pa of the chosen isophotal fit may not be reliable . however , the disk high inclination ( @xmath16=54 ; rc3 ) is not consistent with the observed nearly circular ring . this strongly suggests that the ring is not in the same plane as the disk . to identify the individual regions forming a given ring , we used the program sextractor ( bertin & arnouts 1996 ) , which builds a catalog of regions and their associated fluxes from an astronomical image . the various parameters used allow for the detection of regions even at very low @xmath0 intensities . we masked out the nucleus and extraneous stars that were clearly not part of the ring . the catalog file produced by the program records the position , area , and flux of each region detected . figure 3 shows the outlines of the fitted regions overlaid on the continuum subtracted @xmath0 images . in the case of ngc 7469 , sextractor was unable to identify any regions due to poor spatial resolution compared to the small angular size of the ring . this galaxy is well known to contain a nuclear ring of 0@xmath175 in radius ( genzel et al . 1995 ; davies , tacconi , & genzel 2004 ; daz - santos et al.2007 ) , which is consistent with our ellipse results ( section 3 ) . the lack of sufficient spatial resolution in our images , however , prevents us from continuing with the analysis of the individual regions in the ring , but we refer to above citations for a detailed study of its sf properties , based on high - resolution imaging and near - ir spectroscopy . to obtain ews of the regions , we measure fluxes in both the @xmath5-band and @xmath0 . we do this by first imposing the position and shape of the apertures from the sextractor analysis of the @xmath0 image onto the @xmath5-band image , making use of the fact that these images have already been registered ( see paper i ) . we then run sextractor in its two - image mode to obtain the corresponding @xmath5-band fluxes . among the many parameters to be defined in sextractor , two are of particular importance , namely the initial box size for the region detections , and the threshold value used to determine the extent of an individual region . for the first parameter we chose a value corresponding to the average seeing value in our images , namely 8 pixels or 0@xmath179 . for the threshold value , we chose twice the standard deviation , @xmath18 , of the combined galaxy and sky background across the whole image . we confirmed our parameter settings by independently varying the threshold value and box size and comparing the resulting ring morphology and ews to our `` reference '' values , which were best representative of the morphology seen in the @xmath0 images . several test cases were run by altering each parameter ( fixing one parameter while modifying the other ) by @xmath1950@xmath1 of the reference values . figure 4 shows the resulting images for ngc 1343 , as a representative example . figures 4a and 4b are the result of fixing the threshold and altering the box size . modifying the box size by @xmath1950@xmath1 of the reference value ( 8@xmath88 pixel box ) to a 12@xmath812 box and 4@xmath84 box , respectively , causes a significant change in the visual appearance of the regions . increasing the box size results in an obvious over - blending of the regions , with the opposite occurring when decreasing the box size . figures 4c and 4d of figure 4 are the products of modifying the threshold by @xmath1950@xmath1 of the reference value ( 2@xmath18 ) , to 1.33@xmath18 and 3@xmath18 , respectively . changing the threshold in either direction causes a significant change in the number of region detections forming the ring . overall , the isophotal detection area is more sensitive to the box size , while the number of detections is more sensitive to the threshold . for small variations of the threshold , the majority of the detected regions remained in tact . added detections were either as a result of further segmenting an already - defined region or adding a new region that was now above the noise level . these parameters are important since the background to be subtracted from the @xmath5-band and @xmath0 data is locally defined by sextractor for each region , and includes both the sky background and the emission from the galactic disk . in general , this empirical analysis gives us confidence that our parameter choices do indeed best represent the region morphologies . to obtain ews for the individual regions , the @xmath0 flux was divided by the _ i_-band flux within the same aperture as defined by sextractor . we describe below some of the intricacies of the ew determination . we specifically discuss the correction of those cases where the @xmath5-band flux of an region is particularly low , as well as the error budget of the ew values . many of the galaxies in our sample ( 16 out of 20 ) contained regions whose @xmath5-band emission was lower than the background due to noise . this resulted in negative @xmath5-band fluxes , which are unphysical and can not be used to determine the ew . rather than changing either the region aperture or specific sextractor parameters , we chose to use a physically meaningful upper limit to the @xmath5-band flux . we did this by taking , for all galaxies , the lowest @xmath5-band flux per pixel ( @xmath20 ) of all of the regions . analyzing the values per pixel versus per region avoids the possible bias introduced by the large range of region sizes , which vary by as much as a factor of two in a given ring . we conservatively assume the lowest value to be representative of a 6@xmath18 deviation from the mean . where @xmath20 is negative , we correct any @xmath5-band flux per pixel lower than the value @xmath21 ( equivalent to a 3@xmath18 adjustment ) to have a corrected region flux @xmath22 . this method , thus , applies a correction to those regions whose emission is positive but very low . see appendix a for details on individual region corrections . to estimate the impact of the 3@xmath18 corrections on the ews , we also corrected these fluxes to a lower level , namely @xmath23 ( equivalent to a 2@xmath18 adjustment ) . the maximum impact on ew that we found in such cases was an increase in ew of order 0.2 in log(ew ) , although in many cases the difference was less than 0.1 . in 10 out of the 16 galaxies affected by negative @xmath5-band emission , 5% , on average , of the regions subject to the 3@xmath18 correction would not be subject to a 2@xmath18 correction . that is , those regions with very low positive @xmath5-band emission that were changed using the 3@xmath18 discriminator were no longer affected when compared to the 2@xmath18 value . the other six galaxies were entirely unaffected by the correction change and can therefore be considered more reliable . we present these findings in figure 5 using a 3@xmath18 correction . regions whose @xmath5-band flux was changed to the 3@xmath18 value but was unaffected by the 2@xmath18 comparison are labeled as a diamond with a cross ; these represent lower - limits in the log(ew ) since we find that these values would have increased if we had chosen a 2@xmath18 correction . all other regions with corrected @xmath5-band fluxes , including the ones that were negative , are labeled as open diamonds . those regions where no corrections were needed are labeled as filled diamonds , and we consider those the most reliable . we estimate typical uncertainties in the ew values by considering uncertainties due to background subtraction and region detection , which both result from the application of sextractor . additional contributions to the overall error budget will come from measurement errors in the standard stars , the calibration uncertainties associated with our method of continuum subtraction ( see paper i ) , and in the filter response curves , but we will assume that they are negligible compared to the two main contributors . to make a conservative estimate of the uncertainty in the background subtraction we consider the @xmath5-band backgrounds . these will be more affected than the h@xmath24 measurements because ( 1 ) the background is higher in @xmath5 than in h@xmath24 , ( 2 ) the contrast between the region and background is much smaller than in h@xmath24 , and ( 3 ) the basic morphology of an region has been set by its h@xmath24 appearance in the @xmath5-band . from our analysis of the negative residual @xmath5-band fluxes , we know that the maximum difference in the log of the ew between the two cases ( i.e. , applying a 2@xmath18 versus 3@xmath18 correction to the affected measurements ) is 0.2 . such a difference of log(ew)=0.2 corresponds to a factor of 1.5 difference in the @xmath5-background subtracted before measuring the flux of an region , where log(ew)=0.1 would correspond to a factor of 1.3 . even considering the small field of view of our @xmath5-band images , these factors cover the uncertainty in the determination of the background . the other main contributor of the ew uncertainties is in the flux measurement of the regions . an rms value , based on gaussian noise statistics , is associated with each detection . on average , we estimate the rms to be 5% of the detection , which equates to 0.02 in the log of the ew . this constant rms value is due to the fact that both flux and source size play a role in the noise ( e.g. , detections with weaker fluxes cover smaller areas ) . we thus arrive at a conservative error budget consisting of two components , 0.2 and 0.02 in the log(ew ) . these combine in a worst case scenario to a factor of 1.6 , or an error budget in log(ew ) of 0.2 . we take 3@xmath18 of the total error budget as a guideline for discriminating between ews ( and thus regions ) . as a consequence , two ew values must be 0.6 apart in the log to make them significantly different . whereas this is by all means a large uncertainty , figure 5 shows that the values can range by as much as 3 in the log for a given ring during the evolution of massive stars so that , in fact , most of our measurements lead to straightforward and believable conclusions on at least relative ages , as discussed in section 6.1 . we stress that the error budget determined here is a very conservative one , and that for most regions the actual uncertainty will typically be significantly smaller . we use starburst99 ( leitherer et al . 1999 ) stellar population modeling to estimate the age of the individual regions . the models present the time evolution of the ew(@xmath0 ) for varying imf properties and metallicities , which will affect the evolution of the population . figure 6 shows the time evolution of ew(@xmath0 ) for both an instantaneous starburst ( i.e. , all stars are formed simultaneously in an instantaneous starburst episode ) and a continuous starburst ( i.e. , the star formation rate is constant in time ) . while the starburst choice can greatly affect the determination of the ages of the population greater than @xmath65@xmath25 , it is not clear which model best represents the stellar activity in nuclear rings . since active massive star formation is prevalent in regions , adopting an instantaneous burst would seem to fit the model predictions of younger ages better over the range of 1@xmath25 - 10@xmath25 . if , on the other hand , there is an abundant supply of gas entering the ring , starbursts may be a composite of episodic bursts ( elmegreen 1997 ; allard et al . 2006 ; sarzi et al . star formation in the ring would cyclically enter an active starburst period until it exhausts the surrounding gas supply , and then cease star formation until enough gas sparks a new starburst period . figure 7 shows starburst99 model curves for an instantaneous and a continuous starburst with different imf slopes , mass cut - offs , and metallicities . we vary the imfs in both scenarios and find that the evolution model is not very sensitive to the upper limit masses ( i.e. , 30@xmath26 and 100@xmath26 ) or the imf slopes ( @xmath27 ) , especially in the instantaneous starburst case . the models also confirm that high metallicity regions show lower ew(@xmath0 ) while low metallicity regions have invariably high ew(@xmath0 ) . regardless of the metallicity or type of burst , however , we find that all of the predicted evolution curves show a significant monotonic drop of ew with increasing age . moreover , the variation shown in the model curves ( fig . 7 ) is most likely an overestimate of the real variations in the nuclear rings under consideration here . for instance , allard et al . ( 2006 ) and sarzi et al . ( 2007 ) found solar metallicity values for all nuclear rings in their sample , and there is no convincing evidence for changes in either the imf slope or upper mass cutoff , except possibly in rather extreme environments such as the galactic center ( e.g. , reviews from elmegreen 2002 and kroupa 2002 ) . since we are more interested in the relative age difference from one region to the next , the sharply declining curve ( in either starburst case ) allows for an unambiguous discussions of age gradients . we adopted an instantaneous stellar starburst with a salpeter ( 1955 ) imf between @xmath28 and @xmath29 , and a solar metallicity , which is a good predictor for circumnuclear regions of barred galaxies ( allard et al . the results from sextractor were used to map the ages of each region according to their azimuthal location in the ring . the right panels of figure 5 show the resolved regions , which are color - coded according to age , in terms of ew . we assigned the physical center of the ring ( as computed with ellipse ) to be ( 0,0 ) , and viewed an region s azimuthal position in a counter - clockwise rotation from north to east . the color ranges are unique to each ring and , thus , do not denote the same ews for all rings . although we did not deproject the rings , all regions have a unique position , which allows us to search for any unambiguous distribution patterns . the pa of the major axes of the ring and the bar are indicated in the image . to complement the images , plots are shown in the left panels , which graph the log(ew ) versus pa , and again indicate the pas of the ring and bar major axes . we use @xmath30 color index maps , shown in figure 8 , as morphological indicators of the ring kinematics . these maps provide direct tracers of the bar dust lanes , which show the flow of gas to the ring . in the cases where the dust lanes were present , the azimuthal direction of the ring rotation was consistent with that of the galaxy in all cases . the contact points of the dust lanes with the ring are clearly visible in most of our images and indicate an offset of @xmath690@xmath31 between the contact points and the bar ( see section 6.3 for details ) . the dust lanes also show directional flow around the ring , assuming that rotation continues in the same direction as seen with the bar dust lanes and spiral arms . the adopted rotational direction of the rings is listed in table 2 and indicated for each ring containing a bar in figure 5 . we find that several galaxies , both barred and non - barred , exhibit relative azimuthal age gradients within the ring , as also observed by ryder , knapen , & takamiya ( 2001 ) and allard et al . ( 2006 ) for ngc 4321 . we only consider age gradients over a significant portion of the ring ( greater than 25% of the ring ) with amplitudes that exceed the uncertainties for differentiating between points over the range of interest . we highlight our findings below , and summarize the characterization of the age distributions for all of the rings ( e.g. , azimuthal gradient , radial gradient , flat distribution , or no recognizable pattern ) in table 3 ( column 2 ) . * ngc 278 ( fig . 5a ) : the entire ring as seen in @xmath0 is very small in size ( 0.2 kpc in radius ) with 20 individual regions detected that comprise the ring . this ring is the smallest one recognized in the galaxy ; a second `` outer '' nuclear ring appears at @xmath6 1 kpc and is discussed in detail in knapen et al . this outer ring is considered non - standard as it is very wide , and more a region of star formation with rather abrupt delimitations on the inner and outer radial range . in addition to this , we were in practice unable to derive a recognizable ring pattern of regions using sextractor , and have thus opted for concentrating on the inner nuclear ring . the host galaxy does not contain a bar in the optical nor near - ir wavelengths ( knapen et al . there is a decrease in the log(ew ) ( i.e. , an increase in age ) over approximately one third of the inner ring ranging from the positions of the regions centered at azimuth 124@xmath32 to 236@xmath32 . this indicates a counterclockwise rotation within the ring . a radial gradient spans the same range , with the younger regions lying on the inner side of the ring . a similar pattern , although not as clearly defined , appears over the range of 12@xmath32 to 100@xmath32 . * ngc 473 ( fig . 5b ) : two - thirds of the ring can be seen in the @xmath0 image , defined by 27 regions . a galactic bar exists at a major axis pa of 164@xmath32 along the major axis . the bar dust lanes indicate a counterclockwise flow of material around the ring . there is a sharply sloped age gradient over the range of 38@xmath32 to 203@xmath32 from younger to older , respectively . * ngc 613 ( fig . 5c ) : there exists a small partial nuclear ring containing 11 individual star - forming regions with a strong nucleus . a fairly flat distribution in the log(ew ) appears around the entire ring with ages averaging no less than 10 myrs . the galactic bar is located at pa = 110@xmath32 along its major axis . although a gradient is not defined , one of the contact points ( at 200@xmath32 ) coincides within 5@xmath32 to the youngest hotspot ( pa = 205@xmath32 ) in its respective hemisphere . * ngc 1300 ( fig . 5d ) : twelve regions form a small complete ring in @xmath0 although there is no clear pattern in ews . * ngc 1343 ( fig . 5e ) : the ring in this barred spiral galaxy is seen as complete in @xmath0 with 20 resolved regions . the mapping of ew to region location indicates a clear bi - polar age pattern around the ring with ew maxima ( i.e. , relative youngest age in the ring ) at pa = 140 and 309@xmath32 . there is a distinct age gradient approximately from one contact point to the other contact point ( located at 172@xmath32 and 352@xmath32 , respectively ) , where the age increases counterclockwise from the bar contact points . * ngc 1530 ( fig . 5f ) : three - quarters of the ring can be seen in @xmath0 containing 9 distinct regions . the galaxy has a large bar at pa = 122@xmath32 and two open spiral arms originating from its ends , rotating clockwise . we find bipolar age gradients , one in each hemisphere , as bisected by the position of the bar . the youngest regions are near the bar pa , ( i.e. , not the contact points ) over the range of 283 to 207 and 115 to 7 , with the ring rotation also clockwise . half of the ring is distinctly younger than the other half , where the turnover points coincide with the ring major axis pa (= 25@xmath32 ) . * ngc 4303 ( fig . 5 g ) : only seven regions define the small ring . the ring is observed to be three - quarters complete in @xmath0 with bipolar age maxima at pa = 114@xmath32 and 304@xmath32 . there are no obvious age gradients in between the maxima , but the latter are offset by less than 25@xmath32 from the bar contact points ( 100@xmath32 and 280@xmath32 ) . * ngc 4314 ( fig . 5h ) : the small ring consists of 10 regions showing a smooth and well - defined age gradient . the age gradient flows from older to younger in the same direction as the presumed ring rotation ( clockwise ) . the youngest region is within 20@xmath32 of one of the contact points ( pa = 45@xmath32 ) . * ngc 4321 ( fig . 5i ) : the well resolved full ring contains 57 detected regions . because missing data prevent the calculation of @xmath0 equivalent widths and ages , we refer to allard et al . ( 2006 ) for a similar analysis in @xmath33 . they find a bipolar age gradient corresponding to the ring pa ( at 170 and 350 ) and the location of the youngest region in each hemisphere . the age increases in a counter - clockwise direction along the direction of flow in the ring . * ngc 5248 ( fig . 5j ) : nineteen regions define the full , elliptically shaped ring . the youngest region is located 4@xmath32 away from of one of the bar contact points ( pa = 47@xmath32 ) . * ngc 5728 ( fig . 5k ) : eighteen regions define the partial ring , which is three - quarters complete . a very steep but small gradient exists from pa = 147@xmath32 to pa = 98@xmath32 from younger to older , according to the clockwise directional flow of the ring . one contact point ( pa = 303@xmath32 ) is within 2@xmath32 of the youngest region in the ring . the second youngest region ( pa = 147@xmath32 ) is within 24@xmath32 of the second contact point ( pa = 123@xmath32 ) , although it does precede that contact point . * ngc 5905 ( fig . 5l ) : the ring is small but complete in @xmath0 . it consists of nine regions and is relatively flat with respect to age ( except for one region ) . there appears to be no correlation in age with respect to the bar contact points . * ngc 5945 ( fig . 5 m ) : only three regions define this patchy ring , which contains low ew values ( all less than 1.5 in the log ) . buta & crocker ( 1993 ) indicate a nuclear ring , but our images do not have enough resolution to confirm this . * ngc 5953 ( fig . 5n ) : the non - barred galaxy has a nuclear ring that appears complete with 23 regions . although we know of no bar to directly fuel the ring , it is well formed and resolved both azimuthally and radially . * ngc 6503 ( fig . 5o ) : the elliptically shaped ring contains 84 resolved regions in this unbarred spiral . no gradients are evident . * ngc 6951 ( fig . 5p ) : the ring is small but complete with eight detected regions . the youngest region ( pa = 339@xmath32 ) is within 16@xmath32 of its respective contact point . the ages steeply increase ( by a factor of 10 ) from 339@xmath32 to 218@xmath32 in a clockwise direction , consistent with the directional flow of the galaxy . * ngc 7217 ( fig . 5q ) : the ring is three - quarters complete in @xmath0 . twenty - two regions were detected in the ring of this unbarred galaxy . ages increase in a clockwise direction from pa = 109@xmath32 to pa = 6@xmath32 . * ic 1438 ( fig . 5r ) : this barred galaxy hosts a complete ring with 11 regions detected . there exists an increasing age gradient from pa = 32@xmath32 to pa = 211@xmath32 in a counterclockwise direction . this direction is consistent with the ring rotational flow . the youngest region is within 2@xmath32 of its respective bar contact point ( pa = 304@xmath32 ) * ngc 7570 ( fig . 5s ) : eighteen regions comprise a mostly complete ring in this barred galaxy . an increasing age gradient ranges counterclockwise from pa = 112@xmath32 to pa = 295@xmath32 . there is also a radial gradient spanning the same range as the azimuthal gradient , with two distinct arcs at nearly identical slopes , and where the inner arc appears consistently older than the outer one . no correlation to the bar contact points is evident . * ngc 7716 ( fig . 5 t ) : thirty regions form a complete ring in this barred galaxy . no gradients or correlations to the bar are observed . * ngc 7742 ( fig . 5u ) : the circular and well resolved ring contains 38 regions . no gradients are observed . while many of the rings contain age gradients , a few show no gradient , namely ngc 613 , ngc 5905 , and ngc 5945 . this type of distribution is qualitatively more consistent with elmegreen s ( 1994 ) predictions , which suggest that the massive star formation in the nuclear ring is shocked into existence due to gravitational instabilities in the nuclear region around the ilrs . therefore , age gradients would not exist . comparisons of ew to age for these three rings also indicate that these starbursts are relatively older than those seen in the other rings , on the order of 10 myrs and higher . elmegreen ( 1997 ) suggests that these rings ( i.e. , with no gradient ) could last for as long as 100 myrs before sf ceases if the inflowing gas rate is high enough . the sfrs of these three rings ( 2@xmath26/yr , 2@xmath26/yr , and 4@xmath26/yr , respectively ; see section 6.2 and table 1 for sfr and inflow discussion ) are among the higher values in our sample but are consistent with other rates documented ( regan et al . 1997 ; jogee et al . . if we take into account that ngc 613 and ngc 5945 are partial rings in the optical , then the sfrs calculated here are lower limits . in this section we explore possible reasons for the result that only three galaxies show bi - polar age gradients around their rings , whereas many of the rings ( 9 of 22 ) show partial gradients along 25% - 30% of their rings , with the rest ( 10 of 22 ) showing no gradient . we find no apparent relationships between the presence or absence of age gradients and the morphology of the rings or their host galaxies . for example , whereas ngc 4321 , with its bi - polar age gradient , has a pair of well - defined dust lanes coming into the ring through the bar , ngc 1300 and ngc 5248 have similar dust lane morphologies yet show no gradients . some of the poorly resolved rings , such as those in ngc 4303 or ngc 5945 , do not show gradients , but given the presence of measured age gradients in the comparably small rings in ngc 6951 and ic 1438 , the lack of spatial resolution in the former two can not easily be used as an argument for the lack of a gradient . but even if we ignore galaxies with either small and badly resolved rings , or rings that do not have the classical aspect of dynamically - induced resonance rings ( ngc 613 and ngc 6503 are in the latter category ) , we are left with a significant number of rings in which we should have detected gradients if present , but which for some reason do not show any . this category includes , for instance , ngc 1300 , very similar to ngc 4314 or ngc 6951 which do show gradients ( all compact well - defined rings with around a dozen regions , sitting in strong bars with well - defined sets of dust lanes ) ; ngc 5248 , similar to ngc 4321 ( complex , rich rings with many regions organized into spiral arms fragments and hosted by a relatively weak bar ) ; ngc 7716 , similar to ngc 7217 which does have a gradient ; and ngc 7742 . for those rings that contain a gradient we also find no correlation between the presence of the gradient and direction of the ring rotation . in relating the location of the youngest regions to those of the contact points between the dust lanes and ring , we have been assuming that massive sf will follow very rapidly once the inflowing gas joins the ring . the appearance of a gradient in the ring may thus depend on whether or not material flowing in will immediately trigger gravitational collapse , and initiate massive sf . this in turn may be due to either a higher mass inflow rate ( or possibly increased clumpiness of the inflow ) and/or an overall higher existing gas density in the ring . each of these possibilities will ensure that soon after a clump of inflowing gas joins the ring , it will start to form massive stars . the alternatives , of low inflow rate and/or overall low gas density in the ring , might lead to a more gradual increase of the gas density , without necessarily leading to immediate gravitational collapse and massive sf on short timescales . in this case , the positions of the regions would bear no relation to the locations of the contact points . the ring would , in such a case , merely increase its gas density gradually , until at some stage , somewhere along the ring , an individual clump would become unstable to collapse ; therefore , no age gradients would ensue . to test this simple model , we must first consider the quantities mentioned above , gas inflow rate and gas density in the ring . these can in principle be determined from interferometric co measurements , but the process is not straightforward . the main issues are the conversion factor of @xmath34 to @xmath35 ( the @xmath36-factor ) , which may well not be equal to the canonical value in the kind of environments we consider here , and the determination of the fractions of detected molecular gas which are in fact flowing into the ring rather than around it ( as a consequence of a `` spray - type '' flow , see , e.g. , regan , vogel & teuben 1997 ; jogee et al . 2002 ) , or with the right physical conditions such as velocity dispersion and clumpiness to take part in the massive sf process . so although co interferometry has been presented for seven of our nuclear rings in the literature ( by regan , vogel & teuben 1997 ; sakamoto et al . 1999 ; jogee et al . 2002 , 2005 ; helfer et al . 2003 ; combes et al . 2004 ; garca - burillo et al . 2005 ) , we will not attempt here to derive the gas inflow rate and gas density in the ring from this collection of data , all with different observational characteristics . for two of our rings , ngc 1530 and ngc 5248 , gas inflow rates have been explicitly determined in the literature . for the former galaxy , regan et al . ( 1997 ) give a value of @xmath37 , whereas jogee et al . ( 2002 ) determine a value of `` a few''@xmath38 for the latter galaxy . considering the uncertainties in these values , we can say little more than that they are consistent with the sfrs determined for our nuclear rings ( see discussion below , and values in table 1 ) . we could assume that the sfr of the ring can act as a tracer of the inflow rate , although that would imply assuming , for instance , that inflowing gas is transformed into massive stars on very short timescales . if we do this , we would expect the age gradients to occur in those galaxies with the higher sfrs , which would be representative of higher gas inflow rates . in contrast , lower sfrs would indicate a low inflow rate so that the gas will gradually fill the ring , leading to sf only if and when the local gas density leads to gravitational collapse . this will occur essentially at random and thus result in a ring with no gradients . kennicutt ( 1998 ) derives a relation between the @xmath0 ionizing flux and the sfr for normal disk star forming galaxies , where the conversion factor is computed using stellar population synthesis modeling . since only stars with masses larger than 10@xmath26 and lifetimes less than 20@xmath25 dominate the integrated ionizing flux , the @xmath0 emission line provides a nearly instantaneous and linear measure of the sfr , independent of the previous star formation history . for solar abundances and our adopted imf ( @xmath39 ; @xmath28 and @xmath29 ) , we can use kennicutt s ( 1998 ) equation : @xmath40 to compute sfrs integrated over the entire ring , which are given in column 10 of table 1 . variable extinction certainly adds uncertainty in the computation of the sfr for circumnuclear regions . from the @xmath41 images , the patchiness of the extinction is very clear in several of our galaxies ( e.g. , ngc 473 , ngc 613 , ngc 1530 , ngc 5945 ) , which reveal only partial rings in @xmath0 . the less obscured regions will contribute more heavily to the line fluxes in these cases . the @xmath41 images clearly indicate that while the dust lanes can be traced nicely in the bar inmost cases ( where there is little star formation ) , the appearance of the nuclear rings is dominated by blue star - forming regions . this type of morphology makes it difficult to precisely model the effect of the dust , but attempts have been made for nuclear regions in particular . osmer , smith , & weedman ( 1974 ) analyzed six galaxies and derived integrated extinction values across the entire circumnuclear region ranging from @xmath42 mag , while hummel , van der hulst & keel ( 1987 ) found an extinction of @xmath43 in the central region of ngc 1097 . kennicutt ( 1998 ) acknowledges the high uncertainty and estimates that the errors in the sfr can be as high as @xmath1950% . a high - resolution survey paper by phillips ( 1996 ) for nine circumnuclear regions confirms the potential for high uncertainty values . following kennicutt , he finds that the sfrs vary by as much as a factor of ten when applying extinctions of @xmath44=1.1 and @xmath44=3.4 mag . he concludes that circumnuclear sf can be a dominant factor to the overall rates of such galaxies . we thus place the same caution that our sfrs contain potentially high uncertainties due to variable extinction throughout the nuclear rings . although there are possible large uncertainties in the sfr due to extinction , using the sfr can provide a possible link to an explanation for the existence of age gradients , via the gas inflow rate . figure 9 plots the integrated sfr over the entire ring for all galaxies and distinguishes those rings that show a gradient and those that do not . there is no obvious correlation between the sfr and the appearance of a gradient , although we can not exclude that such a connection does exist but is obscured by high dust extinction . the average mean for those without gradients ( @xmath45 , where the uncertainty is the standard error ) is slightly larger for those with a gradient ( @xmath46 ) , but given the large mean deviations , we can not distinguish in a statistical sense between the two cases . this could be due to several factors . all of the gradients , except for those in ngc 1343 and ngc 4321 , are partial gradients . that is , their gradient does not range continually from one contact point to the next . ngc 1343 and ngc 4321 , which represent the``ideal '' scenario , do have high sfrs ( @xmath47 for both cases ) when compared to the rest of the sample . computing the average sfr of the ring also is insensitive to the steepness of the gradient , the size of the ring , the size of the individual regions , or the patchiness of the ring , which all add complexity to the interpretation of the data . from our analysis of age gradients we thus conclude that although age gradients along the ring are detected in more than half of the rings , all other rings show random variations of region ew , and thus age , with azimuth . we can not pinpoint any morphological parameters that might distinguish the host galaxies and bars of those rings with from those without gradients . another strong conclusion from our analysis is that the kind of clear bi - polar age gradient seen before in ngc 4321 ( ryder , knapen , & takamiya 2001 ; allard et al . 2006 ) is rare - we only see a similar bi - polar gradient in one more galaxy , ngc 1343 . it is not clear why these two galaxies show their gradients , whereas other galaxies with very similar bar - ring systems show gradients that are not bi - polar , or no gradients at all . it is agreed that the presence of a galactic bar can have significant influence on angular momentum and re - distribution of gas , including inflow towards the central regions ( e.g. , combes & gerin 1985 ; athanassoula 1992 , 2000 ; knapen et al . 1995a ; heller & shlosman 1996 ; maciejewski et al . 2002 ) . observed kinematics reveal that the shock - induced flows move gas radially inwards along the bar dust lanes and tend to collect at the intersection of the dust lanes with the ring . these contact points are thought to be approximately perpendicular ( within the plane of the disk ) to the pa of the bar major axis ( regan & teuben 2003 ) but have been seen to lead the bar by as little as 60@xmath31 in numerical models ( heller & shlosman 1996 ) . the pile up of gas at these two points in the ring can be expected to spark sf . therefore , one could predict to find the youngest stellar regions in the ring near the contact points . for the barred spiral galaxies in our sample , we compare the location of the youngest regions in the ring to the derived location of the contact points . we first identified the youngest region in each hemisphere , where the hemispheres are separated by the bar pa . taking into account the direction of flow around the ring ( from our @xmath48 images ) , we calculate the pa offset between the region and the contact points . table 2 shows the pas of both the contact point and the youngest region in each hemisphere , along with the directional flow of the ring . we find that in two - thirds of the galaxies , one of the bar contact points is aligned with the youngest region in the associated hemisphere within 20@xmath32 of the contact point . two galaxies , ngc 4314 and ngc 5248 , have the youngest region in both hemispheres aligned with the contact points . our results for both of these rings are consistent with those of benedict et al . ( 2002 ) and maoz et al . ( 2001 ) , respectively , who also find that the youngest stars were located at the contact points . figure 10 shows a histogram of the distribution of the angular offsets between the bar major axis and the locations of the two youngest hii regions in each ring . a peak in frequency around the contact point locations is seen . we next calculate the times from the bar major axis pa to the location of the contact points with those of the youngest region in each hemisphere , assuming we azimuthally move in the direction of ring flow . we compute the period of the rings using measurements of the ring radius and the rotational velocity at the ring radius . we chose the major axis length of each ring as derived from ellipse ( see section 3 and table 1 ) , and obtained the rotational velocity values from the literature and from our own densepak ifu data taken in 2003 & 2004 ( l. m. mazzuca et al . 2006 , in prep . ) . table 3 shows the major axis ring radius , the adopted rotational velocity along with its reference , and the period for each ring . since we predict the contact points to be @xmath690@xmath32 from the bar major axis , we can compute their location with respect to the bar pa in each hemisphere , and thus calculate the amount of time it would take to travel to that location . we compare this travel time to that associated with the location of the youngest region in each hemisphere . overall , the expected travel times to the contact points coincide with the times computed from the bar pa to the youngest regions , which further strengthens the theory that the youngest regions are located near the contact points . we find only one region , in ngc 1530 , that deviates from the general trend . in this case , the youngest region in one of the hemispheres is more aligned with the pa of the ring than that of the bar . the analysis described above considers merely the rotational velocity of the gas in the ring region while deriving travel times . there is an additional contribution from the bar , which rotates with a certain bar pattern speed ( assuming that the ring co - rotates with the bar ) . estimating the relative contributions of these two rotation components , we consider parameters for a typical ring in our sample ( see table 1 ) of @xmath49kpc , @xmath50kms@xmath51kpc@xmath51 , and @xmath52,kms@xmath51 , which yields a ratio of the contributions of circular to bar pattern speeds into the velocities around the ring of @xmath53 . given that the rotational velocity is dominant , that the uncertainties in determining the bar pattern speed ( and , in fact , the direction of bar rotation ) are significant ( see elmegreen 1996 and knapen 1999 for reviews ) , and that the uncertainties from other sources , as described above , in the determination of travel times around the ring are already rather large , we have not pursued to derive bar pattern speeds and incorporate them into our analysis . from a larger @xmath0 imaging survey we identified 22 galaxies that contain a nuclear ring . to identify their sizes and shapes , we fitted each ring with an ellipse based on the peak isophotal intensity across the radial range of the ring . the rings vary in size from 0.2 kpc to 1.7 kpc in major axis radius . we compare the ring morphologies to those of their host galaxies . in all but one case the ellipticity and pa of the rings approximately match that of their respective disk , which corroborates the idea that the ring is in the same plane as the disk . the exception is ngc 7570 , which shows a distinct and self - consistent difference in both the pa and eccentricity in the ring versus the disk . all but one of the rings had sufficient spatial resolution for us to identify individual regions . for each region , we computed the ew and converted differences between them into relative ages and age gradients using the starburst99 evolutionary models . although we adopted an instantaneous starburst model , we acknowledge that the rings may be more representative of a hybrid episodic scenario where star formation toggles between latent and active periods corresponding to the time it takes for gas build - up to reach critical density . three rings , ngc 1343 , ngc 1530 , and ngc 4321 , contain bipolar age gradients , where the ages increased from each of the contact points . many of the rings in our sample exhibit a well - defined age distribution pattern , but not throughout the entire ring . this can be a result of the combination of bar - induced dynamics and gravitational instabilities that are occurring on an intermittent basis . we calculated the sfr in each ring from the integrated luminosity contributed by all of the regions forming the ring , and compared those rates to the appearance of a gradient within each ring ( or lack thereof ) . the large uncertainties associated with variable extinction across the rings and the large dispersion of sfr values for both cases yields no clear correlation between the sfr and the emergence ( or lack thereof ) of an age - dependent azimuthal distribution . we do see a connection between the location of the bar contact points and the youngest regions in two - thirds of the galaxies in our sample . @xmath54 we thank dr . isaac shlosman for inflow rate discussions and dr . torsten b " oker for input into earlier stages of the research . we also thank the referee for the thorough review of this paper . jhk acknowledges support from the leverhulme trust in the form of a leverhulme research fellowship . the wht is operated on the island of la palma by the isaac newton group in the spanish observatorio del roque de los muchachos of the instituto de astrofsica de canarias . this research has made use of the nasa / ipac extragalactic database ( ned ) , which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . .morphological characteristics of the galaxies in the observed sample . galaxies are listed in order of increasing ra ( col . 1 ) ; morphological type as stated in the rc3 ( col . 2 ) ; ellipticity of the nuclear ring as derived by using the iraf ellipse task ( col . 3 ) ; pa of the ring major axis , calculated counterclockwise from north to east ( ellipse ; col . 4 ) ; ring radius in arcseconds ( col . 5 ) and kpc ( col . 6 ) ; pa of the galactic disk from rc3 ( a dash represents a circular ring with no defined pa ; col . 7 ) , except for ngc 1530 ( regan et al . 1996 ) which is based on kinematical data ; pa of the bar major axis , except for ngc 1530 ( regan et al . 1996 ) and ngc 7469 ( davies , tacconi , & genzel 2004 ) which are based on kinematical data ( col . 8) ; and the ring sfr ( a dash represents lack of data necessary to calculate the sfr ; col . 9 ) is derived from the total ring luminosity using the kennicutt ( 1998 ) @xmath0-sfr relation . [ cols="<,^,^,^,^,^,^,^,^,^ " , ] allard , e. l. , peletier , r. f. , & knapen , j. h. 2005 , , 633 , l25 allard , e. l.,knapen , j. h. , peletier , r. f. , & sarzi , m. 2006 , submitted to mnras athanassoula , e. 2000 , stars , gas and dust in galaxies : exploring the links , 221 , 243 athanassoula , e. 1992 , , 259 , 345 benedict , g. f. , et al . 1993 , , 105 , 1369 bertin , e. , arnouts , s. 1996 , , s117 , 393 blackman , c. p. 1981 , , 195 , 451 bottema , r. 1989 , , 221 , 236 buta , r. & combes , f. , 1996 , fund . of cosmic phys . 17 , 95 buta , r. & crocker , d. a. 1993 , aj , 105 , 1344 combes , f. , & gerin , m. 1985 , , 150 , 327 combes , f. , et al . 2004 , , 414 , 857 davies , r. i. , tacconi , l. j. , genzel , r. 2004 , apj , 602 , 148 daz - santos , t. , alonso - herrero , a. , colina , l. , ryder , s. d. , knapen j. h. 2007 , , 661 , 149 elmegreen , b. 1996 , iau colloq . 157 : barred galaxies , 91 , 197 elmegreen , b. g. 2002 , modes of star formation and the origin of field populations , 285 , 425 elmegreen , b. 1997 , rmxac , 6 , 165 garca - burillo , s. , combes , f. , schinnerer , e. , boone , f. , & hunt , l. k. 2005 , , 441 , 1011 genzel , r. , weitzel , l. , tacconi - garman , l. e. , blietz , m. , cameron , m. , krabbe , a. , lutz , d. , & sternberg , a. 1995 , , 444 , 129 helfer , t. t. , thornley , m. d. , regan , m. w. , wong , t. , sheth , k. , vogel , s. n. , blitz , l. , & bock , d. c .- j . 2003 , , 145 , 259 heller , c. h. , & shlosman , i. 1994 , , 424 , 84 heller , c. h. , & shlosman , i. 1996 , , 471 , 143 hummel , e. , van der hulst , j. m. , & keel , w. c. 1987 , , 172 , 32 jogee , s. , shlosman , i. , laine , s. , englmaier , p. , knapen , j. h. , scoville , n. , & wilson , c. d. 2002 , , 575 , 156 jogee , s. , scoville , n. , & kenney , j. d. p. 2005 , , 630 , 837 kenney , j. d. p. , wilson , c. d. , scoville , n. z. , devereux , n. a. , & young , j. s. 1992 , , 395 , l79 kenney , j. d. p. , carlstrom , j. e. , & young , j. s. 1993 , , 418 , 687 kennicutt , r. c. 1998 , , 36 , 189 knapen , j.h . , beckman , j. e. , heller , c. h. , shlosman , i. , de jong , r.s . 1995a , apj , 454 , 623 knapen , j. h. , beckman , j. e. , shlosman , i. , peletier , r. f. , heller , c. h. , & de jong , r. s. 1995b , , 443 , l73 knapen , j. h. 1999 , the evolution of galaxies on cosmological timescales , 187 , 72 knapen , j. h. , laine , s. , & relao , m. 1999 , , 269 , 605 knapen , j. h. , shlosman , i. , heller , c. h. , rand , r. j. , beckman , j. e. , & rozas , m. 2000 , , 528 , 219 knapen , j. h. , whyte , l. f. , de blok , w. j. g. , & van der hulst , j. m. 2004 , , 423 , 481 knapen , j. h. 2005 , , 429 , 141 knapen , j.h . , mazzuca , l.m . , bker , t , shlosman , i. , colina , l. combes , f. , axon , d.j . 2006 , , 448 , 489 kormendy , j. , kennicutt , r. c. jr . 2004 ara&a , 42 , 603 kroupa , p. 2002 , modes of star formation and the origin of field populations , 285 , 86 leitherer , c. , et al . 1999 , apjs , 123 , 3 maciejewski , w. , teuben , p. j. , sparke , l. s. , & stone , j. m. 2002 , , 329 , 502 maoz , d. , barth , a. j. , ho , l. c. , sternberg , a. , & filippenko , a. v. 2001 , , 121 , 3048 martinet , l. 1995 , fundamentals of cosmic physics , 15 , 341 moiseev , a. v. , valds , j. r. , & chavushyan , v. h. 2004 , , 421 , 433 oke , j . , 1990 aj 99,5 osmer , p. s. , smith , m. g. , & weedman , d. w. 1974 , , 192 , 279 peterson , c. j. , roberts , m. s. , rubin , v. c. , & ford , w. k. 1978 , , 226 , 770 phillips , a. c. 1996 , asp conf . ser . 91 : iau colloq . 157 : barred galaxies , 91 , 44 piner , b. , stone , j. , teuben , p. 1995 , apj,449 , 508p pogge , r. w. 1989 , , 345 , 730 regan , m. w. , teuben , p. j. , vogel , s. n. , & van der hulst , t. 1996 , , 112 , 2549 regan , m. w. , & teuben , p. j. 2004 , , 600 , 595 regan , m. w. , vogel , s. n. , & teuben , p. j. 1997 , , 482 , l143 regan , m. w. , teuben , p. 2003 apj , 582 , 723 rubin , v. c. 1980 , , 238 , 808 rubin , v. c. , kenney , j. d. p. , & young , j. s. 1997 , , 113 , 1250 ryder , s. d. , knapen , j. h. , & takamiya , m. 2001 , , 323 , 663 sakamoto , k. , okumura , s. k. , ishizuki , s. , & scoville , n. z. 1999 , , 525 , 691 salpeter , e. e. 1955 , apj , 121 , 161 sarzi , m. , allard , e.l . , knapen , j. h. , mazzuca , l. m. , mnras in press schwarz , m. p. 1981 , , 247 , 77 shlosman , i. , begelman , m. c. , & frank , j. 1990 , , 345 , 679 sheth , k. , regan , m. w. , vogel , s. n. , & teuben , p. j. 2000 , , 532 , 221 smith , d. a. , herter , t. , haynes , m. p. , & neff , s. g. 1999 , , 510 , 669 storchi - bergmann , t. , wilson , a. s. , & baldwin , j. a. 1996 , , 460 , 252 tully , r. b. 1988 british astronomical association journal , 98 , 6

a microscopic configuration - interaction ( ci ) methodology is introduced to enable bottom - up schrdinger - equation emulation of unconventional superconductivity in ultracold optical traps . we illustrate the method by exploring the properties ofli atoms in a single square plaquette in the hole - pairing regime , and by analyzing the entanglement ( symmetry - preserving ) and disentanglement physics ( via symmetry - breaking , associated with the separation of charge and spin density waves ) of two coupled plaquettes in the same regime . the single - occupancy rvb states contribute only partially to the exact many - body solutions , and the ci results map onto a hubbard hamiltonian , but not onto the double - occupancy - excluding- one . for the double - plaquette case , effects brought about by breaking the symmetry between two weakly - interacting plaquettes , either by distorting , or by tilting and detuning , one of the plaquettes with respect to the other , as well as spectral changes caused by increased coupling between the two plaquettes , are explored .

1703.10128

rapid experimental advances in the creation of finite systems of ultracold atoms using few optical traps ( bottom - up ) @xcite or through the use of extended optical lattices ( top - down ) @xcite are promising approaches toward realization of feynman s vision of a quantum simulator @xcite , capable of finding solutions to systems that are otherwise numerically and/or analytically intractable . however , apart from a few double - well investigations @xcite , to date model - independent microscopic studies of multiwell systems providing theoretical insights and guidance to experimental efforts are largely lacking . here we introduce a configuration - interaction ( ci ) schrdinger - equation methodology @xcite for exploring finite plaquette systems assembled from individual optical traps ; for a brief description of the ci method , see appendix [ cime ] . these systems are fundamental stepping stones toward bottom - up realization of large scale checkerboard or square ulracold atom lattices which are promising candidates for emulating the physics underlying @xmath3-wave high - t@xmath4 superconductivity @xcite in optical lattices @xcite . the work described here , demonstrating the feasibility of such exact ci calculations for interatomic contact interactions , can be extended to electronic plaquettes , i.e. , to quantum - dot - arrays , governed by long - range coulomb interactions . the plan of the paper is as follows : in section [ sp ] we explore first the properties of ultracold fermionic atoms ( @xmath0li ) confined in a single square plaquette ( 4-sites ) in the regime of hole pairing , and subsequently analyze the entanglement physics of two coupled plaquettes in the hole - pairing regime . in section [ dp ] , for the `` hole - doped '' coupled plaquettes ( 8-sites , six atoms ) , we analyze the wave function anatomy of the entangled @xcite ( schrdinger - cat ) two almost - degenerate manifolds , comprising : ( @xmath5 ) the ground state ( gs ) and 1st excited ( 1ex ) state , and ( @xmath6 ) the two higher excited states ( 2ex and 3ex ) . the almost - degenerate states have good , but opposite , parities . when symmetry - broken ( sb ) either by superposing @xmath7 the degenerate pair in each manifold , or via offsetting the energies ( tilting ) of the two plaquettes the sb states coming from @xmath5 are characterized by a particle ( `` charge'')-density modulation ( `` wave '' ) , i.e , a cdw portraying the hole - paired , ( 4,2 ) or ( 2,4 ) , components , whereas the sb states originating from @xmath6 remain in a ( 3,3 ) particle distribution ( each of the plaquettes having an unpaired hole ) , but exhibit a spin - polarizarion density modulation ( `` wave '' ) , i.e. , a sdw . for the the double - plaquette case , we further explore effects brought about by breaking the symmetry between two weakly - coupled plaquettes , either by distorting one of the plaquettes , or by tilting and detuning one of the plaquettes with respect to the other . spectral changes caused by increased tunnel coupling ( e.g. , by decreased inter - plaquette distance ) between the two plaquettes are also considered . noteworthy is our finding that the gs @xmath3-wave resonating valence bond ( rvb ) state contributes only partially to the exact many - body wave function i.e. , double occupancies ( referred also as doublons ) need be included . indeed , our microscopic results map properly onto a hubbard hamiltonian ( including extended hubbard models @xcite depending upon the range of the experimental parameters ) , but not onto the double - occupancy - excluding @xmath1-@xmath2 model @xcite . our conclusions agree with recent @xcite observations of doublons in two - dimensional ( 2d ) optical lattices . the rest of the paper comprises the summary ( sec . [ summ ] ) and three appendices , concerning a brief description of the ci method ( appendix [ cime ] ) , the mathematical definitions of single - particle densities and two - body and @xmath8-body conditional probability distributions ( appendix [ defs ] ) , and the specification of the correspondence between spin and rvb functions and ci many - body wave functions ( appendix [ spin ] ) . the short - range 2-body repulsion in the hamiltonian is described by a gaussian of width @xmath9 , i.e. , by @xmath10 @xmath11 is the strength parameter . here and throughout the paper : @xmath12 @xmath13 m , where the oscillator length @xmath14 ) , @xmath15 being the @xmath0li mass and @xmath16 khz being the trapping frequency of the plaquette potential wells @xcite . this form of interaction provides a good approximation for the atom - atom interactions and avoids the peculiarities of the delta function in two dimensions @xcite . the potential surface of the 4-site ( or 8-site ) plaquette is constructed with the help of the two - center / smooth - neck oscillator hamiltonian which was previously introduced in refs . the smooth - neck interwell barrier @xmath17 can be varied independently and is controlled by the parameter @xmath18 , where @xmath19 is the intersection height of the two bare potential parabolas from neighboring sites , i.e. , prior to inserting the smooth - neck potential contribution @xcite . for a graphical illustration of the two - center - oscillator ( tco ) double - well potential , including the definitions of @xmath17 and @xmath19 , see fig . [ fig1_t ] . here and throughout the paper : the intersite distance in a single plaquette is @xmath20 @xmath13 m and @xmath21 ( yielding @xmath22 khz ) , unless noted otherwise . the potential surfaces of the 4-site and 8-site plaquettes is constructed by combining such tco potentials along the @xmath23 and @xmath24 directions ( see supplemental material for details @xcite ) . . ( b ) case without a tilt . @xmath25 is the trapping frequency ( given in energy units of khz in this paper ) of the bare harmonic oscillator ( dashed lines ) for each site . , width=264 ] ( a , b , c , e ) and @xmath26 ( d , f ) @xmath0li atoms in a 4-site plaquette . ( a ) energy spectrum . @xmath22 khz . ( b ) gs spd [ red star in ( a ) ] . ( c , d ) sr - cpd2s for gs and 2ex . up black arrows : up - spin at the observation point . red humps : probability for down spin atoms . note the double - occupancy hump at the observation point . ( e , f ) gs total energies ( @xmath27 , in units of @xmath1 ) for the ci , hubbard , and @xmath1-@xmath2 hamiltonians ; ci double - occupancy ( doc ) in red . inset ( i ) illustrates the 4-site external potential . , width=302 ] the ci single - plaquette low - energy spectrum as a function of the two - body repulsion strength @xmath11 , is displayed in fig . [ fig2_t](a ) , with the single - particle density ( spd ) for the gs ( see red star ) , or for the second excited state ( blue star ) , being shown in fig . [ fig2_t](b ) ; for the definition of the spd in the ci approach , see appendix [ defs ] . a total - spin projection @xmath28 has been assumed . the ci wave functions depend both on the atom ( continuous ) position , @xmath29 , and the spin , @xmath30 , @xmath31 , variables . however , for plaquettes with sufficiently high interwell barriers , the microscopic structure of a ci state @xmath32 can be mapped onto the usual hubbard hilbert space involving the superposition of many primitive basis functions @xmath33 ; @xmath3 here denotes a doubly occupied site , @xmath34 and @xmath35 denote the site s spin occupancy , and 0 denotes an empty site . inset ( i ) in fig . [ fig2_t ] describes the 4-site external potential . the coefficients @xmath36 of the primitives @xmath37 in the expansion of @xmath32 , can be extracted from the ci wave functions with the help of the spin - resolved conditional probability distributions ( sr - cpds ) @xcite . two sr - cpds are used here : one ( sr - cpd2 ) , expressed as an expectation value over the many - body wavefunction , describes the space and spin correlation of a particle pair , for a given location and spin of one of them ( referred to as an observation point ) and the other being any other particle of the @xmath38 remaining ones ( with a specified spin ) . the other probability distribution ( sr - cpdn ) , given by the modulus square of the many - body function , expresses the spatial probability of finding the @xmath8-th particle with a specified spin when one fixes the positions and spins of the other @xmath38 particles ( for the mathematical definition of sr - cpds , see appendix [ defs ] ) . ( [ dgap ] ) ] for a single 4-well plaquette with @xmath39 @xmath13 m ( a , b ) and @xmath40 @xmath13 m ( c , d ) and variable @xmath41 . positive values ( blue to yellow ) indicate pair - binding . in ( a , c ) the horizontal axis is the repulsion strength @xmath11 , and in ( b , d ) the extended - hubbard parameter ratio @xmath42 . in ( a , c ) @xmath43 is in units of hz , and in ( b , d ) in units of the corresponding hopping parameter @xmath1 between nearest - neighbor wells . parameters used in the calculations for the ground states of @xmath44 , @xmath45 , and @xmath46 fermions are : @xmath47 khz , @xmath48 @xmath13 m , @xmath49 @xmath13 m , and @xmath50 , @xmath28 . @xmath17 varies in the ranges from @xmath51 khz ( @xmath52 ) to @xmath53 khz ( @xmath54 ) in ( a , b ) , and @xmath55 khz ( @xmath56 ) to @xmath57 khz ( @xmath58 ) in ( c , d ) . insets show cuts for an intersite barrier @xmath59 in ( a , b ) and @xmath60 in ( c , d ) . , width=279 ] figs . [ fig2_t](c , d ) display the sr - cpd2s of a 4-fermion plaquette for the ci states ( gs and 2ex ) associated with the two stars in fig . [ fig2_t](a ) . the gs sr - cpd2 in fig . [ fig2_t](c ) reveals the presence ( 89.4% ) of a @xmath3-wave rvb component , and the 2ex sr - cpd2 in fig . [ fig2_t](d ) reveals the presence ( 84.2% ) of an @xmath61-wave rvb component ( see the analysis in appendix [ spin ] ) ; these percentages are derived from the double - occupancy fraction at the fixed - point site [ i.e. , the hump volume above the arrows in figs . [ fig2_t](c , d ) ] . the above sr - cpd2 analysis points to a deficiency of the @xmath1-@xmath2 model because it excludes @xcite double occupancies . we further illustrate the limitations of the @xmath1-@xmath2 model for the full range of the repulsive interaction by plotting for all three approaches ( ci , hubbard , @xmath1-@xmath2 ) the 4-fermion and 2-fermion total energies ( @xmath27 ) for a single plaquette , as well as the double occupancy ( @xmath62 ) , as a function of @xmath42 ; see figs . [ fig2_t](e ) and [ fig2_t](f ) , respectively . evidently , the ci and hubbard energies , apart from a constant shift , are in very good overall agreement , whereas the @xmath1-@xmath2 values deviate greatly . the exact ci results provide a here - to - date lacking benchmark for assessments of the validity of the @xmath1-@xmath2 model and its variants @xcite , as well as the hubbard - calculated double occupancies . the hole - pairing gap may be defined in terms of the energies of a 4-site plaquette , as follows @xcite : @xmath63 , \label{dgap}\ ] ] where @xmath8 is the total number of fermions on the plaquette . ci - calculated pairing gaps are shown in fig . [ fig3_t ] for two interwell distances : @xmath64 @xmath13 m [ in ( a , b ) ] , and @xmath65 @xmath13 m [ in ( c , d ) ] . fig . [ fig3_t](a ) displays @xmath43 as a function of the repulsion sterngth , @xmath11 , and the interwell parameter @xmath41 in a single plaquette . the gap maximum ( @xmath66 0.72 hz ) occurs at ( @xmath67 @xmath68 , @xmath69 ) . to compare with the results of the hubbard model , fig . [ fig3_t](b ) displays the same ci results for @xmath43 , but with all energies expressed in units of the intersite tunneling parameter @xmath1 and the interaction strength @xmath11 expressed as the ratio of @xmath42 , with @xmath70 being the hubbard on - site repulsion extracted from the ci calculation . the maximum of @xmath43 is now @xmath71 , occurring at ( @xmath72 , @xmath73 ) . as a function of the inter - well barrier - height ratio @xmath41 . the distance between the wells is @xmath39 @xmath13 m , which results in @xmath74 khz . the remaining trap parameters are @xmath47 khz , @xmath48 @xmath13 m , and @xmath49 @xmath13 m ( note the logarithmic energy scale ) . , width=294 ] when @xmath75 @xcite , the maximum value ( i.e. , @xmath71 ) of @xmath43 and the range @xmath76 , corresponding to @xmath77 ( i.e. , hole pairing ) , agree with those found for a single - plaquette pure hubbard model @xcite . the additional dependence of @xmath43 on @xmath41 [ see fig . [ fig3_t](b ) ] can not be described by the standard hubbard model ; it reflects the effect of hamiltonian terms that are present in the ci calculation , but are absent in the standard hubbard model , pointing to possible applications of the extended hubbard hamiltonian @xcite in the optical traps assemblage . particularly relevant here is the off - site repulsion @xmath78 which effectively reduces @xcite the on - site @xmath70 . however , @xmath78 decreases strongly for increasing intersite barrier heights @xmath41 ( see fig . [ fig4_t ] ) , whereas @xmath70 is highly insensitive . when @xmath78 becomes sufficiently small relative to @xmath70 , the standard hubbard single - plaquette results are recovered [ see inset in fig . [ fig3_t](b ) ] . @xmath13 m and @xmath79 @xmath13 m , ( i ) illustration of the external potential for the double plaquette . ( a ) spectrum in the extended interval @xmath80 . ( b ) enlarged spectrum in the limited interval @xmath81 . , width=287 ] @xmath70 in fig . [ fig4_t ] is the on - site repulsion as determined from the ground state energy evolution as a function of the interaction strength @xmath11 in a single well . @xmath78 is the inter - site repulsive interaction , determined through the matrix element of @xmath82 ( see eq . [ 2bi ] ) taken for the ground state wavefunctions , described by gaussians determined through fitting to the ci spds . @xmath1 is the inter - well tunneling parameter , as determined through the energy difference between the two lowest energy states in the @xmath24 direction ( the tunneling split ) . note the insensitivity of @xmath70 vs @xmath41 compared to the strong variation in @xmath78 . the tunneling parameter @xmath1 exhibits a reduced sensitivity compared to @xmath78 . the @xmath78 term constitutes a perturbation to the pure hubbard model . when the @xmath78 term becomes of the order of magnitude of @xmath43 , i.e. , @xmath83 , the pairing gap vanishes ; this happens for values of @xmath84 for the plaquette considered in this figure . the role played by the off - site interaction @xmath78 is further illustrated in figs . [ fig3_t](c , d ) , where @xmath43 is plotted for the case when the intersite distance in the single plaquette is @xmath40 nm [ compared with @xmath39 nm in fig . [ fig3_t](a , b ) ] . clearly much higher values of the intersite potential barriers ( i.e. , @xmath85 ) are required to reach the @xmath77 region . the pairing - gap behavior estimated from a single 4-well optical plaquette should be reflected in the properties of a two - plaquette molecule ( tpm ) when the two plaquettes are weakly coupled ( wc ) . for the single plaquette [ eq . ( [ dgap ] ) and fig . [ fig3_t ] , @xmath43 indicates that the gs of the wc - tpm is associated with a ( left , right ) ( 4,2 ) or a ( 2,4 ) particle distribution , with the equal - particle arrangement between the ( left , right ) plaquettes , i.e. , ( 3,3 ) , corresponding to an excited state ; @xmath86 indicates reverse energy ordering . the gs and lowest excited state wave functions of the wc - tpm show complex behavior corresponding to entangled two mirror - reflected charge - density - wave ( cdw ) or spin - density - wave ( sdw ) symmetry - broken configurations . experimental probing and quantitative analysis of such entangled states has recently been demonstrated @xcite , based on inducing particle escape from the optical wells by lowering one side of the trapping potential . [ fig5_t ] displays ci results for the spectra of @xmath87 @xmath0li atoms in two weakly - coupled plaquettes separated by @xmath88 @xmath13 m . the well - separation within the identical plaquettes is @xmath39 @xmath13 m [ fig . [ fig5_t](i ) ] , so that the intraplaquette barriers are sufficiently high to yield a vanishing off - site @xmath78 , see above . the corresponding low - energy spectrum is plotted versus @xmath89 ; the entire spectrum is plotted in fig . [ fig5_t](a ) and an enlarged view of the spectrum in the interval @xmath90 is displayed in fig . [ fig5_t](b ) . we focus on the two ( lowest ) pairs of degenerate states denoted as ( @xmath91 , @xmath92 ) and ( @xmath93 , @xmath94 ) in fig . [ fig5_t](b ) . the ci method preserves all the quantum numbers , which are explicitly denoted in fig . [ fig5_t ] ; i.e. , ( @xmath95 ) associated with the parities @xmath96 and @xmath97 along the @xmath23 and @xmath24 directions , and the total spin values @xmath50 ( for @xmath91 , @xmath92 , @xmath94 ) or @xmath98 ( for @xmath93 ) . the pairs @xmath5 and @xmath6 constitute the four lowest - in - energy states in the whole range @xmath80 that we investigated [ fig . [ fig5_t](b ) ] . the crossing at @xmath99 is reflected in the modified spectrum generated by subtracting the energy @xmath100 at each point @xmath89 ; see the pocket in fig . [ fig6_t](a ) , where the horizontal axis is expressed now in units of @xmath42 and the crossing occurs at @xmath101 . for @xmath102 , the @xmath5 pair is lower in energy , with the @xmath103 , @xmath104 , and @xmath50 state ( denoted as @xmath91 ) being the ground state ; for @xmath105 , the @xmath6 pair is lower in energy , with the @xmath103 , @xmath104 , and @xmath98 state ( denoted as @xmath93 ) being the ground state . to investigate the microscopic structure of the states in the @xmath5 and @xmath6 pairs , we first display the ci spds [ @xmath106 and spin - polarization densities [ spols , @xmath107 for the states @xmath91 and @xmath93 in fig . [ fig6_t](b ) ; see definitions in appendix [ defs ] . these ci spds and spols are identical for both states , since both parities @xmath96 and @xmath97 , as well as the total spin @xmath108 , are preserved . in particular , the spds exhibit eight humps of equal height , integrating to a total number of @xmath87 particles . furthermore , the spols are structureless plane surfaces . the spds and spols for the ci states @xmath92 and @xmath94 exhibit a similar behavior . , or ( e - h ) by distorting the internal structure of one of the plaquettes . for the chosen value of @xmath42 in ( a ) with @xmath109 , no - pair - binding is observed as evident from the spd in ( c ) . however , for a larger tilt , @xmath110 , the spectrum in ( b ) and spd in ( d ) indicate gs pair - binding , reflecting a ( 4,2 ) configuration in ( k ) . similarly , changing the intra - plaquette energy barrier in the right plaquette from @xmath59 to @xmath111 ( e ) results in no pair binding [ see also spd in ( g ) ] for the chosen value of @xmath42 , whereas a slightly larger barrier , @xmath112 ( f ) induces pair - binding as evident from the spectrum in ( f ) and spd in ( h ) . , width=302 ] direct insight into the nature of the gs structure is afforded by the sr - cpdn shown in fig . [ fig6_t](c ) , where the appearance of a peak in the left plaquette ( see red peak ) , integrating to very - close to unit probability for finding an up - spin particle there when fixing the positions and spins of the other five atoms ( see black arrows ) , indicates a dominant ( 4,2 ) wc - tpm configuration . little double occupancy is portrayed by small red peaks on diagonal sites ( left plaquette ) . although the ci - calculated densities appear uniform across the two plaquettes [ see fig . [ fig6_t](b ) , the microscopic structures of the @xmath5 and @xmath6 states are different : they involve different superpositions of many primitive basis functions , @xmath37 , as exemplified by the @xmath32 superpositions in figs . [ fig6_t](i ) and [ fig6_t](ii ) . these superpositions contain also primitives with doubly - occupied sites ( denoted as @xmath3 earlier , not shown explicitly ) . for illustrative purposes , the pictorial representations in these panels describe visually one of these primitives associated with the ( 4,2 ) [ fig . [ fig6_t](i ) ] and the ( 3,3 ) [ fig . [ fig6_t](ii ) ] configurations . as aforementioned , detailed information can be extracted from the ci wave functions with the help of the sr - cpds @xcite . however , for @xmath87 fermions in @xmath113 sites , calculating the large number of needed cpds is time consuming . nevertheless , key features of the hidden anatomy of the entangled states in the @xmath5 and @xmath6 pairs can be revealed through the implementation of a forced breaking of their @xmath96 parity symmetry , i.e. , by subtracting and adding their member states and constructing the four auxiliary states @xmath114 and @xmath115 . the expectation value of @xmath96 in the @xmath116 and @xmath117 states vanishes ; this symmetry - breaking results in different characteristics of the above two symmetry - broken states . the spd for @xmath118 exhibits a ( 4,2 ) particle configuration [ fig . [ fig6_t](d ) , left ] , while the associated spol is featureless [ fig . [ fig6_t](d ) , right ] . the spd for @xmath119 corresponds to the mirror - reflected ( 2,4 ) configuration [ fig . [ fig6_t](e ) , left ] , with the associated spol remaining structureless [ fig . [ fig6_t](e ) , right ] . the two fermions ( or two holes ) in the two - particle plaquette are not localized on specific sites . indeed after the symmetry breaking , the wave functions are still an entangled superposition of many @xmath37 s , thus distributing the particles ( or holes ) with equal probability over all sites in each plaquette . in a plaquette with two fermions , the volumes under the four density humps integrate to @xmath46 , whereas the volumes under the four density humps in a 4-fermion plaquette integrate to @xmath44 . naturally , the two broken - symmetry states @xmath116 can be characterized as a charge density wave ( cdw ) , because they exhibit a modulation in the spds , but none in the spin polarizations . the spds and spin polarizations of the @xmath117 states are displayed in figs . [ fig6_t](f ) and [ fig6_t](g ) . in both cases , the spds are symmetric with respect to the left and right plaquettes . furthermore , the volumes under the humps in each plaquette integrate to @xmath120 , indicating that these states have a ( 3,3 ) configuration . however , the left - right asymmetry ( due to the broken @xmath96 symmetry ) emerges now as an asymmetry in the spin polarization . the three fermions ( with total spin @xmath121 ) in one plaquette have a spin peojection @xmath122 , while the remaining three fermions in the other plaquette have the opposite spin projection @xmath123 . the @xmath117 broken - symmetry states exhibit spin - density wave ( sdw ) characteristics , see right panels in figs . [ fig6_t](f , g ) . @xmath13 m and @xmath20 @xmath13 m showing degeneracy splitting between the gs and 1ex [ compare with fig . [ fig6_t](b ) , @xmath88 @xmath13 m ] . ( b , c ) the sr - cpdns corresponding to the gs and 1ex in ( a ) ( see stars respectively ) were evaluated for @xmath124 . the sr - cpdn observation points are marked by black arrows with indicated spin directions . the predicted sr - cpdns are denoted in red . , width=302 ] in addition to the above analysis , the ( 4,2 ) or ( 2,4 ) components of the @xmath5 states , but not the ( 3@xmath34,3@xmath35 ) or ( 3@xmath35,3@xmath34 ) components of the @xmath6 states can be separated in actual experiments by lifting the left / right degeneracy with the help of two processes : ( 1 ) tilting one plaquette with respect to the other [ fig . [ fig7_t](a - d ) ] , and ( 2 ) distorting one of the two plaquettes fig . [ fig7_t](e - h ) ] . from the ci spectrum depicted in fig . [ fig7_t](b ) , the degeneracy of the states in the _ a _ pair is lifted for a tilt @xmath110 , with one of the states becoming the gs at @xmath125 ( see vertical bar ) . moreover the corresponding ci spd plotted in fig . [ fig7_t](d ) shows that the tilt induced the symmetry breaking discussed earlier , i.e. , a ( 4.2 ) configuration ; for smaller tilt results , see figs . [ fig7_t](a , c ) . similar symmetry - breaking results ( i.e. , spectral energy gap corresponding to the emergence of a ( 3,3 ) , cdw configuration ) are found also for distortions of the plaquette - landscape [ see figs . [ fig7_t](e - h ) ] . while we focused here on conditions allowing exploration of pure interplaquette entanglement effects , the sensitivity of the energy spectrum and entanglement characteristics to the interplaquette distance ( i.e. , increasing tunneling between the plaquettes ) is illustrated in fig . [ fig8_t](a - c ) , where the spectrum for a double - plaqutee with @xmath20 @xmath13 m and @xmath126 @xmath13 m is displayed . the spectrum in fig . [ fig8_t](a ) shows formation of an energy gap between the two lowest states [ compare with fig . [ fig6_t](a ) ] . the sr - cpdn in fig . [ fig8_t](b ) , corresponding to the gs at @xmath124 ( see red star ) , reflects contributions from both the ( 4,2 ) and ( 3,3 ) configurations ( see the large red peak in the left plaquette and a smaller one in the right plaquette , corresponding to the probabilities of finding spin - up atoms in these locations ) . on the other hand the result [ fig . [ fig8_t](c ) ] for the 1ex state [ gray star in fig . [ fig8_t](a ) ] portrays formation of a pure ( 4,2)@xmath127(2,4 ) configuration [ compare to the ( 4,2)@xmath128(2,4 ) gs of the wc - tpm in fig . [ fig6_t](c ) ; the small difference between the @xmath128 and @xmath127 cases is not visible in the plots ] . in this work we have developed and implemented a configuration - interaction - based computational methodology for obtaining exact solutions to the microscopic many - body quantum hamiltonian describing ulracold fermionic atoms ( here @xmath0li atoms ) moving under the influence of an optically - induced multi - well confining potential surface and with short - range interatomic repulsive interactions ; we remark here that the same ci methodology can be extended in a straightforward manner to treat interatomic dipolar interactions , or confined electrons interacting via long - range coulomb interactions . a similar type of potential energy surface , made of the assembly of plaquettes each comprising four sites ( potential wells ) arranged in a square geometry has been introduced in early investigations @xcite of electronic high - temperature superconductivity in cuprate materials , in the context of the two - dimensional hubbard model often mentioned as a starting point for formulating a theoretical understanding of unconventional superconductivity @xcite . the hubbard model has been originally derived for the description of the behavior of strongly correlated electrons in solids @xcite . however , variants of this model have also been implemented , for over a decade , targeting investigations of strongly - interacting ultracold atoms ( bosonic or fermionic ) in optical lattices , with the interest in such systems originating from remarkable advances in atom cooling and optical trapping techniques , that opened heretofore untapped prospects of preparing , emulating , and measuring the behavior of strongly - interacting quantum many - body systems under pristine , defect - free , environments . analysis of such ultracold - atom emulations of interacting quantum many - body systems revealed on various occasions @xcite that many hubbard models that were simulated with ultracold atoms were found not to have a standard form meaning that the corresponding hamiltonians , required for analysis of these results , frequently had to incorporate terms that are not included in the standard hubbard model and its customary variants . these extra terms may include correlated and occupation - dependent tunneling within the lowest band , as well as correlated tunneling and occupation of higher bands @xcite . the above findings connote that progress in developing future ultracold - atom emulations of interacting quantum many - body systems ( including quantum magnetism and the origins of unconventional pairing mechanisms ) , and theoretical methods for the analysis of such emulations ( including the development and implementation of effective models , such as extended and non - standard hubbard hamiltonians ) , require benchmark exact calculations ( such as the ones introduced in this work ) , which owing to their ab - initio nature incorporate ( with no restriction or approximation ) all contributions to the many - body microscopic hamiltonian . with such computational methodology , one can then proceed to construct , assess , and improve effective models for a reliable analysis of the new forthcoming data . indeed , in this work we focused on providing benchmark exact quantum - mechanical solutions , aiding and enabling a bottom - up approach aimed at ultracold atoms experiments , and their analysis , starting from a single ultracold atom plaquette as a building block , and progressing in a systematic manner to double plaquettes ( with variable inter - plaquette couplings ) and larger , more complex , systems . our exact schrdinger - equation ultracold - fermionic - atoms - plaquette simulations demonstrated entangled @xmath3- and @xmath61- wave rvb states coexisting with partial double occupancies , uncovered hole - pairing phase diagrams , and explored the robustness of the energy spectrum and entanglement of a double - plaquette system ( assembled from entangled - multipartite - plaquettes ) through symmetry - breaking via interplaquette tilting and distortion and as a function of the strength of interplaquette coupling . these calculations may serve as a launchpad for further exact calculations , the development of approximate treatments @xcite , and experiments ( including the use of site resolved microscopy enabling direct observation of charge and spin correlations @xcite ) on systems with hierarchically increasing complexity @xcite . acknowledgment : work supported by the air force office of scientic research under award no . fa9550 - 15 - 1 - 0519 . calculations were carried out at the gatech center for computational materials science . we thank r. schumann for correspondence concerning the analytic hubbard - model solutions of the 4-site plaquette . as aforementioned , we use the method of configuration interaction for determining the solution of the many - body problem specified by the @xmath8-fermion general hamiltonian @xmath129 where @xmath29 , @xmath130 denote the vector positions of the @xmath131 and @xmath132 fermions ( e.g. , @xmath0li atoms ) . this hamiltonian is the sum of a single - particle part @xmath133 , which includes a kinetic energy term and a single - particle external confinement potential ( see fig . [ fig1_t ] ) that expresses formation of an interwell barrier between the individual wells , and the two - particle interaction @xmath82 . for the case of 2d ultracold atoms , the two - body repulsion is taken as a gaussian given by eq . ( [ 2bi ] ) . in the ci method , one writes the many - body wave function @xmath134 as a linear superposition of slater determinants @xmath135 that span the many - body hilbert space and are constructed out of the single - particle _ spin - orbitals _ @xmath136 and @xmath137 where @xmath138 denote up ( down ) spins . namely @xmath139 where @xmath140 and the master index @xmath141 counts the number of arrangements @xmath142 under the restriction that @xmath143 . of course , @xmath144 counts the excitation spectrum , with @xmath145 corresponding to the ground state . the many - body schrdinger equation @xmath146 transforms into a matrix diagonalizatiom problem , which yields the coefficients @xmath147 and the eigenenergies @xmath148 . because the resulting matrix is sparse , we implement its numerical diagonalization employing the well known arpack solver @xcite . the matrix elements @xmath149 between the basis determinants [ see eq . ( [ detexd ] ) ] are calculated using the slater rules @xcite . naturally , an important ingredient in this respect are the matrix elements of the two - body interaction , @xmath150 in the basis formed out of the single - particle spatial orbitals @xmath151 , @xmath152 . in our approach , these matrix elements are determined numerically and stored separately . the slater determinants @xmath153 [ see eq . ( [ detexd ] ) ] conserve the third projection @xmath154 , but not the square @xmath155 of the total spin . however , because @xmath155 commutes with the many - body hamiltonian , the ci solutions are automatically eigenstates of @xmath155 with eigenvalues @xmath156 . after the diagonalization , these eigenvalues are determined by applying @xmath155 onto @xmath157 and using the relation @xmath158 \psi^n_i,\ ] ] where the operator @xmath159 interchanges the spins of fermions @xmath131 and @xmath132 provided that their spins are different ; @xmath160 and @xmath161 denote the number of spin - up and spin - down fermions , respectively . when there is no tilt [ i.e. , @xmath162 , see fig . [ fig1_t](b ) ] , the @xmath163-parity operator associated with reflections about the origin of the axes is defined as @xmath164 and has eigenvalues @xmath95 . with the separable in @xmath23 and @xmath24 basis that we use , it is easy to calculate the parity eigenvalues for the slater determinants , eq . ( [ detexd ] ) , that span the many - body hilbert space . the many - body hamiltonian used in this paper ( without an applied magnetic field ) conserves also the partial @xmath165 and @xmath166 parities . denoting the ci wave function as @xmath167 , the single - particle density ( spd ) is defined as @xmath168 furthermore the single - particle spin polarization ( spol ) is defined as @xmath169 where @xmath34 and @xmath35 denote the up and down values , respectively , of the @xmath9 spin variable . the spol is the difference between the up and down spin densities . we probe the intrinsic structure of the ci eigenstates using the spin resolved conditional probability distribution @xmath170 defined by the expression @xcite @xmath171 where @xmath172 is the dirac delta function , @xmath8 the total number of particles , @xmath173 are particle indices , @xmath174 and @xmath175 represent the spins of particles @xmath131 and @xmath132 . the position of the fixed point is @xmath176 and the spin of the fixed particle is @xmath177 ; @xmath178 ( together with the associated spin at that location , @xmath177 ) is referred to as the observation point . @xmath9 is the spin of the particle whose spatial distribution we want to know . since the spin - resolved conditional probability distribution [ in eq . ( [ cpd ] ) ] is represented by a two - body operator we denote it as sr - cpd2 . however , we found that sometimes sr - cpd2s alone are not sufficient to fully decipher the intrinsic configuration of the emerging quantum states . in such a case , one needs to calculate higher correlation functions . in this paper , we use the sr - cpdn ( @xmath8-point correlation function ) defined as the modulus square of the full many - body ci wave function , i.e. , @xmath179 where one fixes the positions and spins of @xmath38 particles and inquiries about the ( conditional ) probability of finding the @xmath8th particle with spin @xmath9 at any position @xmath180 . in this section we provide a detailed description of the procedure used to identify the d - wave rvb state in our spin resolved sr - cpd2s . as first described in ref . @xcite one can map sr - cpd2s to spin eigenfunctions by analyzing the volumes underneath the sr - cpd2s . the versatility of this method has been demonstrated several times @xcite and detailed explanations of the methodology can be found in @xcite . this procedure was carried out by us to identify the rvb states within our plaquettes . the starting point is the general spin eigenfunction for four spin @xmath181 fermions trapped in a single plaquette with quantum numbers @xmath182 , which is given as @xcite : @xmath183 this function is parameterized by the angle @xmath184 and yields the quantum numbers @xmath182 for all values of @xmath184 . the angle @xmath184 is a free parameter that for certain values gives spin functions with specified characteristics ( e.g. , the orthogonal functions mentioned below ) , or can be treated as a `` fitting '' parameter ( see the discussion below eq . ( [ eq : partial_conditional_prob ] ) in connection with @xmath185 ) . ( [ eq : generalspinfun ] ) expresses in a compact form the fact that the dimension of the total - spin space for @xmath186 and @xmath50 , @xmath28 is 2 ; see the branching diagram in ref . @xcite and ref . for example , two orthogonal basis functions can be obtained by setting @xmath187 and @xmath188 . from the general spin function , one can read off the spin components contributing to a specific sr - cpd2 . for instance , for a spin down sr - cpd2 with a fixed spin - up fermion on position 1 ( counting from left to right in the corresponding kets ) , the spin components contributing to the conditional probability densities are the following three : @xmath189 this is referred to as the partial conditional probability @xmath190 ( partial because it corresponds to a part of the full sr - cpd2 , i.e. one peak for a specific spin configuration ) . using a normalized sr - cpd2 . one can directly equate the volume underneath the hump in position 2 to @xmath190 and determine the angle @xmath185 . due to the involved squares , this procedure is not necessarily unique . the unique solution can be found by comparing another hump ( i.e. , at position 3 ) to the corresponding partial conditional probability @xmath191 . however it is important to note that this procedure is only exact as long as ( i ) the overlap between sites is sufficiently small and ( ii ) the amount of double occupancy is small . from fig . [ fig2_t](c , d ) , it is clear that ( i ) is fulfilled . to assure that the possible error due to ( ii ) is as small as possible , we minimize @xmath192 where @xmath193 represents the volume under the sr - cpd2 hump at position @xmath131 . for the ground state sr - cpd2 shown in fig . [ fig2_t ] , this procedure yielded an angle @xmath194 . it is known that the spin function for the @xmath3-wave rvb is given as @xcite @xmath195 which corresponds to an angle of @xmath196 . we can therefore conclude that our ci results show the presence of a @xmath3-wave rvb state with @xmath197 fidelity . however while the ci result shows the presence of an rvb with high fidelity , the sr - cpd2 also undoubtedly shows the presence of double occupancy . the volume of the hump at position 1 amounts to @xmath198 . this might seem at first contradictory but is in fact the correct solution for an interaction strength of @xmath199 . to show this we calculated the ground state solution of the hubbard model @xmath200 which is given as @xmath201 the first two lines of eq . ( [ eq : hubbard_dwave_rvb ] ) are the @xmath3-wave rvb component and appear with dominant coefficients , however the wavefunction clearly contains contributions from doubly occupied states . summing the squared coefficients of the spin primitives that contain doubly occupied sites yields a double occupancy of @xmath202 in excellent agreement with the ci result . the same analysis can be performed for the @xmath61-wave rvb state . the spin function for an @xmath61-wave rvb state is given in literature as@xcite : @xmath203 . the angle determined from our ci is @xmath204 corresponding to a fidelity of @xmath205 however just like the ground state sr - cpd2 , the sr - cpd2 for the second excited state shows a non - zero double occupancy . therefore we computed the hubbard model solution for the 2nd excited state at @xmath199 : @xmath206 this eigenfunction predicts a double occupancy on site 1 of @xmath207 . the volume of the hump on position 1 indicating double occupancy in our ci amounts to @xmath208 in good agreement with the hubbard result . 99 g. zrn , f. serwane , t. lompe , a. n. wenz , m. g. ries , j. e. bohn , and s. jochim , fermionization of two distinguishable fermions , phys . lett . * 108 * , 075303 ( 2012 ) . s. murmann , a. bergschneider , v. m. klinkhamer , g. zrn , t. lompe , and s. jochim , two fermions in a double well : exploring a fundamental building block of the hubbard model , phys . * 114 * 080402 ( 2015 ) . s. murmann , f. deuretzbacher , g. zrn , j. bjerlin , s. m. reimann , l. santos , t. lompe , and s. jochim , antiferromagnetic heisenberg spin chain of a few cold atoms in a one - dimensional trap , phys . 115 , 215301 ( 2015 ) . a. m. kaufman , b. j. lester , c. m. reynolds , m. l. wall , m. foss - feig , k. r. a. hazzard , a. m. rey and c. a. regal , two - particle quantum interference in tunnel - coupled optical tweezers , science * 345 * , 306 ( 2014 ) . a. m. kaufman , b. j. lester , m. foss - feig , m. l. wall , a. m. rey and c. a. regal , entangling two transportable neutral atoms via local spin exchange , nature * 527 * 208 ( 2015 ) . r. a. hart , p. m. duarte , yang , x. liu , th . paiva , e. khatami , r. t. scalettar , n. trivedi , d. a. huse , and r. g. hulet , observation of antiferromagnetic correlations in the hubbard model with ultracold atoms , nature * 519 * , 211 ( 2015 ) . m. boll , t. a. hilker , g. salomon , a. omran , j. nespolo , l. pollet , i. bloch , and c. gross , spin- and density - resolved microscopy of antiferromagnetic correlations in fermi - hubbard chains , science * 353 * , 1257 ( 2016 ) . m. f. parsons , a. mazurenko , c. s. chiu , g. ji , d. greif , and m. greiner , site - resolved measurement of the spin - correlation function in the fermi - hubbard model , science * 353 * , 1253 ( 2016 ) . l. w. cheuk , m. a. nichols , k. r. lawrence , m. okan , h. zhang , e. khatami , n. trivedi , t. paiva , m. rigol , and m. w. zwierlein , observation of spatial charge and spin correlations in the 2d fermi - hubbard model , science * 353 * , 1260 ( 2016 ) . r. p. feynman , simulating physics with computers , int . . phys . * 21 * , 467 ( 1982 ) . b. b. brandt , c. yannouleas , and u. landman , double - well ultracold - fermions computational microscopy : wave - function anatomy of attractive - pairing and wigner - molecule entanglement and natural orbitals , nano lett . * 15 * , 7105 ( 2015 ) . m. a. garca - march , a. yuste , b. juli - daz , and a. polls , mesoscopic superpositions of tonks - girardeau states and the bose - fermi mapping , phys . rev . a * 92 * , 033621 ( 2015 ) . c. yannouleas , b. b. brandt , and u. landman , ultracold few fermionic atoms in needle - shaped double wells : spin chains and resonating spin clusters from microscopic hamiltonians emulated via antiferromagnetic heisenberg and @xmath1-@xmath2 models , new j. phys . * 18 * , 073018 ( 2016 ) . c. yannouleas and u. landman , symmetry breaking and quantum correlations in finite systems : studies of quantum dots and ultracold bose gases and related nuclear and chemical methods , rep . . phys . * 70 * , 2067 ( 2007 ) . ying li , c. yannouleas , and u. landman , artificial quantum - dot helium molecules : electronic spectra , spin structures , and heisenberg clusters , phys . b * 80 * , 045326 ( 2009 ) . a. szabo and n.s . ostlund , _ modern quantum chemistry _ ( mcgraw - hill , new york , 1989 ) . p. w. anderson , p. a. lee , m. randeria , t. m. rice , n. trivedi , and f. c. zhang , the physics behind high - temperature superconducting cuprates : the `` plain vanilla '' version of rvb , j. phys . cond . * 16 * , r755 ( 2004 ) ; p. w. anderson , who or what is rvb ? , phys . today * 61 * , 8 ( 2008 ) . m. ogata and h. fukuyama , the @xmath1-@xmath2 model for the oxide high - t@xmath4 superconductors , rep . * 71 * , 036501 ( 2008 ) . d. j. scalapino and s. a. trugman , local antiferromagnetic correlations and @xmath209 pairing , philos . mag . b * 74 * , 607 ( 1996 ) . tsai and s. a. kivelson , superconductivity in inhomogeneous hubbard models , phys . b * 73 * , 214510 ( 2006 ) . w .- f . tsai , inhomogeneous hubbard models ( ph . d. dissertation , university of california los angeles , 2008 ) . w. hofstetter , j. i. cirac , p. zoller , e. demler , and m. d. lukin , high - temperature superfluidity of fermionic atoms in optical lattices , phys . lett . * 89 * , 220407 ( 2002 ) . s. trebst , u. schollwck , m. troyer , and p. zoller , @xmath3-wave resonating valence bond states of fermionic atoms in optical lattices , phys . lett . * 96 * , 250402 , ( 2006 ) . a. m. rey , r. sensarma , s. flling , m. greiner , e. demler , and m. d. lukin , controlled preparation and detection of @xmath3-wave superfluidity in two - dimensional optical superlattices , eur . . lett . * 87 * , 60001 ( 2009 ) . b. paredes and i. bloch , minimum instances of topological matter in an optical plaquette , phys . a * 77 * , 023603 ( 2008 ) . s. nascimbne , y .- a . chen , m. atala , m. aidelsburger , s. trotzky , b. paredes , and i. bloch , experimental realization of plaquette resonating valence - bond states with ultracold atoms in optical superlattices , phys . lett . * 108 * , 205301 ( 2012 ) . l. m. falicov and r. a. harris , two - electron homopolar molecule : a test for spin - density waves and charge - density waves , j. chem . * 51 * , 3153 ( 1969 ) . o. dutta , m. gajda , ph . hauke , m. lewenstein , d .- s . lhmann , b. a. malomed , t. sowiski and j. zakrzewski , non - standard hubbard models in optical lattices : a review , rep . 78 * , 066001 ( 2015 ) . f. c. zhang and t. m. rice , effective hamiltonian for the superconducting cu oxides , phys . b * 37 * , 3759(r ) ( 1988 ) . the mass of @xmath0li is taken to be 10964.90@xmath210 , with @xmath210 being the free - electron mass . this gives @xmath48 @xmath13 m for @xmath211 khz . r. a. doganov , s. klaiman , o. e. alon , a. i. streltsov , and l. s. cederbaum , two trapped particles interacting by a finite - range two - body potential in two spatial dimensions , phys . a * 87 * , 033631 ( 2013 ) . c. yannouleas and u. landman , spontaneous symmetry breaking in single and molecular quantum dots , phys . rev . lett . * 82 * , 5325 ( 1999 ) ; erratum * 85 * , 2220 ( 2000 ) . see supplemental material ( in appendices d and e ) . l. o. baksmaty , c. yannouleas , and u. landman , rapidly rotating boson molecules with long- or short - range repulsion : an exact diagonalization study , phys . a * 75 * , 023620 ( 2007 ) . a. paramekanti , m. randeria , and n. trivedi , high - t@xmath4 superconductors : a variational theory of the superconducting state , phys . b * 70 * , 054504 ( 2004 ) . j. spaek , m. zegrodnik , and j. kaczmarczyk , universal properties of high - temperature superconductors from real - space pairing : @xmath1-@xmath2-@xmath70 model and its quantitative comparison with experiment , phys . b * 95 * , 024506 ( 2017 ) . r. schumann , thermodynamics of a 4-site hubbard model by analytical diagonalization , ann . ( leipzig ) * 11 * , 49 ( 2002 ) . p. a. lee , n. nagaosa , and x. g. wen , doping a mott insulator : physics of high - temperature superconductivity , rev . 78 * , 17 ( 2006 ) . p. a. lee , from high temperature superconductivity to quantum spin liquid : progress in strong correlation physics , rep . . phys . * 71 * , 012501 ( 2008 ) . j. hubbard , generalized wigner lattices in one dimension and some applications to tetracyanoquinodimethane ( tcnq ) salts , phys . b * 17 * , 494 ( 1978 ) . d. baeriswyl d. eichenberger , and m. menteshashvili , variational ground states of the two - dimensional hubbard model , new j. of phys . * 11 * , 075010 ( 2009 ) . h. yokoyama , m. ogata , and y. tanaka , mott transitions and @xmath3-wave superconductivity in half - filled - band hubbard model on square lattice with geometric frustration , j. phys . japan * 75 * , 114706 ( 2006 ) . f. mezzacapo , a. angelone , and g. pupillo , two holes in a two - dimensional quantum antiferromagnet : a variational study based on entangled - plaquette states , phys . b * 94 * , 155120 ( 2016 ) . c. regal , bringing order to neutral atom arrays , science * 354 * , 972 ( 2016 ) . d. barredo , s. de lsleuc , v. lienhard , th . lahaye , and a. browaeys , an atom - by - atom assembler of defect - free arbitrary two - dimensional atomic arrays , science * 354 * , 1021 ( 2016 ) . m. endres , h. bernien , a. keesling , h. levine , e. r. anschuetz , a. krajenbrink , c. senko , v. vuletic , m. greiner , m. d. lukin atom - by - atom assembly of defect - free one - dimensional cold atom arrays , science * 354 * , 1024 ( 2016 ) . r. b. lehoucq , d. c. sorensen , and c. yang , _ arpack users guide : solution of large - scale eigenvalue problems with implicitly restarted arnoldi methods _ ( siam , philadelphia , 1998 ) . ying li , confined quantum fermionic systems , ( ph . d. dissertation , georgia institute of technology , 2009 ) . r. pauncz , _ spin eigenfunctions , construction and use _ ( plenum press , new york , 1979 ) . * supplemental material * + ( references are renumbered ; see list at the end ) for our single plaquette calculations we combine the double well external potentials from @xcite in both the @xmath23 and @xmath24 directions . @xmath212 where @xmath213 is the double well external potential ( see fig . 1 in the main paper ) . this results in a two - dimensional confining potential that has four minima and smooth connecting necks between them . we have full control over the distances between the minima in @xmath23 and @xmath24 and over the barriers between them . the set of parameters used to control the double - well potential are @xmath17 , @xmath214 , and @xmath215 , where @xmath17 is the height of the smooth inter - well barrier , @xmath214 is the tilt between the left and right well , and @xmath215 is the distance between the wells @xcite . for a graphical illustration of the double - well potential , including the definitions of @xmath17 , @xmath216 , and @xmath19 , see fig . 1 of the main paper . for our double - plaquette calculations we use a repetition of the single plaquette external potential from sec.[sec : single_plaquette_hamiltonian ] . we keep the potential along @xmath24 as is . for the potential along @xmath23 we start out with three double - well potentials and use heaviside theta functions @xmath217 to cut the inward facing potential walls to the right and/or left of the minima . we then combine the resulting potentials . this procedure is illustrated in fig . [ fig : shematicquadruplewellassembly ] . the above procedure applies directly to large inter - plaquette distances . for smaller inter - plaquette distances end - effects of the two - plaquette assembly are alleviated through `` symmetrization '' of the trapping potential at the left and right edges of the double - plaquette assembly so that it takes the shape of the intraplaquette interwell potential from the bottom of the well to the midpoint between sites . just like in sec . [ sec : single_plaquette_hamiltonian ] this gives us full control over the positions of the minima in @xmath23 and @xmath24 and allows us to independently vary the barriers in the @xmath23 and @xmath24 directions . unless otherwise stated the parameters for our double - plaquette are : @xmath218 khz , @xmath48 @xmath13 m , @xmath39 @xmath13 m ; ( intraplaquette ) , @xmath22 khz ( intraplaquette ) @xmath219 , @xmath220 @xmath13 m for all potential wells , and total - spin projection @xmath28 . the @xmath17 between the plaquettes ( interplaquette ) varies depending on the interplaquette distance ( @xmath21 is kept constant ) . for rapid convergence of the full ci calculations , it is often beneficial to use a basis that already diagonalizes the single particle part of the hamiltonian . since the single - particle hamiltonians used in this paper are separable in @xmath23 and @xmath24 we have to solve two one - dimensional schrdinger equations . a software package that does so , with high performance and high accuracy , is chebfun @xcite . it uses chebychev polynomials as an efficient and highly accurate approximation of the true wave functions . from the documentation : `` the implementation of chebfun is based on the mathematical fact that smooth functions can be represented very efficiently by polynomial interpolation in chebyshev points , or equivalently , thanks to the fast fourier transform , by expansions in chebyshev polynomials '' ; and `` chebfun makes use of adaptive procedures that aim to find the right number of points automatically so as to represent each function to roughly machine precision , that is , about 15 digits of relative accuracy . '' we have confirmed the accuracy of these solutions with both symbolic and numeric calculations in mathematica and the agreement was excellent at significantly higher performance . to obtain numerical solutions to the hubbard model we use the software package sneg @xcite , which is a `` mathematica package for symbolic calculations with second - quantization - operator expressions '' . from the documentation : `` sneg library is a mathematica package that provides a framework for performing calculations using the operators of the second quantization with an emphasis on the anti - commuting fermionic operators in the context of solid - state and atomic physics . it consists of a collection of transformation rules that define the algebra of operators and a comprehensive library of utility functions . '' 29 ying li , c. yannouleas , and u. landman , artificial quantum - dot helium molecules : electronic spectra , spin structures , and heisenberg clusters , phys . b * 80 * , 045326 ( 2009 ) . b. b. brandt , c. yannouleas , and u. landman , double - well ultracold - fermions computational microscopy : wave - function anatomy of attractive - pairing and wigner - molecule entanglement and natural orbitals , nano lett . * 15 * , 7105 ( 2015 ) . c. yannouleas , b. b. brandt , and u. landman , ultracold few fermionic atoms in needle - shaped double wells : spin chains and resonating spin clusters from microscopic hamiltonians emulated via antiferromagnetic heisenberg and @xmath1-@xmath2 models , new j. phys . * 18 * , 073018 ( 2016 ) . t. a. driscoll , n. hale , and l. n. trefethen , _ chebfun guide _ ( pafnuty publications ) . r. itko , sneg mathematica package for symbolic calculations with second - quantizationoperator expressions , comp . . commun . * 182 * , 2259 ( 2011 ) . schumann , r. , thermodynamics of a 4-site hubbard model by analytical diagonalization , ann . phys . , * 11 * : 4988 ( 2002 ) .

based on the beltrami - de sitter spacetime , we present the newton - hooke model under the newton - hooke contraction of the spacetime with respect to the transformation group , algebra and geometry . it is shown that in newton - hooke space - time , there are inertial - type coordinate systems and inertial - type observers , which move along straight lines with uniform velocity . and they are invariant under the newton - hooke group . in order to determine uniquely the newton - hooke limit , we propose the galilei - hooke s relativity principle as well as the postulate on newton - hooke universal time . all results are readily extended to the newton - hooke model as a contraction of beltrami - anti - de sitter spacetime with negative cosmological constant . + + keywords : newton - hooke space - time , galilei - hooke s relativity principle , universal time postulate , contraction , beltrami - de sitter spacetime

hep-th0403013

it is well known that there are different kinds of kinematics on homogeneous ( 3 + 1)-d space - times @xcite , and all of them can be contracted from minkowski , de sitter or anti - de sitter spacetimes under group contraction @xcite , respectively . among them , newton - hooke ( nh ) space - times @xmath1 introduced in @xcite are of special interest as a kind of non - relativistic cosmological models in @xcite . in this paper , we present a model for the nh space - time @xmath2 based on the beltrami - de sitter ( @xmath0 ) spacetime @xcite , denoted by @xmath3 . we find that in @xmath2 there exist a kind of special coordinate systems in which the test particles , which none force acts upon , move at constant coordinate velocities along straight lines . in the coordinates , these particles look like free particles rather than the particles driven by a repulsive force provided by the positive cosmological constant in other coordinates , say , the ones contracted from the _ static _ @xmath4 universe @xcite . at the first glance , this property seems very strange . why in @xmath2 contracted from the @xmath4 spacetime with constant curvature there exist such a kind of uniform - velocity motions ? the basic reason is , in fact , that there exist a kind of inertial - type motions in the beltrami model of the @xmath4 spacetime . such an inertial - type property comes in @xmath2 from the @xmath0 spacetime @xmath3 . among various models of @xmath4 spacetimes , the beltrami model is an important one in which @xmath4 spacetime is in analog with minkowski spacetime in the sense that there are inertial - type coordinates , inertial - type motions and they are invariant under fractional linear transformations of a 10-parameter group @xmath4 group @xmath5 . it is precisely the beltrami model @xcite of a 4-hyperboloid @xmath6 in the 5-d minkowski spacetime . in @xmath3 , a set of beltrami coordinate systems covers @xmath6 patch by patch and test particles and light signals move along the timelike and null geodesics , respectively , at _ constant _ coordinate velocities . therefore , they look like in free motions in a spacetime without gravity . thus , the beltrami coordinates and observers @xmath7 in these systems may be regarded as of global inertial - type " . this is why there is such a merit of the nh contraction from @xmath0 spacetime that in @xmath8 there remain inertial - type coordinates and inertial - type observers . this also shows that the usual algebraic definition of the nh limit can not determine the contraction of geometry uniquely because of free choice of coordinates if coordinates have no fundamental physical meaning . in particular , for the same kind of coordinates the contraction depends on what dimension of the coordinates might be chosen . in order to determine the unique nh limit , we should supplement certain physical conditions . we argue that the suitable physical conditions should be the counterpart of galilean relativity principle in @xmath2 , which is named the galilei - hooke s relativity principle , and the postulate on newton - hooke universal time @xmath9 , which satisfies @xmath10 and the latter is invariant under the newton - hooke contraction of @xmath11 , i.e. the limit of infinite signal - velocity @xmath12 and the curvature radius @xmath13 of the @xmath0 spacetime . the newton - hooke universal time is similar to the newton s universal time in newtonian mechanics in the sense that it is separated from space in the metric . we also find that there is something interesting in @xmath2 , which is closely related to the violation of euclid s fifth axiom . in newtonian mechanics and special relativity , a particle in inertial motion can be stationary in only one inertial frame . if an inertial frame @xmath14 has a relative velocity in another one , @xmath15 , the relative velocity will be constant at every time and every point . in @xmath2 , however , it is possible that a particle in inertial - type motion can be at rest in two different inertial - type frames . in this case the velocity of @xmath14 relative to @xmath15 will be different from point to point . this paper is organized as follows . in sec . [ bds ] , we briefly review @xmath0 spacetime . in sec . [ nhlimit - im ] we study the contraction of transformation group and geometry of @xmath0 spacetime in the nh limit . we also study motion of particles and light signals in the nh limit of @xmath0 spacetime . in sec . [ unique - rp ] , we show why the usual limiting procedure can not lead to the unique nh limit . we propose the galilei - hooke s relativity principle and the postulate on newton - hooke universal time to uniquely determine the contraction of geometry in the nh limit . in sec . [ sect : gref ] , we discuss the behaviors of inertial frames in @xmath2 that are different from newtonian mechanics and special relativity . geometric diagrams are also given in this section . finally , we end with a few concluding remarks . in appendix [ appdx : conncurvt ] , we list some results of the connection and curvature of @xmath0 spacetime and their contraction in the nh limit . in appendix [ appdx : limit ] we use the 5-d minkowski spacetime to show why the nh limit is not unique . we start with a 4-d hyperboloid @xmath6 embedded in a 5-d minkowski spacetime with @xmath16 : [ 5sphr ] s _ : & & ^_ab ^a ^b= -r^2 , + [ ds2]&&ds^2=^_ab d^a d^b , where @xmath17 , @xmath18 , and @xmath19 is the cosmological constant . clearly , eqs.([5sphr ] ) and ( [ ds2 ] ) are invariant under @xmath4 group @xmath20 . the @xmath0 spacetime @xmath21 is a @xmath4 spacetime defined by the following beltrami coordinates on @xmath22 patch by patch @xcite . clearly , @xmath23 . for intrinsic geometry of @xmath3 , there are at least eight patches @xmath24 . in @xmath25 , for instance , the beltrami coordinates are [ u4]&&x^|_u_4 = r ^/^4,=0 , , 3 ; + & & ^4=((^0)^2-_a=1 ^ 3(^a)^2 + r^2 ) ^1/2 > 0 . it is important that the coordinate transformation in each intersection is a fractional linear transformation . for example , let @xmath26 ( @xmath27 1 , 2 , 4 ) be the coordinates on @xmath28 , then , in @xmath29 , the transformation is @xmath30 ( @xmath31 ) , @xmath32 , which is induced by @xmath33 . in each patch , there are condition and beltrami metric @xmath34dx^\mu dx^\nu.\end{aligned}\ ] ] under fractional linear transformations of @xmath35 @xmath36 where @xmath37 in @xmath38 , condition ( [ domain ] ) and metric ( [ bhl ] ) are invariant . note that eqs.([domain])-([g ] ) are defined on @xmath39 patch by patch . this is , in fact , a cornerstone for the special relativity - type principle . in addition , at the origin of the coordinate system , the metric ( [ bhl ] ) becomes minkowskian . the generators of @xmath35 in beltrami coordinates are expressed as @xmath40 hereafter , we use bold roman letters to denote generators of lie group . later we shall use a bold italic letter @xmath41 to denote the triple @xmath42 ( see , eq.([eml ] ) , for example ) . they form an @xmath43 algebra : @xmath44 & = & r^{-2 } \mathbf{l}_{\mu\nu } \nno\\ { [ } \mathbf{l}_{\mu\nu},\mathbf{p}_\rho { ] } & = & \eta_{\nu\rho } \mathbf{p}_\mu - \eta_{\mu\rho } \mathbf{p}_\nu \label{so(1,4)}\\ { [ } \mathbf{l}_{\mu\nu},\mathbf{l}_{\rho\sigma } { ] } & = & \eta_{\nu \rho } \mathbf{l}_{\mu\sigma } - \eta_{\nu \sigma } \mathbf{l}_{\mu\rho } + \eta_{\mu \sigma } \mathbf{l}_{\nu \rho } - \eta_{\mu\rho } \mathbf{l}_{\nu \sigma } , \nno\end{aligned}\ ] ] as expected . the casimir operators of @xmath35 are ( see , e.g. , @xcite)[c1 ] _ 1 : = _ ^ -1 2r^2 _ ^ , _ 2 : = _ ^-1 r^2 ^2 , where[sw ] ^&= & ^ _ , ^=^ ^ _ , + _ & = & 1 2 _ ^^ , ^=^ _ , + & = & 1 8 _ ^ ^. in eq.([sw ] ) , @xmath45 is 4-d levi - civita symbol in flat spacetime with @xmath46 . according to the spirit of einstein s special relativity , one can define simultaneity such that two events @xmath47 and @xmath48 are simultaneous if and only if the beltrami time coordinates @xmath49 s for the two events coincide , [ s1]a^0:=x^0(a ) = x^0(b)=:b^0 . it defines the laboratory time in one patch . the simultaneity also defines a 3 + 1 decomposition of spacetime ds^2 = n^2 ( dx^0)^2 - h_ij ( dx^i+n^i dx^0 ) ( dx^j+n^j dx^0 ) with the lapse function , shift vector , and induced 3-geometry on 3-hypersurface @xmath50 in one coordinate patch @xmath51\}^{-1/2 } , \nonumber \\ & & n^i = x^0 x^i [ r^2-(x^0)^2]^{-1 } , \\ & & h_{ij}=\dl_{ij } \si_{\si}^{-1}(x)- { [ r\si_{\si}(x)]^{-2 } \dl_{ik } \dl_{jl}}x^k x^l , \nonumber\end{aligned}\ ] ] respectively , where @xmath52 , @xmath53 is the kronecker @xmath54-symbol , @xmath55 . in particular , at @xmath56 , @xmath57 . in @xmath3 , the simultaneity can also be defined with respect to the proper time of a clock rested at the spatial origin of the coordinate system @xcite : @xmath58 if this proper time is chosen as the time coordinate , the metric ( [ bhl ] ) becomes the robertson - walker - like one as follows @xcite : @xmath59 where l[spacelike ] dl__t^2:=__t^-2(x ) ( _ _ t(x ) _ ij -1 r^2 _ ik_jlx^k x^l ) dx^i dx^j , + _ _ t(x , x):=1 + 1r^2_ijx^i x^j>0 . the metric eq.([dsrw ] ) is closely linked with the cosmological principle . the two kinds of simultaneity eqs.([s1 ] ) and ( [ ptime ] ) indicate that in @xmath21 there is a relation between the special relativity - type principle , which could be introduced in the following subsection , and the cosmological principle . it is important to see the advantage of beltrami coordinates both test free massive particles and light signals in @xmath3 are moving along timelike straight world lines and the null ones , respectively . in terms of straight lines , we refer to those curves with equations in a linear form . this indicates that both their motions and the beltrami coordinate systems are of the inertial - type . furthermore , these properties are invariant under the fractional linear transformations ( [ g ] ) . in @xmath3 geodesics , which satisfy [ geod ] + ^ _ = 0with ^ _ = 1r^2(x ) ( x_^_+ x_^ _ ) , [ gamma ] are lobachevski - like straight world lines @xcite . in particular , timelike geodesics have the form @xmath60 with the initial condition @xmath61 and the constraint @xmath62 where @xmath63 is a new parameter such that [ w1 ] w(s ) = . null geodesics have the form [ bdsnullg ] x^= c^w + b^ , with the initial condition @xmath64 and the constraint @xmath65 where @xmath66 is an affine parameter and @xmath67 is a new parameter such that [ w2]w ( ) = with @xmath68 namely , free particles and light signals move along straight lines with a uniform component velocity @xmath69 which implies that the beltrami coordinate system is of physical meaning as inertial - type . furthermore , for a free particle 4-momentum and 4-angular momentum , @xmath70 and @xmath71 , defined by p^&:= & , + [ angular4]l^&:=&x^p^-x^p^ , are conserved along a geodesic , i.e. & = & 0 , + & = & 0 . they constitute a 5-d angular momentum @xmath72 for a free particle in @xmath6 in such a way that @xmath73 , @xmath74 , and it is conserved along the geodesic , = 0 . the einstein s famous formula can be generalized in @xmath39 : [ eml ] & -12r^2l^abl_ab = e^2-^2 c^2 + ^2- ^2=m_0 ^ 2 c^4 , & + & e c = p^0 , p^i = p^i , k^i=1 c l^0i , j^i=1 2 ^ijkl^jk , & where @xmath75 and @xmath76 is totally anti - symmetric with @xmath77 . as mentioned in the previous subsection , in this paper we use a bold italic letter , @xmath78 , say , to denote the triple @xmath79 . notations such as @xmath80 , @xmath81 and @xmath82 are abbreviations of @xmath83 , @xmath84 and @xmath85 , respectively . in this sense , sometimes we do not care whether the summed indices are one upper and one lower , as done in newtonian mechanics . in addition , eq.([eml ] ) gives rise the energy of photons , e_= . in any case , these offer a consistent way to define the observables for free particles and this kind of definitions differ from any others in @xmath4 space . of course , these issues significantly indicate that the motion of a free particle in @xmath3 should be of inertial - type in analog to newton s and einstein s conception for the inertial motion of a free particle with constant velocity . consequently , the coordinate systems with beltrami metric should be linked with inertial - type frames and corresponding observer should be of inertial - type as well . the deviation of nearby , ` same - directed ' , straight world lines is governed by the geodesic deviation equation [ gde ] + r^ _ ^ = 0,where @xmath86 is the deviation vector of nearby geodesics such that [ dv ] g _ ^ = 0 on the straight world lines . for simplicity , consider two static particles in a beltrami coordinate system , the zero - component of the deviation vector of which is identical to zero . then , eqs.([gde ] ) and ( [ dv ] ) reduce to - 2 r^2 = 0and _ ijx^j ^i = 0 . they give rise to [ bgde ] = 0 . notice that for the two static particles @xmath87 vainishes initially . thus , measured by use of beltrami coordinates , the deviation of two static particles in the same system will be unchanged in the evolution@xcite . the conclusion is not only valid in a vicinity of the spatial origin of a system , but also valid at other places . as was noted in @xcite , all above results indicate that a principle similar to the principle of the special relativity for the inertial motion with respect to the inertial coordinates in special relativity could be set up in the beltrami - de sitter spacetime @xmath3 . it is named the beltrami - de sitter relativity principle . in order to consider the non - relativistic limit of @xmath39 , we set @xmath88 and @xmath89 , and correspondingly , we assume that @xmath90 and @xmath91 with @xmath92 and @xmath93 being finite under the limit . naively , this non - relativistic limit may be attained when we simply let @xmath94 , but this limit is not well defined if @xmath13 remains finite ( for algebraic reasons , see , e.g. @xcite ) . in order to obtain the meaningful non - relativistic limit , one has to consider the so - called newton - hooke limit , which is defined as [ nu ] c , r , = c r is a positive , finite constant . @xmath95 has the same dimension as time . furthermore , @xmath9 , @xmath96 and the primed quantities correspondingly are assumed to be finite under the limit . this is a crucial requirement for the newton - hooke limit . otherwise , the limit will not be unique . in sec.[unique - rp ] , we shall discuss the problem in details . under the nh limit , [ sigma ] _ c , r ( x ) = 1 - ^2 t^2:=_+(t ) , _ c , r ( a , x ) = 1 - ^2t_a t:=_+(t_a , t ) . the condition @xmath97 can be assured by [ t ] - < t < . it can be shown that the transformations ( [ g ] ) become [ gnh ] t = , x^i= ( x^j - a^j - u^jt ) , o^i _ j so(3 ) , which is called a _ newton - hooke transformation_. the first equation of eq.([gnh ] ) shows that @xmath9 is separated from space in the transformation and that the simultaneity for @xmath9 is absolute . in this sense , we call @xmath9 the newton - hooke universal time . an nh transformation has inverse transformation , reading t = , x^i = o^ i_j ( x^j - a^j - u^j t ) with t_a = - t_a , a^i = - o^i_j , u^i = - o^i_j . in the inverse transformation and follows , @xmath98 are the entries of the inverse matrix of @xmath99 , satisfying @xmath100 so the inverse transformation of an nh transformation is still an nh transformation . let @xmath101 be the set of all the nh transformations . then it follows that @xmath101 is a subset of the group of diffeomorphisms . it can be further shown that an nh transformation with parameters @xmath92 , @xmath102 , @xmath103 and the rotation @xmath104 followed by another with parameters @xmath105 , @xmath106 , @xmath107 and the rotation @xmath108 is equivalent to an nh transformation with the following parameters , @xmath109 in short , the composition of two nh transformations are still an nh transformation . since @xmath101 , the subset of the group of diffeomorphisms , is closed under the inverse operation and multiplication , it follows that @xmath101 is a subgroup . we call @xmath101 the newton - hooke group . it is a fractional linear realization of a lie group , having 10 free parameters , @xmath110 , and three parameters in @xmath111 . note that the transformation for time coordinate is independent of space coordinates , and that the time transformation is still fractional linear while the transformation for space coordinates are linear among themselves . in other words , if two events have the same value for coordinate @xmath9 , they have the same value for @xmath112 . thus the simultaneity defined by the ` time ' coordinate ( either @xmath9 or @xmath112 ) is absolute under the action of the nh group . such a space - time is called the newton - hooke space - time , denoted by @xmath2 as before . from the transformation ( [ gnh ] ) , the ` time - translation ' generator with respect to @xmath92 reads @xmath113 similarly , the ` space - translation ' generators with respect to @xmath102 and boost generators with respect to @xmath103 can be easily obtained as * p*_i=_i , _ i = t_i , respectively . the space - rotation generators are the generators of @xmath114 as usual . from the above explicit forms of generators , we can readily write down the following lie algebra : [ eq : liealg]&[*h*,*p*_i]=^2 * k*_i,= * p*_i , & + & [ * j*_i,*p*_j]=_ijk * p*_k,=_ijk * k*_k,=_ijk * j*_k & + & and other lie brackets vanish , & where @xmath115 is totally anti - symmetric with @xmath116 . the lie algebra may also be reached by taking the nh contraction of the lie algebra ( [ so(1,4 ) ] ) on @xmath0 noting = c*p*_0 , _ i=1 c * l*_0i , * j*_i=1 2 _ ijk * l*_jk , and it is exactly the same as @xmath117 in eq.(13 ) of @xcite as long as setting @xmath118 . the first casimir operator is [ eq : c1 ] _ 1 = * p*_i * p*_i-^2 * k*_i * k*_i , which can also be obtained from the contraction of @xmath119 from eq.([c1 ] ) . the second casimir operator tends to zero as @xmath12 ( and @xmath13 ) @xmath120 . if the nh quantum mechanics and other physical aspects are concerned , we must consider the central extension of @xmath117 and the corresponding casimir operators . we will examine these aspects in ref . obviously , the nh limit will not alter the definition of simultaneity , eq.([s1 ] ) . under the nh limit , eq.([lapseshift ] ) turns out to be @xmath121 and the invariant beltrami metric ( [ bhl ] ) now splits to two metrics for time and space , respectively . they are [ tmetric ] d^2=1 _ + ^2(t)dt^2= , and in a hypersurface of simultaneity [ dl_3]d l^2=_+^-1(t)dl_0 ^ 2 , d l_0 ^ 2 : = _ ij dx^i dx^j . note that under the nh limit the second definition of simultaneity with respect to the proper time ( [ ptime ] ) of a clock rested at the spatial origin of the coordinate system in @xmath3 now coincides with the first one . this implies that @xmath122 on the whole @xmath2 and @xmath123 on each hypersurface of simultaneity are defined intrisically . namely , they are invariant under the nh transformations . we can examine the fact , using eq.([gnh ] ) and formulae @xmath124 @xmath125 and in the hypersurface of simultaneity ( @xmath126 ) @xmath127 the fact can also be checked directly by calculating the lie derivatives of @xmath122 and @xmath123 with respect to all the generators . under the nh limit , the robertson - walker - like metric ( [ dsrw ] ) in @xmath3 also splits to two parts . the detailed discussion is left in the next section . it should be noted that the space - time in the nh limit is not a 4-d metric spacetime because there is no invariant metric on it . in spite of that , a connection still exists as the contraction of the connection on @xmath3 . let the coefficients of the connection on @xmath2 be denoted by @xmath128 , where the indices take values from @xmath9 , @xmath129 , @xmath130 and @xmath131 . then a direct calculation results in all of them , among which only the following are nonzero : @xmath132 a further discussion on connection and curvature in @xmath2 is made in appendix [ appdx : conncurvt ] . under the nh limit , the geodesic equation ( [ geod ] ) becomes [ geod+t ] & + & ^_tt = 0 , + [ geod+i ] & + & 2^i_tj = 0 . the general solution of eq.([geod+t ] ) is @xmath133 where @xmath134 and @xmath135 are two integral constants . depending on whether @xmath136 is zero or not in the initial condition , this can be reduced to @xmath137 for convenience , we call a curve in @xmath2 timelike if @xmath136 is nonzero at every parameter @xmath138 along it , and spacelike if it is a curve in the hypersurface @xmath139 . a timelike curve is also called a world line of a particle , as did in relativity . for a free particle with mass @xmath140 , the second expression of eq.([tau - t ] ) can be directly obtained from eq.([tmetric ] ) under the initial condition @xmath141 when @xmath142 . eq.([geod+i ] ) becomes & + & 2 ( ) = 0 , which can be integrated out : x^i= v^i + b^i . this is the solution with the initial condition [ nhini]t|_=0=0 , x^i|_=0=b^i , . |_=0=v^i . by use of the second expression of eq.([tau - t ] ) , the geodesic appears in an explicit form of a straight world line @xmath143 this property is in analog with the straight line in the beltrami model of lobachevski plane @xcite . the parameter @xmath144 can also be obtained from eq.([w1 ] ) under the initial condition eq.([nhini ] ) as long as @xmath145 . ( the situation corresponding to @xmath145 does not appear in the nh limit . ) in the nh limit , free particles move along straight lines at constant velocities @xmath146 eq.([gnh ] ) gives rise the velocity addition law v^i : = = [ _ + ( t_a , t ) v^j - u^j+^2t_a ( x^j - a^j ) ] . in particular , for free particles the above expression becomes [ v+ ] v^i = [ v^j - u^j+^2 t_a(b^j - a^j ) ] . it is easy to see that free particles remain in uniform - velocity motions along straight lines under time translation , spatial translation , spatial rotation , and boost transformation . the non - relativistic energy for a particle can be obtained in standard way . to order @xmath147 , @xmath148 , and @xmath149 , ( x)=1-^2 t^2 + , and ( x)for a test particle . the energy and 3-momentum of a particle are e=^-1(x)m_0 c^2 m_0c^2 + 1 2 ( 1-^2 t^2 ) m_0 ^2 + m_0 ^ 2 t- 1 2 m_0 ^ 2 ^2 , and [ pi ] p^i = ^-1(x ) m_0 m_0v^i , respectively . the non - relativistic energy for a particle is then [ enr ] e_nr & & 1 2m_0 ^2 - ( t -m_0 ) ^2 = 1 2m_0 ( ^2 -^2 ^2 ) , where the second expression in eq.([enr ] ) can be obtained directly from the generalized einstein formula eq.([eml ] ) . therefore , the kinetic energy in the nh limit is still @xmath150 and the non - relativistic energy for a particle is equal to the sum of its kinetic energy and the ( negative ) energy contributed from boost . in particular , for a free particle , the non - relativistic energy reduces to [ enrfree ] e_nr 1 2 m_0 ^2 -1 2 m_0 ^ 2 ^ 2 , where @xmath151 is the initial position of the free particle in eq.([tp ] ) . as in newtonian mechanics , time in beltrami coordinate system is absolute in the nh limit . one only needs to consider the geodesic deviation under condition @xmath152 . hence , given a timelike geodesic @xmath153 in @xmath21 , we can construct a congruence of geodesics @xmath154 , @xmath155 . the deviation is given by ^i(t ) = also be obtained from the contraction of eq.([gde ] ) . it is invariant from one inertial - type frame to another inertial - type frame . similarly , a light signal moves globally along a null geodesic in @xmath3 . the null geodesic equations formally still have the forms of eqs.([geod+t ] ) and ( [ geod+i ] ) , but now = 0 , where @xmath156 is an affine parameter . therefore , a null geodesic in @xmath3 becomes a spacelike geodesic in @xmath2 , which can be integrated as a straight line [ nhnullg ] x^i = c^i + b^i from eq.([geod+i ] ) under the initial condition t|_=0=0 , x^i|_=0=b^i , . |_=0=c^i . eq.([nhnullg ] ) can be obtained from eq.([bdsnullg ] ) ) in the limit of @xmath157 in eq.([w2 ] ) . ( again , the situation for finite @xmath158 in eq.(2.25 ) does not appear in the nh limit . ) now , the geodesic deviation becomes = 0 . usually , the nh limit is defined by both @xmath12 and @xmath13 @xmath159 , but @xmath160 keeping fixed @xcite . it should be noted , however , that such a definition of limit is not well defined because this contraction depends on the realization of the @xmath4 group acting on the @xmath4 spacetime . even for the same kinds of coordinates , it also depends on whether the spatial coordinates are dimensional or dimensionless . the followings are several possible nh limits of de sitter spacetime . it has been shown that the @xmath0 spacetime , based on the definition of proper - time simultaneity , has the form in ( [ dsrw ] ) . under the nh limit , the geometry becomes d^2= dt^2 , and d l^2 = ^2(t)dl^2__t . in this case , time spans a 1-d euclidean metric space while the spatial space is conformal to the 3-d euclidean metric space . the conformal factor depends on time @xmath161 . as was noted previously , in @xmath21 spacetime we can define two different kinds of simultaneity . in the nh limit , however , the two kinds of definition of simultaneity coincide as was mentioned in the previous section because of eq.([tmetric ] ) . if one introduces dimensionless spatial beltrami coordinates such that x^i= 1 r x^i , then _ = 1 - x^ix^i / r^2=1-x^i x^i and the robertson - walker - like metric can be rewritten as ds^2 = c^2 dt^2-r^2 ^ 2(ct / r)dl^2 , where @xmath162 is the line - element on the unit 3-sphere , which is dimensionless . now , taking the nh limit , in which the dimensionless coordinates @xmath163 are kept finite , we get [ nrds ] d^2= dt^2- 1 ^2 ^2 ( t)dl ^2 . in this case , the concept of spacetime and de sitter group as well remain even in the nh limit ! namely , the ` contraction ' is trivial in the group , algebra and geometry aspects ! it is also the case for the beltrami metric ( [ bhl ] ) if the spatial coordinates chosen as of dimensionless . similar uniqueness problem also appears in other forms of de sitter spacetime . for example , for @xmath164 de sitter metric , [ nhk0 ] & & ds^2 = c^2 d|t^2 - e^2c|t / r(d|r^2 + |r^2 d^2 ) + & & d^2=d|t^2 , d l^2 = e^2|t(d |r^2 + |r^2 d^2 ) , & |r keeps finite . + d^2 = d|t^2 - 1 ^2 e^2|t(d^2 + ^2 d^2 ) , & = |r / r keeps finite . for the static de sitter metric , [ sds ] & & ds^2 = ( 1-r^2/r^2)c^2 dt^2 - -r^2 d^2 + & & d^2=dt^2 , dl^2 = dr^2+r^2 d^2 , & r keeps finite . + d ^2 = ( 1-^2)dt^2 - 1 ^2 ( + ^2 d^2 ) , & r = r/r keeps finite . if one requires @xmath165 , and @xmath166 keep finite or requires one or two of spatial coordinates @xmath167 keep finite , then one may get different contractions of the same geometry ( see , appendix [ appdx : limit ] ) . it should be noted that the starting point of ref.@xcite is the static de sitter metric . under the nh limit in the first manner in eq.([sds ] ) , @xmath112 is equal to the proper time and the spatial coordinates @xmath168 satisfy @xmath169 . when such a coordinate system is chosen , free particles in the nh limit of @xmath4 spacetime obeys [ gnhgeod ] = ^2 q^i . it can be proved that it is the contraction of geodesic equation in the static @xmath4 spacetime , too @xcite . from eq.([gnhgeod ] ) , one can not read out that free particles move along straight lines at constant velocities . the relation between coordinates used in @xcite and the beltrami coordinates is given by the second expression of eq.([tau - t ] ) or = 1 ^-1 t , and q^i= . it should be noted that in terms of the beltrami coordinates @xmath170 eq.([gnhgeod ] ) turns to be eq.([nuvi ] ) of the inertial - type motion and the non - relativistic energy of a particle , eq.(5 ) in ref . @xcite , turns to be eq.([enr ] ) . these examples show that the condition eq.([nu ] ) is not enough to determine the limit . is it possible that there exist one or more physical principles to determine the nh limit uniquely ? the answer is yes . in the following subsections , we propose two principles , galilei - hooke s relativity principle and postulate on newton - hooke universal time @xmath171 , and show that these two principles should fix the nh limit procedure . recall that in newtonian mechanics , the galilean relativity principle is respected . the principle states that the laws of mechanics are the same in all inertial frames . in particular , the first law of newtonian mechanics , which says that a body at rest or in uniform - velocity motion along a straight line remains its state as long as no force acts upon it , is the same in all inertial frames . therefore , an inertial observer can not determine by local mechanic experiments whether he is at rest or in a uniform - velocity motion along a straight line . in the previous discussion on the nh limit , the relativity principle is not considered even though the inertial - type motion has been introduced in both @xmath0 spacetime and nh space - time . we have seen that if in @xmath2 and the constant curvature spacetime resulted from the trivial contraction in previous subsection , the inertial - type motion , the motion at uniform - velocity along straight line in sec . [ motion ] , is regarded as a genuine inertial motion , if a beltrami coordinate system is regarded as a true inertial coordinate system , and if a set of particles static in a beltrami coordinate system is regarded as an inertial reference frame , then the inertial motion is independent of the choice of inertial frames . the statement may serve as the counterpart of the first law of newtonian mechanics in @xmath2 and can be expressed in the same way as the first newton s law of mechanics : in @xmath2 a body at rest or in uniform - velocity motion along a straight line remains its state as long as no force acts upon it . the counterpart of the second newton s law of mechanics may also be set up in @xmath2 as [ 2ndlaw ] = f^i . in @xmath2 , both sides of eq.([2ndlaw ] ) should transform in the same manner under the transformation eq.([gnh ] ) . now , we may introduce a relativity principle parallel to the galilean relativity principle in newtonian mechanics as follows : _ the laws of mechanics in the newton - hooke space - time are the same in all inertial frames . _ + we name it galilei - hooke s relativity principle in honor of galilei and hooke . ( from now on , the ` -type ' in phrases , such as , inertial - type motion , inertial - type observer , inertial - type frame , are removed . ) by use of galilei - hooke s relativity principle , the contraction like the third example in appendix [ appdx : limit ] can be excluded . furthermore , the limit that @xmath172 while @xmath13 keeps finite can also be excluded by the galilei - hooke s relativity principle . therefore , the requirement @xmath173 in the non - relativistic approximation of special - like relativity on @xmath21 @xcite can be relaxed because it is the result of the galilei - hooke s relativity principle . it should be noted that even when the galilei - hooke s relativity principle is respected , it is still not enough to fix the nh limit uniquely . in order to fix the contraction procedure , it is needed to explore what should happen at any given point as the origin of a coordinate system chosen . in the case of @xmath0 , the metric at origin is minkowskian and under the nh contraction , it splits to two parts . the similar situation also happens in the case of static @xmath4 metric . since both spacetime @xmath21 and space - time @xmath2 are homogeneous and isotropic , all points should follow this behavior . note that for the metric at the origin , there is no room for @xmath13 . thus , this may fix the contraction procedure . this may also be seen from both the light - cone and the homogeneous lorentz group at the origin . in fact , we should define the inertial motion , inertial coordinates as well as inertial observers and require two first principles for @xmath0 : the relativity principle and the postulate on universal constants , which require that there exist two universal constants of @xmath12 with dimension of velocity and @xmath13 with dimension of length . as was indicated in @xcite , the second postulate implies that the light cone at the origin is the same as that in a minkowski spacetime . the lack of the two principles seems to indicate in mathematics that the contraction procedure may not be fixed by only considering algebraic contraction . therefore , we propose the second physical principle , the postulate on the newton - hooke universal time : _ there exits a bounded newton - hooke universal time @xmath174 under the nh limit of @xmath11 , which measures the time of physical processes . _ thus , under the postulate , especially at origin of the ( beltrami ) coordinates , the metric splits to two parts under the limit of @xmath175 . in brief , only considering algebraic contraction is not enough to uniquely determine the nh limit . one has to supplement two physical principles . one is the galilei - hooke s relativity principle and the other is the postulate on newton - hooke universal time . in newtonian mechanics and special relativity , if two particles have the same velocity relative to one inertial frame , they have the same velocity relative to every inertial frame . as a consequense , there can not be a particle that is at rest at the same time in two different inertial frames . in addition , in newtonian mechanics and special relativity , any two inertial frames @xmath15 and @xmath14 have the property that all particles being at rest in @xmath14 have the same velocity relative to @xmath15 . in the following , we shall see these are not the facts in @xmath2 . similar to what in newtonian mechanics and special relativity , the equation of motion for a free particle is eq.([tp ] ) . especially , when @xmath176 , two coordinate systems for @xmath15 and @xmath14 related by the transformation ( [ gnh ] ) have the same relation as that in newtonian mechanics , except the coordinate time @xmath9 taking value in @xmath177 in @xmath2 . when @xmath178 , however , the velocities of free particles at rest in @xmath14 are different from one another in @xmath15 . this can be seen from the equation of motion of a static particle in @xmath14 in terms of @xmath15 , @xmath179\,t.\ ] ] obviously , when @xmath178 , there is a special fixed spatial point @xmath180 in @xmath14 , whose velocity relative to @xmath15 is also zero ! namely , its spatial coordinates in @xmath15 is also fixed at @xmath181 therefore , when @xmath178 , there is always a free particle @xmath182 at rest in both frames @xmath15 and @xmath14 . let us select another two distinct particles @xmath183 and @xmath184 such that @xmath183 is static in @xmath15 while @xmath184 is static in @xmath14 . from the point of view of an observer in @xmath15 , @xmath182 and @xmath183 have the same velocity while @xmath182 and @xmath184 have different velocities . from the point of view of an observer in @xmath14 , it is on the contrary : @xmath182 and @xmath184 have the same velocity while @xmath182 and @xmath183 have different velocities . so , in this case , the observation that two particles have the same velocity is a phenomenon depending on reference frames : two frames could share a common particle that is static in both of them . for such a particle it could say that another remote particle is static relative to it while , at the same time , it could deny this statement . the answer depends on which reference frame it thinks itself belongs to . customarily , we use an inertial observer to take the place of an inertial frame in which he is static . this is due to the properties of inertial frames in newtonian mechanics and special relativity . from the above discussion we find that this is misleading in @xmath2 , because the same inertial observer can be static in distinct inertial frames . in fact , as we can see more clearly in next section , the choice of the time origin of nh inertial coordinate systems has much more nontrivial meaning than the newtonian case . especially , the concept of static state is dependent on this choice . at last , we point out that , whenever @xmath178 , we can shift the spatial origins of @xmath15 and @xmath14 , respectively , so that the transformation from @xmath15 to @xmath14 have the standard form as @xmath185 in this subsection we try to give a clear geometric picture to the nh transformations . we can see the shadow of projective geometry . in @xmath2 a 2-d surface @xmath186 is called a timelike plane and a 2-d surface @xmath187 is called a spacelike plane , where @xmath188 and @xmath189 are parameters describing the surfaces and @xmath190 , @xmath191 s and @xmath102 are some constants . since eqs.([tlp ] ) and ( [ slp ] ) remain the same forms , respectively , under the nh transformations , the above concepts are well defined . it can be shown that straight lines on the planes in @xmath2 , which can be expressed by @xmath192 in terms of the parameters @xmath188 and @xmath189 , are also straight lines ( i.e. geodesics ) in @xmath2 . in eq.([lp ] ) , @xmath193 , @xmath194 and @xmath195 are some constants , and @xmath188 takes value in @xmath196 for timelike planes and @xmath197 for spacelike planes . two geodesics in a plane are called parallel lines if they have no common points . it is obviously that euclid s fifth axiom is valid on spacelike planes but invalid on timelike planes . applying the same ideas to discuss spacetimes of newtonian mechanics or special relativity , we can find that euclid s fifth axiom is valid for every plane . this is one of the main difference between @xmath2 and the spacetime in newtonian mechanics or the special relativity . the concept of straight line , plane and parallel are independent of the choice of beltrami coordinate systems . that is , if a curve is a straight line in one coordinate system , it is a straight line in all coordinate systems ; if two straight lines are parallel in one coordinate system , they are parallel in all coordinate systems . for an inertial frame @xmath15 , the world lines @xmath198 and @xmath199 are straight lines that are parallel to each other . from the point of view other inertial frame , @xmath14 , say , these world lines are still parallel straight lines . it does represent the invalidity of euclid s fifth axiom that there exists an observer who can sit statically in both @xmath15 and @xmath14 . to be clearer , we give some diagrams to the standard transformation ( [ eq : std ] ) . define @xmath200 it is meaningful for all nonzero @xmath92 and can be generalized to the case @xmath201 , that can be regarded as a special case in which @xmath202 is at infinity on the time axis . it should be noted that @xmath202 is not a possible coordinate time because @xmath203 for simplicity we only consider a plane @xmath204 containing the world line of the spatial origin of the systems , namely , @xmath205 and @xmath206 if the world lines of @xmath207 are shown as in fig.[fig : s ] , the world lines of @xmath208 will be shown , in the same coordinate system , as in fig.[fig : sprime ] . in this figure , these world lines are parallel to each other and are focused on the point @xmath209 on the time axis . conversely if a congruence of geodesics is focused on a point in @xmath210 ( not necessarily on the time axis ) , we can say that they are the world lines of rest particles in certain an inertial frame . due to the nature of fractional linear transformations , the cases in fig.[fig : s ] and fig.[fig : sprime ] are symmetric . if we draw the world lines in the @xmath211-@xmath112 coordinate system , then the world lines @xmath208 will look like those in fig.[fig : s ] , while the world lines @xmath207 will be focused on @xmath212 . . horizontal lines are @xmath139.,width=377,height=302 ] . horizontal lines are @xmath213,width=377,height=302 ] fig.[fig : t ] is a graphic illustration for the transformation of coordinate time . in the diagram , the horizontal @xmath9-axis and the vertical @xmath112-axis form a cartesian coordinate system . the point @xmath214 is located at @xmath215 where @xmath202 is given in eq.([eq : dualt ] ) . choose points @xmath161 and @xmath216 such that they are collinear with @xmath214 . let the coordinates of @xmath161 and @xmath216 be @xmath217 and @xmath218 , respectively . then , @xmath9 and @xmath112 can be linked by the transformation ( [ eq : std ] ) . from eq.([eq : std ] ) and the diagram , we can see @xmath219 are invariant time in @xmath2 . compared with special relativity in which there exists an invariant velocity @xmath12 , the nh mechanics is a kind of relativity due to the existence of finite extremum time . we have shown that in @xmath3 , the transformation group between beltrami coordinate systems contracts to @xmath101 group in the nh limit . the geometry of @xmath0 spacetime splits time and spatial space in the nh limit . the former may be regarded as a subspace of @xmath220 with metric , while the latter is conformal to a 3-d euclidean space . the conformal factor depends on time only . test particles and light signals without force acting upon them move along straight lines at uniform velocity in the nh limit , as expected . the behavior of uniform - velocity motion along straight lines is independent of the inertial frame which observer is in and the velocity addition law is obtained . this can be regarded as the key point of galilei - hooke s relativity principle , which together with the postulate on newton - hooke universal time determines the nh limit uniquely . the contraction of @xmath0 spacetime shows that the beltrami coordinate systems should be regarded as inertial coordinate system both in relativistic and non - relativistic levels since test particles and signals without force acting upon them all move along straight lines at uniform velocities . this implies that the mechanics in non - zero constant - curvature space - time may be set up in parallel to the one in flat space - time , starting from the relativity principle and inertial law . the dynamics both at classical and quantum level will be studied elsewhere . it is remarkable that unlike in newtonian mechanics and special relativity , a free particle in @xmath2 ( and @xmath21 ) can be static at the same time with respect to two different inertial frames which have relative velocity . this property is closely related to the violation of euclid s fifth axiom . in this sense , the nh limit of @xmath0 spacetime can be regarded as the version of the newtonian mechanics without euclid s fifth axiom . another remarkable property of @xmath2 is that there is a universal time . it is similar to the newton s universal time in newtonian mechanics in the sense that it is separated from space in the metric and that the simultaneity is absolute . however , in any inertial coordinate system , this universal time as the coordinate time should be in the range @xmath221 . compared with special relativity in which there exists an invariant velocity @xmath12 , the nh mechanics looks like a kind of relativity due to the existence of finite extremum universal time . it is related to the fact that the generators of spatial translation and boost along the same direction form an @xmath222 vector under the time translation . [ see , eqs.([eq : mult ] ) , ( [ eq : liealg ] ) , and ( [ eq : c1 ] ) . ] further studies on space - time , group structures , and kinematics and dynamics in the nh limit are needed . it should be noted that when the nh limit in @xmath3 is taken , the zeroth coordinate @xmath223 . in order eq.([5sphr ] ) to be meaningful , one of @xmath224 , at least , should tend to @xmath225 . for example , in @xmath226 , @xmath227 while @xmath228 keep finite so that @xmath229 from eq.([u4 ] ) . the intersection of @xmath28 and @xmath226 , @xmath230 , can only appears at @xmath231 ( and thus @xmath232 ) @xmath159 . when @xmath233 , namely , @xmath173 faster than @xmath234 , all results become the correspondences in flat space - time . in particular , the galilei - hooke s relativity principle reduces to the galilean one . non - relativistic energy of a test particle reduces to the one in the newtonian mechanics . the nh group reduces to the galilei group . the newton - hooke universal time becomes newton s universal time . in the present paper , we only discuss the contraction of @xmath0 spacetime which has a positive cosmological constant . all results are readily extended to the contraction of beltrami - anti - de sitter spacetime which has negative cosmological constant . as pointed out in sec . [ sect : bds ] and in @xcite , in the beltrami - de sitter spacetime @xmath3 the metric @xmath235 and its inverse @xmath236 are invariant under the fractional linear transformation ( [ g ] ) . consequently , the connection coefficients eq.([gamma ] ) and the components of the curvature tensor @xmath237 have the same property . since the coordinate transformation from one patch to another is a special case of the fractional linear transformation ( [ g ] ) , the forms of these quantities are the same in all the coordinate patches and the expressions of the geodesics have the same form in all the patches , too . the connection on @xmath3 is torsion free . the connection on @xmath2 is torsion free , too . since not a particular beltrami coordinate system is specified in obtaining eq.([cs ] ) from the contraction of coefficients of connection , eq.([cs ] ) should remain the same form under the nh transformations . in fact , we can define a new set of symbols @xmath238 among which only the following are nonzero : @xmath239 then it can be verified that @xmath128 and @xmath238 satisfy the standard transformation relation for coefficients of connection if @xmath240 is an nh transformation . so , the connection in @xmath2 is well defined . the nonzero components of the curvature tensor and ricci tensor calculated from eq.([cs ] ) are , respectively @xmath241 it follows that the mathematical forms of the curvature tensor and ricci tensor are also invariant under the nh transformation . and , on the other hand , they can also be obtained by contracting the curvature tensor and ricci tensor on @xmath3 , respectively . finally , we re - study the deviation equation of geodesics in @xmath2 in terms of 4-d language . given a timelike geodesic @xmath153 in @xmath2 , we can construct a family of geodesics , @xmath242 @xmath243 , labeled by the parameter @xmath244 . for convenience , it will be denoted by @xmath245 with @xmath246 , @xmath129 , @xmath130 and @xmath131 and @xmath247 . assume that the geodesic labeled by @xmath248 is @xmath153 . then , for each geodesic , eqs.([geod+t ] ) and ( [ geod+i ] ) are equivilent to @xmath249 where @xmath250 , etc . the deviation @xmath86 is a vector field along @xmath153 , defined by @xmath251 obviously , @xmath252 . if we take the partial derivative with respect to @xmath244 on both sides of eq.([eq : gfam ] ) then set @xmath248 , we can obtain the derivation equation @xmath253 where @xmath254 is the tangent vector of @xmath153 . using the expression of connection coefficients and the fact that @xmath252 , we can verify that the @xmath9-component of the covariant derivative of @xmath86 is always zero . it follows that , when @xmath246 , eq.([eq : der ] ) is an identity . when @xmath255 , it is equivalent to @xmath256 we have seen that the result of the limit ( [ nu ] ) is not unique . in addition to eq.([nu ] ) , we need to specify the limits of the coordinates . we should have also noticed that the topological structure of contraction may not be the same as that of @xmath257 . here , we try to give a geometric point of view of the contraction . first , for a given positive @xmath258 , @xmath257 is viewed as a hyperboloid of the 5-d minkowski spacetime @xmath259 , as shown in section [ sect : bds ] . when @xmath260 , or @xmath13 , runs over all the positive values , a family of beltrami - de sitter spacetimes are obtained . obviously , through each spacelike point in @xmath259 , there is one and only one @xmath0 spacetime in the family . roughly speaking , to specify a particular limit evolving @xmath261 , one must specify a region @xmath262 consisting of spacelike points of @xmath259 and a congruence of curves in @xmath263 satisfying the following conditions . + ( 1 ) through each spacelike point in @xmath263 there is one and only one curve in the congruence . + ( 2 ) at each spacelike point in @xmath263 , the corresponding curve in the congruence is not tangent to @xmath21 passing through the same point . consequently , there is a direction along each of these curves such that the parameter @xmath13 increases strictly . that is , @xmath13 can be viewed as the parameter for each curve in the congruence . + ( 3 ) on each curve , the parameter @xmath13 can take any sufficiently large positive value . + ( 4 ) the congruence is smooth in the following sense . the congruence of curves can be described using four parameters @xmath9 , @xmath96 ( @xmath264 , 2 , 3 ) , say , such that @xmath9 , @xmath96 and @xmath13 is a set of local coordinates on @xmath263 when @xmath13 is sufficiently large . + given such a region @xmath263 and a congruence @xmath265 of curves in @xmath263 , which will be denoted by @xmath266 , these curves can be viewed as the orbits of an associated limit process . thus , on a given curve @xmath153 in the congruence @xmath265 , points with different values of the parameter @xmath13 are viewed as the same " point in the limit process . each of these curves can be identified with the limit point of the process . and the parameters @xmath9 , @xmath96 can be regarded as the coordinates of the limit point . the first example is the nh limit obtained in section [ nhlimit ] . the region @xmath263 is the set of spacelike points @xmath267 in @xmath259 such that @xmath268 . obviously , @xmath263 is homeomorphic to @xmath269 . the congruence @xmath270 consists of curves @xmath271 with \{^0 , ^i,^4 } = ( 1 - ^2 t^2 + r^-2||^2)^-1/2\{t r , x^i , r } , where @xmath13 is the parameter on the curve , @xmath9 and @xmath41 are parameters describing curves in @xmath270 , and @xmath272 is a parameter to distinguish one congruence from another . equivalently , the limit process associated to the above congruence is to keep @xmath273 and @xmath274 finite as @xmath13 tends to infinity . at the same time , @xmath275 also tends to infinity . for the second example , let the region @xmath263 be the set of all spacelike points in @xmath259 and @xmath276 with \{^0 , ^i , ^4}=(1 - ^2 t^2 + ||^2)^-1/2\{t r , r^i , r } then the congruence consisting of curves @xmath277 corresponds to the limit process as shown in section [ unique ] . in this limit process , @xmath275 tends to infinity while @xmath161 and the dimensionless spatial coordinates @xmath278 are kept finite . obviously this also meets the requirement of nh limit . the resulted spacetime is , once again , a de sitter spacetime . see , section [ unique ] . now let us consider the last example . the region @xmath263 will be the set of spacelike points in @xmath259 such that @xmath279 . in the limit process , both @xmath13 and @xmath12 tend to infinity with @xmath280 kept fixed . the region @xmath281 where @xmath282 , for example , is the subset of @xmath263 on which @xmath283 is positive / negative . beltrami coordinates can be defined on each of these subsets . for example , on @xmath282 , the beltrami coordinates can be defined to be @xmath284 note that @xmath285 , @xmath286 and @xmath287 are dimensionless while @xmath288 has a dimension of length . we assume that all the coordinates are kept finite in the limit process . in other words , we are considering a congruence of curves , @xmath289 and @xmath288 could only describe part of the curves in the congruence . these parameters corresponds to the beltrami coordinates on @xmath282 . there exist some curves in the congruence that must described by parameters corresponding to the coordinates on @xmath290 , @xmath291 , respectively . ] @xmath292 on @xmath263 with \{^0,^1,^2,^3,^4}=^-1/2(t , x^1 , x^2 , z)\{t r , x^1 r , x^2 r , z , r } , where @xmath293 the parameters @xmath9 , @xmath289 , @xmath288 and @xmath13 can be regarded to be coordinates on @xmath294 . the 5-d minkowski metric is then @xmath295 where @xmath296 to obtain the limit , we first consider the induced metric on a hypersurface @xmath297 , on which @xmath298 . then we take the limit of @xmath299 as @xmath261 . obviously , its limit is @xmath300 where @xmath301 and @xmath302 take values from 0 to 2 , with @xmath303 and @xmath304 needless of speaking , we obtain a space - time having the structure of @xmath305 , where @xmath306 is a unit 3-d beltrami - de sitter spacetime , while @xmath307 is labeled by the coordinate @xmath288 . the authors would like to thank professors y. h. gao , q. k. lu , x. c. song and y. q. yu for valuable discussions . special thanks are given to prof . j. z. pan who made contribution to part of this work . this work is partly supported by nsfc under grants nos . 90103004 , 10175070 , 10375087 , 10347148 . h. bacry and j. m. lvy - leblond , possible kinematics , j. math . phys . 9 ( 1968 ) 1605 ; h. bacry and j. nuyts , classification of ten - dimensional kinematical groups with space isotropy . 27 ( 1986 ) 2455 ; j. r. derome and j. g. dubois , hooke s symmetries and nonrelativistic cosmological kinematics . il nuovo cimento , * 9 * ( 1972 ) 351 . guo , c .- g . huang , z. xu and b. zhou , on beltrami model of de sitter spacetime , _ mod . _ * a19 * , ( 2004 ) 1701 - 1709 ; arxiv : hep - th/0311156 ; on special relativity with cosmological constant , _ phys . a _ * 331 * , ( 2004 ) 1 - 7 ; arxiv : hep - th/0403171 ; three kinds of special relativity via inverse wick rotation ; _ chin . * 22 * , ( 2005 ) 2477 - 2480 ; hep - th/0508094 ; h .- y . guo , c .- g . huang , y. tian , z. xu and b. zhou , beltrami - de sitter spacetime and de sitter invariant special relativity , _ acta phys . _ * 54 * , ( 2005 ) 2494 - 2504 ( in chinese ) ; h .- y . guo , c .- g . huang , and b. zhou , temperature at horizon in de sitter spacetime , _ europhys . lett . _ * 72 * , ( 2005 ) 1045 - 1051 ; hep - th/0404010 ; c. zhe , s .- x . chen , and c .- g . huang , absence of gzk cutoff and tests of de sitter invariant special relativity , _ chin . * 22 * , ( 2005 ) , 791 . k. h. look ( qi - keng lu ) , why the minkowski metric must be used ? ( unpublished , 1970 ) ; k.h . look , c.l . tsou ( z.l . zou ) and h.y . kuo ( h.y . guo ) , the kinematic effects in the classical domains and the red - shift phenomena of extra - galactic objects , _ acta physica sinica _ * 23 * ( 1974 ) 225 . k.h . look , c.l . tsou ( z.l . zou ) and h.y . kuo ( h.y . guo ) , relativistic principle of space - time with constant curvature and its cosmological significance for einstein s centenary , _ nature _ ( shanghai , suppl . ) ( in chinese ) .

we simulate the performance of a new type of instrument , a superconducting multi - object spectrograph ( supermos ) , that uses microwave kinetic inductance detectors ( mkids ) . mkids , a new detector technology , feature good qe in the uvoir , can count individual photons with microsecond timing accuracy and , like x - ray calorimeters , determine their energy to several percent . the performance of giga- , a supermos designed for wide field imaging follow - up observations , is evaluated using simulated observations of the cosmos mock catalog with an array of 100,000 r=e/e=30 mkid pixels . we compare our results against a simultaneous simulation of lsst observations . in three years on a dedicated 4m - class telescope , giga- could observe billion galaxies , yielding a low resolution spectral energy distribution ( sed ) spanning 350 - 1350 nm for each ; 1000 times the number measured with any currently proposed lsst spectroscopic follow - up , at a fraction of the cost and time . giga- would provide redshifts for galaxies up to with magnitudes , with accuracy.03 for the whole sample , and.007 for a select subset . we also find catastrophic failure rates and biases that are consistently lower than for lsst . the added constraint on dark energy parameters for wl+cmb by giga- using the fomswg default model is equivalent to multiplying the lsst fisher matrix by a factor of=1.27( ) , 1.53( ) , or 1.98( ) . this is equivalent to multiplying both the lsst coverage area and the training sets by , and reducing all systematics by a factor of 1/ , advantages that are robust to even more extreme models of intrinsic alignment .

1307.5066

the accelerated expansion of the universe @xcite is commonly attributed to a negative pressure component dubbed dark energy , making up approximately 73% of the energy content of the universe @xcite . the nature of dark energy remains a mystery , though it can be probed through its effect on the growth of structure over cosmic time . as a result , the experiments aimed at understanding dark energy are quickly growing in number . galaxy surveys to map large scale structure and probe cosmology are becoming increasingly ambitious both in terms of the cosmological volumes they probe , as well as in the development of technological advances necessary for more precise and efficient measurement of galaxy spectral energy distributions ( seds ) . for example , the large synoptic survey telescope ( lsst ; * ? ? ? * ) plans to image @xmath1210 billion galaxies to @xmath4@xmath1326.5 with data in the @xmath14 photometric bands over @xmath1220,000deg@xmath15 of the southern sky . similar current and future wide field imaging experiments include the dark energy survey ( des ; * ? ? ? * ) , euclid @xcite , and kids @xcite . traditionally , sources selected by color and/or magnitude from initial imaging data in a handful of frequency bands were followed up with conventional dispersed spectrographs in order to obtain accurate redshifts . however , in the coming data rich era , this approach is not possible . even the largest planned fiber - fed multi - object spectrographs can not hope to follow up even 1% of the lsst catalog @xcite . one of the most important lsst science goals uses independent probes to measure the effect of dark energy : weak gravitational lensing ( wl ) , baryon acoustic oscillations ( baos ) and galaxy clusters ( * ? ? ? * and references therein ) . all of these techniques , however , rely on the precise determination of redshifts for as many galaxies and quasars as possible @xcite , most of which will be faint given the steeply rising number counts towards fainter magnitudes ( e.g. , * ? ? ? redshifts estimated from galaxy colors in a handful of broad bands have significant problems @xcite since photometric accuracy depends on spectral coverage , resolution and signal - to - noise ( s / n ) . the biases and the high catastrophic failure rates that result from redshift determination using standard photometry add significant errors to the dark energy measurements @xcite . this naturally leads away from broad band imaging towards massively multiplexed low resolution spectroscopy or spectrophotometry . we consider a new instrument and survey , giga-@xmath0 , that will take low resolution spectra and find the redshifts of two billion objects in the lsst field down to @xmath525@xmath16 magnitude in @xmath17 band . this survey , when combined with lsst imaging , will enable unique galaxy science . giga-@xmath0 is made possible by optical through near - ir microwave kinetic inductance detectors ( mkids ; * ? ? ? * ) , a low temperature detector ( ltd ) developed at ucsb that can detect the energy and arrival time of each incoming photon without the use of bandpass filters or dispersive optics @xcite . mkids , described in section [ sec : mkids ] , are nearly ideal , noiseless photon detectors , as they do not suffer from read noise or dark current , and have nearly perfect cosmic ray rejection . in giga-@xmath0 , described in section [ sec : expt ] , mkids will be used in a configuration similar to a conventional multi - object spectrograph , but without the use of a wavelength dispersive element . giga-@xmath0 could be on the sky by 2020 , and with 3 years on a 4-m telescope could improve on the lsst constraints for @xmath18 , the dark energy equation of state parameter , and @xmath9 , its evolution , and in conjunction with lsst map the distribution of dark matter ( e.g. , * ? ? ? * ; * ? ? ? the rest of this paper is organized as follows . section [ sec : sim_obs ] explains the development of mock catalogs from simulated observations for both lsst and giga-@xmath0 . section [ sec : photozs ] describes the redshift estimation , and compares results for the two experiments , as well as a summary of statistics for current or planned survey projects with similar science goals . we explore dark energy parameter constraints in section [ sec : de ] , and conclude in section [ sec : conc ] . large formats , high quantum efficiency , and low readout noise make semiconductor detectors the most popular type of detector used in the optical and near - ir wavelength regime . however , thermal noise from their high ( @xmath12100k ) operating temperatures and the semiconductor band gap place fundamental limits . reducing gap parameters by a factor of a thousand can be achieved with cryogenic superconducting detectors , operating at around 100mk . a superconducting detector can count single photons with no false counts while determining the energy ( to a few percent ) and arrival time ( to roughly 1@xmath19s ) of the incoming photon . since the photon energy is always much greater than the gap energy , much broader wavelength coverage is possible , enabling observations at infrared wavelengths that are vital to understanding the high redshift universe . a cryogenic detector technology with sensitivity and ease of multiplexing initially demonstrated at millimeter wavelengths @xcite are mkids @xcite . intrinsic frequency domain multiplexing allows thousands of pixels to be read out over a single microwave cable @xcite . they can count individual photons with no false counts and determine the energy and arrival time of every photon with good quantum efficiency @xcite . their physical pixel size and maximum count rate are well matched with large telescopes . these capabilities enable powerful new astrophysical instruments usable from the ground and space . the mkids described here are sensitive to 0.15@xmath19 m wavelength radiation ( with cutoffs imposed by the sky count rate and the properties of the materials being used ) but are optimized for near infrared ( nir ) and optical wavelengths ( 3501350 nm ) . the array camera for optical to near - ir spectrophotometry ( arcons ) is the first ever optical / nir mkid camera . it was commissioned in july , 2011 at the palomar 200 inch telescope and as of december 2012 , has now observed over a combined 24 nights on the lick and palomar telescopes @xcite . some of the science targets observed include interacting binaries ( am cvns , lmxbs , short period eclipsing sources ) , qsos ( for low resolution redshift measurements ) , supernovae ( type ia and type ii ) and the crab pulsar . arcons , representing the current state of optical mkids , houses a 2,024 detector array ( figure [ fig : mkids ] ) , making it the largest optical / uv camera based on low temperature detectors by an order of magnitude . the energy resolution of the devices , r(=e/@xmath2e ) , currently about 20 at 254 nm ( or about 12 at 423 nm ) , can reasonably be expected to continue to improve towards the theoretical limit of 150 at 254 nm over the next several years as designs and materials evolve . furthermore , the parallel technologies of infrared - blocking filters , broadband antireflection coatings , and detector quantum efficiency continue to develop , which will increase the performance of arcons and giga-@xmath0 . conventional multi - object spectrographs employ a mask inserted at the focal plane to pass light from targets through the slits ( or apertures ) , blocking background sky and other nearby source photons to reduce sky noise and contamination . a dispersive element such as a diffraction grating or prism then spreads the light as a function of wavelength on a detector . the supermos concept uses the same mask - based approach to reduce sky background and contamination from other sources , but uses the intrinsic energy resolving capability of each mkid detector to measure the spectrum . since each mkid pixel provides spectral information the focal plane is used much more efficiently , yielding a simple and compact system . a very simple implementation for giga-@xmath0 is shown in figure [ fig : optics ] , envisioned as an instrument for the cassegrain or naysmith focus of a dedicated 4 m class telescope . aside from the inherent energy resolution of mkids , giga-@xmath0 is enabled by the large pixel counts possible with mkids . one square degree field of view can be divided among 100,000 detectors , each fed by a macropixel covering 10@xmath2010 of the sky to be able to cover 20,000 square degrees in a reasonable amount of time ( see figure [ fig : pix ] ) . galaxy number counts in @xmath21 band to 24.5@xmath16 magnitude ( e.g. , * ? ? ? * ) ensure that @xmath2280% of the macropixels will contain a galaxy at each pointing . a mask cut using pre - existing lsst ( or earlier dark energy survey ) imaging would permit light from one celestial source per macropixel into a reimaging system that focused the light onto the corresponding large plate scale mkid located directly below . fov is divided into 100,000 10x10 macropixels . the background hubble udf image is for illustrative purposes and not to scale , with macropixels delineated in blue . existing catalogs will be used to select a target for each macropixel , and a corresponding hole drilled into a metal mask ( purple circle ) , with the diameter allowed to vary depending on the object size and shape . source light passing through the mask will land on the corresponding mkid with the same plate scale as the macropixel . some subset of macropixels for each field will be selected to monitor the sky background ( yellow circles ) . ] we note two potential drawbacks to this aperture masking technique : @xmath320,000 precut masks are required , and it limits the galaxy sampling to a relatively uniform spacing , making observations of galaxy clusters more difficult . however , a dedicated laser mask - milling facility can address the first issue ( such masks have been made for e.g. , * ? ? ? * ; * ? ? ? * ) , and careful survey design incorporating fields with multiple visits can ameliorate the second . in @xmath3 15 minutes adequate s / n can be achieved to determine the redshift of galaxies with magnitudes @xmath525 ( section [ sec : photozs ] ) . assuming 80% of the macropixels contain a source , giga-@xmath0 would acquire @xmath3320,000 spectra per hour . rapid mask changes can be performed with pre - loaded cartridges , and the photon counting nature of the mkids allows the mask to be aligned using real time feedback from the science array . in one night with 8 hours of observing , this equates to @xmath3 2.5 million spectra per night . at this rate , the entire lsst field could be covered in about three years . lsst will provide galaxy shapes and radii . assuming that poisson statistics from the sky dominates errors , the mask hole radius that maximizes s / n is proportional to the encircled energy squared over the area of the hole , which depends upon the light profile of the target . for an object with a gaussian profile , for example , the s / n is maximized by capturing @xmath372% of a galaxy s light . for more realistic profiles , it is slightly less than this . the cosmos mock catalog ( cmc ; section [ sec : cmc ] ) indicates that the majority of galaxies out to high redshift have half light radii equal to less than half an arcsecond . this scale translates to hole diameters of @xmath340@xmath19 m in the design presented here , well within the limits of current laser drilling technology . seeing conditions at the site may broaden galaxy profiles , and therefore must also be taken into account when determining hole size . mkids have a maximum count rate which can be tuned to some degree during fabrication to fall within the range @xmath31,00010,000 counts / pixel / s . the hole sizes for very bright or large galaxies would likely require accounting for the maximum allowable photon count rate . when working into the near - ir , sky subtraction becomes a dominant concern . for giga-@xmath0 , concurrent with galaxy target selection from the lsst imaging will be the selection of known dark areas of the sky . approximately 1020% of the macropixels in a typical observation , greater than 10,000 mkids , will collect approximately 1,000 photons per second from the sky ( based on the gemini south model ) , with each photon individually time tagged to within a microsecond . this sky background data can then be used to build up a map consisting of spectra as a function of time at every point on the array , facilitating the subtraction of the sky background to the poisson limit over the entire spectral range of the detectors . in @xmath23 band , for example , with a sky brightness of roughly 16.6 magnitudes per square arcsecond , a 24.5@xmath16 magnitude galaxy with about half of its light falling in 1 square arcsecond would have a contrast ratio of @xmath35e-4 . figure [ fig : sedswptsjy ] shows this measured at a few sigma in a 15 minute exposure . different mask hole positions within a macropixel illuminate the mkid very slightly differently , so both the throughput and quantum efficiency of each mkid will need to be calibrated as a function of mask hole position . stray light , for example , may be an important factor , involving cross - talk from one mask hole to another s corresponding detector . these differences can be calibrated through laboratory and on - sky testing . no fringing effects have been observed with arcons . the cosmos mock catalog ( cmc ; * ? ? ? * ) makes use of the latest survey data gathered through deep extragalactic surveys . it was specifically designed to be used to forecast the yields of future dark energy surveys , by converting the observed properties of each cosmos galaxy into simulated properties that can then be viewed using any instrument configuration . thus we combine the synthetic galaxy spectra from the cmc with our instrument throughput model to generate catalogs of simulated giga-@xmath0 ( and lsst ) observations . the cmc is based on the cosmic evolution survey ( cosmos ) observations @xcite which cover approximately 2 square degrees with 30-band photometry from multiple instruments spanning x - ray through radio frequencies . the subset used as input for the cmc is the photometric redshift catalog , covering a central 1.24 square degree patch fully covered by hst / acs imaging and not masked , providing a sample of 538,000 objects down to @xmath24@xmath1326.5 @xcite . for each galaxy , agn , or star , a best fit template is assigned based on the 30-band photometry . the same template is used for the redshift fit , and comes from a composite library of elliptical , spiral and starburst galaxy templates and stellar templates used by the @xmath25 photo-@xmath0 code . the best fit templates are redshifted and scaled to be in the observer s frame , assuming a perfect instrument ( perfect efficiency , delta function psf , etc . ) . two possible downsides to using these observations are the potential bias due to faint agn contribution , and the possible bias in the redshift distribution at @xmath0@xmath221.25 , where the photo-@xmath0 s from the observed catalog become degraded . to ensure a representative population , the cmc was compared with catalogs from the hst ultra deep field , goods and vvds - deep surveys for galaxy count , color , redshift and emission - line distributions , and found to be consistent . the cmc was updated in 2011 ( version july 15 , 2011 ) to improve the estimation of emission line fluxes , incorporated in the simulated galaxy seds , which we have used for the simulations presented here . table [ tab : cmc_models ] shows the breakdown by galaxy type of the cmc . the newest version of the catalog contains 646,706 objects , to a limiting simulated subaru @xmath17 band ab magnitude of 26.5 over 1.24 square degrees of sky . this catalog has already had most stars and agn removed . for our simulations , we use cmc objects with @xmath17@xmath1325 ( a complete sample ) , and remove the 916 agn and 1373 point - like objects with no radius solution ( likely stars ; there are 2 objects which overlap the agn / point - like object designations ) , which results in 219,759 galaxies for simulating mock observations , with redshifts up to @xmath36 . lccc type & n@xmath26 & @xmath2725 & non - agn + ell - s0 & 14927 & 11060 & 10985 + sa - sc & 38246 & 7651 & 7471 + sd - sdm & 35704 & 10290 & 10045 + starburst & 557829 & 193045 & 191258 + total & 646706 & 222046 & 219759 + to simulate the performance of giga-@xmath0 , realistic models for filters , optical throughput , device quantum efficiency ( qe ) , telescope reflectivity and sky background were generated . a locale with conditions similar to the cerro pachn , chile , with access to the southern sky is assumed . we take optical throughput to be 0.7 , accounting for @xmath34% loss at each of five lenses and @xmath310% loss at an ir - blocking filter . since the qe for the mkid detectors in arcons varied between roughly 73% at 200 nm and 22% at 3@xmath19 m @xcite , and there is significant room for improvement ( section [ sec : mkids ] ) , we assume a constant mkid qe of 0.75 for giga-@xmath0 . a model for bare aluminum reflectivity squared to account for two reflections at the primary and secondary mirrors is used to account for losses at reflective surfaces . to model the atmospheric transmission at cerro pachn , chile , we combine the extinction curve taken from the gemini website in the optical , merged with transmission in the nir . to model the sky background for simulating measurement errors and estimates of s / n , we merge the optical and near infrared sky backgrounds given for gemini south@xmath28 assuming an airmass of 1.5 and water vapour column of 4.3 mm . this is likely an overestimate of the far - red continuum brightness ( e.g. , * ? ? ? * ) , but a subdominant effect at r@xmath29=30 , as oh lines will be the primary contributor and thus we deem this a conservative estimate . the lunar phase assumed is within 7 nights from new moon ( `` dark '' , but not `` darkest '' ) , where the @xmath30 magnitude of the sky is @xmath320.7mag / arcsec@xmath15 . in practice , grey time can likely be used without significant degradation . mkids do not require the use of filters for information about the energy of the incoming photons . however , in order to compare our results with the lsst simulated photometry as transparently as possible and to be able to use existing redshift estimation codes without having to develop our own at this time , we can simulate effective photometric bands as independent spectral resolution elements . collected photons are separated into energy bins , thus an effective filter would take the form of the error distribution of the energy determination convolved with a tophat , which represents the quantization of binning . since photon energies are determined by a fit to the phase shift of the mkid with time , we can choose bins small compared to the true resolution so as not to impose any additional degradation . for the analysis presented here , each `` filter '' is a gaussian , centered in wavelength one full width half maximum ( fwhm ) away from its neighbors , where @xmath31 r@xmath32 is equal to 30 at the fiducial @xmath33 which we take here to be 423 nm , and decreases linearly with increasing wavelengths . figure [ fig : through ] ( bottom panel ) illustrates the effective filter set in the context of total throughput , using a normalization that ensures no double - counting of photons . cutoffs are imposed at 350 and 1350 nm , where the sky brightness and atmospheric transmission present practical limits . it is important to note that although we distinguish frequency `` filters '' here , all wavelengths are observed simultaneously by giga-@xmath0 , resulting in extremely efficient use of exposure time . as well , each pseudo - filter sees the same observing conditions with time , simplifying analysis considerably , a second major advantage over usual multi - filter photometry . in practice , for an mkid array , a more optimal solution would be to take the photon events and construct a maximum likelihood algorithm to reconstruct the spectrum . combining these throughput models as a function of wavelength for the various loss mechanisms gives the total expected system throughput depicted in figure [ fig : through ] ( bottom panel ) . for the final simulated observations , noise was added to the observed fluxes according to the properties of our system . ( bottom ) experiments , accounting for optical element ( filter and lens ) throughputs , mirror reflectivity , detector quantum efficiency and atmospheric transmission . the lsst @xmath14 filter implementation ( see section [ sec : lsst ] ) is based on the @xcite publication . the effective giga-@xmath0 filter set is for devices with energy resolution r=30 at 423 nm . cutoffs are imposed at 350 and 1350 nm due to the effects of sky brightness and atmospheric transmission constraints . ] we also simulated observations of the cmc for an lsst 3-year stack . the lsst design , as outlined in the lsst science book @xcite , incorporates 3 lenses , each with a projected @xmath33% loss , giving a throughput of @xmath391% . they use a 3 mirror telescope , which we model using the same aluminum reflectivity model but cubed . we note that this is pessimistic compared with the proposed lsst design , that uses a multilayer mirror coating to improve on the far - red performance of the aluminum , without the @xmath34 band absorption of silver . @xcite predict that the devices used in the lsst experiment will be very similar to the red wavelength - enhanced charge coupled devices ( ccds ) developed at the massachusetts institute of technology ( mit ) lincoln laboratory , with a qe shown in figure [ fig : lsst_qe ] . table [ tab : lsst_filts ] lists the lsst filter band centers and bandwidth , as well as the exposure time for each filter . in the lsst case , only one filter may be used at a time . cccc filter & central @xmath35 & bandwidth & exposure time + & [ nm ] & [ nm ] & [ s ] + @xmath34 & 360.0 & 80.0 & 700 + @xmath36 & 476.0 & 152.0 & 1000 + @xmath37 & 621.0 & 69.0 & 2300 + @xmath17 & 754.5 & 63.5 & 2300 + @xmath0 & 870.0 & 52.0 & 2300 + @xmath38 & 1015.0 & 65.0 & 2000 + qe . ] figure [ fig : ptsrc ] illustrates our simulated 5@xmath39 and 10@xmath39 magnitude limits for the lsst 3 year stack and giga-@xmath0 experiment , accounting for optical element ( filter and lens ) throughputs , mirror reflectivity , detector quantum efficiency and atmospheric transmission . note that each lsst filter encompasses 25 giga-@xmath0 pseudo - filters . if the giga-@xmath0 filters corresponding to each lsst filter were combined , the resulting magnitude limits would be more equivalent . and 10@xmath39 magnitude limits for a 3 year lsst stack and giga-@xmath0 experiment . ] we show the simulated photometric measurements by lsst and giga-@xmath0 for four example cmc mock galaxy spectra in figure [ fig : sedswptsjy ] . orange points denote the lsst mock observations with errorbars , and green points those predicted for the giga-@xmath0 experiment . errors are simply derived from poisson statistics , and optimal sky background subtraction has been assumed . though the s / n is typically lower per filter for giga-@xmath0 , the wavelength coverage is greater , and the exposure time is much smaller for giga-@xmath0 . indeed , as will be seen in section [ sec : photoz_res ] , photometric depth does not equate to photometric redshift accuracy . despite high s / n , the lsst filter set ( r@xmath35 ) inevitably leads to more color - redshift degeneracies , making unambiguous redshift determination impossible for the majority of galaxies ( e.g. , * ? ? ? very accurate photo-@xmath0 s are required to optimally exploit the expected data sets of dark energy surveys such as des , lsst and euclid ( e.g. , * ? ? ? however , photo-@xmath0 accuracy is greatly affected by experimental observing strategy , filter set and photometric sensitivity , which will impact the determination of redshifts from broadband seds . the best photometric results arise when strong continuum breaks in a galaxy spectrum fall between two instrument filters , and are therefore well constrained . one such typical feature of early - type galaxies is the 4000break which arises from the onset of stellar photospheric opacity due to the absorption of mainly ionized metals ( e.g. caii ) in the atmospheres of late - type stars . the balmer break at 3646 marks the termination of the hydrogen balmer series , and is indicative of younger stellar populations and more recent star - formation . the lyman break is another pronounced continuum discontinuity at 912 , observed in star - forming galaxy spectra . it is produced both in the stellar atmospheres of massive stars as a result of the hydrogen ionization edge and by photoelectric absorption by interstellar and intergalactic hi gas . lastly , sources at high redshifts have spectra that exhibit a suppressed continuum blueward of 1216 ( ly-@xmath7 ) due to additional opacity from line blanketing by intervening gas clouds along the line of sight . by @xmath0@xmath31.25 , the 4000 break has shifted to @xmath30.9@xmath19 m , out of the range of typical optical filters such as those used by lsst . not until @xmath0@xmath223 , does the redshifted lyman break re - enter the visible , so nir or uv data becomes imperative to get reliable redshifts in this `` redshift desert '' . the wavelength coverage of giga-@xmath0 is continuous and dense from @xmath3350nm1.35@xmath19 m ( figure [ fig : through ] ) , narrowing the redshift desert to 2.25@xmath5@xmath0@xmath53 . we therefore expect fewer photo-@xmath0 degeneracies and a lower catastrophic failure rate for giga-@xmath0 than for lsst . @xcite compared 17 photo-@xmath0 estimation methods used currently in the literature through blind tests on both simulated data and real data from the great observatories origins deep survey ( goods ; using each code , accuracies were determined for global photo-@xmath0 bias , scatter , and outlier rates . differences between codes stemmed mainly from whether they were empirical or template - fitting , the training set in the former case and the template set in the latter , the use of priors , handling of the lyman-@xmath7 forest and the benefit of adding mid - ir photometry . for a detailed discussion of photo-@xmath0 methods and their common elements see @xcite . the three photo-@xmath0 codes which ranked highest were @xmath25 @xcite , bayesian photo-@xmath0 s @xcite and eazy @xcite . the best results were achieved by using an empirical code ( smaller biases ) or optimizing templates , and correcting for systematic offsets . one reason to favor a template - based photo-@xmath0 code without any required training for this study is that the galaxies probed by upcoming surveys will span a cosmological volume and parameter space much greater than what can be well represented currently through spectroscopy . even very small mismatches between the mean photometric target and the training set can induce photo-@xmath0 biases large enough to corrupt derived cosmological parameters significantly @xcite . hence we proceed with the template - based photo-@xmath0 code eazy , and interpret the results presented here as what could be achieved prior to comprehensive spectroscopic surveys . its ease of use and demonstrated improvement in photo-@xmath0 s with the inclusion of ir data are also factors in why eazy is commonly used ( e.g. , * ? ? ? * ) , facilitating our aim of a side - by - side comparison between lsst and giga-@xmath0 . secondly , emission lines can change the colors of objects significantly , and are present in real observations . the treatment of emission lines can improve the photo-@xmath0 accuracy by a factor of @xmath32.5 @xcite , as they can be critical for minimizing systematic errors ( such as aliasing in the redshifts ) . both the cmc ( section [ sec : cmc ] ) and templates provided with eazy include emission lines . lastly , we note that the cmc synthetic spectra from which our mock observations are derived were generated using the @xmath25 code and therefore it is more instructive to use a different code for our analysis . eazy was developed to be specifically optimized for samples of galaxies with a limited amount of or biased ( e.g. , band - selected ) spectroscopic information available @xcite . combining the functionality of several pre - existing redshift estimation codes , eazy allows the user to fit linear combinations of templates and the choice of using a prior ( flux- and redshift - based ) through a parameter file . an error function can be used to downweight spectra to account for wavelength - dependent template mismatch such as in the uv where dust extinction is strongest and most variable , and in the nir where thermal dust emission and stochastic pah line features begin to appear . furthermore , the user may apply @xcite igm absorption to templates . by default , a probability - weighted integral is taken over the full redshift grid in order to assign an object redshift , marginalizing over the posterior redshift probability distribution ( in lieu of e.g. assigning the single most likely redshift by @xmath40 minimization , although the user has control over which ) . though this does not permit simple spectral classification , the increased photo-@xmath0 precision and ability to reproduce complex star - formation histories by fitting non - negative linear combinations of the templates may allow for better physical separation of photometric samples . one particular feature is the applicability to a wider range of redshifts and intrinsic colors than would be possible with an empirical photo-@xmath0 code , as no representative training set exists . the default eazy template set was generated using the @xcite non - negative matrix factorization method to reduce the stellar population synthesis code pgase model library @xcite to five `` principle component '' spectral templates of the calibration catalog . the templates are calibrated with a catalog @xcite derived from the @xcite semi - analytic models based on the @xcite millennium simulation . theoretically , this simulated 1 deg@xmath15 light cone contains a more realistic distribution of galaxies over 0@xmath13@xmath0@xmath54 than the more local spectroscopy most codes are trained on . one additional dusty starburst template was added , however , to account for extremely dusty galaxies which appear to be lacking representation in the semi - analytic models . a newer version of eazy ( v1.1 ; private communication with g. brammer ) implements emission lines following the prescription of @xcite ( after * ? ? ? * ) . these template spectra are shown in the top panel of figure [ fig : templates ] . a second set of templates provided with eazy that were used in this analysis are a grid of single pgase models ( which we will refer to as `` pegase13 '' ) that provide a self - consistent treatment of emission lines . they were designed to match the set described by @xcite - constant star formation rate models with additional dust reddening following the @xcite law . one tenth of the 260 pegase13 spectral templates are plotted in the bottom panel of figure [ fig : templates ] . the redshift grid for template fitting was done in steps of 0.005 on a log(1+@xmath0 ) scale . this imposes a limit on redshift resolution that becomes important on scales of @xmath41/(1+@xmath0)@xmath130.01 , as features can be recovered if their wavelength is greater than two grid steps . as our total sample redshift accuracy is larger than this , our sampling step is not the dominant contributor to systematic errors . in order to make the fairest comparison possible , we did not employ techniques which might aid in decreasing scatter in photo-@xmath0 estimates in a biased way . for example , we perform simple cuts , but we do not use priors in this analysis as they may improve results in a way that is preferential for one experiment . to minimize obscuration , we present only basic first order results that could be improved upon in the future when the goal is to get the most out of data . the eazy estimated redshifts for both giga-@xmath0 and lsst were compared with the input redshifts from the cmc catalog , and the statistics used to quantitatively assess their quality are given in table [ tab : zstats ] . the difference between input catalog and eazy estimated redshift is @xmath42=@xmath43-@xmath44 . the distribution for @xmath42 is typically non - gaussian , with extended tails and secondary peaks due to catastrophic outliers , which we arbitrarily define here as objects for which @xmath45@xmath460.15 . we quantify bias as @xmath47 , where the factor ( 1+@xmath48 ) accounts for the scaling of redshift errors with the stretching of rest - frame spectral features . one common method for estimating the redshift accuracy from @xmath49 uses the normalized median absolute deviation ( nmad : * ? ? ? * after @xcite ) , defined here as @xmath50 less sensitive to outliers , by this definition @xmath51 is equal to the standard deviation for a gaussian distribution , directly comparable to other papers which quote @xmath52/(1+@xmath0 ) . figure [ fig : gigazvslsst ] shows the distribution of @xmath42/(1+@xmath48 ) for the pegase13 templates , as a function of the known input catalog redshift and as a function of the object s observed @xmath17 band magnitude . the top line of plots are for our lsst simulations , and the bottom line for giga-@xmath0 . the horizontal lines at @xmath530.15 demarcate catastrophic failures from the central swath . we find that these statistics are most sensitive to the following : * n@xmath54 * wavelength coverage * redshift * s / n * width of the redshift probability distribution , p(@xmath0 ) * the number of spectral types being fit for * template set lccccccccc[h ! ] eazy & 0.061 & 25.4 & -0.011 & 100 & & 0.038 & 22.7 & 0.001 & 100 + pegase13 & 0.041 & 18.4 & -0.014 & 100 & & 0.030 & 18.7 & -0.008 & 100 + @xmath1324.5 & 0.037 & 17.4 & -0.017 & 71.4 & & 0.023 & 15.8 & -0.008 & 64.4 + @xmath1324 & 0.035 & 17.2 & -0.016 & 49.8 & & 0.018 & 14.0 & -0.008 & 43.9 + @xmath1322.5 & 0.032 & 12.5 & -0.012 & 14.1 & & 0.012 & 10.1 & -0.007 & 12.5 + 0.5@xmath13@xmath0 & 0.033 & 10.3 & -0.008 & 76.7 & & 0.026 & 10.3 & -0.002 & 76.7 + 0.5@xmath13@xmath0@xmath132.25 , 3@xmath13@xmath0@xmath136 & 0.031 & 8.7 & -0.008 & 68.7 & & 0.025 & 8.7 & -0.004 & 68.7 + @xmath55 @xmath132.5 & 0.035 & 15.4 & -0.015 & 77.7 & & 0.023 & 14.1 & -0.007 & 72.2 + mag@xmath1322.5 , @xmath56@xmath460.17 & 0.029 & 5.4 & -0.008 & 13.0 & & 0.010 & 0.3 & -0.006 & 11.2 + pegase13 ellipticals subset & 0.028 & 2.0 & 0.003 & 2.2 & & 0.007 & 2.5 & -0.001 & 2.7 + lrrccrc[h ! ] combo 17 & @xmath5710,000 & @xmath570.25 & @xmath5824 & 17 & 0.06 & @xmath55% + cosmos & @xmath57100,000 & 2 & @xmath5924 & 30 & 0.06 & @xmath5720% + & @xmath5730,000 & 2 & @xmath6022.5 & 30 & 0.007 & @xmath131% + cfhtls - deep & 244,701 & 4 & @xmath6124 & 5 & 0.028 & 3.5% + cfhtls - wide & 592,891 & 35 & @xmath6122.5 & 5 & 0.036 & 2.8% + primus & 120,000 & 9.1 & @xmath6223.5 & r@xmath29@xmath5790 & @xmath63 & @xmath572% + wigglez & 238,000 & 1,000 & 20@xmath13@xmath37@xmath1322.5 & r@xmath29=845 & @xmath50.001 & @xmath530% + alhambra & 500,000 & 4 & @xmath6425 & 23 & 0.03 & + boss & 1,500,000 & 10,000 & @xmath65@xmath6619.9 & r@xmath29@xmath571600 & @xmath50.005 & @xmath572% + des & 300,000,000 & 5,000 & @xmath6724 & 5 & 0.1 & + euclid & 2,000,000,000 & 15,000 & y , j , k@xmath524 & 3@xmath68 & @xmath50.05 & @xmath510% + & 50,000,000 & 15,000 & h@xmath69@xmath703e-16erg / s / cm@xmath15 & r@xmath72@xmath57250 & @xmath50.001 & @xmath1320% + lsst & 3,000,000,000 & 20,000 & @xmath7326.5 & 6 & @xmath50.05 & @xmath510% + giga-@xmath0 & 2,000,000,000 & 20,000 & @xmath7325.0 & r@xmath29=30 & 0.03 & @xmath5719% + & 224,000,000 & 20,000 & @xmath7322.5 & r@xmath29=30 & 0.01 & 0.3% + therefore we have included in table [ tab : zstats ] the effect of making cuts on the parameters not dictated by experimental design . in particular , we note that as scatter is correlated with how many spectral types one must fit for , increasing the number of objects in the catalog does not reduce the systematic error scatter or incidence of catastrophic failures . however , better sampling statistics will improve the mean uncertainty in each redshift bin . we tested this by analyzing only a randomly chosen fraction of our catalog , where the fraction was one of [ 3/4 , 2/3 , 1/2 , 1/3 , 1/4 ] . the statistics presented here are robust to this type of selection to less than 0.1% . as expected , the effect of increasing the mkid detector resolution to r@xmath29=60 or increasing the exposure time per target from 15 to 30 minutes decreased both the scatter ( by @xmath3@xmath74 ) and the catastrophic failure rate . eazy allows for the construction of a quality factor , q@xmath75 , with each computed redshift that depends on the @xmath40 value as well as the 99% confidence interval and the integrated probability , as a metric for the reliability of the redshift estimates that does not preferentially select out high redshift sources . however , we found that q@xmath75 was not a good predictor of redshift accuracy . many objects had multiply peaked redshift probability distributions , and/or were highly non - gaussian . we have made one broad cut on width of the p(@xmath0 ) , but more complicated functions of the probability distribution characteristics such as peak height might be useful to pursue in future studies . the estimated redshift results which gave the least scatter were obtained using the pegase13 templates , fit one at a time because the solution time quickly becomes prohibitively large as the number of templates increases . however , we found that the overall bias was reduced when using the eazy templates . the template - fitting redshift estimation method is susceptible to the fact that template colors and redshift are often degenerate . empirical redshift estimation codes may produce smaller biases ( by a factor of @xmath32 ) , since the model will match the data better by construction , suggesting systematic inaccuracies in most template sets . a sufficient training set , however , could be used to recalibrate templates , thereby reducing the inaccuracy ( e.g. , * ? ? ? spectroscopic calibration samples themselves , however , may lack spectra of some subset of rare galaxies that otherwise may not be easily identified and removed ( see e.g. @xcite , for a discussion ) . contamination occurs predominantly in two `` islands '' . one is in the redshift desert , at 2.25@xmath5@xmath0@xmath53 for the giga-@xmath0 filters . the second is at @xmath48@xmath50.7 . these are most likely caused by attributing the high redshift lyman break to the low redshift 4000 break , and vice versa . of course , seds not well represented by the template set will have issues , as will very blue galaxies with featureless seds . this simulation does not incorporate a magnitude prior , although doing so may reduce the size of these islands . our findings for lsst of scatters of @xmath334% in @xmath42/(1+@xmath0 ) are consistent with their studies @xcite , although we find higher outlier rates , ranging from @xmath3520% except for the most extreme cuts , more in alignment with the findings of @xcite . the redshift estimates for giga-@xmath0 are superior , in terms of dispersion , bias , and catastrophic failure rate by up to a factor of over 3 over all parameter cuts , highlighting how giga-@xmath0 s @xmath3 log spectral resolution improves on the commonly used optical filters . the inclusion of nir data could improve both the lsst and giga-@xmath0 results since it is then possible to simultaneously constrain both the lyman and 4000 breaks . though it is difficult to make strict comparisons with other experiments , we list some of the salient statistics for various similar multi - band photometric or multi - object spectroscopic experiments , both past and planned , in table [ tab : compare ] . though not comprehensive , it gives an idea of the state of the field . because all giga-@xmath0 `` filters '' observe simultaneously , giga-@xmath0 does not suffer from the trade - off between photometric depth ( or number density ) and higher spectral resolution that the experiments using more , narrower filters or spectroscopy will face . finally , we mention the cleaning of outliers to yield lower outlier rates . depending on the science application such a filtering can be effective , for example , for dark energy studies with weak lensing that do not rely on complete galaxy samples . however , some science applications do rely on redshifts for all objects not allowing for filtering . for those kind of applications the results reported in this section are particularly informative . baryon acoustic oscillations ( bao ) are ripples that appear in the spatial pattern of galaxies , exhibiting coherence on the particular co - moving scale of @xmath3150mpc separation , determined from cosmic microwave background ( cmb ) observations @xcite . they appear in the galaxy distribution as a `` bump '' @xcite , or `` wiggles '' in the matter fluctuation power spectrum @xcite , analogous to a low redshift cmb power spectrum . since the physical size is known , bao serve as a ruler with which to measure the geometry of the universe in both the radial and angular directions . furthermore , bao in these orthogonal directions are subject to different systematics , which can be used as a cross check . indeed , bao may have the lowest level of systematic uncertainty of all current dark energy probes @xcite . reaping the potential of bao , however , requires redshift estimates more accurate than what was found with our main galaxy population in section [ sec : photoz_res ] . this is similarly true for wl analyses , in order to separate galaxies into redshift slices and correct for intrinsic galaxy alignment contamination . traditionally , spectroscopy has been used to ascertain galaxy redshifts from which distances are derived , probing cosmology through an averaged three - dimensional bao measurement . more recently , multi - object spectroscopic experiments such as the wigglez survey @xcite , hetdex @xcite , boss @xcite , and big boss @xcite have been conducted or are planned , with the aim of improving on the current bao measurements . however , at present , uncertainties in the dark energy constraints set by bao are limited by data volume . to fully realize the potential of this method , larger numbers of objects ( yet with precise redshift estimation ) are needed . luckily , for wl or bao dark energy analyses , it is not necessary to have complete galaxy samples . therefore low resolution galaxy surveys can optimally select a subset of galaxies so as to meet the redshift accuracy criterion for the sample with minimal impact on the derived cosmological constraints ( e.g. , * ? ? ? luminous red galaxies ( lrgs ) are a homogenous subset of the main galaxy population , predominantly massive early - type galaxies . they are strongly biased , mapping the observable galaxy distribution to the underlying mass density distrubition . intrinsically luminous , they are excellent tracers of large scale structure , and an observational sample can be selected using photometric colors relatively easily to high redshift due to the strong 4000 break in their seds ( e.g. , * ? ? ? * ; * ? ? ? @xcite show that @xmath6@xmath30.003 is sufficient precision to measure the bao in the radial direction , which has more stringent requirements than the angular direction . doing better than this merely oversamples the bao bump , and at this level , redshift space distortions and nonlinear effects can produce comparable errors . this limit corresponds to @xmath315mpc / h , the intrinsic co - moving width of the bump along the line of sight at @xmath0=0.5 , caused primarily by silk damping @xcite . this is the mean redshift for the pau lrg survey which will cover the northern skies over an area of 8000deg@xmath15 , sampling cosmological volume v@xmath310h@xmath76gpc@xmath77 @xcite . by attempting reconstruction , aided by the need for fewer fitting templates , it is possible to do even better , as was done for boss @xcite . in section [ sec : photoz_res ] we showed that limiting the galaxy catalog to a sample of elliptical galaxies reduces the scatter in redshift error to @xmath78=0.007 with a catastrophic failure rate and average bias of 2.5% and -0.001 respectively . in practice , lrg selection is based on color and luminosity selection , but this is not usually difficult for lrgs . with an energy resolution of r@xmath1=30 and the number density of objects catalogued by lsst , giga-@xmath0 should be able to achieve the necessary redshift estimation accuracy for lsst lrgs . the large survey volume probed will ensure an error not dominated by sampling limits , and shot noise should be comparable or less to that of the pau experiment . in addition , since giga-@xmath0 would probe the southern hemisphere , it allows for a joint analysis of the data sets . quasars are extremely luminous objects , believed to be accreting supermassive black holes . type i quasars , due to the high velocities of accretion disk material , are observed to have characteristic broad ( @xmath31/201/10 fwhm ) emission lines in their seds @xcite . they are uv dropout objects , and thus broadband filters will only begin to see the ly-@xmath7 break ( @xmath79@xmath31200 ) at @xmath0@xmath222.2 . though their number density is small compared to ordinary galaxies , quasars are more biased tracers of the matter distribution , and their bias increases with redshift . visible out to high redshifts , the cosmological volume they can be used to probe is much greater than with galaxies , and since sample variance and shot noise decrease as the square root of the volume , they have the potential to measure large scale structure even better than lrgs around the peak of the matter power spectrum in the range @xmath0@xmath313 . furthermore , quasars can also be used to measure bao at high redshift where systematic effects such as redshift distortions and nonlinearities have less influence . @xcite used the sloan digital sky survey ( sdss ) , 2df qso redshift survey ( 2qz ) , and 2df - sdss lrg and qso ( 2slaq ) quasar catalogs to measure bao features , but these were detected only at low statistical significance . however , they estimate that a quarter million @xmath0@xmath132.2 quasars over 3000deg@xmath15 would yield a @xmath33@xmath39 detection of the bao peak . lsst predicts that they will produce a catalog of roughly 10 million quasars . likely these will be identified using photometric selection through color - color and color - magnitude diagrams as was done for sdss @xcite . however , above @xmath0@xmath32.5 , selection becomes much more difficult as quasar colors become indistinguishable from that of stars . in red . this qso has a redshift of 1.703 . ] with the r@xmath1=30 resolution of giga-@xmath0 , the broad emission lines of quasars will be resolved , and their redshifts can be estimated using these spectral features . the high number counts from lsst and redshift accuracy enabled by giga-@xmath0 could , with negligible cost , provide low resolution spectroscopy for the lsst quasar candidates , yielding a precision measurement of the matter power spectrum as well as bao at high redshift . furthermore , quasars could be unambiguously told apart from stars . besides bao and measurements of the distribution of structure to @xmath0@xmath36 , giga-@xmath0 could also be used to study the quasar luminosity function , quasar clustering and bias , set limits on the quasar duty cycle , and improve our understanding of these objects and their evolution and co - evolution with their host galaxies . while we do not explicitly predict the redshift accuracies achievable with giga-@xmath0 here , leaving it for a future investigation , @xcite show that with the j - pas instrument , with 42 contiguous 118 fwhm filters spanning 430815 nm , they could extract photo-@xmath0 s of type i quasars with ( rms ) accuracy @xmath80@xmath3@xmath81 . they show that it is possible to obtain near - spectroscopic photometric redshifts ( which suffer from intrinsic errors due to line shifts ) for quasars with a template fitting method , with a negligible number of catastrophic redshift errors . higher resolution spectra or greater s / n would mostly serve to bring down the number of catastrophic errors . we now estimate the impact of the giga-@xmath0 photometric redshifts on constraints from weak lensing with lsst . in general such a forecast requires the selection of a galaxy sample and a model for statistical and systematic errors . we describe both of these , presenting results appropriate for 3 years of data for both instruments . we first construct the subset of the cosmos mock catalog that has successful shape measurements and photo-@xmath0s . there are several options for doing this with varying levels of aggressiveness depending on s / n cuts , definition of a `` resolved '' galaxy , and photo-@xmath0 quality . to construct the galaxy sample , we first consider the objects resolved by lsst that would have measured shapes . the specific cuts applied were : @xmath82 the resolution factor res@xmath460.4 . the resolution factor is defined using the @xcite convention : @xmath83 where @xmath84 and @xmath85 are the half - light radii of the galaxy and the point - spread function ( psf ) respectively . this cut prevents source galaxies small compared to the psf from being used . the detection signal - to - noise ratio s / n@xmath4618 . the ellipticity measurement uncertainty @xmath86@xmath130.2 ( per component ) . again the ellipticity definition ( and calculation of @xmath86 ) follows @xcite : @xmath87=@xmath88 , where @xmath89 and @xmath90 are the major and minor axes . alternative cuts are possible , which may lead to a larger or smaller source sample . for lsst , we assumed @xmath85=0.39arcsec , which is obtained for a kolmogorov profile with a full width at half maximum of 0.69arcsec . the depth of the imaging data was taken to be @xmath37=26.8 and @xmath17=26.2 ( 5@xmath39 point source ; this is after 3 years , the final lsst dataset will be deeper ) . galaxy shape catalogs were generated separately in the @xmath37 and @xmath17 filters , and their union taken . the density of objects with successful shape measurements is 14.9gal / arcmin@xmath15 . not all of the objects with successful shape measurements can be used : they must also have reliable photo-@xmath0s . to account for this , we split the galaxies into photo-@xmath0 bins of width @xmath91=0.1 . if photo-@xmath0s were always a good tracer of the true redshift , we could use all of the galaxies in each bin . in practice , the redshift outliers must be removed as they contribute to pernicious systematic errors , even if the probability distribution @xmath92 were exactly known . for example , intrinsic galaxy alignments which are known to exist for red galaxies and may contaminate the true lensing signal at the level of up to 23% for blue galaxies @xcite can in principle be removed via the redshift dependence of the signal ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . however , these removal methods do not work if the redshift outliers are themselves intrinsically aligned . the only safe approach is to reduce the outlier rate . thus we impose a requirement on the outlier rate of @xmath135% , so that if their intrinsic alignments are @xmath572% of the lensing signal , the overall contamination is at no more than the part - per - thousand level . this requirement is achieved by the following method : @xmath82 for each galaxy , we compute @xmath55 , the 99% confidence width of the posterior redshift distribution from eazy . in each photo-@xmath0 bin , we impose a cut @xmath93 on @xmath55 . this cut is reduced until the 5% outlier rate is met . in some cases , no cut on @xmath55 can reduce the outlier rate below 5% : in this case that photo-@xmath0 bin is completely removed from the sample . this results in a culled galaxy catalog with both measured shapes and reliable photo-@xmath0s . the size of the culled catalog increases as the photo-@xmath0 performance improves . the unweighted density of galaxies @xmath94 in this catalog is 12.2gal / arcmin@xmath15 ( lsst photo-@xmath0 ) versus 13.0gal / arcmin@xmath15 ( giga-@xmath0 photo-@xmath0 ) . once the galaxy catalog is constructed , the effective source density is obtained via @xmath95 where the sum is over galaxies in the catalog , and the factor in the sum down - weights galaxies whose measurement uncertainty is significant compared with the intrinsic rms dispersion of @xmath570.4 . the effective source densities are 14.5 ( all shapes ) , 11.9 ( lsst photo-@xmath0 ) , and 12.7 ( giga-@xmath0 photo-@xmath0 ) gal / arcmin@xmath15 . cosmological parameter constraints were estimated using the weak lensing fisher matrix code from the figure of merit science working group ( fomswg ; * ? ? ? * ) . this code includes constraints from the shear power spectrum and the geometrical part of galaxy - galaxy lensing ( i.e. ratios of the signals at various redshifts , which depend only on the background cosmology and not the relationship between the galaxies and the mass ; @xcite ) . the inner workings are described at length in the fomswg report ( * ? ? ? * appendix a2 ) and will not be repeated here except for the intrinsic alignment models , which have been updated . in addition to statistical errors , the fomswg code enables several systematic errors to be included : ( i ) shear calibration errors ; ( ii ) photo-@xmath0 biases ; ( iii ) non - gaussian contributions to the covariance matrix of lensing power spectra ; and ( iv ) intrinsic alignments . we turn off ( iii ) here since recent studies have suggested that the effect can be mitigated by nonlinear transformations on the shear map that remove the non - gaussian tails of the lensing convergence distribution contributed by the most massive halos ( e.g. * ? ? ? the original fomswg forecasting software did not include photo-@xmath0 outliers and so we revisit the issue here . fomswg assumed that the systematic uncertainty in the photo-@xmath0s could be captured in a photo-@xmath0 bias @xmath96 where @xmath17=@xmath97 indicates a redshift slice index . fomswg then assumed that a complete spectroscopic survey of @xmath98 source galaxies would be conducted to calibrate the photo-@xmath0 error distribution . if the fraction of source galaxies in the @xmath99 redshift slice is @xmath100 , then it follows that there would be @xmath101 spectroscopic galaxies in the @xmath99 slice . then the spectroscopic survey would enable us to impose a prior on the photo-@xmath0 bias of width : @xmath102 here we extend this approach to include a nonparametric description of redshift outliers ( e.g. * ? ? ? we suppose that a fraction , @xmath103 , of the galaxies in the @xmath99 photo-@xmath0 bin are actually outliers and lie in the @xmath104 true redshift bin . only the true outliers , as defined by @xmath105@xmath46@xmath106 , are included here the `` core '' of the photo-@xmath0 error distribution is modeled using the offset parameters @xmath96 . by construction , @xmath107=0 for all @xmath17 ( a galaxy in the correct bin is not an outlier ) but note that in general @xmath103@xmath108@xmath109 ( the outlier scattering between redshift slices need not be symmetric ) . the shear power spectrum @xmath110 between the @xmath17 and @xmath111 photo-@xmath0 slices then differs from the true power spectrum @xmath112 according to @xmath113 \nonumber \\ & & + \sum_k \epsilon_{jk } [ c_\ell^{ik}({\rm true})-c_\ell^{ij}({\rm true } ) ] , \end{aligned}\ ] ] to first order in @xmath114 . in analogy to eq . ( [ eq : prior - pz ] ) , we treat the suite of @xmath115 parameters @xmath116 as additional nuisance parameters , and impose a prior of the form @xmath117 we consider here both the intrinsic alignment model used by the fomswg and several variations . the key issue is the treatment of the correlation between gravitational lensing and intrinsic alignments ( the `` gi '' term ) , which in the fomswg was parameterized by the matter - intrinsic ellipticity cross - power spectrum , @xmath118 . in general , the matter density @xmath119 , galaxy density @xmath36 , and galaxy ellipticity @xmath87 have a joint symmetric @xmath120 power spectrum matrix : @xmath121 where the @xmath90s represent bias cofficients and @xmath37s represent correlation coefficients . then @xmath122 is a function of @xmath123 and @xmath0 . alternatively , since @xmath123 maps into a given multipole in accordance with the limber formula @xmath124 , they may be considered to take on values in each of the @xmath125 angular scale bins and each of the @xmath126 redshift bins . the fomswg default model ( `` model iii '' here ) imposed a prior on @xmath122 of the following form : @xmath127 that is , fomswg assumed that at large scales ( linear scales , roughly @xmath128 ) galaxies trace the matter density well enough that the observable @xmath129 could be used to estimate @xmath130 and remove the gi signal . however at the nonlinear scales , a weak prior was applied to prevent @xmath131 from exceeding the observed value of @xmath132 estimated from sdss @xcite ; also now see the wigglez survey results ( * ? ? ? * table 4 ) , which combined with sdss give @xmath133 ( @xmath134 ) at @xmath135 for blue galaxies . ; as noted there the passive evolution model would predict @xmath136 . note that the error bar presented in @xcite is @xmath137 . ] ( the square root of the number of bins is inserted to prevent the prior from being `` averaged down '' over many bins . ) the following modifications have been considered here , ordered from optimism to pessimism : : the gi term is ignored entirely , i.e. @xmath138 . this is an unrealistically optimistic model , in that it assumes that a combination of theoretical modeling and observations of the galaxy density - galaxy ellipticity correlation will allow us to compute and subtract off the gi term even in the nonlinear regime and at all redshifts . the fomswg default model , but with the prior coefficient reduced from 0.003 to 0.001 . this model assumes that _ either _ further studied will reduce the upper limits on intrinsic alignments for blue galaxies , _ or _ that advances in modeling galaxy bias in the nonlinear regime will allow us to convert @xmath129 into @xmath130 with @xmath139% uncertainty . the fomswg default model . the fomswg default model , but with the prior coefficient increased to 10 to effectively force future wl data to constrain the gi power spectrum and its redshift evolution in the nonlinear regime , while simultaneously fitting the cosmological parameters . no prior on @xmath122 is applied : the future wl data must now constrain the full gi power spectrum at all scales and redshifts , while simultaneously fitting the cosmological parameters . this model is designed to be overly pessimistic . in light of the rapidly improving observational constraints on intrinsic alignments , and the substantial modeling effort in both intrinsic alignments and galaxy biasing , model ii may be a reasonable ( though not assured ) forecast for the theoretical uncertainty by the time of the lsst weak lensing analysis . table [ tab : deresults ] shows the results for the lsst and giga-@xmath0 weak lensing cases including intrinsic alignment terms , assuming a training sample of 25,000 spectroscopic redshifts , and galaxy shapes from lsst photometry . @xmath140 is the uncertainty on @xmath18 if it is assumed to be a constant , @xmath141 the uncertainty on the rate of change of @xmath18 . @xmath142 parameterizes the rate of the growth of structure , and @xmath143 the curvature of spacetime . the planck cmb data are included in all of the models shown , since without the cmb constraints the `` standard '' cosmological parameters ( @xmath144 , @xmath145 , ... ) can be altered to accommodate a fit of the wl data with almost any smooth dark energy model . lcccc & @xmath140 ' '' '' & @xmath141 ' '' '' & @xmath146 ' '' '' & @xmath147 ' '' '' + & + lsst photo - z & ' '' '' 0.0271 & 0.494 & 0.158 & 0.0246 + giga - z photo - z & ' '' '' 0.0246 & 0.405 & 0.124 & 0.0200 + & + lsst photo - z & ' '' '' 0.0373 & 0.671 & 0.204 & 0.0251 + giga - z photo - z & ' '' '' 0.0341 & 0.562 & 0.157 & 0.0204 + & + lsst photo - z & ' '' '' 0.0382 & 0.695 & 0.221 & 0.0252 + giga - z photo - z & ' '' '' 0.0348 & 0.576 & 0.168 & 0.0205 + & + lsst photo - z & ' '' '' 0.0396 & 0.743 & 0.273 & 0.0258 + giga - z photo - z & ' '' '' 0.0364 & 0.627 & 0.206 & 0.0211 + & + lsst photo - z & ' '' '' 0.0503 & 1.053 & 0.330 & 0.0270 + giga - z photo - z & ' '' '' 0.0450 & 0.912 & 0.279 & 0.0223 + for the `` fomswg default '' model ( model iii ) , the improvement in parameter constraints from the giga-@xmath0 photo-@xmath0s for wl+planck is equivalent to multiplying the lsst fisher matrix by a factor of @xmath148 ( @xmath8 ) , 1.53 ( @xmath9 ) , or 1.98 ( @xmath142 ) . note that e.g. for @xmath9 the improvement is equivalent to both multiplying the lsst coverage area and the training sets by 1.53 and reducing all systematics by a factor of @xmath149 . in general , we see larger improvements for the cases with more complex redshift dependence , such as changing @xmath18 or the rate of growth of structure . the improvement for the rate of growth of structure one of the most important constraints for weak lensing since it can not be probed by supernovae is particularly impressive : in model iii , adding the giga-@xmath0 photometric redshifts would be as valuable for @xmath150 as adding a northern hemisphere lsst . while the choice of intrinsic alignment model affects the cosmological constraints ( as expected , the more optimistic models that assume better knowledge of the intrinsic alignments lead to smaller error bars ) , the advantages of giga-@xmath0 are robust to even the more extreme models presented here . for @xmath150 , the `` advantage factor '' @xmath7 defined in the previous paragraph varies as 1.82 , 2.00 , 1.98 , 1.94 , or 1.49 ( models i v respectively ) ; for @xmath9 it is 1.60 , 1.51 , 1.53 , 1.45 , or 1.38 ( models i v respectively ) . several dark energy probes rely on photometric redshift estimates that are accurate and exhibit little bias . the des , lsst , euclid , kids and other wide field imaging experiments will survey much of the sky in the usual photometric bands , but to fully realize their potential , the photo-@xmath0 scatter and biases must be well calibrated . we have simulated realistic observations with both a 3 year lsst stack and a proposed experiment , giga-@xmath0 , and compared the results side - by - side . the mock catalog used , based on cosmos observations , is deep , complete , and representative of the real span of galaxies we might expect these experiments to observe , including objects from the `` redshift desert '' . by construction , this mock catalog likely contains objects with unusual seds . we have shown that giga-@xmath0 , with r@xmath1=30 spectral resolution , spatial resolution , and continuous wavelength coverage between 350 and 1350 nm can efficiently and effectively obtain spectrophotometry of a much larger and deeper sample of galaxies than is possible with current spectrographs . from our simulations , we predict redshift estimate accuracies of @xmath6@xmath30.03 for the whole sample , and @xmath6@xmath30.007 for a select subset , which in turn adds constraint on dark energy parameters for wl + cmb ( planck ) . in particular , for the rate of growth of structure , one of the most important constraints for weak lensing since it can not be probed by supernovae , for the default fomswg model , adding the giga-@xmath0 photometric redshifts would be equivalent to doubling the lsst footprint ( e.g. , by running a second complete lsst survey in the north ) . this data could be obtained inexpensively compared with most current and future surveys . with des set to come online imminently , giga-@xmath0 would be able to use des catalogs to inform a first pass , and operate in parallel with lsst and other wide field imaging surveys . d. m. was supported by a grant from the keck institute for space studies . c. h. was supported by doe doe.de-sc0006624 , and the david and lucile packard foundation . , m. p. , & tyson , j. a. 2002 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 4836 , society of photo - optical instrumentation engineers ( spie ) conference series , ed . j. a. tyson & s. wolff , 240246 , b. a. , obrien , k. , mchugh , s. , bumble , b. , moore , d. , golwala , s. , & zmuidzinas , j. 2010 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 7735 , society of photo - optical instrumentation engineers ( spie ) conference series

we employed a real - space formulation of orbital - free density functional theory using finite - element basis to study the defect - core and energetics of an edge dislocation in aluminum . our study shows that the core - size of a perfect edge dislocation is around ten times the magnitude of the burgers vector . this finding is contrary to the widely accepted notion that continuum descriptions of dislocation energetics are accurate beyond burgers vector from the dislocation line . consistent with prior electronic - structure studies , we find that the perfect edge dislocation dissociates into two shockley partials with a partial separation distance of . interestingly , our study revealed a significant influence of macroscopic deformations on the core - energy of shockley partials . we show that this dependence of the core - energy on macroscopic deformations results in an additional force on dislocations , beyond the peach - koehler force , that is proportional to strain gradients . further , we demonstrate that this force from core - effects can be significant and can play an important role in governing the dislocation behavior in regions of inhomogeneous deformations . density functional theory , real space , dislocation , core size , core energy

1406.3741

dislocations are line defects in crystalline materials which play an important role in governing the deformation and failure mechanisms in solids . the energetics of dislocations and their interactions with other defects solute atoms , precipitates , grain boundaries , surfaces and interfaces significantly influence the mechanical properties of crystalline materials ( cf . e.g. @xcite ) . for instance , the kinetic barriers for dislocation motion dislocation glide and climb and their dependence on crystallographic planes and directions govern ductility and creep in metals @xcite . interaction of dislocations with vacancies , solute atoms and precipitates results in solid - solution strengthening / softening , precipitate hardening and aging in metals @xcite . further , dislocation interactions with grain boundaries and surfaces are responsible for the observed strengthening mechanisms like the hall - petch effect @xcite , and enhanced yield strength in surface dominated nanostructures @xcite . the behavior of dislocations ( nucleation , kinetics , evolution ) in crystalline materials is governed by physics on multiple length - scales . in particular , a dislocation produces elastic fields that are long - ranged , and through these elastic fields interacts with other defects and external loads at macroscopic scales . on the other hand , the quantum - mechanical and atomistic scale interactions play an important role in governing the nucleation and kinetics of these defects . while atomistic scale interactions can significantly influence the behavior of dislocations , these are localized to a region around the dislocation line referred to as the dislocation - core . thus , the energy of a dislocation comprises of the stored elastic energy ( @xmath2 ) , associated with the elastic fields outside the dislocation - core , and a core - energy ( @xmath3 ) associated with quantum - mechanical and atomistic scale interactions inside the dislocation - core . continuum theories based on elastic formulations have been widely used to study deformation and failure mechanisms mediated through dislocations ( cf . e. g. @xcite ) , where the energetics of dislocations are solely determined by the elastic energy and the core - energy is often assumed to be an inconsequential constant . in order to overcome the inability of continuum theories to describe the dislocation - core , explicit atomistic calculations based on empirical interatomic potentials have also been employed to study deformation mechanisms mediated by dislocations ( cf . e. g. @xcite ) , and have provided many useful insights . however , interatomic potentials , whose parameters are often fit to bulk properties , may not accurately describe the defect - core which is governed by the electronic - structure ( cf . e. g. @xcite ) . electronic - structure calculations using plane - wave implementations of density functional theory ( dft ) have been employed to study the dislocation core - structure in a wide range of crystalline materials ( cf . e. g. @xcite ) and the energetics of dislocation - solute interactions in metals with different crystallographic symmetries @xcite . as the displacement fields produced by isolated dislocations are not compatible with periodic boundary conditions , these calculations have either been restricted to artificial dipole and quadrapole configurations of dislocations or free - surfaces have been introduced to contain isolated dislocations . recent efforts have also focused on the development of flexible boundary conditions by extending the lattice green s function method to electronic - structure calculations @xcite . flexible boundary conditions accurately account for the long - ranged elastic fields of an isolated dislocation @xcite , however , the electronic - structure in these studies is computed by introducing free surfaces to accommodate the restrictive periodic boundary conditions associated with plane - wave based dft implementations . while these aforementioned studies have provided useful insights into the dislocation core - structure , a direct quantification of the dislocation core - energy solely from electronic - structure calculations and its role in governing dislocation behavior has remained elusive thus far . we note that some prior _ ab initio _ studies using a dipole or quadrapole configuration of dislocations ( cf . e.g. @xcite ) , have attempted to indirectly compute the core - energy of an isolated dislocation by subtracting from the total energy the elastic interaction energy between dislocations in the simulation cell and their periodic images . this approach assumes the spacing between dislocations is large enough that the dislocation - cores do not overlap . however , these prior studies have been conducted on computational cells containing a few hundred atoms , which , as demonstrated in this work , are much smaller than the core - size of an isolated perfect edge dislocation in aluminum . in this work , we conduct large - scale electronic - structure calculations using orbital - free density functional theory to study an edge dislocation in aluminum . in our study , we use the wgc kinetic energy functional @xcite which has been shown to be in good agreement with kohn - sham dft for a wide range of material properties in aluminum @xcite . we employ a local real - space formulation of orbital - free density functional theory @xcite , where the extended interactions are reformulated as local variational problems . this real - space formulation of orbital - free density functional theory is used in conjunction with the finite - element basis that enables the consideration of complex geometries and general boundary conditions , which is crucial in resolving the aforementioned limitations of plane - wave basis in the study of energetics of isolated dislocations . we begin our study by computing the size of the dislocation - core for a perfect edge dislocation in aluminum . to this end , we consider a perfect edge dislocation with the atomic positions given by isotropic elasticity theory . for these fixed atomic positions , we identify the region where the perturbations in the electronic - structure arising from the defect - core are significant and have a non - trivial contribution to the dislocation energy . this allows us to unambiguously identify the dislocation - core from the viewpoint of energetics . our study suggests that the dislocation core - size of a perfect edge dislocation is about @xmath4 , where @xmath5 denotes the burgers vector . this estimate is much larger than conventional estimates based on displacement fields , which suggest a dislocation core - size of @xmath6 @xcite , and underscores the long - ranged nature of the perturbations in the electronic fields arising from defects . we note that a similar long - ranged nature of the electronic fields was observed in recent studies on point defects @xcite . as a next step in our study , we allow for atomic relaxations , and the perfect edge dislocation dissociates into shockley partials with a partial separation distance of @xmath1 . the dislocation energy per unit length of the relaxed shockley partials in the simulation domain corresponding to the identified core - size , which denotes the dislocation core - energy , is computed to be @xmath7 @xmath8 . we next study the role of macroscopic deformations on the dislocation core - energy and core - structure . in particular , we considered a wide range of macroscopic deformations including : ( i ) equi - triaxial strains representing volumetric deformations ; ( ii ) uniaxial strains along the the burgers vector , normal to the slip plane , and along the dislocation line ; ( ii ) equi - biaxial strains along these crystallographic directions ; ( iii ) shear strains along these crystallographic directions , excepting the one which results in dislocation glide . interestingly , we find that the dislocation core - energy is significantly influenced by applied macroscopic deformations . in particular , we find that this core - energy dependence on macroscopic strains is non - linear and non - monotonic . further , this core - energy dependence on macroscopic deformations suggests that the dislocation experiences an additional configurational force beyond the peach - koehler force . this additional force arising from core - effects , and referred to as the core - force , is proportional to the strain gradients and can be significant in regions of inhomogeneous deformations . in particular , we estimate that , in the case of a dislocation dipole with the two dislocations aligned at @xmath9 to the slip direction , the core - force has a glide component which is greater than the peierls nabarro force even at a distance of @xmath10 . while we observe a significant influence of macroscopic deformations on the core - energy , the core - structure , studied using differential displacement plots , did not change significantly in response to the applied macroscopic uniaxial , biaxial and triaxial strains . however , the shear strain which exerts a glide force on the screw components of the shockley partials showed a significant influence , as expected , on the core - structure . the remainder of this paper is organized as follows . section [ sec : ofdft ] presents an overview of the local real - space formulation of orbital - free dft employed in this work . section [ sec : results ] presents our electronic - structure study of the edge dislocation in aluminum , along with a discussion of the new findings and its implications . we finally conclude with an outlook in section [ sec : conclusions ] . in this section , for the sake of completeness and to keep the discussion self - contained , we provide an overview of the local real - space formulation of orbital - free dft employed in the present work and the finite - element discretization of the formulation . we refer to @xcite for a comprehensive discussion on these topics . the ground - state energy in density functional theory @xcite for a materials system containing @xmath11 nuclei and @xmath12 electrons is given by @xmath13 where @xmath14 denotes the ground - state electron density and @xmath15 denotes the vector collecting the positions of atoms . in the above , @xmath16 denotes the kinetic energy of non - interacting electrons ; @xmath17 denotes the exchange - correlation energy ; @xmath18 denotes the hartree energy or the electrostatic interaction energy between the electrons ; @xmath19 denotes the electrostatic interaction energy between the electrons and nuclei ; and @xmath20 denotes the nuclear - nuclear repulsion energy . in the kohn - sham formalism , the kinetic energy of non - interacting electrons , @xmath16 , is evaluated using the single - electron wavefunctions computed from the kohn - sham equations . in orbital - free dft , in order to circumvent the cubic - scaling ( @xmath21 ) complexity in evaluating the single - electron wavefunctions or associated density matrix in the kohn - sham formalism , the kinetic energy of non - interacting electrons is modeled as an explicit functional of electron density @xcite . this reduces the problem of evaluating the ground - state energy and electron density to a variational problem that scales linearly with system size . the widely used model for @xmath16 , which has been shown to be a transferable model , in particular , for al , mg and al - mg materials systems @xcite , is the wang - govind - carter ( wgc ) kinetic energy functional @xcite given by @xmath22 where @xmath23 in equation , the first term denotes the thomas - fermi energy with @xmath24 , and the second term denotes the von - weizs@xmath25cker correction @xcite . the last term denotes the density dependent kernel energy , @xmath26 , where the kernel @xmath27 and the parameters @xmath28 and @xmath29 are chosen such that the linear response of a uniform electron gas is given by the lindhard response . in the wgc functional , the parameters are chosen to be @xmath30 , and the density dependent kernel is expanded as a taylor series about the bulk average electron density to second order , thus yielding a series of density independent kernels @xcite . the second term in equation , @xmath17 , denoting the exchange - correlation energy , includes all the quantum - mechanical interactions and is modeled in density functional theory . widely used models for the exchange - correlation energy , especially for solid - state calculations of ground - state properties , include the local density approximation ( lda ) and the generalized gradient approximation ( gga ) @xcite , and have been demonstrated to be transferable models for a range of materials systems and properties . the last three terms in equation constitute the classical electrostatic energy between electrons and the nuclei . we note that all the terms in the energy functional are local , excepting the electrostatic energy and the kernel energy , which are extended in real - space . as a local real - space formulation is desirable for scalable numerical implementations as well as the development of coarse - graining techniques , a local reformulation of these extended interactions has been proposed and implemented recently @xcite . by noting that the extended interactions in electrostatics are governed by the @xmath31 kernel , which is the green s function of the laplace operator , the electrostatic interaction energy can be reformulated as the following local variational problem : @xmath32 where @xmath33 denotes the charge distribution corresponding to the ionic pseudopotential of the @xmath34 nucleus , @xmath35 denotes the electrostatic potential corresponding to the total charge distribution comprising of the electrons and the nuclei , and @xmath36 denotes a suitable function space incorporating appropriate boundary conditions for the problem being solved . in the above , @xmath37 denotes the self energy of the nuclear charge distributions which can be evaluated , again , by taking recourse to the poisson equation ( cf . @xcite ) . a local reformulation of the extended interactions in the kernel kinetic energy functional has been proposed in @xcite , and has been subsequently recast into a variational problem in @xcite . it was demonstrated in @xcite that the series of density independent kernels obtained from the taylor series expansion of the wgc density dependent kernel can each be approximated in the fourier - space using a sum of partial fractions . using this approximation , a real - space reformulation for these extended interactions has been proposed by taking recourse to the solutions of a series of complex helmholtz equations . in particular , if @xmath38 denotes a density independent kernel , and is approximated in the fourier - space using the approximation @xmath39 , the local reformulation of the kernel energy is given by the following saddle point problem @xcite : @xmath40 where @xmath41d\br\bigr\}\ , . \end{split}\ ] ] in the above , @xmath42 and @xmath43 denote the vector of potential fields given by @xmath44 and @xmath45 , and @xmath46 denotes a suitable function space incorporating appropriate boundary conditions for the problem being solved . the auxiliary potential fields , @xmath47 and @xmath48 for @xmath49 , introduced in the local reformulation of kernel energy are referred to as the _ kernel potentials_. using the reformulations in equations and , the extended interactions in electrostatics and the kernel kinetic energy functional can be recast into local variational problems in auxiliary potential fields electrostatic potential and kernel potentials . thus , using these local reformulations , the problem of evaluating the ground - state energy and electron density reduces to a local saddle point problem in the _ electronic fields _ comprising of electron density , electrostatic potential and kernel potentials ( cf . @xcite ) . such a local reformulation in real - space is crucial for employing non - homogenous dirichlet boundary conditions for studying the energetics of isolated defects in bulk , which is exploited in this work ( elaborated subsequently in section [ sec : results ] ) . plane wave basis has been the popular choice for the numerical discretization of the orbital - free dft variational problem @xcite owing to the ease of computing the extended electrostatic interactions and the kernel kinetic energy functional in fourier - space using fourier transforms . however , fourier - space methods are limited to periodic systems , which poses a severe restriction in studying defects in materials . in particular , the geometry of an isolated dislocation is not periodic , and , thus , orbital - free dft simulations on dislocations have either been limited to an artificial dipole / quadrapole arrangements of dislocations , or free - surfaces are introduced to study isolated dislocations @xcite . while these studies have provided useful insights into the core - structure of dislocations , a quantification of the core - energies has been beyond reach . in order to overcome these limitations , recent efforts have also focused on developing real - space discretization techniques for the orbital - free dft problem , which include finite difference discretization @xcite and the finite - element discretization @xcite . in particular , the finite - element discretization is attractive as it can handle complex geometries and boundary conditions , which is a crucial aspect that is exploited in this study . further , recent investigations have demonstrated the accuracy and computational efficiency afforded by higher - order finite - element discretization for density functional theory calculations @xcite , which , along with the good parallel scalability of finite - element discretizations , make the finite - element basis a desirable choice for real - space orbital - free dft calculations . we refer to @xcite for a comprehensive discussion on the numerical analysis of the finite - element discretization of the orbital - free dft problem , and the numerical aspects of the solution of the discrete orbital - free dft problem . we now present our study of an edge dislocation in aluminum using large - scale electronic - structure calculations . in particular , we unambiguously identify the size of the dislocation core , directly compute the dislocation core - energy , study the effect of macroscopic deformations on dislocation core - energy and core - structure , and investigate its implications on dislocation behavior . in this study , we employ orbital - free density functional theory calculations using the wang - govind - carter ( wgc ) kinetic energy functional @xcite ( second order taylor expansion of the density dependent kernel , cf . @xcite ) , a local density approximation ( lda ) for the exchange - correlation energy @xcite , and goodwin - needs - heine pseudopotential @xcite . while orbital - free dft uses an approximation for the kinetic energy functional , it has been demonstrated that the wgc kinetic energy functional is in good agreement with kohn - sham dft for a wide range of material properties in aluminum , which include bulk and defect properties @xcite . further , orbital - free dft is inherently linear - scaling in the number of electrons and enables the consideration of large system sizes , which is necessary , as will be demonstrated in this work , for an accurate study of the energetics of dislocations . in the present work , we employ recently developed local real - space formulation of orbital - free dft ( cf . section [ sec : ofdft ] and @xcite ) and a finite - element discretization of the formulation @xcite to compute the energetics of an edge dislocation in aluminum . in this real - space formulation , all the extended interactions , which include the electrostatic interactions and the kernels energies in the wgc kinetic energy functional , are reformulated into local variational problems in auxiliary potential fields that include the electrostatic potential and kernel potentials . the finite - element basis employed in the present work enables consideration of complex geometries and general boundary conditions ( dirichlet , periodic and semi - periodic ) , which is another crucial aspect that enables an accurate quantification of the energetics of isolated dislocations . the accuracy of the local real - space formulation of orbital - free dft and the convergence of the finite - element discretization of the formulation have been demonstrated for both bulk and defect properties in aluminum @xcite . in the present study , we use quadratic hexahedral finite - elements in all our calculations , where the basis functions correspond to a tensor product of quadratic polynomials in each dimension . the finite - element discretization and other numerical parameters quadrature rules and stopping tolerances on iterative solvers are chosen such that the errors in the computed dislocation energies are less than @xmath50 @xmath51 . in all our calculations , the atomic relaxations are performed till the maximum force on any atom is less than @xmath52 @xmath8 . in order to compute the dislocation core - energy , we begin by first estimating the dislocation core - size . continuum theories estimate the dislocation core - sizes based on displacement fields , and prior studies suggest the dislocation core - size to be @xmath53 @xmath54 @xmath55 @xcite , where @xmath5 denotes the burgers vector . while the displacement and strain fields may be well described by elastic formulations outside of a @xmath54 @xmath55 core region , the perturbations in electronic - structure due to the dislocation may be present on a larger region . the energetic contributions from these electronic - structure effects can not be captured in continuum elasticity theories or atomistic calculations using empirical potentials . thus , in the present work , we consider the dislocation - core to be the region where electronic - structure effects and their contribution to the defect energy are significant . we consider a perfect edge dislocation in face - centered - cubic aluminum to determine the dislocation core - size . we align our coordinate system , @xmath56 axes ( or equivalently @xmath57 ) , along @xmath58-[1\,\overline{1}\,1]-[1\,\overline{1}\,\overline{2}]$ ] crystallographic directions , respectively . we begin by considering a perfect crystal of size @xmath59 where @xmath60 denotes the lattice parameter and @xmath61 is an integer - valued scaling factor used to consider a sequence of increasing simulation domain sizes . a perfect edge dislocation with burgers vector @xmath62 $ ] is introduced at the center of the simulation domain by removing two consecutive half - planes normal to @xmath63 $ ] and applying the continuum displacement fields of an edge dislocation to the positions of atoms . we note that this approach of creating the edge dislocation is devoid of the asymmetry that can otherwise arise by applying the volterra solution in the most obvious manner by creating slip ( cf . @xcite ) . in this work , the displacement fields from isotropic elasticity @xcite are employed as anisotropic effects are not strong in aluminum . upon the application of displacement fields , the geometry of the computational domain is no longer cuboidal , and , thus , the use of a finite - element basis which can accommodate complex geometries is crucial to the present study . in order to simulate a perfect edge dislocation , we hold the positions of the atoms fixed and compute the electronic - structure using orbital - free dft . we employ dirichlet boundary conditions on the electronic fields comprising of electron density , electrostatic potential and kernel potentials in the @xmath64 and @xmath65 directions and use periodic boundary conditions along the @xmath66 direction . the dirichlet boundary conditions are determined under the cauchy - born approximation , where the values of electronic fields on the dirichlet boundary are computed from periodic unit - cells undergoing the deformations produced by the edge dislocation . the electronic - structure , thus computed , represents an isolated edge dislocation in bulk with the electronic - structure perturbations from the edge dislocation vanishing on the dirichlet boundary . the local real - space formulation of orbital - free dft and the finite - element basis are key to employing these boundary conditions , which are not possible to realize in fourier - space based formulations . we computed the electronic - structure and ground - state energy of the perfect edge dislocation for varying simulation domains with @xmath67 . figure [ fig : contours ] shows the contours of the electron density for @xmath68 . we note that a scaling factor @xmath61 corresponds to a domain - size where the distance from dislocation line to the boundary along @xmath63 $ ] is @xmath69 . the dislocation energy ( @xmath70 ) for these various simulation domains is computed as @xmath71 where @xmath72 denotes the energy of the @xmath11 atom system comprising of the dislocation and occupying a volume @xmath73 , and @xmath74 denotes energy of a perfect crystal containing the same number of atoms and occupying the same volume . we first computed the dislocation energy at equilibrium volume ( i.e. @xmath75 ) for the various domain - sizes considered in this study , and the results are presented in table [ tab : elasticelec ] . the computed dislocation energy increases with increasing domain - size , and has an asymptotic logarithmic divergence as expected from continuum theories . in order to understand the extent of electronic relaxations , we consider the change in dislocation energy by increasing the domain - size for instance , from a domain - size of @xmath76 to @xmath77and denote this change by @xmath78 . this change in the energy has two contributions : ( i ) the increase in the elastic energy due to the increase in the domain - size , which we denote by @xmath79 ; ( ii ) electronic contribution from perturbations in the electronic - structure , which we denote by @xmath80 . as the elastic energy contribution to the dislocation energy is due to the elastic deformation fields produced by the dislocation , we estimate @xmath79 by integrating the elastic energy density in the region of interest . the elastic energy density at any point , in turn , is computed from an orbital - free dft calculation on a unit - cell undergoing the macroscopic deformation produced by the edge dislocation at that point . we verified that the discretization errors in the computation of @xmath79 are of the order of @xmath81 . upon computing @xmath79 , we infer @xmath80 from @xmath78 ( @xmath82 ) . the computed @xmath79 and @xmath80 are reported in table [ tab : elasticelec ] . it is interesting to note from the results that @xmath80 is comparable to @xmath79 up to a domain - size of @xmath4 , suggesting that the electronic - structure perturbations are significant up to distances as far as @xmath4 from the dislocation line . this result is in strong contrast to conventional estimates of core - sizes to be @xmath83 . further , we note that a core - size of @xmath4 contains over @xmath84 atoms , which is much larger than the computational cells employed in most prior electronic - structure studies of dislocations , and underscores the need to consider sufficiently large simulation domains to accurately compute the energetics of dislocations . in the remainder of this work , we consider @xmath4 to be the core - size of edge dislocation , and the dislocation energy corresponding to this core - size as the dislocation core - energy . for a perfect edge dislocation , the computed core - energy is @xmath85 , or , equivalently , the core - energy per unit length of dislocation line is @xmath86 . a plot of the dislocation energies ( @xmath70 ) reported in table [ tab : elasticelec ] as a function of domain - size is provided in figure [ fig : cellsize ] , which shows the expected asymptotic logarithmic divergence of the dislocation energy . ( corresponding to a domain of @xmath87 along @xmath63-[1\overline{1}1]-[1\overline{1}\overline{2}]$]).,scaledwidth=65.0% ] as a next step in our study , we allowed for internal atomic relaxations in the simulation domain corresponding to @xmath4 by holding the positions of atoms fixed on the boundary . upon atomic relaxations , the perfect edge dislocation split into shockley partials . figure [ fig : dd ] shows the edge and screw components of the differential displacements @xcite and indicates the location of the shockley partials . the partial separation distance computed from the edge - component differential displacement plot is @xmath88 ( @xmath89 ) , and that computed from the screw - component differential displacement plot is @xmath90 ( @xmath91 ) . this is in good agreement with other dft studies on an edge dislocation in aluminum which have reported partial separation distances between @xmath92 @xcite . the core - energy of shockley partials is computed to be @xmath93 , or , equivalently , the core - energy per unit length of dislocation line is @xmath94 . the core - energy of the relaxed shockley partials , as expected , is significantly lower than that of the perfect edge dislocation . in the remainder of this paper , all the reported core - energies correspond to the relaxed shockley partials . we note that holding the positions of atoms fixed on the boundary , prescribed by the displacement fields of a perfect edge dislocation , can result in finite cell - size effects that influence the core - structure of the shockley partials and the core - energy . in order to investigate the influence of this possible finite cell - size effect , we considered simulation domains corresponding to @xmath95 and @xmath96 and computed the core - structure in these simulation domains , and compared it with the core - structure obtained from the @xmath4 simulation domain . the partial separation distance of both the edge- and screw - components did not change by increasing the simulation domain . further , we computed the difference in the positions of atoms ( in the region corresponding to @xmath4 simulation domain ) obtained from the @xmath96 and @xmath4 simulation domains . the maximum difference in the positions of atoms is @xmath97 , which is close to the tolerance associated with force relaxations . these results suggest that the finite cell - size effects arising from the fixed spatial boundary conditions on atomic positions are not important in simulation domains of size @xmath4 and beyond . while the computed partial separation distance is in good agreement with prior electronic - structure studies , the partial separation distances are over - predicted in dft compared to experimental investigations . the partial separation distance of the edge dislocation in aluminum is estimated to be @xmath98 using weak - field tem @xcite . one plausible cause of this deficiency in dft calculations can be the use of pseudopotentials , especially since the stacking fault energy is sensitive to the choice of the pseudopotential @xcite . while pseudopotential dft calculations have been shown to be accurate for the prediction of bulk properties in solids , their transferability to accurately predict defect properties is yet to been ascertained to this end , there is a growing need for the development of efficient large - scale real - space methods for all - electron kohn - sham dft calculations , and we refer to @xcite for recent efforts in this direction . most continuum and mesoscale models describing deformation and failure mechanisms in crystalline solids account for the role of defects through the elastic interactions between various defects , as well as , elastic interactions between defects and macroscopically applied loads . however , recent electronic - structure studies on point defects @xcite suggest that the defect - core can play a significant role in governing the overall energetics of these defects , and that there is a strong influence of macroscopic deformations on the electronic - structure of the defect - core and its energetics . further , a recent electronic - structure study also reported a sizeable effect of pressure on the dislocation - core properties in semiconductor materials @xcite . thus , as a next step in our study , we investigate the effect of macroscopic deformations on the core - energy and core - structure of an edge dislocation in aluminum . we begin with a perfect edge dislocation in @xmath4 simulation domain , which corresponds to the core - size of the edge dislocation . we subsequently apply an affine deformation on the simulation domain corresponding to a macroscopic strain @xmath99 , and , by holding the positions on the boundary fixed , compute the electronic - structure as well as the atomic structure of the relaxed shockley partials . in these simulations , as discussed earlier , dirichlet boundary conditions are applied on the electronic fields in the @xmath64 and @xmath65 directions , where the boundary values are computed using the cauchy - born hypothesis . we note that the applied macroscopic strain manifests into the boundary conditions on the positions of atoms , as well as , the values of the electronic fields on the dirichlet boundary . following equation , the dislocation energy of this @xmath4 simulation domain ( dislocation core - energy ) as function of macroscopic strain is computed as @xmath100 in the above expression , @xmath101 denotes the ground - state energy ( after internal atomic relaxations ) of the @xmath4 simulation domain , which contains the dislocation , under an affine deformation corresponding to macroscopic strain @xmath102 , and @xmath103 denotes the energy of a perfect crystal under the same affine deformation , containing the same number of atoms and occupying the same volume . we begin our investigation by studying the effect of macroscopic volumetric strain ( @xmath104 ) , corresponding to equi - triaxial strain , on the core - energy of shockley partials and the core - structure . in this study , we consider volumetric strains of @xmath105 , @xmath106 , @xmath107 , @xmath108 , @xmath109 and @xmath110 . the computed core - energy demonstrated a strong dependence on volumetric strain , and is shown in figure [ fig : volumetric_strain ] . the core - energy ( per unit length of dislocation line ) changed from @xmath111 at @xmath105 volumetric strain to @xmath112 at @xmath110 volumetric strain , and this change corresponds to a significant fraction ( @xmath113 ) of the core - energy at equilibrium . this finding is in sharp contrast to the assumptions in continuum formulations that ignore the core - energy in determining the energetics of dislocations and their interactions . the core - structure , analyzed using the differential displacement ( dd ) plots , is also found to be marginally influenced by the applied macroscopic volumetric strain . in particular , for the range of volumetric strains considered , the partial separation in the edge - component dd plots is in the range of @xmath114 , and the partial separation in the screw - component dd plot is in the range of @xmath115 . we next study the effect of uniaxial strains along the coordinate directions , @xmath58-[1\,\overline{1}\,1]-[1\,\overline{1}\,\overline{2}]$ ] , on the dislocation core - energy and core - structure . we consider uniaxial strains of @xmath116 , @xmath117 , @xmath118 , @xmath119 , @xmath120 and @xmath121 along each of the coordinate directions . figures [ fig : uniaxial1 ] , [ fig : uniaxial2 ] , and [ fig : uniaxial3 ] show the core - energy dependence on @xmath122 ( uniaxial strain along @xmath58 $ ] ) , @xmath123 ( uniaxial strain along @xmath124 $ ] ) and @xmath125 ( uniaxial strain along @xmath126 $ ] ) , respectively . as in the case of volumetric strains , the core - energy monotonically decreases from compressive strains to tensile strains ( for the range of strains considered ) for uniaxial strains @xmath122 ( along the burgers vector ) and @xmath125 ( along the dislocation line ) . however , the core - energy dependence is non - monotonic and non - linear for @xmath123 uniaxial strain ( along the normal to the slip plane ) . these results show that the core - energy of an edge dislocation can have a complex and non - linear dependence on macroscopic strains . as in the case of volumetric strains , the core - structure is found to be marginally sensitive to uniaxial strains . for the range of uniaxial strains considered in this study , the partial separation in the edge - component dd plots is in the range of @xmath114 and the partial separation in the screw - component dd plots is in the range of @xmath127 . the fact that the core - energy shows a strong dependence on the volumetric and uniaxial strains , while the core - structure is only marginally influenced by these strains , suggests that the electronic - structure at the core can play a dominant role in governing the energetics of dislocations and that the core - structure alone may not completely characterize the defect - core . we next investigate the effect of biaxial strains on the edge dislocation . to this end , we consider equi - biaxial states of strain along the coordinate directions , and computed the core - energies and core - structures under these applied macroscopic strains . figures [ fig : biaxial12 ] , [ fig : biaxial13 ] , [ fig : biaxial23 ] show the dependence of core - energy on equi - biaxial strains @xmath128 , @xmath129 and @xmath130 , respectively . as in the case of uniaxial strains , the core - energy dependence on biaxial strains is found to be non - monotonic when there is macroscopic strain along @xmath124 $ ] , the direction normal to the slip plane ( cf . figures [ fig : biaxial12 ] , [ fig : biaxial23 ] ) . these results again underscore the highly non - linear nature of core - energy dependence on macroscopic strains . the biaxial strains also influenced the core - structure , and to a greater extent than uniaxial strains . for the range of uniaxial strains considered in this study , the partial separation in the edge - component dd plots is in the range of @xmath131 and the partial separation in the screw - component dd plots is in the range of @xmath132 . we next consider shear strains @xmath133 and @xmath134 and their influence on the core - energy and core - structure . in the present work , we do not study the influence of @xmath135 strain , as , even at small values of shear stress resulting from this shear strain ( @xmath136 , cf . @xcite ) , the edge dislocation glides by overcoming the peierls barrier . the computed core - energy dependence on @xmath133 and @xmath134 shear strains is shown in figure [ fig : shear_strain ] . it is interesting to note that the core - energy dependence is symmetric in @xmath133 , to within the discretization errors of @xmath137 . on the other hand , the core - energy dependence on @xmath134 strain is asymmetric . this qualitative change in core - energy dependence can be rationalized from the forces created by these macroscopic shear strains on the shockley partials . the shear stress associated with @xmath133 results in climb forces , in opposite directions , on the screw components of shockley partials . upon changing the sign of the shear stress , the direction of the climb forces on each of the shockley partials is reversed , but these configurations are symmetry related and thus the core - energy variation is symmetric with respect to @xmath133 . in contrast , the shear stress associated with @xmath134 results in glide forces , in opposite directions , on the screw component of shockley partials . thus , the sign of this shear stress has an effect of either increasing or decreasing the partial separation distance between the shockley partials , which is the source of the asymmetry in core - energy dependence on @xmath134 . this is also further corroborated from the core - structure , which shows a significant change in the partial separation distance in the screw - component dd plot in response to @xmath134 strain . in particular , the partial separation in the screw - component of the dd plot changes from @xmath138 at @xmath139 to @xmath140 at @xmath141 ( cf . figure [ fig : dd_screw_e23 ] ) . further , the partial separation in the edge - component of the dd plot also changes from @xmath89 at @xmath139 to @xmath142 at @xmath141 . on the other hand , the core - structures are found to be similar for both negative and positive @xmath133 strains of a given magnitude , as expected from symmetry . for the range of @xmath133 shear strains considered in the present work , the partial separation in the edge - component dd plots is in the range of @xmath114 and the partial separation in the screw - component dd plots is in the range of @xmath127 . the study on the effect of macroscopic strains revealed a strong dislocation core - energy dependence on macroscopic deformations . interestingly , the slope of the core - energy dependence on macroscopic strains is non - zero at zero strain . this suggests that the energetics of dislocation - cores play a non - trivial role in governing the behavior of dislocations even at small strains and at macroscopic scales . in order to elucidate the role of dislocation core - energy in influencing dislocation behavior , we consider the force on a unit length of dislocation line segment resulting from external loads or other defects . the classical force on the dislocation arising from elastic interactions is given by the peach - koehler force @xcite : @xmath143 where @xmath144 denotes the stress tensor , @xmath5 denotes the burgers vector and @xmath145 denotes the unit vector along the dislocation line . however , since the core - energy is dependent on macroscopic strains at the dislocation core , there is an additional configurational force associated with this dependence , referred to as the core - force , and is given by @xmath146 given the non - linear nature of the core - energy dependence on macroscopic strains , @xmath147 , in general , is a function of the macroscopic strain at the dislocation core . however , for typical strains of @xmath148 occurring in most deformations in solids , @xmath147 can be computed , to leading order , from the slopes of the core - energy dependence on various uniaxial and shear strains at zero strain ( cf . figures [ fig : uniaxial1]-[fig : uniaxial3 ] , [ fig : shear13]-[fig : shear23 ] ) , i.e. , @xmath149 . it is interesting to note that , while the peach - koehler force depends on the stress - tensor ( which is related to the strain - tensor ) , the core - force depends on the gradient of strain - tensor and can be significant in regions of inhomogeneous deformations . we note that past studies ( cf . e.g. @xcite ) , employing elastic formulations to model the defect - core using dislocation dipoles and line forces , have investigated the interaction of dislocation cores with applied pressure and have suggested corrections to the peach - koehler force based on these interactions . however , a direct quantification of this core - force solely from electronic - structure calculations has been beyond reach heretofore . in order to understand the implications of the core - energy dependence on macroscopic strains on the behavior of dislocations , we consider the interaction between two dislocations . in particular , we consider a dislocation dipole , with one dislocation ( dislocation @xmath150 ) along the @xmath151axis located at @xmath152 with a burgers vector ( @xmath5 ) of @xmath153 $ ] and another dislocation ( dislocation @xmath154 ) , also along @xmath151axis , located at @xmath155 with a burgers vector of @xmath156 $ ] ( @xmath157 ) . the peach - koehler force @xcite on a unit length of dislocation @xmath154 due to the elastic fields from dislocation @xmath150 is given by @xmath158 where @xmath159 denotes the stress tensor associated with the elastic fields produced by dislocation @xmath150 at the location of dislocation @xmath154 , and @xmath160 denotes the unit vector along the dislocation line of @xmath154 ( unit vector along @xmath151axis ) . we note that the peach - koehler force ( @xmath161 ) decays as @xmath162 , where @xmath163 denotes the distance between the two dislocations , which follows from the asymptotic behavior of the elastic field of a dislocation . on the other hand , the core - force force resulting from the macroscopic strain dependence of the core - energy is given by @xmath164 where @xmath165 denotes the core - energy of dislocation a in the elastic field of b ( @xmath166 ) , and @xmath167 denotes the core - energy of dislocation b in the elastic field of a ( @xmath168 ) . for the dipole configuration , we note that @xmath169 @xcite , and , from symmetry , @xmath170 . thus , @xmath171 noting that @xmath172 is @xmath173 , the core - force for dipole interactions decays as @xmath173 . thus , this is a shorter - ranged force in dislocation - dislocation interactions compared to the peach - koehler force . to the glide - plane.,scaledwidth=100.0% ] in order to further investigate the quantitative aspects of this core - force and its implications on dislocation - dislocations interactions , we consider two specific cases as shown in figure [ fig : casestudy ] . in the first case , the dislocations are aligned along the glide - plane , whereas in the second case they are at an angle of @xmath9 . for these two cases , we computed the peach - koehler force and the core - force on dislocation b. _ case ( i ) : _ for the first case , using the stress tensor associated with the elastic fields of a stationary edge dislocation @xcite , the peach - koehler force is given by @xmath174 where @xmath175 , @xmath176 , @xmath177 denote unit vectors along @xmath56 axis , and @xmath178 . the corresponding core - force , using the strain field of a stationary edge dislocation , is given by @xmath179 the second equality in the above expression for core - force uses the fact @xmath180 ( from the strain field of a stationary edge dislocation ) , and @xmath181 , from symmetry , for an edge dislocation . it is interesting to note that , for this case , while the peach - koehler force is a glide force on the dislocation , the core - force is a climb force . _ case ( ii ) : _ for the second case , the peach - koehler force and the core - force are given by @xmath182 @xmath183 the above expression for core - force uses the fact @xmath180 ( from the strain field of a stationary edge dislocation ) , and @xmath181 from symmetry . it is interesting to note that , for this case , while the peach - koehler force is a climb force on the dislocation , the core - force has both glide and climb components . we next investigate the relative importance of the core - force in governing the dislocation behavior . to this end , we estimated the glide component of the core - force for _ case ( ii ) _ from equation , and compared it to the peierls - nabarro force the critical force that will cause dislocation glide . for these calculations , we used the computed elastic constants @xmath184 . using @xmath185 for the peierls stress @xcite , we computed the peierls - nabarro force ( from equation [ eqn : pk ] ) on a @xmath186 length of dislocation line to be @xmath187 @xmath8 . using @xmath188 and @xmath189 , estimated from the data in figures [ fig : uniaxial1 ] and [ fig : uniaxial2 ] , the glide component of the core - force from case ( ii ) on a @xmath186 length of dislocation line is computed to be @xmath190 @xmath8 where @xmath163 is the distance between the dislocations @xmath150 and @xmath154 measured in angstroms . it is interesting to note that even at a distance of @xmath10 , the magnitude of core - force is larger than the magnitude of peierls - nabarro force . these results suggest that the short - ranged core - force , resulting from the dependence of core - energy on macroscopic deformations , is significant up to tens of nanometers and can play an important role in governing the behavior of dislocations , especially in regions of inhomogeneous deformations . in the present work , we employed a real - space formulation of orbital - free dft with finite - element discretization to study an edge dislocation in aluminum . the local real - space formulation and the finite - element discretization are crucial in accurately quantifying the energetics of an isolated dislocation , which is otherwise not accessible using a fourier - space based formulation . in this study , we computed the core - size of an edge dislocation by identifying the region where the perturbations in the electronic - structure arising from the defect - core , which can not be accounted for in continuum theories , are significant and have a non - trivial contribution to the dislocation energy . this allows us to unambiguously identify the core - size of an isolated dislocation from the viewpoint of energetics . we estimated the core - size of an edge dislocation in aluminum to be about @xmath4 , which is much larger than conventional core - size estimates of @xmath83 . the relaxed core - structure showed two shockley partials with a partial separation distance of @xmath191 that is consistent with prior electronic - structure studies of an edge dislocation in aluminum . the core - energy energy of relaxed shockely partials corresponding to a core - size of @xmath4per unit length of dislocation is computed to be @xmath192 . we subsequently investigated the effect of macroscopically applied deformations on the core - energy and core - structure of an isolated edge dislocation . interestingly , our studies suggest that the dislocation core - energy is significantly influenced by the applied macroscopic deformations . this core - energy dependence on macroscopic deformations results in an additional configurational force on the dislocation , beyond the peach - koehler force , and is referred to as _ core - force _ in this work . this core - force is proportional to the gradient of the strain - tensor and can be significant in regions of inhomogeneous deformations . we quantified this core - force for a dislocation dipole for two specific configurations of the dipole : ( i ) dislocations aligned along the slip plane ; ( ii ) dislocations at a @xmath193 angle to the slip plane . interestingly , for the first configuration , where the peach - kohler force is a glide force on the dislocation , the core - force is a climb force on the dislocation . for the second configuration , while the peach - koehler force is a climb force on the dislocation , the core - force has both glide and climb components on the dislocation . we estimated the glide component of the core - force for the second configuration and compared it to the peierls - nabarro force , which is the critical force required for dislocation glide . surprisingly , even up to @xmath10 the glide component of the core - force was greater than the peierls - nabarro force , suggesting that the core - force can play an important role in governing dislocation behavior , possibly even at mesoscopic scales . investigations on the influence of macroscopic deformations on the core - structure revealed that the separation between the shockley partials is only marginally influenced by uniaxial , biaxial and triaxial strains . however , the core - structure is significantly influenced by the shear strain in the plane containing the dislocation line and the direction normal to the slip plane . this is expected as the resulting shear stress exerts a glide force on the screw - component of the partials , subsequently leading to a significant change in the partial separation . the present work has provided the framework for accurately quantifying the energetics of isolated dislocations using electronic - structure calculations . while the present study focussed on an edge dislocation in aluminum , concurrent efforts are also focussed on studying a screw dislocation . further , ongoing and future efforts are also aimed at extending the computational methods to enable similar studies using kohn - sham dft , which is an important step towards considering materials systems where available orbital - free kinetic energy functionals do not offer the desired accuracy . investigating the role of core - force in governing the mechanical response at the mesoscopic scale , possibly using dislocation dynamics simulations , also presents an important direction for future research . we thank dr . roman gr@xmath194ger for assistance with the differential displacement plots of shockley partials . we are grateful to the support of national science foundation ( grant number cmmi0927478 ) under the auspices of which part of the work was conducted . the study on the effect of macroscopic deformations on dislocation energetics was conducted as part of research supported by the us department of energy , office of basic energy sciences , division of materials science and engineering under award number de - sc0008637 that funds the predictive integrated structural materials science ( prisms ) center at university of michigan . v. g. gratefully acknowledges the support of alexander von humboldt foundation through a research fellowship , and is grateful to the hospitality of the max - planck institute for mathematics in sciences while completing this work . banerjee , s. , ghoniem , n. , lu , g. , kioussis , n. , 2007 . non - singular descriptions of dislocation cores : a hybrid ab initio continuum approach . phil . mag . 87 , 4131 - 4150 . blase , x. , lin , k. , canning , a. , louie , s.g . , chrzan , d.c . , 2000 . structure and energy of the @xmath195 partial dislocation in diamond : a combined _ ab initio _ and elasticity theory analysis . 84 , 5780 - 5783 . bulatov , v.v . , yip , s. , argon , a.s . , 1995 . atomic modes of dislocation mobility in silicon . philos . mag . a 72 , 453 - 496 . carling , k.m . , carter , e.a . , 2003 . orbital - free density functional theory calculations of the properties of al , mg and al - mg crystalline phases . model . mater . sc . choly , n. , kaxiras , e. , 2002 . kinetic energy functionals for non periodic systems . solid state commun . 121 , 281 - 286 . clouet , e. 2011 . dislocation core field . i. modeling in anisotropic linear elasticity theory . phys . rev . b 84 , 224111 . clouet , e. , ventelon , l. , willaime , f. , 2009 . dislocation core energies and core fields from first principles . 102 , 055502 . fleck , n.a . , muller , g.m . , ashby , m.f . , hutchinson , j.w . , strain gradient plasticity : theory and experiment . acta metallurgica et materialia 42 , 475 - 487 . frederiksen , s.l . , jacobsen , k.w . density functional theory studies of screw dislocation core structures in bcc metals . phil . mag . 83 , 365 - 375 . garcia - cervera , c.j . , 2007 . an efficient real space method for orbital - free density functional theory . phys . 2 , 334 - 357 . gavini , v. , 2008 . role of macroscopic deformations in energetics of vacancies in aluminum . lett . 101 , 205503 . gavini , v. , 2009 . role of the defect core in energetics of vacancies . soc . a 465 , 3239 - 3266 . gavini , v. , bhattacharya , k. , ortiz , m. , 2007a . quasi - continuum orbital - free density - functional theory : a route to multi - million atom non - periodic dft calculation . solids 55 , 697 - 718 . gavini , v. , bhattacharya , k. , ortiz , m. , 2007b . vacancy clustering and prismatic dislocation loop formation in aluminum . phys . rev . b 76 , 180101(r ) . gavini , v. , knap , j. , bhattacharya , k. , ortiz , m. , 2007c . non - periodic finite - element formulation of orbital - free density - functional theory . j. mech . solids 55 , 669 - 696 . gavini , v. , liu , l. , 2011 . a homogenization analysis of the field theoretical approach to the quasi - continuum method . j. mech . phys . solids 59 , 1506 - 1535 . gehlen , p.c . , hirth , j.p . , hoagland , r.g . , kanninen , m.f . , 1972 . a new representation of the strain field associated with the cube - edge dislocation in a model of a @xmath28-iron . j. appl 43 , 3921 - 3933 . ghoniem , n.m . , tong , s .- h . , sun , l.z . parametric dislocation dynamics : a thermodynamics - based approach to investigations of mesoscopic plastic deformation . phys . rev . b 61 , 913 - 927 . goodwin , l. , needs , r.j . , heine , v. , 1990 . a pseudopotential total energy study of impurity - promoted intergranular embrittlement . j. phys : condens . matter 2 , 351 - 365 . greer , j.r . , nix , w.d . , 2006 . nanoscale gold pillars strengthened through dislocation starvation . phys . rev . b 73 , 245410 . gumbsch , p. , daw , m.s . , interface stresses and their effects on the elastic moduli of metallic multilayers . phys . rev . b 44 , 3934 - 3938 . gumbsch , p. , gao , h. , 1999 . dislocations faster than the speed of sound . science 283 , 965 - 968 . hansen , n. , 2004 . hall - petch relation and boundary strengthening . scripta mater . 51 , 801 - 806 . henager , c.h . , hoagland , r.g . dislocation and stacking fault core fields in fcc metals . phil . mag . 85 , 4477 - 4508 . hllerbauer , w. , karnthaler , h.p . , 1981 . hirth , j.p . , lothe , j. , 1968 . theory of dislocations , mcgraw - hill , new york . ho , g. , ong , m.t . , caspersen , k.j . , carter , e.a . , 2007 . energetics and kinetics of vacancy diffusion and aggregation in shocked aluminum via orbital - free density functional theory . chem . phys . 9 , 4951 - 4966 . huang , c. , carter , e.a . transferable local pseudopotentials for magnesium , aluminum and silicon . 10 , 7109 - 7120 . hung , l. , huang , c. , shin , i. , ho , g. , ligneres , v.l . , carter , e.a . 2010 . introducing profess 2.0 : a parallelized , fully linear scaling program for orbital - free density functional theory calculations . comm . , 181 , 2208 . ismail - beigi , s. , arias , t.a . , _ ab initio _ study of screw dislocations in mo and ta : a new picture of plasticity in bcc transition metals . 84 , 1499 - 1502 . iyer , m. , pollock , t.m . , gavini , v. , 2014 . energetics and nucleation of point defects in aluminum under extreme tensile hydrostatic stresses . b 89 , 014108 ( 2014 ) . kabir , m. , lau , t.t . , rodney , d. , yip , s. , van vliet , k.j . , 2010 . predicting dislocation climb and creep from explicit atomistic details . 105 , 095501 . kelchner , c.l . , plimpton , s.j . , hamilton , j.c . , 1998 . dislocation nucleation and defect structure during surface indentation . phys . rev . b 58 , 11085 - 11088 . leyson , g.p.m . , curtin , w.a . , hector , l.g . , woodward , c.f . , 2010 . quantitative prediction of solute strengthening in aluminium alloys . nat . mater . 9 , 750 - 755 . li , j. , wang , c .- z . , chang , j .- p . , cai , w. , bulatov , v.v . , ho , k .- , yip , s. , 2004 . core energy and peierls stress of a screw dislocation in bcc molybdenum : a periodic - cell tight - binding study . b. 70 , 104113 . li , j. , van vliet , k.j . , zhu , t. , yip , s. , suresh , s. , 2002 . atomistic mechanisms governing elastic limit and incipient plasticity in crystals . nature 418 , 307 - 310 . lu , g. , kaxiras , e. , 2002 . can vacancies lubricate dislocation motion in aluminum ? 89 , 105501 . lu , g. , kioussis , n. , bulatov , v.v . , kaxiras , e. , 2000 . generalized stacking - fault energy surface and dislocation properties of aluminum . phys . rev . b 62 , 3099 - 3108 . marian , j. , cai , w. , bulatov , v.v . dynamic transitions from smooth to rough to twinning in dislocation motion . 3 , 158 - 163 . marian , j. , knap , j. , ortiz , m. , 2004b . nanovoid cavitation by dislocation emission in aluminum . 93 , 165503 . martin , r.m . electronic structure : basic theory and practical methods , cambridge university press , cambridge . meyers , m.a . , mishra , a. , benson , d.j . mechanical properties of nanocrystalline materials . 51 , 427 - 556 . motamarri , p. , iyer , m. , knap , j. , gavini , v. , 2012 . higher - order adaptive finite - element methods for orbital - free density functional theory . 231 , 6596 - 6621 . motamarri , p. , nowak , m.r . , leiter , k. , knap , j. , gavini , v. , 2013 . higher - order adaptive finite - element methods for kohn - sham density functional theory . 253 , 308 - 343 . motamarri , p. , gavini , v. , 2014 . a subquadratic - scaling subspace projection method for large - scale kohn - sham dft calculations using spectral finite - element discretization . b 90 , 115127 . parr , r. , yang , w. , 2003 . density - functional theory of atoms and molecules , oxford university press , oxford . peach , m. , koehler , j.s . the forces exerted on dislocations and the stress fields produced by them . phys . rev . 80 , 436 - 439 . peierls , r. , 1940 . the size of a dislocation . proceedings of the physical society 52 , 34 . perdew , j.p . , zunger , a. , 1981 . self - interaction correction to density - functional approximations for many - electron systems . b 23 , 5048 . pizzagalli , l. , demenet , j .- l . , rabier , j. , 2009 . theoretical study of pressure effect on the dislocation core properties in semiconductors . phys . rev . b 79 , 045203 . pollock , t.m . , argon , a.s . creep resistance of cmsx-3 nickel base superalloy single crystals . acta metallurgica et materialia 40 , 1 - 30 . radhakrishnan , b. , gavini , v. , 2010 . effect of cell size on the energetics of vacancies in aluminum studied via orbital - free density functional theory . b 82 , 094117 . rice , j.r . , dislocation nucleation from a crack tip : an analysis based on the peierls concept . solids 40 , 239 - 271 . shin , i. , carter , e.a . , 2012 . orbital - free density functional theory simulations of dislocations in magnesium . simul . mater . eng . , 20 , 015006 . shin , i. , carter , e.a . , 2013 . possible origin of the discrepancy in peierls stresses of fcc metals : first - principles simulations of dislocation mobility in aluminum . b. 88 , 064106 . shin , i. , ramasubramaniam , a. , huang , c. , hung , l. , carter , e. a. , 2009 . orbital - free density functional theory simulations of dislocations in aluminum . phil . mag . 89 , 3195 - 3213 . sinclair , j.e . , gehlen , p.c . , hoagland , r.g . , hirth , j.p , 1978 . flexible boundary conditions and nonlinear geometric effects in atomic dislocation modeling . j. appl . phys . 49 , 3890 - 3897 . suryanarayana , p. , phanish , d. , 2014 . augmented lagrangian formulation of orbital - free density functional theory . arxiv:1405.6456 . tadmor , e.b . , ortiz , m. , phillips , r. , 1996 . quasicontinuum analysis of defects in solids . a 73 , 1529 - 1563 . trinkle , d.r . , woodward , c. 2005 . the chemistry of deformation : how solutes soften pure metals . science 310 , 1665 - 1667 . trinkle , d.r . , 2008 . lattice green function for extended defect calculations : computation and error estimation with long - range forces . phys . rev . b 78 , 014110 . wang , y.a . , govind , n. , carter , e.a . , 1999 . orbital - free kinetic - energy density functionals with a density - dependent kernel . b 60 , 16350 . weinberger , c.r . , cai , w. , 2008 . surface - controlled dislocation multiplication in metal micropillars . 105 , 14304 - 14307 . woodward , c. , trinkle , d.r . , hector , l.g . , olmsted , d.l . prediction of dislocation cores in aluminum from density functional theory . 100 , 045507 .

we apply a three - dimensional ( 3d ) implementation of the time - dependent restricted - active - space self - consistent - field ( td - rasscf ) method to investigate effects of electron correlation in the ground state of be as well as in its photoionization dynamics by short xuv pulses , including time - delay in photoionization . first , we obtain the ground state by propagation in imaginary time . we show that the flexibility of the td - rasscf on the choice of the active orbital space makes it possible to consider only relevant active space orbitals , facilitating the convergence to the ground state compared to the multiconfigurational time - dependent hartree - fock method , used as a benchmark to show the accuracy and efficiency of td - rasscf . second , we solve the equations of motion to compute photoelectron spectra of be after interacting with a short linearly polarized xuv laser pulse . we compare the spectra for different ras schemes , and in this way we identify the orbital spaces that are relevant for an accurate description of the photoelectron spectra . finally , we investigate the effects of electron correlation on the magnitude of the relative time - delay in the photoionization process into two different ionic channels . one channel , the ground state channel in the ion , is accessible without electron correlation . the other channel is only accessible when including electron correlation . the time - delay is highly sensitivity to the choice of the active space , and hence to the account of electron - electron correlation .

1703.06022

the development of short and intense laser pulses has opened the possibility to control and observe electronic and nuclear motion on the attosecond time scale @xcite . the new light sources are key to probe and manipulate the electronic structure and dynamics in many - electron atoms @xcite and molecules @xcite . for instance , few photon ionization using short pulses results in ejected electrons whose spatial distribution depends on the energy spectrum and angular momentum of the initial state as well as the ionic channels involved @xcite . likewise it has become clear that accurate modelling of time - delays in photoionization require careful consideration of electron - correlation effects ( see , e.g. , refs . @xcite and the recent review ref . accordingly , there is a range of processes that require explicit time - dependent methods capable of treating electron correlations beyond mean - field and single - active electron approximations . in the case of he , it is possible to solve numerically the time - dependent schrdinger equation ( tdse ) ( see , e.g. , refs . @xcite ) . for larger systems , and for photon energies that only affect the valence shell , one can consider two electrons immersed in the mean - field potential produced by the inner electrons @xcite . in general , however , in order to consider more than two electrons in an atom and to make the tdse tractable it is mandatory to use approximations to the wave function and hence the tdse . to this end , the time - dependent configuration - interaction ( td - ci ) method consists of expanding the many - electron wave function as @xmath0 where @xmath1 denotes the full ci hilbert space of all accessible configurations . it is the finite size of @xmath1 that introduces approximations . in eq . , @xmath2 is the time - dependent coefficient of the configuration @xmath3 , formed by a set of spin - orbitals . in this formalism , the tdse corresponds to a set of first - order differential equations for @xmath2 . the description of the continuum , however , demands a large number of orbitals and configurations , which makes the ci approach numerically intractable with increasing number of electrons . to overcome this limitation , there are several methods such as the time - dependent configuration - interaction singles ( td - cis ) @xcite , the time - dependent restricted - active - space configuration - interaction ( td - ras - ci ) @xcite and the time - dependent generalized - active - space configuration - interaction ( td - gas - ci ) method @xcite which impose restrictions on the allowed excitations and the active orbital spaces . these approximations still require many configurations in addition to a careful design of the partitions in the active space . it is also very challenging to extend those methodologies to situations with more than a single electron in the continuum , as is also the case with the @xmath4-matrix method @xcite although some progress has been reported in that direction @xcite . despite the difficulties , it is nevertheless attractive to explore a wave function approach , because of , e.g. , the unambiguous extraction of observables . the theory should ideally reduce as much as possible the number of orbitals needed for an accurate description of the configuration space . a breakthrough along those lines came with the multiconfiguration time - dependent hartree ( mctdh ) method @xcite and the mctdh - fock ( mctdhf ) method ( see , e.g. , refs . @xcite ) , where time - dependent spin - orbitals are introduced in the ansatz , making @xmath5 time - dependent @xmath6 the main advantage of this method is that the use of time - dependent spin - orbitals makes it possible to describe the wave function and , in particular , the continuum with a smaller number of orbitals and configurations than with the time - independent orbitals used in the ci approach corresponding to eq . . the mctdhf method has , e.g. , been applied to describe high - harmonic generation ( hhg ) in low dimensions @xcite , polarization of the continuum @xcite and to calculate cross sections of atomic @xcite and molecular systems @xcite . since the spatial orbitals are time - dependent , the one- and two - body operators must be updated at each time step , leading to a high numerical cost , especially in the 3d case . moreover , as in the ci picture , the use of many orbitals may still imply an intractable number of accessible configurations . there are several strategies to diminish the numerical effort without loss of accuracy , such as the time - dependent occupation - restricted multiple - active - space ( td - ormas ) @xcite , the time - dependent complete - active - space self - consistent - field ( td - casscf ) @xcite and the time - dependent restricted - active - space self - consistent - field ( td - rasscf ) @xcite methods . in particular , the td - rasscf method benefits from the ras to diminish the accessible configurations and hence allows the consideration of only a subset @xmath7 of the configurations in @xmath8 by dividing the active space into two or more parts . among these parts electron excitations take place with certain restrictions @xcite , specified at will , and most often chosen by physical insight into the problem at hand . in the present work we apply the td - rasscf method with double excitations ( td - rasscf - d ) to address the role of electron correlation in be in the ground state , in photoionization spectra induced by short xuv pulses and in time - delay studies . the td - rasscf - d method was previously shown to be accurate and computationally efficient in 1d cases @xcite , and the td - rasscf theory was recently extended with a space partitioning concept @xcite . the main advantage of the approach resides in the possibility of selecting an appropriate ras which captures the most important configurations for a given system and physical process . for a very recent discussion of these aspects in the case of cold atomic bosons see ref . @xcite . the flexibility in choosing the ras is remarkable in the imaginary time propagation ( itp ) , where we show that the active space concept facilitates the convergence . by identifying the most important active orbitals , the td - rasscf - d method can be as accurate as the mctdhf method , but with a smaller number of configurations highlighting the important role of double excitations for the ground state . the application of this method to 3d systems is , however , still computationally challenging due to the large number of non - zero matrix elements of the two - body operator . to overcome this issue , we develop and use the _ coupled basis _ method to diminish the number of operations required to evaluate the two - body operator . in addition to the itp studies for the ground state , we present studies of be subject to short linearly polarized xuv laser pulses . in general the photoelectron energy spectra ( pes ) reveal detailed properties of atomic , molecular or solid targets , including energies , structural and symmetrical properties of the states of the system . for instance , the directional distribution of the ejected electrons depends on the angular momentum of the final and initial states and the number of absorbed photons @xcite . as for the ground state studies , the present td - rasscf - d method allows the identification of the most important active orbitals for an accurate description of the photoelectron spectra . furthermore , the electronic dynamics of the remaining ion during the photoionization has an impact on the outgoing electrons , enclosed , in particular , in the apparent time of ionization @xcite . the experimental and theoretical computation of this quantity is very sensitive to the method , and constitutes a fundamental probe of the many - body dynamics @xcite . in this work we investigate the relative time - delay in ionization between the channel ( i ) be[@xmath9)@xmath10s@xmath11 @xmath12 [ be@xmath13(@xmath14 ) + @xmath15 ] @xmath10p@xmath16 , and the channel ( ii ) be[@xmath9)@xmath10s@xmath11 @xmath12 [ be@xmath13(@xmath17 ) + @xmath18 ] @xmath10p@xmath16 with @xmath19 . as is evident from dipole selection rules , process ( i ) is possible without electron - electron correlation , while in process ( ii ) ionization is accompanied by a shakeup in the ion and requires electron - electron correlation . the study shows that an accurate description of time - delays in be requires a treatment beyond the mean - field time - dependent hartree - fock theory , and identifies the ras scheme that provides an accurate value for the delay . the paper is organized as follows . in sec . [ sec : the_system_and_the_method ] we describe the hamiltonian , the td - rasscf - d method and the numerical techniques used , including the coupled basis method . in sec . [ sec : results ] we present the results for be , both concerning the ground state studies ( sec . [ sec : results].a ) , the pes ( sec . [ sec : results].b ) and the relative time - delays in ionization ( sec . [ sec : results].c ) . [ sec : conclusions_and_outlook ] summarizes the main findings and concludes . atomic units are used throughout unless indicated otherwise . the td - rasscf theory for fermions was presented in refs . @xcite , and the similarities and differences between the fermionic and bosonic cases were discussed in ref . @xcite , so the presentation here will be brief . we consider an @xmath20 electron atom with a nuclear charge @xmath21 interacting with a linearly polarized laser pulse of short duration . we take the spin - restricted ansatz of the many - electron wave function as @xmath22 with the configurations @xmath23 constructed from @xmath24 spatial orbitals which can each hold a spin up and a spin down electron , i.e. , @xmath25 time - dependent spin - orbitals , @xmath26 , which form an orthonormal basis in the @xmath27 space , with the @xmath28 space being the rest of the single - particle hilbert space in fig . [ fig : fig1 ] . restricted - active - space ( ras ) associated with the present considerations for the be atom . the ras is divided in two partitions : @xmath29 which is formed by @xmath30 spatial orbitals ( @xmath31 ) and @xmath32 by @xmath33 spatial orbitals ( @xmath34 ) . the figure illustrates ( a ) all the electrons in the lowest - energy configuration and ( b ) an example of a doubly - excited configuration . the indexes to the left indicate the notation used to label the orbitals in the different spaces . the formalism also allows for a core with always occupied time - dependent orbitals corresponding to a @xmath35 space @xcite . in this work we do not invoke the @xmath35 space and therefore @xmath35 is not shown in the figure . ] the sum in eq . runs over all accessible configurations specified by the hilbert space with restrictions on the active space , @xmath7 . in the mctdhf method @xcite the active space , formed by the @xmath24 time - dependent spatial orbitals , constitutes the full @xmath27 space , i.e. , the configurations considered in mctdhf are all the combinations of orbitals allowed by the pauli principle . in contrast , in the td - rasscf approach we divide the active space in @xmath29 and @xmath32 , which contain @xmath30 and @xmath33 spatial orbitals , respectively , as shown in fig . [ fig : fig1 ] . the ras concept can be extended by a @xmath35 space describing a core where the time - dependent orbitals are always occupied . in this paper we consider td - rasscf - d without a core , that is , we allow double - electron excitations from @xmath29 to @xmath32 . the td - rasscf - d method is numerically efficient and accurately accounts for the two - body interactions . these properties are related to brillouin s theorem , which establishes that a single excitation in a self - consistent method does not improve the hf solution @xcite . also note that configurations with double excitations with time - dependent orbitals include non - vanishing projections on the singles space spanned by time - independent orbitals . to apply the td - rasscf - d method , it is convenient to write the hamiltonian in second quantization @xmath36 where @xmath37 and @xmath38 are fermionic annihilation and creation operators of an electron in the time - dependent spin - orbital @xmath39 . we use the notation @xmath40 for all the orbitals and @xmath41 for the orbitals in @xmath27 space , where we specify by single ( @xmath42 ) or double prime ( @xmath43 ) that they belong to different partitions if necessary . in eq . the matrix elements are given by @xmath44 where @xmath45 denotes space and spin degrees of freedom , @xmath46 is the mean - field operator @xmath47 and @xmath48 is the one - body hamiltonian in the length gauge @xmath49 with @xmath21 the nuclear charge , @xmath50 the magnitude of the position vector , and @xmath51 the electric field of the laser pulse . we now briefly describe the derivation of the equations of motion . the td variational principle @xcite establishes that the best approximation to the time - dependent schrdinger equation ( tdse ) is a stationary point of the functional @xmath52&= & \int_0^t \left [ \langle\psi(t)|\left(i\cfrac{\partial}{\partial t}-h\right)|\psi(t)\rangle \right . \\ & + & \left . \sum_{ij}\epsilon_j^i(t)\left ( \langle\phi_i(t)|\phi_j(t)\rangle-\delta_{ij } \right ) \right]\mathrm{d}t , \end{aligned}\ ] ] where the lagrange multipliers @xmath53 ensure that the orbitals in the @xmath27 space are orthonormal during the time interval @xmath54 $ ] . to simplify the notation , we do not indicate the time , space or spin dependence of the amplitudes and the orbitals . then , we seek stationary points , @xmath55 , and in the td - rasscf - d case the stationary points fulfil the equations @xcite @xmath56 @xmath57 @xmath58 with @xmath59 , @xmath60 @xmath61 , @xmath62 and @xmath63|\psi\rangle$ ] . the amplitude equation describes the time evolution of the expansion coefficients @xmath2 of eq . , and the orbital equations - describe the projection of the time derivative of the orbitals on the @xmath28 and @xmath27 spaces , respectively . the eqs . - constitute a set of coupled non - linear differential equations and the strategy to solve them is the following : i ) first , we use eq . to solve for @xmath64 , ii ) then , we can solve the amplitudes and the @xmath28 space eqs . ( see the discussion in ref . @xcite ) . on the one hand , let us consider the case of mctdhf , that is , if we take into account all the combinations of the orbitals to construct the slater determinants , then eq . becomes an identity @xcite . therefore , in this case we only need to solve eqs . and , and we can set @xmath65 to be any antihermitian matrix @xcite and in practice often @xmath66 is used . on the other hand , for the td - rasscf - d method including only double excitations @xcite , the constraints on the active space imply that we only have the freedom to choose a fixed value for @xmath67 if both the orbitals @xmath68 and @xmath69 belong to either @xmath29 or @xmath32 . the calculation of @xmath70 for @xmath71 and @xmath72 belonging to different partitions implies solving eq . , together with eqs . and . the main advantage of the td - rasscf - d method compared with mctdhf is that it is possible to include only the relevant configurations for a given physical observable or process . the numerical effort is therefore smaller for the same number of orbitals @xcite , a point we will discuss further below . this method shares with the mctdhf the numerical instability in the @xmath28 space equation induced by the singularity of the inverse of the reduced one - body density matrix , @xmath73 . in particular , a system with low entanglement would require only a small number of orbitals , and the addition of more leads to small eigenvalues of the one - body density matrix @xcite . to address this problem , the one - body density matrix is usually regularized @xmath74 , where @xmath75 is a regularization parameter @xcite . in this work we set @xmath76 . here we discuss the single - orbital basis and for completeness we collect the formulas needed for its construction . to describe the orbitals of the many - electron wave function , we use a basis set expansion in a finite element discrete variable representation ( fe - dvr ) for the radial grid @xcite and spherical harmonics for the angular part , i.e. , @xmath77 where @xmath78 are the coefficients of the expansion , @xmath79 are the fe - dvr functions and @xmath80 are the spherical harmonics . to define the fe - dvr functions we divide the radial grid @xmath81 $ ] in @xmath82 finite elements with @xmath83 nodes in each element . then , the collective subscript @xmath84 in eq . denotes the node @xmath85 in the element @xmath86 . the fe - dvr function is defined using gauss - lobatto functions @xcite @xmath87 where @xmath88 are the nodes and @xmath89 the weights of the associated gauss - lobatto quadrature . the fe - dvr functions have the following properties @xcite @xmath90 we obtain the radial part of the kinetic energy term by calculating the matrix element of the second derivative in the radial coordinate @xcite @xmath91 where @xcite @xmath92 we introduce the bridge functions @xcite @xmath93 which ensure the continuity of the functions between adjacent elements , and fulfil eqs . - . the matrix elements of the two - body operator need to be updated at each time step and this update is the main bottleneck of time - dependent scf methods . the two - body electron - electron interaction reads in the multipole expansion @xmath94 where @xmath95 ( @xmath96 ) is the largest ( smallest ) between @xmath97 . the electron - electron coulomb repulsion is not diagonal in the angular coordinates , because , in the evaluation of the matrix elements , each spherical harmonic in eq . couples with two angular functions coming from the product of two orbitals . this leads to numerous non - vanishing matrix elements which require both large memory for storage and long cpu time for updating the orbitals . there are several proposals to overcome these issues . on the one hand , we can reduce the number of applications of the two - body operator using the constant mean - field approximation @xcite or by restricting the number of operations by allowing only a small number of @xmath98 in the single - orbital expansion @xcite , or both . on the other hand , we can make the two - body operator sparse by expanding it in a pseudo - dvr basis @xcite for the angular grid . in this work , we apply what we call the _ coupled basis _ method . we use that the electron - electron interaction commutes with the total angular momentum and its projection on the @xmath99 axis for two electrons , i.e. , @xmath100=\left[(l_{z,1}+l_{z,2}),1/|\vec{r_1}-\vec{r_2}|\right]=0 $ ] . this implies that @xmath101 conserves the coupled @xmath102 in a basis of two - orbital functions in the coupled representation . at each time step , we therefore transform the product of two orbitals into a two - orbital function in the coupled representation @xmath103 where we used the fe - dvr property that @xmath104 , and where we defined @xmath105 with @xmath106 the gaunt coefficients @xmath107 from eq . we now see that the two - body matrix elements in the coupled basis form a block diagonal matrix in @xmath108 . using this property , the mean - field operator in eq . reads as @xmath109 where @xmath110 . it is useful to rewrite eq . in the fe - dvr basis . after some algebra we find @xmath111 with @xmath112 where @xmath113 is approximated by @xcite @xmath114^{-1}+\cfrac{r_\alpha^l r_\gamma^l}{r_\text{max}^{2l+1}}\right),\ ] ] where @xmath115 is twice the kinetic energy matrix . now , we can obtain the terms needed to solve eqs . and , such as the application of the mean - field operator on a single spatial orbital @xmath116 similarly we obtain for the two - body operator @xmath117 to highlight the benefits from using the coupled basis method , we compute the number of operations performed at each update . for simplicity , we assume that @xmath118 and take into account all possible combinations of spherical harmonics , even if the associated clebsch - gordan coefficient is zero . first , the number of operations required for the transformation to the coupled basis is @xmath119 , where @xmath120 is the number of radial functions , @xmath121 and @xmath122 are the numbers of angular functions in the single - orbital and coupled basis , respectively . second , the evaluation of the mean - field operator needs @xmath123 operations , and the cost of its application to each orbital requires @xmath124 operations . finally , the calculation of the matrix element of the two - body operator requires @xmath125 operations . the most important achievement of this method is that the number of operations of the mean - field operator scales linearly with the number of angular functions , contrary to the numerical effort required if we use directly the product basis , where the scaling goes as @xmath126 . to describe the electronic structure accurately we require a dense grid in the radial coordinate around the nucleus , which in momentum space implies highly oscillating functions with large momenta . these functions introduce rapid oscillations in time which require a very small time step to be resolved and this stiffness leads to instabilities in eqs . - @xcite . in many cases , however , these high - energy states do not contribute to the dynamics of interests and they can therefore be removed by a suitable projection . the calculation of the eigenstates of the full hamiltonian is not possible in the present framework , and even worse , in the presence of the laser , the hamiltonian would have to be diagonalized at each time step . for these reasons , we consider only the high energy states of the one - body hamiltonian , which is often a good approximation @xcite . taking this consideration into account , we apply the energy subspace projection described in refs . first , we diagonalize the field - free one - body hamiltonian to obtain the high - energy one - body states @xmath127 . second , we define the projector @xmath128 to remove the contributions with eigenenergies @xmath129 from the time - dependent many - body hamiltonian , @xmath130 , obtaining the stiffness - free hamiltonian @xmath131 . the many - body nature of the problem makes it difficult to apply @xmath132 on the hamiltonian , therefore , we perform the projection on the wave function @xmath133 , which consists on projection out the high energy contribution from each orbital . the energy cutoff is set between 350 - 1500 a.u .. in this section we describe a procedure to obtain the photoelectron spectrum . often used methods consists of projecting on scattering waves , coulomb waves or plane waves @xcite . while the projection on scattering states can be applied immediately after the end of the pulse , the projection on coulomb or plane waves requires the propagation over some periods after the laser is switched off to allow the ejected electron to arrive at the outer region , where the interaction of the outgoing electron with the atom can either be neglected or approximated by the coulombic monopole term . in this work , we proceed as follows . first , we integrate the one - body density over @xmath134 electrons and second , we project on coulomb or plane waves in the outer region , as specified below , in the remaining coordinate . the photoelectron momentum distribution ( pmd ) reads @xmath135 where @xmath136 is defined just after eq . ( [ eq : p_space ] ) and @xmath137 is defined as @xmath138 where @xmath139 is an outgoing scattering wave function and @xmath140 is a window function introduced to remove boundary effects related to the outer region ( @xmath141 ) and the end of the box ( @xmath142 ) @xcite @xmath143 in the present work we use @xmath144 and @xmath145 . in the case of plane waves @xmath146 , we use the expansion in terms of the spherical bessel functions , @xmath147 @xmath148 to perform the integral in eq . . in the case of coulomb wave functions @xmath149 @xcite , we have @xmath150 where @xmath151 is the regular coulomb function , @xmath152 $ ] is the coulomb phase shift and @xmath153 . the photoelectron spectrum is obtained by integration over the angular coordinates in momentum space , @xmath154 @xmath155 and the triply differential energy and angular resolved probability is obtained as @xmath156 the validity of the projection on plane waves to describe the spectrum relies on assuming that the wave packet is far from the atom and that it is not affected by the atomic potential @xcite . this assumption is valid for the laser parameters and the range of the photoelectron spectra analyzed in this work , as we validated by comparison with the results obtained by projection on coulomb waves . in this section , we describe the impact of electron correlation on the ground state of be and the photoionization process , including time - delays , due to the interaction with a linearly polarized xuv laser pulse of short duration . we use be for these illustrative calculations because it allows for converged mctdhf reference data . in this way we can identify the most important part of the orbital space for a given physical observable . we describe the radial coordinate from @xmath157 to @xmath158 using 8 fes of length @xmath159 . from @xmath160 to the end of the box we add elements of length @xmath161 , with @xmath162 . to calculate the ground state we perform an imaginary - time propagation ( itp ) of an initial guess function in a box from @xmath163 . the dynamics due to the interaction with the laser is subsequently described performing a real - time propagation ( rtp ) , where we set the end of the box to @xmath164 by adding 43 elements of length 4 with 8 nodes in each . for the itp , the angular part of the orbitals is described with a maximum orbital angular momentum @xmath165 and magnetic quantum numbers @xmath166 for 7 or less orbitals and @xmath167 and @xmath168 otherwise . for the rtp , we use @xmath167 and @xmath168 in all the cases , and this is sufficient to obtain convergence for the xuv pulses considered in the present work . for the itp and rtp we use an adaptative runge - kutta propagator . the typical time step @xmath169 ranges from @xmath170 atomic units . .[tab : ground_state_energies ] ground state energies of be for several ras schemes . @xmath30 and @xmath33 denote the numbers of spatial orbitals in @xmath29 and @xmath32 ( fig . [ fig : fig1 ] ) , respectively . when all the orbitals are in @xmath29 ( @xmath171 and @xmath172 ) , the td - rasscf approach is equal to the mctdhf approach with @xmath173 orbitals . when , for be , @xmath174 , the td - rasscf approach is equal to the tdhf approach . the @xmath175 entry denoted by _ random _ is an example , where we use a random initial wave function . in the other cases the initial guess wave function was designed as described in the text . [ cols="^,^,^,^,^ " , ] to obtain the ground state we perform an itp for an appropriate initial guess function @xcite . the choice of the initial guess function is crucial , since there exists an infinite number of non - physical standing wave solutions which are local minima of the td variational principle @xcite . to ensure that the ground state obtained in the itp is not affected by the selection of the initial guess function we may choose to take an initial random wave function @xcite . the coefficients @xmath176 of the orbitals [ eq . ( [ eq : phip ] ) ] are chosen random , and the amplitudes @xmath177 are taken as @xmath178 , where @xmath179 is a random number @xmath180 and @xmath181 the number of configurations . this strategy , however , turns out to be problematic for many ras partitions , including the mctdhf case , due to the many local minima of the energy located in the manifold @xcite . then a more careful consideration of the design of the initial guess wave function becomes mandatory in order to reach a good approximation , ideally the global minimum , for the ground state @xcite . one of the great advantages of the td - rasscf method is that we can choose the partition in the active space which induces the most important slater determinants for a given physical process . in the 3d case this choice becomes very important , since an appropriate set of configurations can facilitate the couplings of the spherical harmonics to the @xmath182 ground state . for example , in the case of be , the main configuration for the ground state is @xmath183 , and an appropriate ras scheme would be two spatial orbitals in the @xmath29 space and the rest in @xmath32 . the orbitals in the @xmath29 space would be close to @xmath184-type orbitals , whereas the orbitals in @xmath32 would be linear combinations of a set of orbitals which together with the orbitals in @xmath29 can couple to @xmath182 . as an example we focus on the case of @xmath175 orbitals , included in table [ tab : ground_state_energies ] , where we show the ground state energy for different ras schemes . on the one hand , we use a random initial guess function for the mctdhf approach with @xmath175 orbitals and we obtain the energy @xmath185 ( the entry in table [ tab : ground_state_energies ] denoted by _ random _ ) . this value of the mctdhf energy is in agreement with the energy previously reported in ref . @xcite . on the other hand we design an initial guess function as follows : i ) we choose two orbitals as the @xmath186 and @xmath187 hydrogenic functions for @xmath188 , ii ) we set the amplitude of the slater determinant which contains them equal to @xmath189 , iii ) we choose the 3rd orbital randomly and iv ) we set the rest of the coefficients of the configurations to the same non - negative value to obtain a normalized wave function . in this case we obtain the lower @xmath175 mctdhf energy @xmath190 shown in table [ tab : ground_state_energies ] . we note that this design procedure is somewhat similar to first performing a hf calculation with @xmath184 electrons , and then adding a third orbital to perform a mcscf calculation ; a strategy often followed in quantum chemistry multi - configurational self - consistent - field calculations . in the case of td - rasscf - d , we impose the most important configurations by setting @xmath191 and @xmath192 and we obtain again @xmath193 now using a random initial guess function , obtained by choosing the coefficients and amplitudes as described in the beginning of this section . the difference between these two approximations to the ground state lies in the orbitals . for the ground state with energy @xmath185 , the angular part of all the orbitals turns out to be @xmath194 , that is , the many - electron wave function corresponds to a multiconfigurational hartree - fock ( mchf ) using @xmath195 orbitals . however , for @xmath193 and ( @xmath196 ) we obtain that the main contribution to the angular part of the two orbitals @xmath197 is spherically symmetric , while the third orbital @xmath198 is a linear combination of @xmath199 and @xmath200 , such that the expansion coefficients [ eq . ] fulfil @xmath201 . this condition ensures that the total magnetic quantum number of the ground state is vanishing , consistent with its @xmath182 term . note that the ground state energies of the ras @xmath202 and @xmath203 are the same because these two schemes are equivalent @xcite . this improvement in the ground state energy for @xmath175 orbitals manifests the importance of the mixing of the orbitals with different values of @xmath98 . if this mixing is not permitted , we would require at least 5 orbitals to improve the hf energy , i.e. , the @xmath204 orbitals , to guarantee the coupling to @xmath182 . let us remark that the ground state energy for mctdhf with 5 orbitals is lower than the energy obtained from time - independent multiconfigurational hartree fock calculations with 5 orbitals with fixed @xmath205 and @xmath98 @xcite , and that the @xmath206 and @xmath207 results , for which case comparison is possible , are in good agreement with mchf results @xcite . for a given number of orbitals @xmath24 , the ras scheme @xmath191 and @xmath208 , may be seen as an improvement to the tdhf solution by adding the possibility for double excitations . the wave function associated with this ras @xmath209 is formed by @xmath210 configurations , whereas the number of configurations of the mctdhf is @xmath211 . for example , for @xmath207 , the number of configurations required for this ras is @xmath212 , whereas it is 1296 for the mctdhf approach . the discrepancy in the energy is , however , only @xmath213 ( see table [ tab : ground_state_energies ] ) . because of the reduction in configurations , the results in table [ tab : ground_state_energies ] are obtained with the td - rasscf - d approach at a reduced computational cost in terms of cpu and memory compared to those obtained with the mctdhf approach . the number of operations scales with the sum of two major contributions : the calculation of the two - body operator and the integration of the amplitude equations @xcite . the numerical effort of the first one is discussed in detail in sec . the integration of the amplitude equations scales with the number of configurations , @xmath214 , as @xmath215 . we will come back to the reduction in computational cost in connection with the photoelectron spectra discussed below . in this section we illustrate the application of the td - rasscf - d method to photoelectron spectra ( pes ) of be after interacting with short linearly polarized xuv laser pulses . we consider pulses described by the vector potential @xmath216 , where the duration of the pulse is @xmath217 and the frequency bandwidth @xmath218 with @xmath219 the number of cycles . for the photon energies considered in this paper , the photoelectrons come from the ionization of be to one of the three low - lying states of @xmath220 : @xmath221 $ ] ( with an ionization potential of @xmath222 ev ) , @xmath223 $ ] ( @xmath224 ev ) , @xmath225 $ ] ( @xmath226 ev ) and the higher lying channel @xmath227 $ ] ( @xmath228 ev ) @xcite , see fig . [ fig : fig2 ] . energy levels of be and @xmath220 involved in the photoionization process . experimental energies @xcite are labelled by their terms and dominant configurations . as indicated by the arrows , we consider lasers with central frequencies corresponding to photon energies of 30 ev and 150 ev . the zig - zag curve above the 30 ev arrow denotes a change in scale beweeen the 30 ev and the 150 ev arrows . ] photoelectron spectra for linearly polarized laser pulses with ( a ) 6 and ( b ) 10 cycles with a central frequency corresponding to 30 ev , and an intensity of @xmath229 w/@xmath230 . the ras schemes are @xmath231 ( black , solid ) , @xmath232 ( red , solid ) , @xmath233 ( blue , long - dashed ) , @xmath234 ( orange , dotted ) , @xmath235 ( dark green , dash - dotted ) , @xmath236 ( black , long dash - double dotted ) and @xmath237 ( purple , thin - dotted ) . ] first , we consider laser pulses with a central frequency corresponding to a photon energy of @xmath238 ev and an intensity of @xmath229 w/@xmath230 . to illustrate the accuracy of the td - rasscf - d method , we show the pes after the ionization with pulses containing @xmath239 [ fig . [ fig : fig3](a ) ] and @xmath240 cycles [ fig . [ fig : fig3](b ) ] for several ras schemes . according to the ionization channels in fig . [ fig : fig2 ] , the peaks of the photoelectron spectrum should be located at @xmath241 , @xmath242 and @xmath243 ev . we first note that the result of the tdhf @xmath244 calculation overestimates the height of the pes compared with the results of the rest of the ras schemes . this behavior of the tdhf result clearly shows the inadequacy of this approach to describe the electronic structure and dynamics of be during the ionization process . we find for mctdhf with 4 orbitals and 6 cycles that the main peak is located at @xmath245 21.02 ev , and ranges from 20.3 - 20.5 ev for the rest of the schemes in fig . [ fig : fig3](a ) . the pes for the 10 cycles pulse in fig . [ fig : fig3](b ) , presents more narrow peaks in the pes , since the bandwidth of the pulse is reduced from @xmath246 ev to @xmath247 ev but the peak positions remain similar to the ones in fig . [ fig : fig3](a ) . for the schemes with more than 4 orbitals , the main peak is located at identical positions at the scale of the figure . also note that the results for mctdhf for 6 , 7 and 9 orbitals are indistinguishable over the entire energy range . in the tail at lower energies , however , the ras schemes @xmath248 and @xmath249 differ from the mctdhf results . this disagreement lies at energies corresponding to the peak of the ionization channel @xmath250\rightarrow\text{be}^+~[(1s^22p)^2\text{p}^o]$ ] , fig . [ fig : fig2 ] . this means that @xmath191 and only two electrons in @xmath32 is not sufficient to describe one electron in an excited orbital and another in the continuum , even for @xmath251 . to isolate the corresponding peak we use that single ionization changes the angular momentum of the system by @xmath252 and @xmath253 . since the ionic channels @xmath254 and @xmath255 have different symmetries , the ejected electron must be @xmath256 for the first ionic channels and @xmath257 for the second channel . the maximum in the probability of the @xmath256 continuum electron lies in the direction parallel to the polarization of the laser , whereas it vanishes in the perpendicular direction . the @xmath184 and @xmath258 photoelectrons can be detected in both directions . then , we can disentangle the contributions of the two peaks by analyzing the momentum distributions in the directions parallel and perpendicular to the laser polarizations , such differential quantities are shown in figs . [ fig : fig4 ] and [ fig : fig5 ] for 6 and 10 cycles , respectively . triply differential probabilities for @xmath259 ( parallel to the laser polarization direction ) @xmath260 ( perpendicular to the laser polarization direction ) and @xmath261 for linearly polarized laser pulses with 6 cycles , central frequency corresponding to 30 ev and an intensity of @xmath229 w/@xmath230 . the ras schemes shown are as in fig . [ fig : fig3 ] . ] in figs . [ fig : fig4](a ) and [ fig : fig4](b ) we show the triply differential probabilities [ eq . ] parallel and perpendicular to the polarization of the laser pulse . we note that the tdhf is not sufficient to obtain the peak in the perpendicular direction , because the channel be@xmath262 $ ] is inaccurately described at that level of approximation . the reason is that the one - body operator can not couple directly to that correlated channel . in fig . [ fig : fig4](a ) the peak along the polarization direction is smaller than for mctdhf with 4 orbitals . by carefully analyzing the triply differential probability we find that the angular distribution for tdhf is wider which explains the large pes . we see that the probability for ionization in the parallel direction [ fig . [ fig : fig4](a ) ] is much higher for mctdhf with 4 orbitals than with the other schemes . around 10 ev there is a small peak which contributes to the main peak , and which comes from the @xmath184 and @xmath258 electrons associated with the be@xmath262 $ ] channel . the energy and angle resolved signals for the other methods overlap and are indistinguishable on the scale of the figure . for 10 cycles there is , however , a small difference among them between approximately 10 and 17 ev , as shown in fig . [ fig : fig5](a ) . compared to the 6 cycles case , for 10 cycles , the bandwidth of the pulse is smaller , @xmath263 ev , and the influence of the @xmath184 and @xmath258 electrons is present in the pes for all the ras schemes considered . to investigate this difference further , we turn to the perpendicular direction . for both 6 and 10 cycles , the mctdhf solution for 6 , 7 and 9 orbitals overlap , [ fig . [ fig : fig4](b ) and fig . [ fig : fig5](b ) ] , whereas they are different from the ras schemes @xmath264 and @xmath265 . for a 10 cycles pulse , the peaks are located at 16.24 ev for mctdhf , 15.7 ev for ( @xmath266 ) and 15.3 ev for ( @xmath267 ) . for the 6 cycle pulse , the peaks are shifted by approximately 0.3 ev for all the active spaces . this effect may be caused by the enhancement of other ionization channels opening up due to the wider bandwidth . triply differential probabilities for @xmath268 and @xmath261 for a linearly polarized laser pulse with 10 cycles , a central frequency corresponding to 30 ev and an intensity of @xmath229 w/@xmath230 . the ras schemes shown are as in fig . [ fig : fig3 ] . ] photoelectron spectrum for linearly polarized laser pulses with ( a ) 6 and ( b ) 10 cycles , a central frequency corresponding to 30 ev , and an intensity of @xmath229 w/@xmath230 . ] to analyze only the influence on the spectra of choosing different ras schemes , we describe the pes for several partition schemes fixing the number of orbitals . in figs . [ fig : fig6 ] and [ fig : fig7 ] we show the pes and the triply differential probabilities , respectively , for several ras schemes for 7 orbitals , i.e. , @xmath269 . for 10 cycles , the peak at 20.3 ev corresponding to ionization into the ionic @xmath254 channel as well as the tail for higher energies coincide for all the schemes , whereas for 6 cycles , the peak is shifted to lower energies , probably due to the ionization into the ionic state @xmath255 . as in the previous cases , the disagreement comes at lower energies corresponding to the ionic state @xmath255 . in this range , the mctdhf and the ras @xmath270 results can not be distinguished on the scale of fig . [ fig : fig6 ] . however , for the ras @xmath271 there is an excess in the signal for energies lower than 17 ev , whereas for larger energies we find a smaller signal than for mctdhf . this effect is more pronounced for @xmath264 . this can be understood in terms of the differential energy distribution in fig . [ fig : fig7 ] . in the parallel direction , the curves for these last two ras schemes overlap , and they differ between the @xmath272 and @xmath273 schemes . in the perpendicular direction we can see differences among all the schemes . the peak of the mctdhf calculation is located at 16.25 ev with a value @xmath274 , which is slightly larger than @xmath275 for @xmath276 . for @xmath277 and @xmath278 , the peak is shifted to lower energies , both lying at 15.24 ev . the height of these peaks is very similar to the heights obtained with the mctdhf results . triply differential probabilities for @xmath279 and @xmath261 for a linearly polarized laser pulse with 10 cycles , a central frequency corresponding to 30 ev and an intensity of @xmath229 w/@xmath230 for a ras scheme with @xmath280 orbitals . ] the peak in the pes coming from the ionic state @xmath225 $ ] is not observed in the present results due to the small cross section of this transition @xcite . for the pulses of finite duration used here , this peak is buried in the tails of the other peaks . the rtp calculations also illustrate the reduction of the numerical effort of the td - rasscf - d approach compared with the mctdhf method . for the ras scheme @xmath248 , 110 cycles of propagation takes 18h in 20 cores compared to 22h for mctdhf . in the case of 9 spatial orbitals the difference is more marked , taking 34h for the ras scheme @xmath281 and 47h for mctdhf . photoelectron spectrum for a linearly polarized laser pulse with 10 cycles , a central frequency corresponding to 150 ev and an intensity of @xmath282 w/@xmath230 . the ras schemes shown are as in fig . [ fig : fig3 ] . ] we analyze the core ionization to the ionic channel @xmath283 $ ] [ fig . [ fig : fig2 ] ] using @xmath284 ev linearly polarized laser pulses for several ras schemes . the pes is shown in fig . [ fig : fig8 ] for a 10 cycles pulse with @xmath285 ev and an intensity of @xmath282 w/@xmath230 . the pes are almost overlapping for all the ras schemes used , which means that the mctdhf and td - rasscf - d methods describe with the same accuracy the ionization into this channel , even for the minimum number of configurations , that is , using the ras scheme @xmath286 . these ras schemes contain the most relevant configurations needed to describe the dynamics of the process , that is , @xmath287s@xmath288 and an electron @xmath15 in the continuum . the peak is located at @xmath245 21.37 ev , corresponding to an ionization energy of 128.63 ev , a bit higher than the experimental value 123.35 ev @xcite , which is in agreement with the @xmath161 ev shift found in ref . @xcite . in recent years there has been a large interest in time - delays in photoionization studies ( see , e.g. , the review ref . @xcite and references therein ) . here we consider the delay between the apparent times of ionization into the channels @xmath289 $ ] @xmath290 and @xmath291 $ ] @xmath292 and we exploit the flexibility of the td - rasscf method to address the role of electron correlation on photoionization time - delays . we do so in the following by considering the angle - resolved radial density after the pulse for several ras schemes . we show in figs . [ fig : fig9](a)-(b ) the density along the polarization direction and in figs . [ fig : fig9](c)-(d ) the density in the perpendicular direction at different times after the end of the laser pulse . as we have discussed in previous sections , the ionization into both channels contributes in the parallel direction , whereas in the perpendicular direction only the @xmath184 and @xmath258 photoelectrons associated with the @xmath291 $ ] channel contribute . let us remark that the results of the mctdhf calculations have also been obtained with 9 spatial orbitals , and on the scale of the figure they coincide with the results of the calculation for 7 orbitals . in the parallel direction , the height of the density decreases as it spreads during the propagation . in the perpendicular direction the outgoing wavepacket is not yet formed at @xmath293 a.u . , therefore the peak in the density increases for later times . the ras @xmath270 reproduces accurately the dynamics for ionization into the channel @xmath289 $ ] [ fig . [ fig : fig9](a ) ] , whereas there are small differences with the @xmath277 scheme , as shown in fig . [ fig : fig9](b ) . the main differences between the different levels of theory are found in the perpendicular direction shown in figs . [ fig : fig9](c ) and ( d ) . in the case of @xmath270 , we find that the heights of the density peaks are smaller than in mctdhf case , but remain on equal positions . in contrast to this case , for @xmath271 the density distributions are similar , but the ras result is shifted in the radial coordinate with respect to the mctdhf result . to analyze independently the dynamics of the ejected electrons from each ionic channels we can use that the ejected @xmath256 electron associated with the ionic channel @xmath289 $ ] only contributes in the direction parallel to the polarization axis of the laser , whereas the @xmath184 and @xmath258 electrons associated with the @xmath291 $ ] channel contribute to both the parallel and perpendicular directions . to distinguish between these two channels by the angle resolved radial density in the parallel and perpendicular directions we benefit from the fact that the influence of the @xmath184 and @xmath258 electrons in the parallel direction is negligible compared to the @xmath256 electron ejected from @xmath289 $ ] , as seen in fig . [ fig : fig9 ] . this allows us to obtain the expectation value of the position of the electron in the outer region , which fulfils ehrenfest s theorem , @xmath294 , with @xmath295 the ionization time . we now calculate the time - delays for td - rasscf and mctdhf and compare the impact of the electron correlation by varying the ras scheme . let us remark that we need to include correlation in the photoionization process of be , because we can not resolve the ionization process @xmath296\rightarrow$ ] @xmath291+e^{-}(s~\text{or}~d)$ ] using tdhf . we find for mctdhf with 7 orbitals a time - delay between the two channels of @xmath297 as , which is a bit smaller than the result for 9 orbitals , @xmath298 as . by reducing @xmath29 and adding two orbitals to @xmath32 , @xmath270 , @xmath299 is @xmath300 as , which further diminishes to @xmath301 as for @xmath271 . the ras scheme @xmath264 leads to @xmath302 as . even if we increase the number of orbitals in @xmath32 to @xmath303 we obtain @xmath304 as , far from the mctdhf value . for these two ras schemes the ionization from the channel @xmath291 $ ] is not well described , leading to a value of @xmath305 which largely differs from the mctdhf result . however , the ionization times of the @xmath184 and @xmath258 electrons are similar for all the ras used , differing less than @xmath306 . in conclusion the relative time - delay of ionization into @xmath291 $ ] and @xmath289 $ ] is around 7 - 8 as . note that this method is sensitive to an error in the calculation of @xmath307 . it can be estimated that for this process , an error of @xmath308 a.u . implies an error in the time - delay of the order of @xmath309 as . triply differential probabilities in position space for @xmath272 , @xmath264 and @xmath310 ras schemes with 7 spatial orbitals after the interaction with a linearly polarized laser pulse with a central frequency corresponding to 30 ev , 10 cycles and an intensity of @xmath229 w/@xmath230 , where ( a ) and ( b ) corresponds to the @xmath311 and ( c ) and ( d ) to @xmath312 . all the panels contain the results for mctdhf with 7 spatial orbitals at @xmath313 ( red , solid ) , @xmath314 ( blue , dashed ) and @xmath315 ( orange , dotted ) , together with @xmath276 in panels ( a ) and ( c ) ( @xmath313 : black , dash - double dotted , @xmath314 : green , dash - dotted and @xmath315 : purple , dotted ) and @xmath277 in panels ( b ) and ( d ) ( @xmath313 : blue , dash - double dotted , @xmath314 : black , dash - dotted and @xmath315 : gray , dotted ) . the instants of time in atomic units are indicated in panel ( d ) . ] in this work we addressed the effect of electron correlation in be in the ground state , in photoelectron spectra and in relative time - delay in photoionization by application of the td - rasscf - d @xcite method , extended in this work to fully 3d systems . we used the coupled basis method on the angular momentum of the single orbital basis to compute the two - body operator . this method reduces the numerical cost since the number of operations to obtain the mean - field operator scales linearly with the radial grid points and quadratically with the number of angular functions . we found that the td - rasscf method including double excitations diminishes the numerical effort compared to the mctdhf by reducing the accessible configurations and that it is accurate mainly due to the importance of the pairwise nature of the electron correlation . furthermore , the selection of the ras makes it possible to identify the most important active space orbitals , which facilitates a convergence to the global ground state . we found that the restriction on the ras scheme permits a random initial guess function to reach a lower ground state energy than with the mctdhf method , unless we start with a designed initial guess function in the latter case . we analyzed the pess resulting from the interaction with a short linearly polarized xuv laser pulse , and found that the mean - field tdhf method is inaccurate in describing single ionization . using the td - rasscf - d method we identified the most important active space orbitals which capture the most relevant configurations . we also computed a time - delay of around 8 as between the ionization channels be@xmath316\rightarrow$ ] be@xmath317+e^-$ ] and be@xmath316\rightarrow$]be@xmath262+e^-$ ] for ionization with a few cycle , xuv laser operated in the perturbative regime . we compared the results for several ras schemes , and observed that their accuracy is related to the ability of the method to describe the peaks in the pes . for example , the tdhf is unable to describe accurately both ionization channels , and therefore , the time - delay . as we include more orbitals in the method we obtain a better description of the ionization into the two channels . the time - delay is very sensitive to the ras scheme used and constitutes a test of the accuracy of the method . over all , we found that the td - rasscf - d methodology constitutes an efficient tool to deal with many - electron atomic systems in the presence of time - dependent interaction as , e.g. , an external xuv field . combined with the coupled basis method , it provides a stable , accurate and efficient procedure to treat orbitals with undefined magnetic quantum number , which is crucial to describe , e.g. , of the interaction with circularly polarized light , where the rotational symmetry is broken . we thank haruhide miyagi and camille l ' ev^ eque for fruitful discussions and comments . this work was supported by the erc- stg ( project no . 277767-tdmet ) and the villum kann rasmussen ( vkr ) center of excellence quscope . the numerical results presented in this work were obtained at the centre for scientific computing , aarhus . 70ifxundefined [ 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/revmodphys.81.163 [ * * , ( ) ] link:\doibase 10.1146/annurev - physchem-032511 - 143702 [ * * , ( ) ] link:\doibase 10.1088/0953 - 4075/49/6/062001 [ * * , ( ) ] link:\doibase 10.1038/nphys1940 [ * * , ( ) ] link:\doibase 10.1126/science.1254061 [ * * , ( ) ] link:\doibase 10.1126/science.aab2160 [ * * , ( ) ] link:\doibase 10.1103/physreva.87.033407 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.115.113004 [ * * , ( ) ] link:\doibase 10.1126/science.1189401 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.105.233002 [ * * , ( ) ] link:\doibase 10.1103/physreva.84.061404 [ * * , ( ) ] link:\doibase 10.1103/physreva.87.023420 [ * * , ( ) ] link:\doibase 10.1088/0953 - 4075/47/24/245003 [ * * , ( ) ] link:\doibase 10.1103/physreva.94.063409 [ * * , ( ) ] \doibase 10.1038/nphys3941 ; http://www.nature.com/nphys/journal/v13/n3/abs/nphys3941.html#supplementary-information [ * * , ( ) ] link:\doibase 10.1103/revmodphys.87.765 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1088/0953 - 4075/34/3/103 [ * * , ( ) ] link:\doibase 10.1103/physreva.77.043420 [ * * , ( ) ] link:\doibase 10.1103/physreva.84.043409 [ * * , ( ) ] link:\doibase 10.1103/physreva.84.013404 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.108.163001 [ * * , ( ) ] link:\doibase 10.1103/physreva.88.053411 [ * * , ( ) ] link:\doibase 10.1103/physreva.89.053407 [ * * , ( ) ] link:\doibase 10.1103/physreva.88.053412 [ * * , ( ) ] link:\doibase 10.1103/physreva.85.023411 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.111.233005 [ * * , ( ) ] link:\doibase 10.1088/1742 - 6596/427/1/012007 [ * * , ( ) ] http://journals.aps.org/pra/abstract/10.1103/physreva.90.062508 [ * * , ( ) ] http://journals.aps.org/pra/abstract/10.1103/physreva.92.063423 [ * * , ( ) ] @noop _ _ ( , ) @noop * * , ( ) link:\doibase 10.1007/s004600050342 [ * * , ( ) ] @noop * * , ( ) http://stacks.iop.org/0953-4075/37/i=4/a=004 [ * * , ( ) ] link:\doibase 10.1103/physreva.71.012712 [ * * , ( ) ] link:\doibase 10.1103/physreva.83.063416 [ * * , ( ) ] link:\doibase 10.1103/physreva.86.013406 [ * * , ( ) ] link:\doibase 10.1103/physreva.91.012509 [ * * , ( ) ] link:\doibase 10.1103/physreva.82.043414 [ * * , ( ) ] link:\doibase 10.1103/physreva.93.023401 [ * * , ( ) ] link:\doibase 10.1063/1.3553176 [ * * , ( ) ] \doibase http://dx.doi.org/10.1103/physreva.91.023417 [ * * , ( ) ] link:\doibase 10.1103/physreva.88.023402 [ * * , ( ) ] link:\doibase 10.1103/physreva.94.023405 [ * * , ( ) ] link:\doibase 10.1103/physreva.87.062511 [ * * , ( ) ] link:\doibase 10.1063/1.4872005 [ * * , ( ) ] link:\doibase 10.1103/physreva.89.063416 [ * * , ( ) ] link:\doibase 10.1103/physreva.95.023415 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physreva.85.033401 [ * * , ( ) ] link:\doibase 10.1017/s0305004100016108 [ * * , ( ) ] http://www.sciencedirect.com/science/article/pii/s0009261400002001 [ * * , ( ) ] link:\doibase 10.1016/s0370 - 1573(99)00047 - 2 [ * * , ( ) ] link:\doibase 10.1088/1742 - 6596/696/1/012009 [ * * , ( ) ] link:\doibase 10.1088/0953 - 4075/37/17/r01 [ * * , ( ) ] link:\doibase 10.1103/physreva.62.032706 [ * * , ( ) ] link:\doibase 10.1088/0953 - 4075/40/23/004 [ * * , ( ) ] link:\doibase 10.1140/epjst / e2014 - 02092 - 3 [ * * , ( ) ] http://journals.aps.org/pra/abstract/10.1103/physreva.93.033420 [ * * , ( ) ] _ _ , http://www.theo-physik.uni-kiel.de/bonitz/theses/hinz{_}13.pdf [ ] , ( ) link:\doibase 10.1021/acs.jpclett.5b01891 [ * * , ( ) ] link:\doibase 10.1103/physreva.76.063407 [ * * , ( ) ] link:\doibase 10.1103/physreva.87.053405 [ * * , ( ) ] link:\doibase 10.1103/physreva.86.053424 [ * * , ( ) ] link:\doibase 10.1007/s00205 - 010 - 0308 - 8 [ * * , ( ) ] @noop _ _ ( , , ) \doibase https://doi.org/10.1103/physreva.35.2429 [ * * , ( ) ] http://stacks.iop.org/0953-4075/26/i=5/a=009 [ * * , ( ) ] \doibase http://dx.doi.org/10.1063/1.555999 [ * * , ( ) ]

in an acyclic multicast network , it is well known that a linear network coding solution over gf( ) exists when is sufficiently large . in particular , for each prime power no smaller than the number of receivers , a linear solution over gf( ) can be efficiently constructed . in this work , we reveal that a linear solution over a given finite field does _ not _ necessarily imply the existence of a linear solution over all larger finite fields . specifically , we prove by construction that : ( i ) for every source dimension no smaller than 3 , there is a multicast network linearly solvable over gf(7 ) but not over gf(8 ) , and another multicast network linearly solvable over gf(16 ) but not over gf(17 ) ; ( ii ) there is a multicast network linearly solvable over gf(5 ) but not over such gf( ) that is a mersenne prime plus 1 , which can be extremely large ; ( iii ) a multicast network linearly solvable over gf( ) and over gf( ) is _ not _ necessarily linearly solvable over gf( ) ; ( iv ) there exists a class of multicast networks with a set of receivers such that the minimum field size for a linear solution over gf( ) is lower bounded by , but not every larger field than gf( ) suffices to yield a linear solution . the insight brought from this work is that not only the field size , but also the order of subgroups in the multiplicative group of a finite field affects the linear solvability of a multicast network . linear network coding , multicast network , field size , lower bound , mersenne prime .

1401.3075

consider a multicast network , which is a finite directed acyclic multigraph with a unique source node @xmath8 and a set @xmath5 of receivers . every edge in the network represents a noiseless transmission channel of unit capacity . the source generates @xmath9 data symbols belonging to a fixed symbol alphabet and will transmit them to all receivers via the network . the maximum flow , which is equal to the number of edge - disjoint paths , from @xmath8 to every receiver is assumed to be no smaller than @xmath9 . the network is said to be solvable if all receivers can recover all @xmath9 source symbols based on their respective received data symbols . when the network has only one receiver , it is solvable by network routing . when @xmath10 , the paradigm of network routing does not guarantee the network to be solvable . the seminal paper @xcite introduced the concept of network coding ( nc ) and proved that the considered multicast network has a nc solution over some infinitely large symbol alphabet . it was further shown in @xcite that _ linear _ nc suffices to yield a solution when the symbol alphabet is algebraically modeled as a sufficiently large finite field , and every intermediate node transmits a linear combination of its received data symbols over the symbol field . since then , there have been extensive studies on the field size requirement of a linear solution for a multicast network . from an algebraic approach , reference @xcite first showed that a multicast network has a linear solution over gf(@xmath0 ) as long as the prime power @xmath0 is larger than @xmath9 times the number @xmath11 of receivers . the requirement of @xmath0 for the existence of a linear solution over gf(@xmath0 ) is further relaxed by @xcite to be larger than @xmath11 , and such a solution can be efficiently constructed by the algorithm proposed in @xcite . meanwhile , the efficient algorithm in @xcite is able to construct a linear solution over gf(@xmath0 ) when @xmath0 is no smaller than @xmath11 , and hence this condition on @xmath0 is slightly relaxed compared with the ones in @xcite and @xcite . this efficient algorithm requires one to initially identify , for each receiver in @xmath5 , @xmath9 edge - disjoint paths starting from the source and ending at it . denote by @xmath12 the maximum number of paths among the @xmath13 paths that contain a common edge . the parameter @xmath12 is always no larger than @xmath11 . by a more elaborate argument , the algorithm in @xcite is refined in @xcite such that it can construct a linear solution over gf(@xmath0 ) as long as @xmath0 is no smaller than @xmath12 . denote by @xmath6 the minimum field size for the existence of a linear solution over gf(@xmath6 ) , and by @xmath14 the maximum field size for the non - existence of a linear solution over gf(@xmath14 ) ( if the network is linearly solvable over every finite field , then @xmath14 is not well defined and we set it to 1 as a convention . ) the algorithm in @xcite implies that @xmath11 is an upper bound on @xmath14 . for the special case that the source dimension @xmath9 is equal to 2 , the upper bound on @xmath14 is reduced to @xmath15 in @xcite . in both cases , the upper bounds additionally guarantee the existence of a linear solution over every gf(@xmath0 ) with @xmath0 larger than the bounds . on the other hand , references @xcite and @xcite independently constructed a class of multicast networks with @xmath6 lower bounded by @xmath7 . on any network in this class , a linear solution exists over every gf(@xmath0 ) with @xmath0 no smaller than @xmath6 , that is , @xmath16 . these results indicate that in many cases , _ a multicast network that is linearly solvable over a given finite field _ gf_(@xmath0 ) is also linearly solvable over every larger finite field . _ to the best of our knowledge , there is no explicit proof or disproof of the above claim for a general multicast network and for the case @xmath17 ; furthermore , all known multicast networks studied in the network coding literature satisfy @xmath16 . although it has been shown in @xcite that the @xmath18-combination ( multicast ) network depicted in fig . [ fig:4 - 2-combination_networks ] has a nonlinear solution over a ternary symbol alphabet but has neither a linear nor a nonlinear solution over any symbol alphabet of size 6 , it does not shed light on disproving the claim @xmath19 because this combination network has a linear solution over every gf(@xmath0 ) with @xmath20 ( see , for example , @xcite . ) moreover , in the case @xmath21 , as revealed in @xcite and @xcite , there exists a linear solution over gf(@xmath0 ) if and only if there exists a @xmath22-vertex coloring in an appropriately defined associated graph . since a @xmath0-vertex coloring in a graph always guarantees a @xmath23-vertex coloring with @xmath24 in the same graph , a multicast network with @xmath21 is linearly solvable over every gf(@xmath23 ) with @xmath25 this evidence seemed to add more support for the correctness of the claim @xmath19 for an arbitrary multicast network . in the present paper , we shall show by constructive proofs that the claim @xmath19 is _ not _ always true . in particular , * we show that there is a multicast network that is linearly solvable over gf(@xmath6 ) but not over gf(@xmath14 ) for each of : ( i ) @xmath26 , @xmath27 ; ( ii ) @xmath28 , @xmath27 ; ( iii ) @xmath29 , @xmath30 ; * we show that for any positive integer @xmath31 with less than 17,425,170 ( base-10 ) digits , there is a multicast network with @xmath26 whereas @xmath32 . * we construct a new class of multicast networks with @xmath6 lower bounded by @xmath7 and with an additional property that not every network in it has @xmath14 smaller than @xmath6 . the insight of our results is that not only the field size but also the orders of the proper multiplicative subgroups in the symbol field affect the linear solvability over the finite field . as we shall see , if a finite field does not contain a large enough proper multiplicative subgroup , or the complement of a large multiplicative subgroup in the finite field is not large enough , it is possible to construct a multicast network that is not linearly solvable over this finite field but linearly solvable over a smaller finite field . in comparison , the characteristic of the symbol field does not appear as important in designing examples here as in the ones in @xcite which show the non - existence of a linear solution for a general multi - source network , because in our exemplifying networks , both @xmath6 and @xmath33 can be of either odd and even characteristic . the classical solvable ( non - multicast ) network that is not linearly solvable , proposed in @xcite , makes use of two types of subnetworks : one is linearly solvable only over a field with even characteristic whereas the other is linearly solvable only over a field with odd characteristic . consequently , the proposed network as a whole is not linearly solvable over any field , even or odd . our results bring about a new facet on the connection between the symbol field structures and network coding solvability problems . the remainder of this paper is organized as follows . section [ sec : fundamental_results ] presents the fundamental results that there exist multicast networks such that @xmath34 . section [ sec : gap ] discusses how large the gap @xmath35 can be . to unify the justification of the linear solvability of different networks presented , section [ sec : general_network ] constructs a general multicast network and obtains an equivalent condition of its linear representability over a field gf(@xmath0 ) . section [ sec : lower_bound ] further constructs a new class of multicast networks with @xmath6 lower bounded by @xmath7 and with the additional property that @xmath34 for some networks in it . section [ sec : summary ] concludes the paper and discusses some interesting problems along this new research thread in network coding theory . a ( single - source ) multicast network is a finite directed acyclic multigraph with a unique source node @xmath8 and a set @xmath5 of receivers . on a multicast network , for every node @xmath36 , denote by @xmath37 and @xmath38 , respectively , the set of its incoming and outgoing edges . every edge in a multicast network represents a noiseless transmission channel of unit capacity . the source generates @xmath9 data symbols belonging to a fixed symbol field and will transmit them to all receivers in @xmath5 . without loss of generality ( wlog ) , assume that @xmath39 , which is referred to as the _ source dimension _ ( otherwise a new source can be created , connected to the old source with @xmath9 edges . ) for an arbitrary set @xmath40 of non - source nodes , denote by @xmath41 the maximum number of edge - disjoint paths starting from @xmath8 and ending at nodes in @xmath40 . each receiver @xmath42 in @xmath5 has @xmath43 . a linear network code ( lnc ) over gf(@xmath0 ) is an assignment of a _ coding coefficient _ @xmath44 to every pair @xmath45 of edges such that @xmath46 when @xmath45 is not an _ adjacent pair _ , that is , when there is not a node @xmath36 such that @xmath47 and @xmath48 . the lnc uniquely determines a _ coding vector _ @xmath49 , which is an @xmath9-dim column vector , for each edge @xmath50 in the network such that : * @xmath51 forms the natural basis of @xmath52 . * @xmath53 when @xmath48 for some @xmath54 . wlog , we assume throughout this paper that * all lncs on a given multicast network have coding coefficients @xmath55 for all those adjacent pairs @xmath45 where @xmath31 is the unique incoming edge to some node . a multicast network is said to be _ linearly solvable _ over gf(@xmath0 ) if there is an lnc over gf(@xmath0 ) such that for each receiver @xmath56 , the @xmath57 matrix @xmath58_{e\in in(t)}$ ] over gf(@xmath0 ) is full rank . such an lnc is called a _ linear solution _ over gf(@xmath0 ) for the multicast network . denote by @xmath6 the minimum field size for the existence of a linear solution over gf(@xmath6 ) , and by @xmath14 the maximum field size for the nonexistence of a linear solution over gf(@xmath14 ) . specific to a finite field gf(@xmath0 ) , let gf(@xmath0)@xmath59 represent the multiplicative group of nonzero elements in gf(@xmath0 ) . @xmath60 when @xmath61 , the efficient algorithm in @xcite can be adopted to construct a linear solution over gf(@xmath0 ) . in the case @xmath21 , it has been shown in @xcite that every linear solution over gf(@xmath0 ) can induce a @xmath22-vertex coloring in an appropriately associated graph of the multicast network and vice versa . since every @xmath22-vertex coloring in a graph can also be regarded as a @xmath62-vertex coloring for @xmath24 , it in turn induces a linear solution over gf(@xmath23 ) when @xmath23 is a prime power . thus , every linear solution over a finite field can induce a linear solution over a larger finite field . all these are tempting facts for one to conjecture that when @xmath63 and @xmath17 , a linear solution over a finite field might also imply the existence of a linear solution over a larger field . the central theme of the present paper is to _ refute _ this conjecture in several aspects . [ theorem:1 ] a multicast network with @xmath64 that is linearly solvable over a finite field is _ not _ necessarily linearly solvable over all larger finite fields . when @xmath65 , this theorem is a direct consequence of the lemma below . assume @xmath66 . expand the network @xmath67 depicted in fig . [ fig : rank_3_networks](a ) to a new multicast network @xmath68 as follows . create @xmath69 new nodes each of which has an incoming edge emanating from the source and an outgoing edge entering every receiver . in @xmath68 , every receiver has the maximum flow from the source equal to @xmath9 . consider an lnc over a given gf(@xmath0 ) . by the topology of @xmath68 , the coding vector for every edge that is originally in @xmath67 is a linear combination of the coding vectors for those edges in @xmath70 that are also in @xmath67 . since the network @xmath67 is linearly solvable over gf(@xmath0 ) with every @xmath71 except for @xmath72 by the lemma below , so is the network @xmath68 . [ lemma : rank_3_network_a ] consider the multicast network depicted in fig . [ fig : rank_3_networks](a ) . denote by @xmath73 , @xmath74 @xmath75 @xmath76 @xmath75 @xmath77 the unique incoming edge to node @xmath78 . for every set @xmath40 of 3 grey nodes with @xmath79 , there is a receiver connected from it . this network is linearly solvable over every finite field gf(@xmath0 ) with @xmath80 except for @xmath81 . a detailed technical proof for lemma [ lemma : rank_3_network_a ] is given in appendix [ appendix_proof_rank_3_network_a ] . in the network in fig . [ fig : rank_3_networks](a ) , observe that every receiver is connected with three nodes @xmath82 where either @xmath83 , @xmath84 , @xmath85 are distinct ( such as @xmath86 ) or two among @xmath83 , @xmath84 , @xmath85 are same ( such as @xmath87 ) . one can then check an lnc over gf(7 ) with @xmath88 = \left [ \begin{array}{ccccccccc } 1 & 1 & 1 & 0 & 0 & 0 & 1 & 1 & 1 \\ 1 & 2 & 4 & 1 & 1 & 1 & 0 & 0 & 0 \\ 0 & 0 & 0 & 1 & 2 & 4 & 1 & 2 & 4 \\ \end{array } \right],\ ] ] where @xmath89 represents the unique incoming edge to node @xmath90 , qualifies to be a linear solution for the network . note that all nonzero entries in this matrix belong to a proper subgroup of order 3 in the multiplicative group @xmath91 . the key reason that results in the network not linearly solvable over gf(8 ) is that there is not a proper subgroup of order at least 3 in the multiplicative group gf(8)@xmath59 . in general , the network in fig . [ fig : rank_3_networks](a ) is linearly solvable over a finite field gf(@xmath0 ) if and only if there exist @xmath92 subject to @xmath93 this is a direct consequence of theorem [ thm : general_network_solvability ] in section [ sec : general_network ] when a more general multicast network depicted in fig . [ fig : general_network ] is to be introduced which subsumes the network in fig . [ fig : rank_3_networks](a ) as a special instance . on a solvable multi - source network , it is well - known that linear network coding ( with linearity in terms of a more general algebraic structure ) is not sufficient to yield a solution . the classical example in @xcite to show this consists of two subnetworks , one is only linearly solvable over a field with even characteristic , whereas the other is only linearly solvable over a field with odd characteristic . more generally , a procedure is introduced in @xcite to construct a ( matroidal ) multi - source network based on a matroid such that the constructed network is linearly solvable over gf(@xmath0 ) if and only if the matroid is representable over gf(@xmath0 ) . this connection is powerful for designing a number of non - linearly solvable networks from a variety of interesting matroid structures ( see the appendix in @xcite for example , ) and the network in @xcite is a instance under this construction . subsequently , the role of characteristics of finite fields on the linear solvability of a general multi - source network is further revealed in @xcite : for an arbitrary finite or co - finite set @xmath94 of prime numbers , a multi - source network can be constructed such that the network is linearly solvable over some field of characteristic @xmath95 for every prime number @xmath95 in @xmath94 , whereas it is not linearly solvable over any finite field whose characteristic is not in @xmath94 . however , the examples presented in this paper can not be established by the procedures introduced in @xcite and @xcite , because they were not designed to construct ( single - source ) multicast networks . moreover , as a result of lemma [ lemma : rank_3_network_a ] and the next lemma , the role of the characteristic of a finite field in the examples designed in the present paper is not as important as in the example in @xcite . [ lemma : rank_3_network_b ] consider the multicast network depicted in fig . [ fig : rank_3_networks](b ) . there are in total 20 grey nodes . for every set @xmath40 of 3 grey nodes with @xmath79 , there is a receiver connected from it . this network is linearly solvable over every finite field gf(@xmath0 ) with @xmath96 except for @xmath97 . a detailed proof of this lemma can be found in appendix [ appendix_proof_rank_3_network_b ] . similar to the network in fig . [ fig : rank_3_networks](a ) , the network in fig . [ fig : rank_3_networks](b ) can also be regarded as a special instance of the general network to be introduced in section [ sec : general_network ] . then , theorem [ thm : general_network_solvability ] to be developed in section [ sec : general_network ] implies that the network in fig . [ fig : rank_3_networks](b ) is linearly solvable over a finite field gf(@xmath0 ) if and only if there is an assignment of @xmath98 , @xmath99 from gf(@xmath0)@xmath59 subject to @xmath100 and @xmath101 for example , when @xmath102 , we can set @xmath103 for all @xmath104 , where @xmath105 is a primitive element in gf(16 ) , and set @xmath106 to be the 10 elements in @xmath107 . such an assignment satisfies the above condition and hence the network is linearly solvable over gf(16 ) . the insight for the network in fig . [ fig : rank_3_networks](b ) to be linearly solvable over gf(16 ) rather than gf(17 ) is that there is such a subgroup in the multiplicative group gf(16)@xmath59 but not in gf(17)@xmath59 that ( i ) the subgroup has order no smaller than 5 ; ( ii ) the complement of the subgroup in the multiplicative group contains at least 10 elements . [ corollary : characteristic ] given a multicast network with @xmath34 , both @xmath6 and @xmath14 can be of either even or odd characteristic . this can be seen from lemma [ lemma : rank_3_network_a ] together with lemma [ lemma : rank_3_network_b ] . to the best of our knowledge , all known multicast networks studied in the network coding literature have the property that @xmath16 . the results in the previous section reveal that it is possible for @xmath34 . however , both examples which illustrate this fact have the special property @xmath108 . a natural question next is how far away can @xmath14 be from @xmath6 . the main result in this section is to show that for some multicast networks , the difference @xmath35 can be extremely large . consider the _ swirl network _ with source dimension @xmath64 depicted in fig . [ fig : swirl_network ] . for every set @xmath40 of @xmath9 grey nodes with @xmath109 , there is a non - depicted receiver connected from it . corresponding to each node @xmath78 of in - degree 2 , where @xmath110 , let @xmath111 denote the two outgoing edges from it . by viewing the swirl network as a special instance of the general network to be introduced in the next section , the following necessary and sufficient condition for the swirl network to be linearly solvable over gf(@xmath0 ) can be readily obtained . a detailed proof is given in appendix [ appendix_proof_swirl_network ] . [ lemma : swirl_network ] the swirl network is linearly solvable over a given gf(@xmath0 ) if and only if there exist @xmath112 subject to @xmath113 under which a linear solution over gf(@xmath0 ) can be subsequently given by prescribing the coding vectors for @xmath114 as @xmath115 \\ % & = \left [ % \begin{matrix } % 1 & 1 & 0 & 0 & \cdots & 0 & 0 & 1 & 1 \\ % 1 & \alpha_1 & 1 & 1 & \ddots & 0 & 0 & 0 & 0 \\ % 0 & 0 & 1 & \alpha_2 & \ddots & 0 & 0 & 0 & 0 \\ % \vdots & \vdots & \vdots & \vdots & \ddots & \vdots & \vdots & \vdots & \vdots \\ % 0 & 0 & 0 & 0 & \cdots & 1 & 1 & 0 & 0 \\ % 0 & 0 & 0 & 0 & \cdots & 1 & \alpha_{\omega-1 } & \delta_1 & \delta_2 \\ % \end{matrix } % \right ] , [ f_{e_{11}}~f_{e_{12}}~f_{e_{21}}~f_{e_{22 } } \cdots f_{e_{\omega1}}~f_{e_{\omega2 } } ] = \left [ \begin{matrix } 1 & 1 & 0 & 0 & \cdots & 0 & 0 & 1 & 1 \\ 1 & \alpha_1 & 1 & 1 & \ddots & 0 & 0 & 0 & 0 \\ 0 & 0 & 1 & \alpha_2 & \ddots & 0 & 0 & 0 & 0 \\ \vdots & \vdots & \vdots & \vdots & \ddots & \vdots & \vdots & \vdots & \vdots \\ 0 & 0 & 0 & 0 & \cdots & 1 & 1 & 0 & 0 \\ 0 & 0 & 0 & 0 & \cdots & 1 & \alpha_{\omega-1 } & \delta_1 & \delta_2 \\ \end{matrix } \right].\ ] ] consider the swirl network with @xmath116 , and an lnc over gf(5 ) with the coding vectors for @xmath117 prescribed by @xmath118 = \left [ \begin{array}{cccccccccccc } 1 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 1 \\ 1 & 4 & 1 & 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 1 & 4 & 1 & 1 & 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 1 & 4 & 1 & 1 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 1 & 4 & 1 & 1 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 4 & 2 & 3 \\ \end{array } \right].\ ] ] apparently condition ( [ eqn : swirl_1 ] ) is satisfied by this lnc . since @xmath119 is a subgroup of gf(5)@xmath59 , it is closed under multiplication by elements in it . moreover , @xmath120 , and so condition ( [ eqn : swirl_2 ] ) is also satisfied . thus , this lnc over gf(5 ) qualifies as a linear solution . on the other hand , since gf(8)@xmath59 does not have a proper subgroup other than @xmath121 , it is not difficult to check that for arbitrary @xmath122 , the set @xmath123 contains at least 6 elements . thus , there are not enough distinct elements in gf(8)@xmath59 to assign for @xmath124 and @xmath125 subject to conditions ( [ eqn : swirl_1 ] ) and ( [ eqn : swirl_2 ] ) . hence , the network is not linearly solvable over gf(8 ) . the argument in the example above can simply be generalized to derive the linear solvability of the swirl network with general source dimension @xmath9 over a given gf(@xmath0 ) . first , assume that the order of the multiplicative group gf(@xmath0)@xmath59 is not prime . this implies @xmath126 . then there is a subgroup @xmath127 in gf(@xmath0)@xmath59 with @xmath128 , and gf(@xmath0)@xmath129 contains no less than 2 elements . let @xmath130 be an element in @xmath127 not equal to 1 and @xmath131 be two distinct elements in gf(@xmath0)@xmath129 . we can assign @xmath132 for all @xmath133 and @xmath134 . such an assignment obeys conditions ( [ eqn : swirl_1 ] ) and ( [ eqn : swirl_2 ] ) . hence , the network is linearly solvable over gf(@xmath0 ) . next , assume that the order of gf(@xmath0)@xmath59 is a prime , in which case @xmath0 can always be written in the form of @xmath135 for some prime @xmath95 . we shall show that for any @xmath136 , the set @xmath137 contains at least @xmath138 elements . this is obviously true when @xmath139 . assume that it is true for @xmath140 and consider the case @xmath141 . note that @xmath142 . if @xmath143 , then so is @xmath144 . assume that @xmath145 . since @xmath146 and @xmath147 is a prime , there is at least one element in @xmath148 that is not in @xmath149 . thus , @xmath150 . as a consequence , we conclude that for any assignment of @xmath151 , where @xmath105 is a primitive element in gf(@xmath135 ) and @xmath152 , the set @xmath153 contains at least @xmath9 elements . in order to further successfully assign @xmath125 subject to ( [ eqn : swirl_1 ] ) and ( [ eqn : swirl_2 ] ) , the size of gf(@xmath135 ) has to be at least @xmath154 . when @xmath155 , we can set @xmath156 for all @xmath157 and @xmath158 , @xmath159 , where @xmath105 is a primitive element in gf(@xmath135 ) , so that conditions ( [ eqn : swirl_1 ] ) and ( [ eqn : swirl_2 ] ) obey . we have proved the following theorem . [ theorem : swirl ] the swirl network with @xmath64 has @xmath26 and @xmath160 . moreover , it is not linearly solvable over all those @xmath161 when @xmath162 is prime . recall that a prime integer in the form @xmath162 is known to be a _ mersenne prime_. then @xmath14 for the swirl network with @xmath64 is equal to one plus the largest mersenne prime no larger than @xmath163 . while whether there are infinitely many mersenne primes is still an open problem , the @xmath164 known mersenne prime ( which is also the largest known prime ) found under the gimps project ( see @xcite ) has a length of 17,425,170 digits under base 10 . thus , when @xmath9 is sufficiently large , the difference @xmath35 for the swirl network is so enormous as to having tens of millions of digits . [ corollary : swirl ] if there are infinitely many mersenne primes , then there are infinitely many multicast networks with @xmath165 , and moreover , the difference @xmath35 can tend to infinity . another interesting consequence of theorem [ theorem : swirl ] is that the swirl network is linearly solvable over gf(@xmath166 ) , but not over a larger gf(@xmath135 ) when @xmath167 is a prime . this unveils that the linear solvability over a finite field does not even guarantee the linear solvability over all larger finite fields of the _ same _ characteristic . moreover , it was pointed out in @xcite , without demonstrating an explicit example or rigorous proof , that a multicast network linearly solvable over gf(@xmath2 ) and over gf(@xmath3 ) may not be linearly solvable over gf(@xmath168 ) . the swirl network is the first class of explicit networks that asserts this conjecture to be correct . [ corollary : swirl_characteristic ] there exists a multicast network such that it is linearly solvable over gf(@xmath2 ) and over gf(@xmath3 ) but _ not _ over gf(@xmath4 ) . assume @xmath169 . since @xmath170 is a prime whereas @xmath171 are not , the swirl network is linearly solvable over gf(@xmath166 ) and over gf(@xmath172 ) but not over gf(@xmath173 ) . * remark . * the inspiration of designing the class of swirl networks stems from a matroid structure referred to as the _ free swirl _ ( see @xcite or chapter 14 in @xcite ) . the matrix in ( [ eqn : swirl_matrix ] ) can be regarded as a ( linear ) representation of the free swirl of rank @xmath9 when conditions ( [ eqn : swirl_1 ] ) and ( [ eqn : swirl_2 ] ) hold . in a multicast network , the coding vectors of an lnc naturally induces a representable matroid on the edge set , and in the strongest sense , thus the induced representable matroid is referred to as a _ network matroid _ @xcite , in which case the linear independence of coding vectors coincides with the independence structure of edge - disjoint paths @xcite . similar to a network matroid , a _ gammoid _ in matroid theory characterizes the independence of node - disjoint paths in a directed graph . it turns out that if the independence structure of a matroid can be reflected on a multicast network , then the matroid should be not only representable , but ( isomorphic to ) a gammoid as well . being representable and a gammoid is the essential structure of the free swirl that guides the construction of multicast networks we look for . in comparison , those representable matroids , such as fano and non - fano matroids that have well - known applications to construct a multi - source network @xcite , are not isomorphic to any gammoid , and hence they do not have corresponding multicast networks reflecting their independence structures . one may observe that the networks presented in the previous two sections with @xmath34 share a similar layered structure . in order to unify the derivation of their respective linear solvability , we next construct a general multicast network with parameters @xmath174 which subsumes the networks depicted in fig . [ fig : rank_3_networks ] and fig . [ fig : swirl_network ] as special cases , and then derive an equivalent condition for the linear solvability of this general network over a field gf(@xmath0 ) . [ alg : general_construction ] given a 3-tuple of positive integers ( @xmath175 ) as input parameters , the procedure below constructs a multicast network consisting of nodes on five layers , which are labeled 1 - 5 from upstream to downstream , and all edges are between adjacent layers . . create a source @xmath8 , which forms the unique node at layer 1 . . create @xmath9 layer-2 nodes , each of which is connected with @xmath8 by an edge . sequentially label these nodes as @xmath176 . . create @xmath9 layer-3 nodes , labeled as @xmath177 . each node @xmath178 , @xmath179 has 2 incoming edges , one leading from @xmath180 and the other from @xmath181 , where @xmath182 will represent @xmath183 . . for each layer-3 node @xmath178 , @xmath157 , create @xmath184 downstream layer-4 nodes @xmath185 , each of which has the unique incoming edge @xmath186 leading from @xmath187 . corresponding to the layer-3 node @xmath188 , create @xmath189 downstream layer-4 nodes @xmath190 , each of which has the unique incoming edge @xmath191 leading from @xmath188 . . for every set @xmath40 of @xmath9 layer-4 nodes with @xmath109 , create a layer-5 node connected from every node in @xmath40 by an edge . set all layer-5 nodes to be receivers . @xmath60 fig . [ fig : general_network ] depicts the general multicast network constructed by algorithm [ alg : general_construction ] with parameters @xmath174 . it can be easily checked that the networks in fig . [ fig : rank_3_networks](a ) and fig . [ fig : rank_3_networks](b ) , as well as the swirl network are all instances of the general network with respective parameters @xmath192 , @xmath193 , @xmath194 . [ thm : general_network_solvability ] consider the general network constructed by algorithm 1 with parameters @xmath174 . the network is linearly solvable over a finite field gf(@xmath0 ) if and only if there exist @xmath195 , @xmath196 , @xmath197 , @xmath198 such that @xmath199 under which a linear solution over gf(@xmath0 ) can be given by prescribing the coding vectors @xmath200 \nonumber \\ \label{eqn : general_network_matrix_reduced } & = \left [ \begin{array}{ccccccccccccc } 1 & \cdots & 1 & 0 & \cdots & 0 & & 0 & \cdots & 0 & 1 & \cdots & 1 \\ \alpha_{11 } & \cdots & \alpha_{1d_1 } & 1 & \cdots & 1 & & \vdots & \ddots & \vdots & 0 & \cdots & 0\\ 0 & \cdots & 0 & \alpha_{21 } & \cdots & \alpha_{2d_1 } & & 0 & \cdots & 0 & 0 & \cdots & 0\\ 0 & \cdots & 0 & 0 & \cdots & 0 & \ddots & 0 & \cdots & 0 & \vdots & \ddots & \vdots \\ \vdots & \ddots & \vdots & \vdots & \ddots & \vdots & & 1 & \cdots & 1 & 0 & \cdots & 0 \\ 0 & \cdots & 0 & 0 & \cdots & 0 & & \alpha_{(\omega-1)1 } & \cdots & \alpha_{(\omega-1)d_1 } & \delta_{1 } & \cdots & \delta_{d_2 } \\ \end{array } \right].\end{aligned}\ ] ] on the network constructed by algorithm [ alg : general_construction ] with parameters @xmath174 , consider an lnc with all coding coefficients being indeterminates . assume the coding vector for the unique incoming edge to node @xmath201 is equal to the @xmath202 @xmath9-dim unit vector . then , the juxtaposition of the coding vectors for edges @xmath186 , @xmath179 can be represented as the @xmath203 matrix @xmath200 \nonumber \\ \label{eqn : matrix_coding_vectors } & = \left [ \begin{array}{ccccccccccccc } \alpha_{11,1 } & \cdots & \alpha_{1d_1,1 } & 0 & \cdots & 0 & & 0 & \cdots & 0 & \delta_{1,1 } & \cdots & \alpha_{d_2,1 } \\ \alpha_{11,2 } & \cdots & \alpha_{1d_1,2 } & \alpha_{21,1 } & \cdots & \alpha_{2d_1,1 } & & \vdots & \ddots & \vdots & 0 & \cdots & 0\\ 0 & \cdots & 0 & \alpha_{21,2 } & \cdots & \alpha_{2d_1,2 } & & 0 & \cdots & 0 & 0 & \cdots & 0\\ 0 & \cdots & 0 & 0 & \cdots & 0 & \ddots & 0 & \cdots & 0 & \vdots & \ddots & \vdots \\ \vdots & \ddots & \vdots & \vdots & \ddots & \vdots & & \alpha_{(\omega-1)1,1 } & \cdots & \alpha_{(\omega-1)d_1,1 } & 0 & \cdots & 0 \\ 0 & \cdots & 0 & 0 & \cdots & 0 & & \alpha_{(\omega-1)1,2 } & \cdots & \alpha_{(\omega-1)d_1,2 } & \delta_{1,2 } & \cdots & \delta_{d_2,2 } \\ \end{array } \right]\end{aligned}\ ] ] where @xmath204 , @xmath205 respectively represent the coding coefficient indeterminates for adjacent pairs @xmath206 and @xmath207 , @xmath157 , @xmath197 , and @xmath208 , @xmath209 respectively represent the coding coefficient indeterminates for adjacent pairs @xmath210 and @xmath211 , @xmath212 . denote by @xmath213 this juxtaposed matrix , and by @xmath214 the set of indeterminates @xmath215 , where @xmath196 , @xmath197 , @xmath198 , and @xmath216 . for every @xmath217 submatrix in @xmath213 , its determinant is a polynomial in indeterminates in @xmath214 . then , for every set @xmath40 of @xmath9 layer-4 nodes , the maximum flow from @xmath8 to @xmath40 is equal to @xmath9 if and only if the juxtaposition of coding vectors for incoming edges to nodes in @xmath40 has a nonzero determinant polynomial . consequently , there is a linear solution over gf(@xmath0 ) if and only if there is a _ matrix completion _ for matrix @xmath213 over gf(@xmath0 ) , that is , an assignment of values in gf(@xmath0 ) to indeterminates in @xmath214 such that for every @xmath217 full - rank submatrix in @xmath213 , the evaluation of its determinant polynomial is nonzero . assume that the network is linearly solvable over gf(@xmath0 ) . then , there is a matrix completion for @xmath213 over gf(@xmath0 ) . consider such a matrix completion that the indeterminates in @xmath214 are assigned to values in gf(@xmath0 ) . observe that the @xmath218 matrix @xmath219 = \left[\begin{smallmatrix } \alpha_{11 , 1 } & \alpha_{12,1 } & 0 & & 0 \\ \alpha_{11 , 2 } & \alpha_{12 , 2 } & \alpha_{21 , 1 } & & \cdots \\ 0 & 0 & \alpha_{21 , 2 } & \cdots & 0 \\ \cdots & \cdots & 0 & & \alpha_{(\omega-1)1 , 1 } \\ 0 & 0 & 0 & & \alpha_{(\omega-1)1 , 2 } \end{smallmatrix}\right]$ ] is full rank before matrix completion , and its determinant is equal to @xmath220 . hence , @xmath221 and @xmath222 . by an analogous argument to all those @xmath218 full - rank submatrices of @xmath213 which are in the form @xmath223 $ ] , where @xmath186 represents @xmath224 when @xmath225 , it can be deduced that * @xmath226 , @xmath227 ; * @xmath228 , @xmath229 ; * @xmath230 , @xmath231 . thus , we can form another matrix @xmath232 by the multiplication of @xmath213 and a diagonal square matrix with diagonal entries equal to @xmath233 . any @xmath9 columns in @xmath213 are linearly independent if and only if the @xmath9 columns in @xmath232 of same indices are linearly independent . now , set @xmath234 , where @xmath157 , @xmath197 , and set @xmath235 , where @xmath198 . then , @xmath232 is in the same form as expressed in ( [ eqn : general_network_matrix_reduced ] ) , and * @xmath236 , @xmath237 ; * @xmath238 , @xmath239 @xmath240 . next , observe that the @xmath218 submatrix @xmath241 = \left[\begin{smallmatrix } \alpha_{11 , 1 } & 0 & & 0 & \delta_{1 , 1}\\ \alpha_{11 , 2 } & \alpha_{21 , 1 } & & \cdots & 0 \\ 0 & \alpha_{21 , 2 } & \cdots & 0 & \cdots \\ \cdots & 0 & & \alpha_{(\omega-1)1 , 1 } & 0 \\ 0 & 0 & & \alpha_{(\omega-1)1 , 2 } & \delta_{1 , 2 } \end{smallmatrix}\right]$ ] of @xmath213 is of full rank before matrix completion . hence , the @xmath242 submatrix in @xmath232 with the same column indices @xmath243 $ ] is of full rank . equivalently , its determinant @xmath244 is nonzero . by the same argument to the full - rank @xmath217 submatrices @xmath245 $ ] , @xmath246 and @xmath247 , it can be deduced that @xmath248 thus , the values @xmath249 induced from the considered matrix completion for @xmath213 over gf(@xmath0 ) satisfies condition ( [ eqn : general_network_linear_solvability ] ) . the proof of the necessity part is complete . to prove the sufficiency part , assume that there exist @xmath195 subject to ( [ eqn : general_network_linear_solvability ] ) . then , it can be carefully checked that the assignment @xmath250 qualifies as a matrix completion for the matrix @xmath213 depicted in ( [ eqn : matrix_coding_vectors ] ) over gf(@xmath0 ) , and hence the matrix in ( [ eqn : general_network_matrix_reduced ] ) is the juxtaposition of coding vectors from a linear solution over gf(@xmath0 ) . the network depicted in fig . [ fig : general_network ] with parameters @xmath174 is linearly solvable over gf(@xmath0 ) if gf(@xmath0)@xmath59 has a proper subgroup @xmath127 such that @xmath251 and @xmath252 . let @xmath127 be a subgroup in gf(@xmath0)@xmath59 such that @xmath251 and @xmath252 . for each @xmath157 , assign @xmath253 to be arbitrary @xmath184 elements in @xmath127 , and @xmath254 to be arbitrary @xmath189 elements in @xmath255 . under such assignments , @xmath256 is always contained in @xmath127 and thus condition ( [ eqn : general_network_linear_solvability ] ) is satisfied and the general network is linearly solvable over gf(@xmath0 ) . in the literature , it is well known that @xmath11 is an upper bound on @xmath6 for multicast networks . to the best of our knowledge , the highest lower bound on @xmath6 for multicast networks is of the order @xmath7 . references @xcite and @xcite independently constructed a class of multicast networks whose @xmath6 is lower bounded by @xmath7 . the two classes of networks are essentially the same , and they are based on the class of @xmath257-combination networks with @xmath21 ( the ( 4 , 2)-combination network is depicted in fig . [ fig:4 - 2-combination_networks ] . ) moreover , they share the same property that @xmath16 . inspired by the results in section [ sec : fundamental_results ] , we shall establish a new class of multicast networks with @xmath6 also lower bounded by @xmath7 whereas _ not _ every network in it has @xmath14 smaller than @xmath6 . let us first revisit the network depicted in fig . [ fig : rank_3_networks](a ) . as proved in lemma [ lemma : rank_3_network_a ] , it is linearly solvable over gf(@xmath6 ) where @xmath28 but not over gf(8 ) . the network is assumed to have a receiver connected from every set @xmath40 of 3 grey nodes with @xmath79 . actually , appropriate deletion of some receivers ( e.g. the ones respectively connected from @xmath258 and @xmath259 ) from the network will not affect the linearly solvability over any particular finite field . stemming from this observation , we shall next construct a new class of multicast networks with @xmath6 lower bounded by @xmath7 by appropriately increasing grey nodes and redefining receivers in the network in fig . [ fig : rank_3_networks](a ) . consider the network with @xmath65 depicted in fig . [ fig : network_lowerbound ] . it has a total of @xmath260 grey nodes , which are classified into three sets as @xmath261 , @xmath262 , and @xmath263 . the set of receivers in this network is prescribed as follows : * type - i : corresponding to every ( unordered ) pair @xmath264 of grey nodes in @xmath265 , there is a receiver connected from @xmath266 , where @xmath267 when @xmath268 and @xmath269 otherwise ; * type - ii : corresponding to every ( unordered ) pair @xmath270 of grey nodes in @xmath271 , there is a type - ii receiver connected from @xmath272 , where @xmath267 when @xmath268 and @xmath269 otherwise ; * type - iii : there is a respective receiver connected from @xmath273 for every @xmath274 , from @xmath275 for every @xmath276 , and from @xmath277 . there is also a respective receiver connected from @xmath278 and from @xmath279 for every @xmath280 ; * type - iv : there is a receiver connected from @xmath281 for every possible @xmath282 . the total number @xmath11 of receivers in the network is then equal to @xmath283 . denote by @xmath73 the incoming edge to node @xmath78 for all @xmath284 . the network depicted in fig . [ fig : network_lowerbound ] can be viewed as an instance of the general network constructed by algorithm 1 with parameters @xmath285 , with some particularly chosen receivers deleted . by noticing that such careful deletion of receivers will not affect the linear solvability of the network , we can directly obtain the next lemma as a corollary of theorem [ thm : general_network_solvability ] . [ lemma : lower_bound ] the network depicted in fig . [ fig : network_lowerbound ] is linearly solvable over a given gf(@xmath0 ) if and only if there is an lnc over gf(@xmath0 ) with coding vectors for @xmath286 prescribed by @xmath287 \\ & = \left [ \begin{matrix } 1 & \cdots & 1 & 0 & \cdots & 0 & 1 & 1 & 1 \\ \alpha_1 & \cdots & \alpha_{m } & 1 & \cdots & 1 & 0 & 0 & 0 \\ 0 & \cdots & 0 & \beta_1 & \cdots & \beta_m & \delta_1 & \delta_2 & \delta_3 \end{matrix } \right ] , \end{aligned}\ ] ] where @xmath288 and @xmath289 in this lemma , prescribed by condition ( [ eqn : lower_bound_1 ] ) , the set @xmath290 contains at least @xmath291 elements . hence , condition ( [ eqn : lower_bound_2 ] ) obeys only if @xmath292 . consider the special case that @xmath293 and @xmath294 is equal to a mersenne prime . since there is a subgroup @xmath127 of order @xmath291 in the multiplicative group gf(@xmath294)@xmath59 , and @xmath295 , we are able to assign @xmath296 and @xmath297 subject to ( [ eqn : lower_bound_1 ] ) and ( [ eqn : lower_bound_2 ] ) by setting @xmath298 and @xmath299 to be arbitrary distinct elements in @xmath300 . therefore , the network is linearly solvable over gf(@xmath294 ) . we next show that the network is not linearly solvable over gf(@xmath301 ) . recall the cauchy - davenport theorem ( see @xcite for example ) which asserts that for any two nonempty subsets @xmath302 and @xmath303 in the additive group @xmath304 , where @xmath95 is a prime , the cardinality of @xmath305 contains at least @xmath306 elements . since the multiplicative group gf(@xmath301)@xmath59 is isomorphic to the additive group @xmath307 , where @xmath294 is prime , there are at least @xmath308 elements in @xmath290 as a direct consequence of the cauchy - davenport theorem . hence , there are only at most 2 elements in @xmath309 , so that condition ( [ eqn : lower_bound_2 ] ) can not be satisfied . the above results are summarized as follows . [ theorem : lower_bound ] for the network depicted in fig . [ fig : network_lowerbound ] , @xmath310 , and hence @xmath6 is lower bounded by @xmath7 . when @xmath293 and @xmath294 is equal to a mersenne prime , @xmath311 whereas the network is _ not _ linearly solvable over gf(@xmath301 ) on an acyclic multicast network , if there is a linear solution over gf(@xmath0 ) , could we claim that there is a linear solution over every gf(@xmath23 ) with @xmath312 ? it would be tempting to answer it positively because by the result in @xcite , the claim is correct when @xmath0 is no smaller than the number of receivers and moreover , as a consequence of the result in @xcite , the positive answer is affirmed for the special case that the source dimension of the network is equal to 2 . in the present paper , however , we show the negative answer for general cases by constructing several classes of multicast networks with different emphasis . these networks are the first ones discovered in the network coding literature with the property that @xmath14 , the maximum field size for the nonexistence of a linear solution over gf(@xmath14 ) , is larger than @xmath6 , the minimum field size for the existence of a linear solution over gf(@xmath6 ) . the insight of various exemplifying networks established in the present paper is that not only the field size of gf(@xmath0 ) , but also the _ order of the proper multiplicative subgroup _ of gf(@xmath0)@xmath59 affects the networks linear solvability over gf(@xmath0 ) . the results in this paper bring about a new thread on the fundamental study of linear network coding , specific to the case of multicast networks , as discussed below : * besides the field size and multiplicative subgroup orders , it is not clear whether there are some other undiscovered inherent structures in a finite field that affects the linear solvability of a multicast network . it deserves further investigation . * all multicast networks presented in this paper that are linearly solvable over a field gf(@xmath0 ) but not over a larger field gf(@xmath23 ) share a common property that for some values @xmath313 , there is a proper subgroup @xmath127 in gf(@xmath0 ) subject to @xmath251 and @xmath252 , but there does not exist a proper subgroup @xmath314 in gf(@xmath23 ) subject to @xmath315 and @xmath316 . thus , when @xmath317 does not contain any proper subgroup other than @xmath121 , _ i.e. _ , @xmath318 is a prime , we can simply set @xmath319 and @xmath320 , such that for any values @xmath313 subject to @xmath321 and @xmath322 , the conditions @xmath323 and @xmath324 hold as well . this leads us to an ambitious conjecture that if a multicast network is linearly solvable over gf(@xmath0 ) where @xmath325 is a prime , then it is linearly solvable over all finite fields of sizes larger than @xmath0 . * as special cases of the previous conjecture , it is of particular interest to study : i ) whether a multicast network linearly solvable over gf(2 ) ( or over gf(3 ) ) is linearly solvable over all larger finite fields ; ii ) whether a multicast network linearly solvable over gf(2 ) and over gf(3 ) is linearly solvable over all finite fields . these two cases are related to other interesting conjectures in the network coding literature : it was conjectured and partially proved in @xcite that every planar multicast network is linearly solvable over gf(3 ) , and certain special planar multicast networks ( including relay - coface networks and terminal - coface networks ) are always linearly solvable over both gf(2 ) and gf(3 ) . the conjectures presented here will further suggest that these special multicast networks may be linearly solvable over all finite fields . stemming from the above discussions , we end this paper by proposing a number of open problems , all of which , except for the first , we conjecture to have positive answers : * for a multicast network , what is the smallest prime power @xmath0 larger than @xmath14 ( such that the network is linearly solvable over all gf(@xmath23 ) with @xmath312 ) ? * can the gap @xmath326 tend to infinity ? * are there infinitely many prime power pairs @xmath327 with @xmath328 such that each @xmath327 corresponds to @xmath329 of some multicast network ? * if a multicast network is linearly solvable over such a gf(@xmath0 ) that gf(@xmath0)@xmath59 does not contain any proper multiplicative subgroup other than @xmath121 , is it linearly solvable over all larger finite fields than gf(@xmath0 ) ? * if a multicast network is linearly solvable over gf(2 ) ( or over gf(3 ) ) , is it linearly solvable over all larger finite fields ? * if a multicast network is linearly solvable over both gf(2 ) and gf(3 ) , is it linearly solvable over all finite fields ? observe that the network in fig . [ fig : rank_3_networks](a ) can be constructed by algorithm 1 introduced in section [ sec : general_network ] with parameters @xmath192 . then theorem [ thm : general_network_solvability ] implies that the network in fig . [ fig : rank_3_networks](a ) is linearly solvable over gf(@xmath0 ) if and only if there exist @xmath330 , @xmath331 such that @xmath332 under constraint ( [ eqn : gf_7_gf_8_1 ] ) , the cardinality of @xmath333 is no smaller than 3 and no larger than 9 . thus , * when @xmath334 , there does not exist any assignment of @xmath335 from gf(@xmath0)@xmath59 satisfying ( [ eqn : gf_7_gf_8_2 ] ) , and hence the network is not linearly solvable over gf(@xmath0 ) ; * when @xmath336 , there are always ways to assign @xmath335 from gf(@xmath0)@xmath59 subject to ( [ eqn : gf_7_gf_8_1 ] ) and ( [ eqn : gf_7_gf_8_2 ] ) , and hence the network is linearly solvable over gf(@xmath0 ) . it remains to consider the case @xmath0 = 7 , 8 , 9 , or 11 . when @xmath0 = 7 , 9 , or 11 , there is a proper subgroup @xmath127 of order at least 3 in the multiplicative group @xmath317 . we can then assign arbitrary three distinct values in @xmath127 to @xmath337 , @xmath338 , @xmath339 and to @xmath340 , @xmath341 , @xmath342 . in this way , the cardinality of @xmath333 is upper bounded by @xmath343 and thus there are at least three values remained in @xmath344 , which @xmath345 , @xmath346 , @xmath347 can be assigned to . in the last case @xmath0 = 8 , since there is no proper subgroup in gf@xmath348 , the method depicted in the previous paragraph does not work any more and an exhaustive search will verify that there does not exist any assignment of @xmath335 from gf(@xmath0)@xmath59 satisfying ( [ eqn : gf_7_gf_8_1 ] ) and ( [ eqn : gf_7_gf_8_2 ] ) . we can now affirm that the network depicted in fig . [ fig : rank_3_networks](a ) is linearly solvable over every gf(@xmath0 ) with @xmath80 except for @xmath72 . observer that the network in fig . [ fig : rank_3_networks](b ) can be constructed by algorithm 1 introduced in section [ sec : general_network ] with parameters @xmath193 . then theorem [ thm : general_network_solvability ] implies that the network in fig . [ fig : rank_3_networks](b ) is linearly solvable over gf(@xmath0 ) if and only if there is an assignment of @xmath98 , @xmath99 from gf(@xmath0)@xmath59 subject to @xmath349 under condition ( [ eqn : gf_16_gf_17_1 ] ) , the cardinality of @xmath350 is lower bounded by 5 and upper bounded by 25 . thus , * when @xmath351 , there does not exist any assignment of @xmath98 , @xmath99 from gf(@xmath0)@xmath59 satisfying ( [ eqn : gf_16_gf_17_1 ] ) , and hence the network is not linearly solvable over gf(@xmath0 ) ; * when @xmath352 , there are always ways to assign @xmath98 , @xmath99 subject to ( [ eqn : gf_16_gf_17_1 ] ) and ( [ eqn : gf_16_gf_17_2 ] ) , and hence the network is linearly solvable gf(@xmath0 ) . it remains to consider the case that @xmath353 . note that when @xmath0 is not equal to 17 or 32 , there is also a proper subgroup @xmath127 in the multiplicative group gf(@xmath0)@xmath59 such that @xmath127 has order no smaller than 5 and the cardinality of gf(@xmath0)@xmath354 is at least 10 . then , we can respectively assign any 5 distinct elements in @xmath127 to @xmath355 and to @xmath356 , and assign any 10 distinct elements in gf(@xmath0)@xmath357 to @xmath358 . such an assignment obeys conditions ( [ eqn : gf_16_gf_17_1 ] ) and ( [ eqn : gf_16_gf_17_2 ] ) . when @xmath359 , denote by @xmath105 be a primitive element in gf(32)@xmath59 . assign @xmath360 for all @xmath104 , and @xmath361 for all @xmath362 . it is easy to check that such an assignment obeys conditions ( [ eqn : gf_16_gf_17_1 ] ) and ( [ eqn : gf_16_gf_17_2 ] ) . when @xmath363 , in order to make @xmath364 contains at least 10 elements , @xmath355 and @xmath356 should be such assigned that the cardinality of @xmath365 is no larger than 6 . to minimize this cardinality , as many as @xmath366 and @xmath367 should be assigned to a same proper subgroup in gf(17)@xmath59 . an exhaustive search will then verify that it is infeasible to assign @xmath355 and @xmath356 from @xmath368 so that @xmath365 contains no more than 6 elements . we can now assert that the network depicted in fig . [ fig : rank_3_networks](b ) is linearly solvable over every gf(@xmath0 ) with @xmath96 except for @xmath363 . observe that the swirl network can be constructed by algorithm 1 in section [ sec : general_network ] with parameters @xmath194 . then according to theorem [ thm : general_network_solvability ] , the swirl network is linearly solvable over gf(@xmath0 ) if and only if there exist @xmath369 , @xmath370 subject to @xmath371 given arbitrary @xmath369 , @xmath370 , define @xmath372 and @xmath373 , @xmath374 . it is straightforward to check that condition ( [ eqn : swirl_proof_1 ] ) holds if and only if @xmath375 moreover , because @xmath376 condition ( [ eqn : swirl_proof_2 ] ) holds if and only if @xmath377 we can now conclude that the swirl network is linearly solvable over gf(@xmath0 ) if and only if there exist @xmath112 satisfying conditions ( [ eqn : swirl_1 ] ) and ( [ eqn : swirl_2 ] ) . the authors would like to thank sidharth jaggi for helping identify and motivate the problem studied in this work . the authors would also appreciate the helpful suggestions by the associate editor as well as anonymous reviewers to help improve the presentation of the paper . s. jaggi , p. sanders , p. a. chou , m. effros , s. egner , k. jain , and l. tolhuizen , `` polynomial time algorithms for multicast network code construction , '' _ ieee trans . inf . theory _ 1973 - 1982 , jun . 2005 . t. xiahou , z. li , c. wu , `` information multicast in ( pseudo-)planar networks : efficient network coding over small finite fields , '' _ international symposium on network coding ( netcod ) _ , calgary , canada , jun .

in recent years , planck - scale modifications to particles dispersion relation have been deeply studied for the possibility to formulate some phenomenology of planckian effects in astrophysical and cosmological frameworks . there are some indications [ arxiv : gr - qc/0611024 ] that finsler geometry can provide some generalization of riemannian geometry which may allow to account for non - trivial ( planckian ) structure of relativistic particles configuration space . we investigate the possibility to formalize planck - scale deformations to relativistic models in curved spacetime , within the framework of finsler geometry . we take into account the general strategy of analysis of modifications of dispersion relations in curved spacetimes proposed in [ arxiv:1507.02056 ] , generalizing to the de sitter case the results obtained in [ arxiv:1407.8143 ] , for deformed relativistic particle kinematics in flat spacetime using finsler formalism .

1611.04995

one of the most important definitions in euclidean geometry concerns the norm of a vector and the distance between two points : they are defined using the pythagoras theorem . in the xix century , riemann generalized ideas of gauss , introduced the concept of manifold and constructed the riemannian geometry , in which pythagoras theorem would be only valid at a point or along a line , but would no longer be the way of measuring distances in an open set in the manifold ; instead , such measure would be performed using a general metric field . euclidean geometry was the background of newtonian mechanics and was invariant by galilean boosts and rotations , but the advent of special relativity and a new kind of invariance group inspired the use of a geometry that unified space and time in a way that could keep invariant the interval between two events : such was minkowski geometry , in which the interval represents a distance in such spacetime , that is measured by a formula similar to the euclidean one , however allowing for positive , null or negative norms of vectors : this is a pseudo - euclidean space . the next generalization , that is responsible for a relativistic description of gravity was performed by einstein by writing general relativity using a generalization of the minkowski space by the same terms of the generalization from euclidean to riemannian geometry : spacetime was defined as a differential manifold endowed with a pseudo - riemannian metric . in the geometries just described , one uses a metric to define the norm of vectors and co - vectors and , then defines the length of curves and the distance between points . an important feature of special relativity that gains a geometrical interpretation is the dispersion relation of particles , _ i d est _ the equation that relates energy , momentum and mass , it is defined as the norm of the four - momentum in minkowski spacetime . therefore the pseudo - euclidean nature of the minkowski metric is what allows us to have a dispersion relation of the form @xmath0 . such relation can be generalized to a curved spacetime , where the dispersion relation is defined as the norm with the curved metric , and the minkowskian form is achieved in normal coordinates . on the other hand , some approaches to quantum gravity phenomenology refers to the possibility of having a modified dispersion relation ( mdr ) as a property of the semi - classical limit of quantum gravity theories , for example @xmath1 , where @xmath2 is a deformation parameter that depends on the model ( which in some cases can be assumed to be proportional to the planck length @xcite ) . so , in this context , an immediate question arises : may a mdr be a manifestation of a departure of the riemannian nature of spacetime at a scale sensible for quantum gravity ? if the situation was the spacetime being non - riemannian in a sense that could afford for a non - quadratic norm of vectors , then we would have a mdr , but the inverse is not necessarily true , therefore we can only speculate about it . a known branch of differential geometry that fits perfectly into this property is the so called _ finsler geometry _ @xcite . basically , it is a rigorous formalization for possible non - quadratic norms of vectors , which is achieved through a four - velocity - dependent metric , but still preserving the parametrization invariance of the arc - length of curves in the manifold through the requirement of having a homogeneous in the velocities metric . the relations between finsler geometry and some approaches to the quantum gravity problem have been widely investigated @xcite , with special interest in models introducing some modification to particles dispersion relation ( see for instance ref.@xcite and references therein ) . in particular , in @xcite girelli , et al . pointed out that phenomenological models that implement some kind of departure from lorentz symmetries can be systematically formalized within the framework of finsler geometry , hypothesizing that such a formalism may be also the correct mathematical framework to describe the so called rainbow metrics " approach @xcite . it is still unclear whether the introduction of momentum - dependent rainbow " metrics imposes some sort of breakdown of lorentz symmetry or otherwise they may be suitable for a scenario with a deformation of ( local ) spacetime symmetries through a deformed poincar group as in the deformed special relativity approach @xcite . this issue was preliminarily treated by the authors of ref.@xcite for what concerns the fate of spacetime symmetries in mdr - inspired finsler geometries and was explicitly considered in ref.@xcite , in which it was shown that it is possible to obtain a description of modified relativistic particle kinematics satisfying both finsler geometry and deformed special relativity prescriptions within the so called @xmath3-poincar framework @xcite , at least at first order in the deformation parameter @xmath4 , where ( in units such as @xmath5 ) we can expect @xmath6 to be of the order of planck scale @xmath7 . however satisfactory for what concerns the description of particles kinematics in the flat - spacetime limit , ref.@xcite leaves open the question whether this approach can be generalized to describe deformed relativity particles phenomenology @xcite within a curved background . we would like here to implement this aspect , using as guidance some previous approaches @xcite to @xmath2-deformed particles kinematics in a totally - symmetric curved spacetime , where we will denote the parameter of spacetime curvature as @xmath8 . a clear aspect presented in @xcite is the lack of a finsler metric that would be probed through spacetime inferences with massless particles , because the massless limit of the presented metric is ill - behaved . for example , this behavior prevents us from having a metric probed by neutrinos , since , even being massive , for most practical reasons they are considered massless due to their tiny rest mass . in this paper , besides generalizing some results of ref . @xcite , we propose a different way of calculating a four - velocity - dependent ( and still partially homogeneous in the velocities ) metric , that can reproduce the main features of the standard finsler case , like the dispersion relation , the geodesics as worldlines , the presence of dsr symmetries and , furthermore , presents a well - behaved massless limit , thus allowing for a _ generalized finsler metric _ for any type of particle . such generalization will allow us to propose an expression that deepens the one of jacob and piran @xcite for the mdr - induced time delay for particles with different energies , and to analyze the finsler nature of spacetime using photons / neutrinos as probes . furthermore , we will study in the last section the law of interaction between elementary particles compatible with dsr models , furnishing a description in terms of the tangent space , thus being compatible with the finsler formalism . in this paper , in order to connect with the previous literature @xcite and express the mathematical and physical concepts in an effective and comprehensible way , we will need to rely on different notations and coordinate sets . + the first one is the comoving relative locality coordinate set for momentum - space @xmath9 and spacetime flat slicing coordinates @xmath10 , which satisfy the relation @xmath11 in this coordinate set the hamiltonian , which formalizes the physics of particles embedded in a de sitter - like curved spacetime , with @xmath8 the parameter of curvature and @xmath2 the deformation parameter due to planck - scale effects , can be written as @xmath12 we stress that for sake of simplicity we are working in @xmath13 dimensions at first order in @xmath2 ( since the same kinematical results are preserved in a higher number of dimensions ) . in order to describe physics in an intuitive way , we will also consider the so called _ conformal time coordinatization _ expressed by coordinates @xmath14 and @xmath15 . the relation between the previous coordinate set and this latter one is @xmath16 one can verify the effectiveness of expressing physics in conformal - time coordinates by noticing that the spacetime line - element in this coordinatization can be expressed simply as @xmath17 . previous works analysing such deformed physical frameworks ( see ref.@xcite ) , express this same hamiltonian ( [ flathamilt ] ) depending on canonical variables @xmath18 , deformation parameters @xmath19 and @xmath20 , deformation parameter @xmath21 and cosmological constant @xmath8 . we define the energy and momentum respectively as @xmath22 and spacetime coordinates as @xmath23 . therefore , in conformal coordinates we have @xmath24 it is also easy to verify that @xmath25 and @xmath26 ( as well as @xmath27 and @xmath28 ) still are conjugate variables . + the last couple of variables which will have some importance along this article is the one composed by the so called natural momenta , _ i d est _ those that have the same functional dependence of the charges of the translation generators of our spacetime : @xmath29 using those variables the expression of the hamiltonian ( [ flathamilt ] ) becomes : @xmath30 the physics of particles described by the deformed hamiltonians defined in this section was widely explored in ref.@xcite , however many aspects of the formalization of those phenomena seem to lack in such approaches to the study of deformed relativistic frameworks , such as metric formalism , killing vectors and particle interactions . we would like here to provide a first exploration on how the integration of elements borrowed by the so called deformed momentum - space framework and finsler geometry could enrich this kind of approaches to the study of deformed relativity theories , providing solid theoretical foundations to the formalism and also new suggestions on the phenomenological side . finsler geometry has been introduced in previous articles @xcite to formalize deformed relativistic frameworks defining a so called finsler norm , a homogenous function @xmath31 on tangent space . starting from the particle action in the hamiltonian formalism in flat spacetime , performing a legendre transform in @xmath32 where @xmath33 is hamiltonian ( [ flathamilt ] ) in the limit @xmath34 , and @xmath35 some lagrange multiplier . in general a semi - classical metric structure is effectively encoded in finsler geometry . on the other hand it turns out that @xmath3-poincar group ( that we are here using as foundation of a planck - scale deformed particle dynamics in a maximally symmetric spacetime ) does describe the symmetries of particles living on a de sitter curved momentum - space @xcite and flat spacetime ( see appendix for more details on momentum - space dynamics ) . the two different metric formalisms finsler metric @xmath36 and momentum - space metric @xmath37 could lead to confusion on the metric structures we refer to when we will introduce concepts such as geodesics and invariant line - element . therefore we may need here to briefly review deformed - relativistic lagrangian formalism and the role that those different metrics play in it , in the massless case for sake of simplicity . + the minimization of ( [ genaction ] ) provide the relation between momentum and four - velocity ( here in 1 + 1d ) : @xmath38 then at first order in @xmath2 is not hard to invert this relation , finding some @xmath39 , which makes possible to re - express the casimir in terms of four - velocities : @xmath40 where @xmath41 is the inverse of momentum - space metric expressed in terms od @xmath42 . it is also possible now to find an explicit solution for @xmath43 , by imposing @xmath44 . in the massive case this leads to @xmath45 in the massless case , in general , it is not possible to solve @xmath44 with respect to @xmath35 , however taking into account ( [ cxpunto ] ) and ( [ lambxpunto ] ) we can find very naturally that in the limit @xmath46 , @xmath47 . in this limit then ( [ hameq ] ) in 1 + 1d become @xmath48 and their inversion gives @xmath49 using those last results we find that the lagrangian of our theory can be expressed as @xmath50 where the lagrangian metric @xmath51 is not univocally determined , as we will better discuss later , and should be chosen according to the requests of the theory . as we can see from ( [ genlag ] ) in this kind of deformed - symmetry theories we loose the uniqueness of the metric for spacetime and momentum - space . those different metrics play a different role in the theory and one should be always careful to use the right one in the right formula . for instance for euler - lagrange derived relations , such as geodesic equations or killing equations we are referring to metric @xmath51 . the light cone structure is instead characterized in terms of momentum - space metric , since the on - shell relation is @xmath52 . intriguingly , this defines the invariant ( see refs@xcite ) element for flat spacetime as @xmath53 . + it may be disappointing to have lost the possibility to express our theory with a unique metric , however from ( [ genlag ] ) we can still identify a couple of interesting relations : @xmath54 which highlight the duality between spacetime and momentum - space of this theory , widely discussed in the literature ( see for instance @xcite ) . + in order to fully understand whether deformed relativistic frameworks and finsler geometry are physically equivalent , in @xcite were introduced symmetry transformations derived within the finsler framework with the ones generated by a deformed - poincar group . we want to show here how those generators can be obtained _ a priori _ using finsler killing equations with metric @xmath51 . for the sake of simplicity , we will here find the boost @xmath55 representation in the limit @xmath34 and then generalize this result to the case at all orders in @xmath8 . hamiltonian ( [ flathamilt ] ) at order zero in @xmath8 is of course invariant under translations since @xmath56 however if we try to find the representation of the boost imposing @xmath57 we obtain two possible solutions : @xmath58 or maybe some combination of the two . which one should we choose ? a solution to this problem can be provided by the powerful finsler formalism already explored in the formalization of deformed relativistic frameworks in ref.@xcite . in finsler geometry it is , in fact , possible to define a deformed killing equation , which for flat spacetime has the form @xmath59 we can solve those equations for a generic four - velocity - dependent metric such as @xmath60 obtaining ( see appendix for further details ) a generic representation of the boost generator of the form @xmath61 whose hamiltonian can be obtained though the contraction of momenta with the metric @xmath62 : @xmath63 the deformed poincar algebra is therefore characterized by the following poisson brackets @xmath64 in equation ( [ gencasimir12 ] ) we can recognize the limit for @xmath34 of ( [ flathamilt ] ) , under the condition @xmath65 , which introduces a freedom that we will have to constrain based on the requests of the theory . in general an approach based on curved momentum - spaces work perfectly fine with a diagonal metric , _ i d est _ @xmath66 as well as with other choices . on the other hand , in order to formalize our theory ( [ hamiltonian ] ) as a finsler geometry we will need a homogeneous - like metric , which will force us to the choice @xmath67 , as we will discuss further in this article . now that we have obtained the representation of the boost generator in the @xmath34 limit , knowing the form of the classical de sitter generator @xmath68 it is not so difficult to find its generalization al first order in @xmath2 , _ i d est _ : @xmath69 therefore the algebra of spacetime symmetry generators @xmath70 at all orders in @xmath8 is characterized by the following poisson brackets : @xmath71 of course those relations are , by definition , coherent with the poisson brackets between generators and hamiltonian ( [ hamtrasl ] ) : @xmath72 it should be noticed that the algebra described in this section is compatible with the well - known q - de sitter algebra @xcite , at first order in @xmath2 , under the choice @xmath73 . we started with a non - trivial local structure of spacetime that could be modeled by a curved momentum - space , which is responsible for the definition of a modified dispersion relation and a deformed picture of interactions ( that will be discussed in section [ comp_rule ] ) . now , we can analyze how a spacetime can emerge from these considerations , _ i d est _ how the local structure can interfere on the geometry of the effective manifold that these high - energetic particles probe . we will follow a procedure that resembles the one presented in refs . @xcite . in these references , the authors perform a legendre transformation that links the action of a particle from the hamiltonian to the lagrangian formalism . since in special and general relativity one identifies the action of a massive particle with the arc - length of its trajectory , the authors generalized this assumption to the case of mdrs , defining a non - quadratic norm of the particle s four - velocity vector . which can be modeled by the well studied finsler geometry , that is defined by a four - velocity - dependent norm . the finsler norm satisfies the usual properties : @xmath74 the second one is rather important , since it expresses the homogeneity property of the norm , which in general implies @xmath75 in @xcite it was formalized that a mdr theory can be well expressed in terms of finsler geometry , identifying the finsler norm from the action integral as @xmath76 where @xmath77 is of course the finsler metric . as we will see , this metric depends on the four - velocity and mass of the particle , however , it will not have a smooth limit for massless particles . therefore , we propose an intermediate procedure that allows to define an action that serves both for the massive and massless cases , thus allowing a well - defined , unique metric that may describe a generalized finsler spacetime , that is probed by high - energetic particles . in order to generalize the procedure just described to fully symmetric curved spacetimes , our starting point is the particle action in conformal coordinates with modified dispersion relation @xmath78 , @xmath79=\int d\tau [ \dot{\eta}\omega+\dot{x}\pi-\lambda ( { \cal h}_\omega - m^2)].\ ] ] this is the simple generalization of covariant mechanics in de sitter spacetime . the equations of motion are @xmath80 following the same procedures of @xcite , _ i d est _ performing the legendre transformation , substituting @xmath81 , the lagrangian is @xmath82 minimizing this lagrangian with respect to @xmath35 we get @xmath83 and thus we can identify the finsler norm related to hamiltonian ( [ hamiltonian ] ) as @xmath84 at this point we can make contact with geometry , identifying the semi - classical metric structure effectively encoded in the finsler formalism . the finsler metric is defined through a homogenous function on tangent space @xmath85 so @xmath86 it furnishes the de sitter metric if @xmath87 and furnishes the metric of paper @xcite for @xmath88 , @xmath89 and @xmath90 . the analysis done so far is valid for the case of massive particles , @xmath91 , since the limit @xmath46 corresponds to a singularity , implying that observables measured with such metric involving observations with particles with tiny mass , like neutrinos , would not be well defined . despite this fact , it is possible to solve the geodesic equation for massless particles by making @xmath92 in the geodesic solution just like was done in refs . @xcite . in fact , now that we have a metric for a massive particle , we can minimize the arc - length to find its world line . this way , just as it is done in @xcite , even though the massless limit does not exist for the definition of the metric , it is well - behaved for the expression of the geodesics . in fact , one could solve the geodesic equation that arises by extremizing the action , which furnishes : @xmath93=\frac{\dot{{\cal f}}}{{\cal f}}\left[\dot{q}^{\alpha}+\frac{1}{2}g^{\mu\alpha}\left(\frac{\partial}{\partial\dot{q}^{\mu}}g_{\gamma\beta}\right)\dot{q}^{\gamma}\dot{q}^{\beta}\right],\ ] ] for @xmath94 using the known expressions from the finsler geometry literature , that follow from euler theorem for homogenous functions @xmath95 we can write the geodesic equation as @xmath96 as this finsler function does not depend on the coordinate @xmath28 , we conclude that @xmath97 is a first integral to this problem , this is equivalent to equation ( [ p1 ] ) . as this formalism is invariant by reparametrization of the solution , we can choose the gauge @xmath98 . doing it and substituting in the first constrain , we can solve the equation @xmath99 integrating this equation with initial condition @xmath100 we obtain its simple solution in which incoming photons have signature @xmath101 : @xmath102 the same worldline with spatial origin in @xmath103 can as well be obtained upon integration from @xmath103 to @xmath104 , imposing the on - shell relation to the finsler norm . for a massless , on - shell particle , @xmath92 , @xmath105 , the above equation is @xmath106 this is exactly the worldline found in the literature from the hamiltonian ( [ hamiltonian ] ) , see ref . @xcite . as this is the straightforward generalization of refs . @xcite , it also possesses the same kind of formal difficulties with respect to the massless metric . however , it has been argued , for example in ref @xcite , that physical observables calculated by a macroscopic observer should be performed using the classical , riemannian metric , and in this case , there would be no problem in having such ill - defined metric , because its effects ( apart from its geodesics ) would be unobservable . but we raise the possibility that the natural metric seen " by a particle , like the one we found with a finsler structure , could be the one considered for the calculus of observables in the new geometrical framework . which allows us to a deeper conclusion that observations performed with particles with a non - trivial dispersion relation ( no matter the origin of this deformation ) would indicate a possible non - riemannian structure of the spacetime in which they propagate . thereby , by generalizing the geometrical structure with which one describes a spacetime , the physical observables would change reciprocally . in the previous section we performed a legendre transformation on the action of a particle with a mdr until we found a version of the action that we could identify with the arc - length of the particle s trajectory in a certain geometry . since this was a non - quadratic function , but still invariant under reparametrizations , we could identify a finsler function and derive a metric from it . for the massive case , there was no problem in the definition of the metric , but for the massless one , the metric was ill - defined . such behaviour is due to the fact that the arc - length is the nambu - goto - like version of the action and it is well - known that it can only properly describe and only makes sense its extremization for the massive case and since the metric that would be found should depend on the mass of the particle , it is natural that such kind of approach would be problematic for @xmath46 ( which was not the case in riemannian geometry . a possible solution for such impasse is the use of an unique action for both cases . therefore , we propose to read a metric from the _ polyakov - like action _ of a point particle @xcite . such formalism was initially explored in ref.@xcite . in fact , by performing the legendre transformation in ( [ ham_action ] ) , we ended up with the lagrangian ( [ lagr1 ] ) . as the lagrangian is an analytic function , the action can be uniquely expressed by a taylor expansion in the four - velocities @xmath107 = \int d\tau\left[{\cal l}{\big |}_{\dot{q } = 0 } + \frac{\partial \cal l}{\partial \dot{q}^{\mu}}{\bigg |}_{\dot{q } = 0}\dot{q}^{\mu } + \frac{1}{2!}\frac{\partial^2 \cal l}{\partial \dot{q}^{\mu}\partial \dot{q}^{\nu}}\bigg|_{\dot{q } = 0}\dot{q}^{\mu}\dot{q}^{\nu } + \frac{1}{3!}\frac{\partial ^3 \cal l}{\partial \dot{q}^{\mu}\partial\dot{q}^{\nu}\partial\dot{q}^{\gamma}}\bigg|_{\dot{q } = 0}\dot{q}^{\mu}\dot{q}^{\nu}\dot{q}^{\gamma } + ... + \lambda m^2 \right],\ ] ] where the zeroth and first order terms vanish as well as those of higher than the third order . therefore the action can be expressed as @xmath107=\int d\tau \left[\frac{1}{4\lambda}\ , g_{\mu\nu}(q,\dot{q},\lambda)\ , \dot{q}^{\mu } \dot{q}^{\nu } + \lambda m^2\right],\ ] ] where @xmath108 and we have identified @xmath109 this defines a metric that depends on the point of the manifold and the four - velocity . as we will see , the field @xmath35 allows to distiguish between the massive and massless cases . if @xmath91 , @xmath35 is found to be ( [ lambda ] ) . if we substitute it into the action , we find @xmath110=m\int d\tau \sqrt{g_{\mu\nu}(q,\dot{q})\ , \dot{q}^{\mu } \dot{q}^{\nu}},\ ] ] where the metric is @xmath111 from this expression , one can already realize that such metric is well - defined in the massless limit , since the extra terms behave like @xmath112 , whereas in the standard finsler case we had @xmath113 , which of course explodes in the massless regime . the extremization of this action furnishes the expected geodesics ( which is necessary for consistency ) , given by eq . ( [ solmassive ] ) , which coincide with those obtained using hamilton equations for the hamiltonian subject to the mass - shell condition @xmath114 , in the same parametrization . for @xmath92 , the lagrange multiplier @xmath35 can not be solved . however , we can absorb it into the definition of the parameter and define the usual _ affine parameter _ as @xmath115 . with this definition , the action assumes the familiar form @xmath116=\frac{1}{2}\int ds\ , g_{\mu\nu}(q , q')q'^{\mu}q'^{\nu},\ ] ] where @xmath117 , generalizing the result obtained in ( [ genlag ] ) to curved spacetimes , for the metric ( [ beta_metric ] ) and @xmath67 . moreover , the extremization of this functional , taking into account the massless condition as a fundamental one , _ i d est _ the on - shell condition @xmath118 written in terms of the four - velocities @xmath119 furnishes the geodesic equation ( [ solmassless ] ) . we have shown that a natural geometrical formalism that describes the spacetime probed by particles with a modified dispersion relation consists in the use of a four - velocity - dependent metric , such that the riemannian structure is recovered in the low - energy limit . to give another appearance to this approach we can use the definition of the conjugate momenta @xmath120 and substitute the four - velocities that appear in the metric by the energy - momentum of the particle . in fact , we end up with @xmath121 where @xmath122 . this is how our formalism takes the form of a rainbow metric " , where the metric that an observer assigns to the spacetime by using high - energetic particles depends on its energy - momentum . as can be seen , once the metric is written this way , it is evident the smooth limit between the massless and massive cases , since the dependence on the mass has disappeared , remaining just the momenta one . this was not the case for the same substitution in the standard finsler case , as can be verified in ref . @xcite already in the flat case , which is still true by adding curvature to the spacetime . the finsler geometry is a formulation in which the arc - length functional plays a central role , and from it , one can derive a metric tensor that depends homogeneously on tangent - space elements . in this paper , we review a way of defining a four - velocity - dependent metric that shares many properties of the finsler one , but is derived from a polyakov - like action for a particle that obeys a modified dispersion relation and generates the same basic features , like the dispersion relation from a norm and geodesics . one of our achievements is the smooth limit that the metric presents when passing from the massive to the massless case . we would like to probe the effects of such metric structure observing the travel - time through cosmological distances of a particle sufficiently energetic . for the sake of measurements , since they are performed using a macroscopic apparatus , usually it is considered the classical riemannian metric for the definition of observables @xcite , for example , the proper time of an observer on earth , which is used to derive the expression for the time - delay of high energetic photons emitted from a grb @xcite . since the metric is riemannian , its components do not depend on the energy / frequency of the probed particle . the formalism of finsler geometry that we just described allows to relax this hypothesis , as we see in this section . the finsler metric is a function that depends on spacetime points and vector fields , _ i d est _ to calculate the inner product of two vectors at a given point , it is also necessary to determine a direction or a third vector at that point . denoting @xmath123 the metric tensor that depends on @xmath124 , then the inner product of vectors @xmath125 and @xmath126 is @xmath127 , it is then , necessary to specify the three vectors @xmath128 to measure the inner product . _ however , for each fixed @xmath129 , we have that @xmath130 is a riemannian metric . _ therefore , it is possible to calculate a vector - directed riemannian scalar product at a given point for a fixed @xmath131 , if this vector is defined at that point . see for instance ref.@xcite for an analysis of vector dependent tensor calculus . as aforesaid , in formalisms with energy - dependent or rainbow " metrics , this metric is considered as the one that a particle probes " , but the one that an observer assigns to the spacetime needs to be the classic one , because if it also was rainbow - like , the observer s metric would depend on the product between the observer s mass and the planck length , which would be in obvious contradiction with the current observations , where @xmath132 is the number of elementary particles that constitute the observer , which gives a negligible contribution . ] . the formalism we describe here allows to circumvent the impossibility of the last paragraph by showing that it is possible to have a riemannian metric , induced by a finsler one by fixing the vector which the metric depends on as being the integral line of the particle that probes the rainbow " metric , _ @xmath133 , where @xmath131 is the particle s four - velocity , is the _ riemannian metric constructed by an observer by means of measurements with a particle with four - velocity @xmath131 . _ to visualize the reasonability of this proposal , consider the following example . suppose that the planck scale was not so far from our experience , such that spacetime inferences could be always performed with particles that manifest modified dispersion relations . since we are hypothesizing that the spacetime probed by such particles is deformed with respect to the standard riemannian one by means of a momenta - dependent metric , if an observer alice that makes inferences using these particles intends to preserve the equivalence principle , she should assign such metric structure to the spacetime manifold . for example , the geometrical locus of the photon sphere of a black hole ( bh ) is a property of the photon s worldlines , and if she wants to preserve the equivalence principle by representing them as geodesics , they would need to be compatible with the metric structure of the spacetime surrounding a bh , which would need to be of rainbow nature . in other words , the map of a spacetime constructed by such observer would need to inform such energy - dependent behavior . so , following this principle , another example consists in the measurement of the proper - time , elapsed on earth from the emission of a particle from a cosmological source up to its arrival at a terrestrial detector . as the spacetime geometry inferred using a particle with the same energy is determined by the deformed metric , the proper - time is calculated as @xmath134 , which in the comoving frame is @xmath135 , where @xmath136 is the rainbow metric that depends on the energy of the detected particle . and considering the standard riemannian metric . ] therefore , for the measurement of the time delay between massless particles from grbs @xcite ( which is one of the standard sources of constrains for mdrs ) , we define it analogously to the shapiro delay @xcite . in this case , the delay was calculated as the difference in the proper time of earth between the time taken for a radar signal to reach a target and return depending if there is a massive object that deforms the spacetime geometry probed by the photon and the observer , _ i d est _ in the sending and receiving of the signal one compares the proper times calculated in first and zeroth order in the mass parameter , for example in a schwarzschild spacetime . equivalently , in this model , depending on the energy of the probed particle , the induced metric will be the ordinary de sitter one induced by a standard dispersion relation or will manifest a first order correction if the particle is sufficiently energetic , induced by the corresponding mdr , which is similar to the first order correction in the mass that deforms the flat metric in the shapiro delay . in our case , the parameter is not the mass of the schwarzschild metric , but it is the planck length @xmath2 . what we define as a time delay for the arrival of photons is the difference between the proper times ( measured on earth ) that record the time elapsed from the emission of massless particles at @xmath137 , to their arrival at the detector , labeled as @xmath138 . for each photon , the proper time elapsed is @xmath139 in this case , @xmath140 is the metric s component in the deformed riemannian geometry induced by a finsler metric that the observer associates to the spacetime from eq . ( [ metricem ] ) , where the photon s four - velocity / momenta is the fixed vector that induces the deformed riemannian metric . in the coordinates that we have been using @xmath141 . integrating the proper time , @xmath142 the redshift @xmath143 is defined as @xmath144 now , consider that we have a hard and a soft massless particle , for which @xmath145 , @xmath146,\\ x_s(\eta)&=&\eta.\end{aligned}\ ] ] suppose that both particles are emitted simultaneously , _ @xmath147 , at the same spatial point @xmath148 ( see fig.[fig : timedelay ] ) . let @xmath138 be the point where the detector on earth is . the elapsed parameter necessary for the soft particle to reach the detector , which is located at the origin is @xmath149 . and @xmath19 is respectively @xmath150 or @xmath151.,title="fig : " ] [ fig : timedelay ] to determine the elapsed parameter , denoted by @xmath152 , necessary for the hard trajectory to reach the same spacial point , we consider @xmath153 since we normalize the scale - factor for @xmath154 , and the redshift @xmath143 is an experimental parameter determined by soft particles , it is natural to define it as @xmath155 so , the proper time elapsed on earth until the arrival of the hard particle can be calculated using equations ( [ x0hb ] ) and ( [ x0sb ] ) as @xmath156.\end{aligned}\ ] ] the time delay is defined as the difference between the proper times @xmath157 and @xmath158 , therefore it is @xmath159.\ ] ] such expression contains the usual one used for the calculus of time - delays presented in @xcite , added to an extra contribution due to the deformed metric that the observer assigns to spacetime when performing measurements with high - energetic particles . the previous approach could not distinguish between the parameters @xmath19 and @xmath20 , since the effect appeared as a factor of @xmath160 . our new proposal allows to distinguish them , and introduces a phenomenological rule for the time - delay in a more flexible shape @xmath161 in our case , @xmath162 and @xmath163 . another interesting correction is the redshift of the frequency of photons in this geometry . different from the previous case in which the metric played a central role , for calculating this quantity it is only necessary to consider the equations of motion of the photon . in fact , in comoving coordinates , the equations of motion for spacetime coordinates and the energy for a massless particle are ( [ uxoux1 ] ) @xmath164 the solution for the energy is @xmath165 where @xmath166 is the energy of the photon for @xmath167 . , while the dotted - blue one for @xmath150 . the red line represents the classical undeformed cosmological redshift effect , for a low - energetic particle.,title="fig : " ] [ fig : lateshift ] for on - shell massless particles we have @xmath118 , which is simply @xmath168.\ ] ] substituting equation ( [ energy_massless ] ) and ( [ on - shell - massless ] ) in the first equation of ( [ uxoux2 ] ) , we find @xmath169,\ ] ] which can be solved for the parameter @xmath170 as @xmath171.\ ] ] substituting this expression in the dependence of the energy with the parameter @xmath170 given by ( [ energy_massless ] ) , we derive the cosmological redshift of the frequency of a photon that obeys a modified dispersion relation ( see fig . [ fig : lateshift ] ) @xmath172 it was explained in @xcite , using an argument already introduced in @xcite , that if one assumes deformed spacetime symmetries , then if we assume that whenever two particles interact , different observers must agree on the existence of the interaction vertex , this latter may be deformed too . this feature is formalized in the @xmath3-poincar framework , introducing a deformed composition law for momenta @xmath173 which we can imagine as describing the decaying of a particle with momentum @xmath174 in two different particles with momenta @xmath175 and @xmath176 . at first order in the deformation parameter @xmath2 we can express all the possible composition rules of momenta ( requiring invariance under the parity transformation @xmath177 , @xmath178 , see @xcite ) relying just on four parameters @xmath179 as @xmath180 such deformed composition rules are known to be related to the coproduct of translation generators . then , in order to determine which coalgebric sector can be compatible with our deformed boost generator ( [ boostrepresentation ] ) , we need to find some condition which would allow us to find some relations between the composition rule parameters . those relations can be found by requiring the composition of the boosted momenta to be invariant under the action of the boost @xmath181 however , before checking whether different ( boosted ) observers agree on the existence of the particles vertex , we should pay attention to the possibility to have the so called _ backreaction _ effect @xcite , since the rapidity parameter @xmath182 can in general change in a momentum - dependent way , compatible with the coproducts of momenta and the action of lorentz transformations on momenta themselves . this effect is defined as the right action @xmath183 , that we can parametrize as @xmath184 therefore , in general we will have @xmath185 now , imposing the complete parametrization of vertex transformation ( [ rightsumrule ] ) with ( [ boostaction ] ) and ( [ rightactions ] ) , we are able to find four relations between our eight parameters @xmath186 , which ensure the existence od the vertex for any observer : @xmath187 those relations are perfectly compatible with the hamiltonian invariance under boost : @xmath188 which shows _ a posteriori _ the coherence of those composition laws with the curved momentum - space framework . the problem is if also finsler geometry can be compatible to such deformed sum " . in finsler geometry , in fact , the particle trajectory depends on spacetime coordinates @xmath189 and four - velocities @xmath190 , which live in a flat tangent space to the curved spacetime . on the other hand , in relative locality @xcite momenta are the fundamental observables and spacetime is defined as the cotangent bundle to the curved momentum - space . if the deformation of momenta composition law fits very well with the curved momentum - space framework , one should not give for granted for this to happen also in the finsler case . however in this section we showed that non trivial composition laws are imposed by the deformation of spacetime symmetries that arise from our theory . the impossibility to express this feature also for finsler formalism would then jeopardise our entire argument . from our perspective , the problem of the tangent space flatness is not a real problem . in fact every local observer is flat , and curvature arises only when confronting observations between different reference frames . assuming then ( [ nontrivcomp ] ) , we are treating different particles as different reference frames in momentum - space , which can be characterized also as a non - trivial composition for four - velocities , formalizing the relations between the different locally - flat reference frames . this feature can be obtained taking into account a two - parameters family of coproducts and momenta compositions that we can obtain from ( [ rappparamgener ] ) : @xmath191 and @xmath192 then , using the expression of four - velocities with respect to momenta ( [ uxoux1 ] ) with ( [ composition0 ] ) and ( [ composition1 ] ) , in the massless case , we can easily obtain the composition laws for @xmath42 : @xmath193 those laws satisfy the relation @xmath194 where the backreaction has the exact same form of the one described in the previous section . a nice property that we can find _ a posteriori _ in the massless case is @xmath195 in which @xmath196 is the velocity that we already found in our massless worldlines ( [ solmassless ] ) . this structure of relations between different tangent spaces allows us to identify the phase - space invariant element as : @xmath197 the doubt we had at the beginning of this section was whether it is possible to express departures from poincar symmetries , often formalized in the literature as a deformed hamiltonian framework ( @xmath2-deformed phase space ) , within the context of the lagrange - finsler geometry ( tangent space ) . as commented in @xcite finsler geometry could be used to describe both breakdown and deformation of spacetime symmetries . however we have showed here that this symmetry deformation implies the necessity to characterize velocities as living in a tangent space with non - trivial translation laws . those laws moreover result to be compatible with the phase - space deformation of the theory ( [ flathamilt ] ) . the question of what may be the nature of the spacetime emerging from a semi - classical limit of a quantum description of the gravitational degrees of freedom is still open . possibly , we may have a dimensional reduction coming from cdt , hoava - lifshitz gravity and others ( see @xcite and references therein ) , non - local metrics @xcite , rainbow - like metrics coming from a quantization of geometrical and matter - like degrees of freedom @xcite and , as we have shown , a possible finsler nature coming from a mdr . in all these cases there exists a common point : the non - riemannian nature of spacetime . which is not a complete surprise , after all there is no _ a priori _ request for nature of spacetime to be limited to riemannian geometry at all scales , despite the abundance of well - defined geometrical structures ( not to mention those that are still to be discovered ) . in this paper , we analyzed the possibility of having a deformed dispersion relation within a curved background , investigating what kind of manifold could be the most natural to `` host '' them . we generalized the results presented in @xcite and we proposed a solution for the massless limit problem , _ i d est _ we found a candidate partially - homogeneous metric using , which are possible to formalize , the most important properties that a proper geometrical model for the description of elementary particles should provide , such as the geodesic equation and the dispersion relation . the presence of such unified finsler structure that an observer would assign to spacetime , when performing measurements observing high - energetic particles , allowed us to formalize the usual phenomenology of time delays of photons from grbs due to the energy - dependent speed of light presented in @xcite , plus a correction of the order of the planck length originated from the non - riemannian structure of spacetime , with these both contributions having the same geometrical origin . we also estimated the correction that such a model could bring to the cosmological redshift effect . in this context , it would be extremely interesting to test such effect using cosmological data , however unfortunately the best observations on some kind of energy dependence for the cosmological redshift yet published @xcite do not have the sensitivity to discriminate the kind of effect we found here . the authors found no energy dependence with a precision of @xmath198 , whereas in the visible range of electromagnetic radiation ( @xmath199 ) the effect we take into account would be of the order of @xmath200 on the redshift magnitude . in any case , we need phenomenological analysis with more energetic particles . the last open issue we explored is the deformation of velocities composition law in the finsler tangent space that symmetrical @xmath2-deformed theories require ( as already pointed out in @xcite ) , in order to allow to different observers to agree on the occurrence ( or not ) of some interaction event . we found that introducing a slightly more complicated formalism inspired by hopf algebras literature it is still possible to obtain a full relativistic description for particles interactions . nl acknowledges support by the european union seventh framework programme ( fp7 2007 - 2013 ) under grant agreement 291823 marie curie fp7-people-2011-cofund ( the new international fellowship mobility programme for experienced researchers in croatia - newfelpro ) , and also partial support by the h2020 twinning project n@xmath201 692194 , rbi - twinning . ipl is supported by the international cooperation program capes - icranet financed by capes - brazilian federal agency for support and evaluation of graduate education within the ministry of education of brazil grant bex 14632/13 - 6 . fn acknowledges support from conacyt grant no . first of all let us define a generic momentum - space metric depending on our two parameters @xmath202 , at first order in the deformation parameter @xmath2 : @xmath203 as well established in literature @xcite we can find the hamiltonian by integrating the momentum - space invariant line - element @xmath204 where the momentum - space geodesics @xmath205 are defined by the geodesic equation : @xmath206 in which the connections are our usual momentum - space christoffel symbols : @xmath207 given our generic metric , then , we just have to solve the following equations : @xmath208 since @xmath209 , we observe that in our case the geodesic can be expressed as @xmath210 let us now impose the initial values for the geodesic as @xmath211 and @xmath212 . it is easy to notice that they imply @xmath213 and @xmath214 . we have now all we need in order to solve our equations , _ i d est _ : @xmath215 and also @xmath216 therefore we are now able to solve integral ( [ intcasimir ] ) , and find the generic hamiltonian expression : @xmath217 in the introductory section we have shown that the lagrangian - hamiltonian legendre transformation defines the relation @xmath218 since eq . ( [ cxpunto ] ) basically states that the first term of [ b1 ] must be invariant under boost , then we have that @xmath219 therefore @xmath220 however as already noticed in @xcite , being spacetime flat we have @xmath221 , and then the second term of last equation is just a total derivative : @xmath222 which can be eliminated during the integration procedure . the killing equation of our theory is then equivalent to ( [ killfinslflat ] ) , being @xmath223 excepted total derivative terms . + given this , from killing equation ( [ killfinslflat ] ) we find the system of differential equations we need to solve in order to find our symmetry generators : @xmath224 we look for perturbative solutions at first order in @xmath2 such as @xmath225 in which the 0th order solutions are the minkowskian ones : @xmath226 this leads to @xmath227 from system ( [ eq : killsysparam ] ) first and second equation we just obtain that the generalized killing vector components should have the form : @xmath228 now using the results obtained ( [ kill1 ] ) with the third equation , re - deriving this one we get @xmath229 then , using ( [ xi1imperfetta ] ) again with the third equation of system ( [ eq : killsysparam ] ) we obtain the relation between @xmath230 and @xmath231 which is : @xmath232 we can now express the complete solutions for the @xmath233 as @xmath234 the natural choice @xmath235 gives us the total charge , in fact it is easy to verify that in this case : @xmath236 the choice @xmath237 works fine in the diagonal case metric . however finsler formalism relies on homogeneous metrics , which in general have non zero off - diagonal terms . the total charge general expression imposing no particular condition to our parameters at first order in @xmath2 has the form @xmath238 in which @xmath239 , @xmath240 and @xmath55 are the deformed - poincar algebra generators in 1 + 1d . in order to obtain a coherent physical framework ( _ i d est _ invariant casimir and particles worldlines ) also in the finsler case , we need then to impose @xmath241 . this choice eliminates any direct sign of the metrics off - diagonal terms in the boost representation ( [ boostrepresentation ] ) . g. amelino - camelia , `` quantum - spacetime phenomenology , '' living rev . * 16 * , 5 ( 2013 ) doi:10.12942/lrr-2013 - 5 [ arxiv:0806.0339 [ gr - qc ] ] . h. rund , _ the differential geometry of finsler spaces . _ springer , berlin ( 1959 ) . n. e. mavromatos , `` lorentz invariance violation from string theory , '' pos qg * -ph * , 027 ( 2007 ) [ arxiv:0708.2250 [ hep - th ] ] . s. i. vacaru , `` finsler branes and quantum gravity phenomenology with lorentz symmetry violations , '' class . * 28 * , 215001 ( 2011 ) doi:10.1088/0264 - 9381/28/21/215001 [ arxiv:1008.4912 [ math - ph ] ] . f. girelli , s. liberati and l. sindoni , `` planck - scale modified dispersion relations and finsler geometry , '' phys . rev . d * 75 * , 064015 ( 2007 ) doi:10.1103/physrevd.75.064015 [ gr - qc/0611024 ] . j. magueijo and l. smolin , `` gravity s rainbow , '' class . * 21 * , 1725 ( 2004 ) doi:10.1088/0264 - 9381/21/7/001 [ gr - qc/0305055 ] . g. amelino - camelia , `` relativity in space - times with short distance structure governed by an observer independent ( planckian ) length scale , '' int . j. mod . d * 11 * , 35 ( 2002 ) doi:10.1142/s0218271802001330 [ gr - qc/0012051 ] . g. amelino - camelia , l. barcaroli , g. gubitosi , s. liberati and n. loret , `` realization of doubly special relativistic symmetries in finsler geometries , '' phys . d * 90 * , no . 12 , 125030 ( 2014 ) doi:10.1103/physrevd.90.125030 [ arxiv:1407.8143 [ gr - qc ] ] . j. lukierski , h. ruegg , a. nowicki and v. n. tolstoy , `` q - deformation of poincar algebra , '' phys . b * 264 * , 331 ( 1991 ) ; j. lukierski , a. nowicki and h. ruegg , `` new quantum poincar algebra and @xmath3-deformed field theory , '' phys . b * 293 * , 344 ( 1992 ) . doi:10.1016/0370 - 2693(92)90894-a . j. lukierski and h. ruegg , `` quantum @xmath3-poincar in any dimension , '' phys . b * 329 * , 189 ( 1994 ) doi:10.1016/0370 - 2693(94)90759 - 5 [ hep - th/9310117 ] ; j. lukierski , h. ruegg and w. j. zakrzewski , `` classical quantum mechanics of free @xmath3-relativistic systems , '' annals phys . * 243 * , 90 ( 1995 ) doi:10.1006/aphy.1995.1092 [ hep - th/9312153 ] . g. amelino - camelia , `` are we at the dawn of quantum gravity phenomenology ? , '' lect . notes phys . * 541 * , 1 ( 2000 ) [ gr - qc/9910089 ] . t. jacobson , s. liberati and d. mattingly , `` a strong astrophysical constraint on the violation of special relativity by quantum gravity , '' nature * 424 * , 1019 ( 2003 ) doi:10.1038/nature01882 [ astro - ph/0212190 ] . g. amelino - camelia , g. gubitosi , n. loret , f. mercati and g. rosati , `` weakness of accelerator bounds on departures from lorentz symmetry for the electron , '' europhys . lett . * 99 * , 21001 ( 2012 ) doi:10.1209/0295 - 5075/99/21001 [ arxiv:1111.0993 [ hep - ph ] ] . g. amelino - camelia , l. barcaroli , g. damico , n. loret and g. rosati , `` icecube and grb neutrinos propagating in quantum spacetime , '' phys . b * 761 * , 318 ( 2016 ) doi:10.1016/j.physletb.2016.07.075 [ arxiv:1605.00496 [ gr - qc ] ] . g. amelino - camelia , a. marciano , m. matassa and g. rosati , `` deformed lorentz symmetry and relative locality in a curved / expanding spacetime , '' phys . rev . d * 86 * , 124035 ( 2012 ) doi:10.1103/physrevd.86.124035 [ arxiv:1206.5315 [ hep - th ] ] ; g. rosati , g. amelino - camelia , a. marciano and m. matassa , `` planck - scale - modified dispersion relations in frw spacetime , '' phys . rev . d * 92 * , no . 12 , 124042 ( 2015 ) doi:10.1103/physrevd.92.124042 [ arxiv:1507.02056 [ hep - th ] ] . u. jacob and t. piran , `` lorentz - violation - induced arrival delays of cosmological particles , '' jcap * 0801 * , 031 ( 2008 ) doi:10.1088/1475 - 7516/2008/01/031 [ arxiv:0712.2170 [ astro - ph ] ] . g. amelino - camelia , m. matassa , f. mercati and g. rosati , `` taming nonlocality in theories with planck - scale deformed lorentz symmetry , '' phys . * 106 * , 071301 ( 2011 ) doi:10.1103/physrevlett.106.071301 [ arxiv:1006.2126 [ gr - qc ] ] . g. amelino - camelia , l. freidel , j. kowalski - glikman and l. smolin , `` the principle of relative locality , '' phys . d * 84 * , 084010 ( 2011 ) doi:10.1103/physrevd.84.084010 [ arxiv:1101.0931 [ hep - th ] ] . g. amelino - camelia , n. loret and g. rosati , `` speed of particles and a relativity of locality in @xmath3-minkowski quantum spacetime , '' phys . b * 700 * , 150 ( 2011 ) doi:10.1016/j.physletb.2011.04.054 [ arxiv:1102.4637 [ hep - th ] ] . g. amelino - camelia , m. arzano , j. kowalski - glikman , g. rosati and g. trevisan , `` relative - locality distant observers and the phenomenology of momentum - space geometry , '' class . quant . * 29 * , 075007 ( 2012 ) doi:10.1088/0264 - 9381/29/7/075007 [ arxiv:1107.1724 [ hep - th ] ] . g. amelino - camelia , l. barcaroli , g. gubitosi and n. loret , `` dual redshift on planck - scale - curved momentum spaces , '' class . grav . * 30 * , 235002 ( 2013 ) doi:10.1088/0264 - 9381/30/23/235002 [ arxiv:1305.5062 [ gr - qc ] ] . n. loret , `` exploring special relative locality with de sitter momentum - space , '' phys . d * 90 * , no . 12 , 124013 ( 2014 ) doi:10.1103/physrevd.90.124013 [ arxiv:1404.5093 [ hep - th ] ] . g. gubitosi and f. mercati , `` relative locality in @xmath3-poincar , '' class . grav . * 30 * , 145002 ( 2013 ) doi:10.1088/0264 - 9381/30/14/145002 [ arxiv:1106.5710 [ gr - qc ] ] . a. ballesteros , n. r. bruno and f. j. herranz , `` on ( 3 + 1 ) anti - de sitter and de sitter lie bialgebras with dimensionful deformation parameters , '' czech . j. phys . * 54 * , 1321 ( 2004 ) doi:10.1007/s10582 - 004 - 9795-x [ hep - th/0408196 ] ; j. lukierski , a. nowicki and h. ruegg , `` real forms of complex quantum anti - de sitter algebra u - q(sp(4:c ) ) and their contraction schemes , '' phys . b * 271 * , 321 ( 1991 ) doi:10.1016/0370 - 2693(91)90094 - 7 [ hep - th/9108018 ] . a. marciano , g. amelino - camelia , n. r. bruno , g. gubitosi , g. mandanici and a. melchiorri , `` interplay between curvature and planck - scale effects in astrophysics and cosmology , '' jcap * 1006 * , 030 ( 2010 ) doi:10.1088/1475 - 7516/2010/06/030 [ arxiv:1004.1110 [ gr - qc ] ] . l. barcaroli and g. gubitosi , `` kinematics of particles with quantum - de sitter - inspired symmetries , '' phys . rev . d * 93 * , no . 12 , 124063 ( 2016 ) doi:10.1103/physrevd.93.124063 [ arxiv:1512.03462 [ gr - qc ] ] . m. assanioussi , a. dapor and j. lewandowski , `` rainbow metric from quantum gravity , '' phys . lett . b * 751 * , 302 ( 2015 ) doi:10.1016/j.physletb.2015.10.043 [ arxiv:1412.6000 [ gr - qc ] ] . j. polchinski , _ string theory , vol . 1 : introduction to the bosonic string _ cambridge university press ( 1998 ) . i. p. lobo , n. loret and f. nettel , `` rainbows without unicorns : metric structures in theories with modified dispersion relations , '' arxiv:1610.04277 [ gr - qc ] . g. g. carvalho , i. p. lobo and e. bittencourt , `` extended disformal approach in the scenario of rainbow gravity , '' phys . d * 93 * , no . 4 , 044005 ( 2016 ) doi:10.1103/physrevd.93.044005 [ arxiv:1511.00495 [ gr - qc ] ] . m. a. javaloyes , anisotropic tensor calculus , " arxiv:1602.05492 [ math.dg ] . j. m. carmona , j. l. cortes and f. mercati , `` relativistic kinematics beyond special relativity , '' phys . d * 86 * , 084032 ( 2012 ) doi:10.1103/physrevd.86.084032 [ arxiv:1206.5961 [ hep - th ] ] . s. majid , `` algebraic approach to quantum gravity . noncommutative spacetime , '' in * oriti , d. ( ed . ) : approaches to quantum gravity * 466 - 492 [ hep - th/0604130 ] . g. amelino - camelia , m. arzano , g. gubitosi and j. magueijo , `` dimensional reduction in the sky , '' phys . d * 87 * , no . 12 , 123532 ( 2013 ) doi:10.1103/physrevd.87.123532 [ arxiv:1305.3153 [ gr - qc ] ] . t. padmanabhan , s. chakraborty and d. kothawala , `` spacetime with zero point length is two - dimensional at the planck scale , '' gen . rel . grav . * 48 * , no . 5 , 55 ( 2016 ) doi:10.1007/s10714 - 016 - 2053 - 2 [ arxiv:1507.05669 [ gr - qc ] ] . i. ferreras and i. trujillo , `` testing the wavelength dependence of cosmological redshift down to @xmath242 , '' astrophys . j. * 825 * , no . 2 , 115 ( 2016 ) doi:10.3847/0004 - 637x/825/2/115 [ arxiv:1605.09380 [ astro-ph.co ] ] .

abell 2256 is a nearby ( ) , rich cluster of galaxies with fascinating observed properties across a range of wavelengths . long believed to represent a cluster merger , recent x - ray and optical results have suggested that in addition to the primary cluster and subcluster there is evidence for a third , poorer system . we present wide - field , high sensitivity 1.4 ghz vla radio observations of abell 2256 in conjunction with optical imaging and additional spectroscopy . over 40 cluster radio galaxies are identified , with optical spectroscopy indicating the emission source ( star formation or agn ) for most of them . while the overall fraction of galaxies exhibiting radio emission is consistent with a large sample of other nearby clusters , we find an increase in the activity level of galaxies belonging to the third system ( hereafter , the `` group '' ) . specifically , the group has relatively more star formation than both the primary cluster and main subcluster . the position of the group is also coincident with the observed cluster radio relic . we suggest that the group recently (.3 gyr ) merged with the primary cluster and that this merger , not the ongoing merger of the primary and the main subcluster , might be responsible for many of the unusual radio properties of abell 2256 . furthermore , the greater star formation activity of the group suggests that the infall of groups is an important driver of galaxy evolution in clusters .

astro-ph0302066

abell 2256 is an historically well - studied cluster of galaxies . it is a rich cluster ( richness class 2 , * ? ? ? * ) , and this fact combined with its proximity and similarity to coma have resulted in numerous observational studies across many wavelengths . investigations of the incidence of radio emission in the direction of abell clusters ( including the first such study done by * ? ? ? * ) included abell 2256 among detections , and subsequent interferometric observations were able to associate radio emission with individual cluster galaxies in abell 2256 ( e.g. , * ? ? ? * ; * ? ? ? * to be referred to as bf76 ) . several of the radio galaxies were shown to be of an unusual class of objects known as `` head - tail '' sources , in which the `` head '' of the radio emission is coincident with the galaxy while the `` tail '' extends out well past the optical boundary of the galaxy . in addition to such galaxies , the cluster also has diffuse radio emission with a very steep spectrum ( @xmath2 , where @xmath3 ) . at x - ray wavelengths , @xcite noted the detection of abell 2256 in the catalogs from the _ uhuru _ mission , placing it among the brighter x - ray clusters of galaxies . it was consequently a popular target for subsequent x - ray missions , including _ einstein , rosat , asca , bepposax _ , and most recently _ chandra_. early work in the optical revealed it to have an unusually large velocity dispersion ( 1351 km s@xmath4 , * ? ? ? the combination of these studies has produced an evolving picture of the nature of abell 2256 . the initial impression of abell 2256 as a prime example of a very rich , dynamically - relaxed system has been dismissed . an increase in the number of measured cluster galaxy velocities indicated probable substructure @xcite , and _ rosat _ resolved the x - ray emission sufficiently to identify two maxima with different temperatures @xcite . a merger scenario would also support the radio observations , as diffuse radio emission apparently unassociated with specific galaxies ( `` halos '' and `` relics '' ) appear to be related to mergers ( e.g. , * ? ? ? * ; * ? ? ? the same is likely true for head - tail radio sources @xcite . recently , the observational evidence that abell 2256 represents multiple clusters in the process of merging has been bolstered further . x - ray observations from _ chandra _ provided much more detail of the infalling subcluster , including its lower temperature than the primary cluster ( * ? ? ? * hereafter smmv02 ) . in fact , these authors noted that the boundary between the infalling subcluster and the primary cluster might be a `` cold front '' @xcite , which occurs when the core of a cooler subcluster has not yet been disrupted by the merger . the authors also noted an x - ray `` shoulder '' coincident with a very steep spectrum radio source to the east of the primary cluster s center . this feature may also be associated with some prior merger activity . * hereafter blc02 ) increased the number of measured cluster velocities from 89 to 277 . the cluster was best decomposed into three separate components , based on both velocity and positional information . the two larger components apparently correspond to the primary cluster and the infalling subcluster , while a third presumably represents a previously unidentified infalling group ( consisting of @xmath5 of the total number of galaxies in the cluster ) . in summary , abell 2256 has been shown to be a dynamically active system involving the merger of multiple systems . consequently , abell 2256 presents an excellent opportunity for study . in this paper , we seek to address two general questions . first , can the dynamics of abell 2256 surmised from the galaxy velocities be related to the radio features of the cluster ? while there are indications that features such as radio halos and relics are associated with cluster mergers , the wealth of data for abell 2256 can potentially elucidate the connection . second , what is the effect of such dynamical activity on member galaxies ? radio and optical studies of several dynamically active clusters have shown that some possess an excess of star - forming galaxies and agn over their relaxed cluster counterparts ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * referred to subsequently as mo03 ) . is there any evidence for more activity either star formation or agn in abell 2256 ? what can abell 2256 tell us about the possible drivers of galaxy evolution in clusters ? in this paper , we present a comprehensive radio and optical investigation of abell 2256 . we have performed wide - field , high sensitivity radio observations of abell 2256 with the national radio astronomy observatory s ( nrao ) very large array ( vla ) . these observations encompassed the entire cluster , from its core out past the abell radius ( an area of over 2 square degrees ) , and identified galaxies whose radio emission amounted to @xmath6 w hz@xmath4 and greater . this corresponds to all agn and galaxies forming stars at rates in excess of 1.2 m@xmath7 yr@xmath4 ( using the relation found in * ? ? ? the radio data are complemented by new wide - field optical images collected in @xmath8 , and optical spectroscopy of many of the radio - selected galaxies . these spectra further increase the velocity database for the cluster , and are used to classify the active galaxies . the paper is organized as follows . in section [ sec : obs ] we give an overview of the observations and data reductions for the radio and optical imaging and the optical spectroscopy . the basic results and analysis are outlined in section [ sec : results ] , followed by their implications in section [ sec : discuss ] . conclusions are summarized in section [ sec : conclude ] . we have adopted @xmath9 km s@xmath4 mpc@xmath4 and @xmath10 , meaning 1 corresponds to 0.98 kpc at abell 2256 . radio observations at 1.4 ghz were made with the vla in two configurations . the higher resolution observations were performed during july 1998 while the vla was in its b configuration , and lower resolution observations were performed during may 1999 in the d configuration . the b configuration provides a resolution of @xmath15 but is insensitive to structures larger than @xmath1120 . abell 2256 includes several extended radio galaxies as well as diffuse emission associated with halo and relic sources ( e.g. , * ? ? ? * ; * ? ? ? * ) , thereby necessitating the inclusion of the lower resolution d array data . the observational strategy paralleled that of other wide - field radio surveys ( e.g. , the nvss and first ; * ? ? ? * ; * ? ? ? a grid of 43 pointings was used to cover the cluster from its core out to large radii ( @xmath1150 , or about 3 mpc ) . these pointings were arranged in a hexagonal pattern , with the primary 37 pointings having a grid spacing of 16 . the remaining 6 pointings formed a smaller hexagon surrounding the central pointing of the main grid . this configuration was adopted to provide near - uniform sensitivity across the cluster , with extra detail in the cluster core . for an excellent overview of the technique , from observation through to construction of the final wide - field image , the reader is directed to condon et al . all of the observations were made in `` line mode '' using seven channels of 3.125mhz bandwidth for each intermediate frequency and polarization . the integration time for each pointing was 17 minutes while in the b configuration , and 8.5 minutes in d. data reduction was performed using nrao s astronomical image processing system ( aips ) , and is described in more detail in mo03 . in practice , the most difficult aspect was determining the optimal set of parameters affecting the raw uv data for each pointing . the goal is a consistent beam shape across all pointings , while adding the least noise to the data and maintaining a beam with good imaging properties ( i.e. , minimal sidelobes ) . the high declination of abell 2256 implies an elongated beam whose position angle changes over the course of the observations . through careful tapering and gridding of the uv data we maintained a beam of @xmath12 across all pointings , and used a circular restoring beam of equal area ( 6.3 ) to produce the final maps for each pointing . the final mosaic had an rms noise of about 40 @xmath13jy beam@xmath4 within 1 mpc of the cluster center , rising to about 60 @xmath13jy beam@xmath4 at the very edges where fewer pointings contribute to the image . these correspond to 5@xmath14 detections of sources down to radio luminosities of @xmath15 w hz@xmath4 and @xmath16 w hz@xmath4 at abell 2256 . in addition , the multi - resolution clean option in the aips task imagr was also used to produce an image including the extended radio emission in the cluster . the result for the central 20 of the cluster is shown in figure [ fig - bigmap ] . the aips task sad ( `` search and destroy '' ) was used to identify sources in the final mosaic . this task was directed to locate all sources with peak fluxes greater than four times the rms noise level determined over the surrounding @xmath17 . sad fits these sources with gaussians and notes their positions and fluxes . it also creates a residual map , which was inspected to locate any extended sources not well fit by gaussians . these were manually added to the lists , with fluxes determined directly by boxing the sources . each radio source was then inspected , and only those with either peak or integral fluxes greater than five times the local noise ( over the surrounding 5 ) were included in the final radio source lists . in total , the radio source list contained about 1300 entries . optical images ( @xmath8 ) of abell 2256 were obtained during july 1999 using the kpno 0.9 meter telescope . these were taken with the mosaic camera , providing a 59 field at a pixel scale of 0.43 . to cover the spacings between the eight individual chips in mosaic , we adopted the standard dither sequence of five images per telescope pointing . the wide field of mosaic allowed abell 2256 to be imaged well past its abell radius in only four pointings . a fifth , deeper pointing was centered on the cluster core . the net exposure time was 750 seconds for each of the four shallower images and 1500 seconds for the deeper central pointing . observations of three landolt selected areas @xcite were performed at a range of airmass ( 1.2 to 2.3 ) to provide photometric calibration . data reduction followed the usual procedures , and details may be found in mo03 . the astrometric registration of the images is good to about 0.4 , as determined from the usno a2.0 catalog @xcite . comparison of the derived magnitudes for the @xmath18 observed standard stars with their published values suggests the photometry is accurate to within 0.05 magnitudes . our galaxy magnitudes were derived using the gunn - oke metric aperture @xcite , which corresponds to a radius of 13.1 kpc using our adopted cosmology . optical galaxies brighter than @xmath19 ( i.e. , @xmath20 using @xmath21 from * ? ? ? * ) were included in the formal list of candidate cluster radio galaxies ( candidate in the sense that they are either cluster members or merely radio galaxies seen in projection ) , as in @xcite . this corresponds to @xmath22 for abell 2256 , adjusted for @xmath23 . the sextractor software package @xcite was used to identify all objects in the fields as well as classify them and calculate their gunn - oke magnitudes . finally , the images were visually inspected to confirm the star / galaxy segregations made by sextractor . the list of optical galaxies was then correlated with the list of radio sources . for this procedure , we adopt a maximum probability for chance coincidence of radio and optical sources of @xmath24 ( for a brief description , see * ? ? ? for abell 2256 , this amounted to a maximum positional separation of just over 10 . three galaxies with separations @xmath258 were removed from the list upon inspection when their radio sources were clearly associated with fainter background galaxies . this meant that all candidate cluster radio galaxies had radio - optical separations of 5 or less . two extended radio galaxies ( the tailed radio galaxies j170330 + 783755 and j170330 + 783955 ) were manually added to the list ( see figures [ fig - koalatad ] and [ fig - flagpole ] ) . the resulting list contained 54 candidate cluster radio galaxies brighter than @xmath26 and within 3 mpc of 17:04:02 + 78:37:55 ( j2000 ) , our adopted center coordinate of abell 2256 . optical spectra of galaxies in abell 2256 were obtained during two separate observing runs in 1999 . in may 1999 , 18 candidate cluster radio galaxies were observed using the kitt peak 2.1-meter telescope and the goldcam spectrograph . using a 2 slit and a 300 line mm@xmath4 grating , spectra covering 37008000@xmath27 at @xmath28 resolution were obtained . the integration time was 900 seconds , and wavelength calibration was achieved using he - ne - ar lamp exposures taken after the science exposures . details may be found in @xcite . numerous cluster galaxies ( including non - radio galaxies ) were also observed using the mx spectrometer @xcite and the steward observatory bok telescope ( 2.3 m ) . this system uses 2 fibers on movable probes to collect up to 32 galaxy spectra in a 45 field , along with 30 sky spectra from fibers `` piggy - backed '' to the target probes . a 400 line mm@xmath4 grating provided wavelength coverage from 4000@xmath27 to 7300@xmath27 at 7@xmath27 resolution . five pointings each of one - hour duration were made . the cluster abell 2255 was also observed during this run , and details for the observations and determination of velocities may be found in @xcite which addresses the kinematics of that cluster . we report velocity measurements for 73 galaxies in table [ tbl - galvel ] . many of these have been published previously in the work done by @xcite and blc02 , but for 22 galaxies the velocities are new . most of these new velocity measurements are at the periphery of the cluster , and in particular to the northwest along the extension of the line connecting the two x - ray peaks . in addition to velocity measurement , we classified the radio galaxies as star forming or agn on the basis of their emission and absorption features . this was achieved using the standard emission line ratio diagnostics ( e.g. , * ? ? ? * ; * ? ? ? * ) after correcting for stellar balmer absorption . our procedures are described in detail in @xcite . unfortunately , the wavelength coverage of the mx spectra did not include [ oii ] for cluster galaxies , preventing us from applying the morphs classification scheme to many of the cluster radio galaxies ( see * ? ? ? * for a description of this classification scheme ) . our velocities and those of blc02 are generally consistent , with our velocities being @xmath29 km s@xmath4 lower on average . as this is within the observational errors , we have not applied any zero point offsets to the velocities in producing a combined velocity database . the cluster includes 294 cluster members , as determined through 3@xmath14 clipping . the systemic velocity was found to be @xmath30 km s@xmath4 , with a velocity dispersion of @xmath31 km s@xmath4 ( using the biweight estimators of * ? ? ? these values were adopted to determine whether the identified radio galaxies were cluster members ; that is , we have accepted all galaxies in the range @xmath32 km s@xmath4 as cluster members . the completeness of our velocity database was determined as a function of both optical magnitude and radial separation from the cluster center , and is depicted in figure [ fig - vcomp ] . for the entire cluster out to 3 mpc in radial extent , we have measured velocities for @xmath33 of those galaxies brighter than the formal @xmath34 limit . a large fraction of the galaxies without velocity measurements are fainter galaxies at the larger radial distances of the study . ulabeledsing a 2 mpc radial limit and ignoring galaxies fainter than @xmath35 , the velocity completeness is about @xmath36 . note that there are an additional 74 galaxies fainter than @xmath34 for which velocities were obtained . using our data supplemented with that of blc02 , 49 of the 54 candidate cluster radio galaxies have measured velocities . of these , 40 are confirmed cluster members . the list of these galaxies and their attributes ( velocities , magnitudes , radio fluxes , and classifications ) may be found in table [ tbl - clusrg ] . prior papers ( bf76 ; * ? ? ? * ; * ? ? ? * ) applied alphabetical designations to the prominent radio features , and we have included these designations in table [ tbl - clusrg ] when they are associated with specific galaxies . in addition , the alphabetical designations have been noted in figure [ fig - bigmap ] . to avoid confusion we will adhere to the following terms and definitions . the radio _ halo _ in abell 2256 is the diffuse emission around the cluster center . its surface brightness is too low for it to be present in our images , although it may be associated with one cluster radio galaxy ( `` d '' in the prior studies ) . the _ relic _ is the diffuse emission to the north of the cluster center , identified as `` g '' and `` h. '' it is often referred to as a halo , although under current usage of the term this is incorrect as it has a high polarization ( up to @xmath37 ; * ? ? ? * ) and is located off the cluster center . lastly , the diffuse emission labeled `` f '' will be referred to as the ultrasteep spectrum source @xcite . it is most likely another relic source . our spectra include 28 of the 40 confirmed cluster radio galaxies , an even split of which are powered by each agn and star formation . among the agn , we associate four galaxies with head - tail radio sources . @xcite identified these four galaxies as definite head - tails ( features a , b , c , and i ; see table [ tbl - clusrg ] and figures [ fig - koalatad ] , [ fig - flagpole ] , and [ fig - flag ] ) , and noted four additional potential head - tails ( f , g , h , and m ) . the resolution of our maps excludes source m as a cluster radio galaxy , and source f ( the ultrasteep spectrum source ; see figure [ fig - f ] ) is unlikely to be a true head - tail source . it is resolved into three components by our observations ( potentially attached with lower surface brightness emission than our observations detect ) , the easternmost of which ( f3 ) is associated with a cluster galaxy ( and probably unassociated with the remainder of the emission of source f , * ? ? ? sources g and h comprise the relic source , with source g apparently associated with a cluster elliptical , as the peak in the radio emission fell only 3.1 from the galaxy ( see figure [ fig - relic1 ] ) . this galaxy may have once been a head - tail radio galaxy which was disrupted by the merger environment ( as suggested by * ? ? ? * and further explored in section [ sec : discuss ] ) . even if only four of the galaxies are associated with head - tail radio sources , this is an unusually high number for a cluster . typical clusters only possess from 02 such sources @xcite . one further cluster agn may be associated with extended emission . the galaxy located at j170448 + 783829 is one of the brighter ellipticals in abell 2256 and is associated with an unresolved 11.4 mjy radio source ( radio feature d in bf76 ; see figure [ fig - center ] ) . it sits in the center of the diffuse cluster halo emission @xcite , and about 1 to the east of the brightest cluster galaxy ( which is not a radio source ) . of the remaining cluster agn for which we have optical spectra , all but one are low - luminosity radio sources with absorption line spectra . the spectrum of j170150 + 790216 is that of a strong liner . the radio emission is unresolved , and the optical image of the galaxy suggests a recent merger or interaction ( see figure [ fig - fig8 ] ) . fourteen cluster radio galaxies are confirmed to be star - forming on the basis of their spectra . the wavelength coverage adopted for the multi - fiber observations did not include cluster [ oii ] emission , preventing us from classifying the galaxies in the morphs scheme . however , we can compensate for this somewhat by assuming that the equivalent width of [ oii ] is @xmath38 that of the h@xmath39-[nii ] equivalent width @xcite . making this assumption , we find that one of the galaxies ( j170534 + 785116 ) is a starburst galaxy , or `` e(b ) '' in the morphs classification scheme . not coincidentally , this galaxy has the largest 1.4 ghz flux of all the cluster star - forming galaxies ( @xmath14.2 mjy ) and was even identified in the early radio study of bf76 ( source `` e '' ) . three of the star - forming galaxies have h@xmath40 absorption sufficient enough to place them in the `` e(a ) '' class ( j165253 + 784629 , j170335 + 785055 , and j171016 + 780207 ) . galaxies with such spectra are likely associated with starbursts in which selective extinction has diminished their [ oii ] emission in dusty hii regions , yet longer - lived a stars have exited ( or dispersed ) these hii regions and produce a strong h@xmath40 absorption feature @xcite . one further galaxy has h@xmath40 absorption which places it close to the e(a ) category ( j170512 + 784814 ) . the remainder of the star - forming galaxies appear to be relatively normal , given the limits of our data . six of the remaining galaxies with measured velocities are background radio galaxies seen in projection on the cluster , and the remaining three are foreground radio galaxies . these galaxies , along with the five candidate cluster radio galaxies which lack velocities , are listed in table [ tbl - morerg ] . relaxing the formal limits of @xmath41 and the radial limit of 3 mpc produces a further four radio galaxies with measured velocities which are also included in table [ tbl - morerg ] . three of these are cluster members , while the fourth is a background galaxy . blc02 note that most of the non - cluster galaxies in their velocity list come from a background group at a velocity of 52,413 km s@xmath4 . three of our background radio galaxies appear to belong to this system , including an apparent fr i source ( j170454 + 782954 , see figure [ fig - backfr1 ] ) . this source has a radio luminosity of @xmath42 w hz@xmath4 and optical magnitude of @xmath43 , typical of fr i sources with wide - angle tail morphology . another obvious radio galaxy can be seen in the figure ( j170515 + 783018 ) , although the velocity of this galaxy is unknown . should it belong to the same system , it would have a radio luminosity of @xmath44 w hz@xmath4 and an absolute magnitude of @xmath45 . lastly , there is faint , diffuse emission just to the north of the fr i source . some of this may be associated with the faint galaxy at the western edge of the emission ( j170450 + 783011 ) , although it is unlikely that this represents a head - tail galaxy at the redshift of the fr i source . if the galaxy were in the same system , its absolute magnitude of @xmath46 would be unusually faint for a head - tail source . how does the radio galaxy population of abell 2256 compare to other nearby clusters ? in mo03 this question was investigated for 20 nearby clusters , including abell 2256 . the analysis used a chi - square test to evaluate the fraction of galaxies with radio emission in a given cluster compared to the same fraction for the composite sample formed by the other 19 clusters . in that paper , the dynamically active cluster abell 2255 was demonstrated to have a highly significant excess of radio galaxies , but abell 2256 appeared relatively normal . this result is not affected by the lack of velocities for five potential members of abell 2256 ; three of these have radio luminosities lower than the limit adopted for the other clusters , and the remaining two do not change any of the statistical results . we note that the lack of an excess of radio galaxies in abell 2256 is consistent with the low blue fraction ( @xmath47 ) for the cluster , reported in @xcite . there is one specific case in which the radio galaxy population of abell 2256 differs from the other clusters , and that is the fraction of optically brightest galaxies ( @xmath48 ) harboring radio sources . there are seven such radio galaxies in abell 2256 , six of which have spectra indicative of agn ( five have absorption line spectra , the sixth is a strong liner ) . while we do not have a spectrum for the seventh galaxy , it is probably also an agn . it is the middle galaxy in a triple galaxy system , the western galaxy of which is a very strong tailed radio galaxy ( see figure [ fig - koalatad ] ) . the excess of these optically - bright radio galaxies is not unexpected , as many prior studies have noted the unusually large number of head - tail radio galaxies in abell 2256 . in addition , the _ chandra _ observations of smmv02 noted at least three x - ray point sources corresponding to cluster elliptical galaxies . abell 2256 is known to contain substructure . blc02 used the kmm algorithm ( kaye s mixture model , see * ? ? ? * ) to decompose the cluster into three separate substructures based on both velocity and position . this is a maximum likelihood algorithm , which fits a specified number of gaussians to the data and determines to which of these gaussians each individual galaxy most likely belongs . roughly 70@xmath49 of the galaxies belonged to the primary cluster ( gaussian # 2 in blc02 , referred to hereafter as the `` primary '' ) , and about 20@xmath49 belonged a second subcluster ( gaussian # 1 , hereafter the `` subcluster '' ) , presumably the infalling subcluster identified in the x - ray observations . the remaining @xmath5 belonged to a third group ( gaussian # 3 , hereafter the `` group '' ) concentrated to the north of the primary center and at higher radial velocity . we have repeated the kmm procedure on our combined velocity list in order to place all galaxies ( including the radio galaxies ) into their respective substructures . our kmm code differs slightly from that of @xcite , in that we assume the covariances among the three attributes ( two position , one velocity ) are zero . of all cluster galaxies . ] thus , we calculate the distance from each galaxy to each specified gaussian , @xmath50 : @xmath51.\ ] ] each galaxy is assigned to the gaussian which produces the smallest distance , and once all galaxies have been assigned the new means and dispersions are calculated . these values are then used in successive iterations until the assignments are stable . in particular , the log likelihood of the fit is described by : @xmath52\ ] ] where the summation over @xmath50 refers to the number of groups to be fit to the data and the summation over @xmath53 refers to the individual galaxies . the factor @xmath54 refers to the probability that a given galaxy @xmath53 belongs to group @xmath50 , which is evaluated based on the values of @xmath55 for that galaxy . the factor @xmath56 is the `` mixing fraction , '' or the fraction of the total population which are members of group @xmath50 . convergence of the algorithm is assumed once @xmath57 has converged . as discussed in @xcite , the significance of a multi - component fit is not easily surmised . consequently , our code also calculates @xmath57 for monte carlo shuffles of the data in order to determine the frequency with which the algorithm can fit a randomized data set as well as it does the real data . in this case , we only shuffled the velocity data so as to avoid issues related to positional sampling of cluster velocities ( our mc test is thus directly analogous to that used to quantify the significance of the dressler - shectman test ) . the results are presented in table [ tbl - kmm ] ; the derived velocity means and dispersions are consistent with those presented in table 8 of blc02 . in total , 186 galaxies were assigned to the primary , 78 galaxies to the large infalling subcluster , and 30 galaxies to the group identified primarily on the basis of its higher radial velocity . only @xmath58 of the mc shuffles had 3-gaussian fits as good as or better than the actual data . note that for the two - component and four - component fits the results were @xmath59 and @xmath60 , respectively , arguing for the three - component model ( when attempting a fifth component , the algorithm failed to converge without removing nearly all of the galaxies from one of the substructures ) . a relatively large number of the radio sources were assigned to the group ( 7 of 30 galaxies with velocities placing them in this structure are in table [ tbl - clusrg ] ) . spatially , this may be seen in figure [ fig - dist ] as an apparent excess of radio galaxies to the north of the cluster center . quantitatively , a chi - square test places this excess of radio sources at only marginal significance , about @xmath36 . individually , there is no statistical difference in the radio galaxy fractions of the primary and the subcluster . if we include the three additional confirmed cluster radio galaxies which were fainter than our @xmath61 limit ( table [ tbl - morerg ] ) , the excess of radio sources in the group is accentuated . one of these sources belongs to the group , increasing the significance of the radio galaxy excess to over @xmath62 . interestingly , the radio galaxies in the group are almost exclusively associated with star formation . our spectra include six of the eight identified radio galaxies in the group , all of which exhibit emission lines . one source , j170150 + 790216 , is a strong liner whose optical morphology suggests a recent galaxy - galaxy interaction ( see figure [ fig - fig8 ] ) . this galaxy is also separated from the other radio detections of the group , lying at a higher declination . the remaining five galaxies have spectra of star - forming galaxies . the propensity for the group members to be radio - detected star - forming galaxies is significant at nearly @xmath63 ( 5 of 30 compared to 9 of 264 ) , although this claim is tempered by our incomplete spectroscopy . however , should all of the group galaxies lacking spectroscopic classification be agn and all the non - group galaxies lacking spectroscopic classification be star forming ( i.e. , a worst case scenario ) , the excess of star - forming galaxies in the group would still be significant at over @xmath64 . star formation therefore seems more prevalent in the group than in the remainder of abell 2256 . additionally , we note that the two group member radio galaxies for which we do not possess spectra appear to be spirals and their radio - to - optical flux ratios are suggestive of star formation rather than agn . conversely , only one of the radio galaxies assigned to the subcluster is associated with star formation . our kmm code suggests that the probability that this galaxy belongs to the subcluster is only slightly greater than the probability that it belongs to the primary . thus , the activity level of the infalling subcluster appears to be quite low . abell 2256 is clearly a dynamically complicated system . attempts to understand its properties have generally relied upon the notion that a major merger was occurring , and have therefore focussed on the interaction of the primary cluster with the large subcluster to the west . however , recent evidence has revealed the presence of a third group . this group was identified in the large number of cluster galaxy velocities obtained by blc02 , and smmv02 found x - ray evidence for a group which presumably had already merged with the primary cluster . we will argue for the importance of this third substructure in understanding some of the radio properties of abell 2256 . the key pieces of evidence for this interpretation are depicted in figure [ fig - relicpos ] . here , the radio relic is shown along with the positions of the group member galaxies . it is remarkable that seven of the eight radio galaxies belonging to the group are located immediately to the north of the relic . with the exception of the two galaxies for which we do not have optical spectra ( j170322 + 784655 and j170455 + 782954 ) , all of these group radio galaxies are powered by star formation . in addition , three further galaxies were assigned to the primary by the kmm algorithm but are arguably associated with the group . the algorithm assigned marginally higher probabilities that j170335 + 785055 , j170448 + 783829 , and j170628 + 784915 were associated with the primary than the group . the first and third of these are star - forming galaxies ( one of which is slightly fainter than our formal @xmath61 limit ) located among the other group star - forming galaxies . the second is the fairly strong radio galaxy located at the rough center of the radio halo ( see also figure [ fig - center ] ) . it is flanked by several fainter galaxies which were assigned to the group . thus , it appears that the galaxy activity associated with the group may be even larger than previously stated . based on this positional evidence , we propose that the group may be responsible for the radio relic source . this was also suggested by blc02 , who noted the higher velocity of the group and placed it between the observer and the primary and currently falling into the cluster . should the merger of the group with the primary be responsible for the radio relic , the group is more likely _ behind _ the primary having recently plunged through the core of the primary . in fact , smmv02 argued that should the x - ray `` shoulder '' identified in the _ chandra _ image be associated with a group , models suggest that the system is viewed about 0.3 gyr after core passage . the passage of the group through the intracluster gas would create a shock , which in turn has powered the relic source . this can also explain the halo , which is centered more on the high - velocity galaxy j170448 + 783829 ( this possibility has also been noted in * ? ? ? it is also interesting to note that the models of @xcite suggested that abell 2256 was better described by two mergers , one of which occurred @xmath65 gyr previously . oddly enough , the premise for this result was the recreation of structure indicated in the _ rosat _ x - ray temperature maps @xcite , an observational finding which was not confirmed by the later satellite observations ( _ asca _ , _ bepposax _ , and _ chandra _ ; * ? ? ? * ; * ? ? ? * ; * ? ? ? several parameters of the relic source are potentially explained by this geometry and scenario . @xcite noted the flat spectral index of the radio emission from the relic , and used this to estimate an age of @xmath66 years . a young age may also be implied by the morphology of the relic in the model of @xcite , `` sheet - like '' relics are younger and have flatter spectral indices . in this model , relics are created when `` fossil radio cocoons '' are adiabatically compressed by a shock front created during a merger . these cocoons are formed by now - defunct radio galaxies which filled small volumes within the intracluster gas with hotter , more diffuse gas and magnetic fields . the emitting electrons remain in these cocoons , having faded from synchrotron emission . the compression increases the energy of the cocoons and increases their magnetic field strengths , thereby once again increasing their synchrotron emission and producing radio relic sources . if we assume this model for abell 2256 , it is possible that the galaxy associated with the peak of the relic emission ( figure [ fig - relic1 ] ) was the progenitor of the cocoon which formed the relic . at early ages , relics produced in this manner can exhibit loop - like structures akin to what is seen along the eastern edge of the relic in abell 2256 , depending on viewing geometry . lastly , the model also predicts polarizations like that in the abell 2256 relic . given the morphology and polarization of the relic , the group would be moving largely along our line of sight but inclined slightly such that it entered from the south . performing a simple calculation wherein this inclination angle is 45@xmath67 , the separation between the star - forming group galaxies and the cluster center implies a travel time of @xmath10.3 gyr at the relative velocity of the group , consistent with the estimate from the x - ray observations . one possible difficulty with this model is the relatively low velocity dispersion of the group . the identified group has a velocity dispersion of around 300 km s@xmath4 , which is consistent with typical values for rich groups ( e.g. , * ? ? ? n - body simulations have demonstrated that groups are effectively dispersed as they pass through the cores of more massive clusters ( e.g. , * ? ? ? * ; * ? ? ? * ) , thereby increasing their velocity dispersions . thus , we would expect a larger velocity dispersion for the group should it be viewed after rather than before core passage . obtaining the true velocity dispersion for the group would require perfect knowledge of the assignments of galaxies to their pre - merger substructures , information which is not available . the low velocity dispersion is apparently a consequence of how the galaxies are assigned . in figure [ fig - trump ] we plot the velocity difference between each galaxy and the velocity center of its associated substructure as a function of radial separation from the center of that substructure . the assignments are heavily influenced by velocity ; the `` trumpet '' shape for a virialized system is not easily apparent for any of the three fitted systems . apparently , only the higher velocity members of the initial group are assigned group membership by the kmm algorithm , thus reducing the fitted dispersion of this substructure . on the other hand , the positions of the group members do resemble the results of n - body simulations viewed shortly after core passage . the group members become more dispersed as one moves along the apparent axis of the merger from the center of the cluster to the northwest ( see figure [ fig - relicpos ] ) . another mystery of the radio population in abell 2256 is the extreme length and straightness of the tailed radio galaxy j170330 + 783955 ( figure [ fig - flagpole ] ) . the tail extends about 300 kpc in our image , although sensitive observations at lower resolution suggest it is potentially twice as long as this @xcite . since the emission is not disrupted , the region must have very low turbulence a requirement which is contrary to the generation of a relic source . however , in the discussion above the relic would be located in the back half of the cluster which would presumably require the head - tail radio galaxy to lie in front of it . it is also likely that the subcluster is entering the primary slightly from behind ( blc02,smmv02 ) . evidence for the group being seen shortly post - merger might also be found in the spectra of group members . @xcite modeled the effect of a cluster - group merger on the star formation history of a disk galaxy initially associated with the infalling group . the time - dependent tidal gravitational field of the merger was able to induce a secondary burst of star formation in the galaxy . the merger also removes a significant fraction of the disk galaxy s halo , thereby reducing its reservoir of gas to replenish the disk . other cluster environmental effects run parallel to this ; ram pressure stripping should also effectively remove gas from infalling galaxies ( e.g. , * ? ? ? * ) . in total , we might expect to see galaxies with starburst and post - starburst features among the spectra of the group members in abell 2256 . for example , this seems to be the case in the coma cluster as post - starburst signatures ( i.e. , strong balmer absorption ) are seen in several galaxies associated with the ngc 4839 system @xcite , which is believed to have recently crossed through the core of the coma cluster @xcite . similarly , in distant clusters a population of starbursts in which selective dust extinction produces strong post - starburst features along with muted star formation features ( i.e. , [ oii ] emission ) has been observed ( the e(a ) galaxies of * ? ? ? * ; * ? ? ? * ) . in some cases , the dust extinction may be severe enough to remove all emission lines and produce a post - starburst spectrum despite ongoing star formation @xcite . there is slight evidence for such an effect in the group members . the lone identified starburst galaxy in abell 2256 belongs to the group , and one of the three identified e(a ) galaxies is the group outlier j170335 + 785055 . additionally , the star - forming galaxy with h@xmath40 absorption just less than the formal limit of the e(a ) class is a group member . unfortunately , the signal - to - noise of our spectra and the frequent lack of [ oii ] in the spectral coverage makes a rigorous evaluation difficult . for the mx observations , the average sn per pixel evaluated over the regions surrounding h@xmath40 is 6.8 , with a minimum of 3.7 and a maximum of 10.9 . improved future spectroscopy would clearly be of great use . it is also interesting to note the lack of star - forming galaxies in the subcluster . only one radio - identified star - forming galaxy was associated with the subcluster , and our kmm code did not place it in the subcluster with high confidence . the galaxy lies along the western edge of the spectroscopically - confirmed cluster members ( j165306 + 783013 ) , and should it belong to the subcluster it trails the core of this system by 1.7 mpc in projected distance . it is possible that this galaxy is just an infalling galaxy unassociated with the subcluster but assigned to it primarily on the basis of its low radial velocity . the dearth of star - forming galaxies does not appear to be limited to those identified through radio selection , either . our spectroscopy database includes eight cluster emission - line galaxies which were not radio detections . in general , these galaxies lie at the periphery of the cluster where the sensitivity of the radio images is lower , thereby explaining their absence from our radio lists . all of these galaxies were assigned to the primary by the kmm code . as the velocity dispersion of cluster star - forming galaxies is generally larger than the velocity dispersion of cluster ellipticals and s0 s , one might expect that the higher radial velocity star - forming galaxies initially associated with the subcluster have been assigned to the primary . however , inspection of the velocity histogram of all galaxies does not indicate that any apparent low radial velocity galaxies were clipped from the overall cluster list . consequently , it appears that the lack of star - forming galaxies associated with the subcluster is real . what is the relationship of this proposed scenario to the findings for other clusters ? of the twenty nearby clusters we have studied , abell 2255 has the most dramatic increase in galaxy activity . in particular , this cluster has a marked excess of optically - fainter ( @xmath68 ) star - forming galaxies . there are some similarities to its dynamical environment with that of abell 2256 , including evidence for a recent merger between the primary cluster and an infalling cluster / group ( with roughly a 8:1 mass ratio ; * ? ? ? * ) and the presence of numerous outlying groups @xcite . mo03 argued that a recent merger , with core passage having occurred about 0.1 gyr previously , was responsible for the increase in the number of cluster star - forming galaxies . the mass ratio between the primary and the group in abell 2256 must be similar to that proposed for abell 2255 , and in each case it appears that core passage apparently occurred relatively recently . the abell 2256 subcluster , however , is larger and seemingly has very few ( if any ) star - forming galaxies which could undergo a burst of star formation . prior to the merger it was likely a fairly isolated , relaxed cluster . any star - forming galaxies associated with it would have simply been nearby field galaxies in the process of infall . presently , such infalling galaxies would more likely be assigned to the primary . hence , the merger of the subcluster with the primary may have very little effect on cluster - wide star formation as it is not likely to introduce a large number of galaxies which might undergo starbursts . consequently , the acquisition of peripheral groups may be more important in understanding galaxy activity in clusters than mergers of larger partners . although the simulations of @xcite suggested that larger mass ratios translated to stronger starbursts during cluster mergers , this was based on the assumption that gas - rich spiral galaxies existed in one of the pre - merger partners . as we have noted , the star - forming galaxy population of the rich subcluster in abell 2256 is quite small , suggesting little available material for star formation induced by the merger . simulations of hierarchical growth of structure also indicate that mergers of comparable - mass partners are much rarer than cluster - group mergers , so increased cluster activity as a function of redshift ( e.g. * ? ? ? * ) can more easily be related to the build up of rich cluster through acquisition of outlying groups . we have presented a multiwavelength observational study of abell 2256 . our deep radio continuum imaging reached an rms sensitivity as low as @xmath69 @xmath13jy beam@xmath4 and encompassed an area from the center of the cluster out to radii of @xmath150 , or about 3 mpc . potential cluster radio galaxies were identifed using optical imaging of the same area , with membership being evaluated on the basis of both new and existing optical spectroscopy . in total , 43 cluster radio galaxies were identified and many of these were characterized as agn or star - forming galaxies on the basis of their spectra . paralleling the study of blc02 , the application of a maximum likelihood algorithm assigned the galaxies to their probable substructures , which correspond to the primary , the infalling subcluster , and a group . taken collectively , the radio galaxy population of abell 2256 is not unusual when compared to a sample of nineteen other nearby clusters . we confirm a statistical excess of the optically brightest radio galaxies , which might have been guessed from the previously noted unusual number of galaxies with extended radio emission in the cluster . outside of this specific population , the radio galaxy fraction of abell 2256 is relatively normal . we do , however , note that the galaxies belonging to the group are more frequently associated with radio emission and nearly all of these galaxies are powered by star formation . the radial velocity of the group is @xmath70 km s@xmath4 greater than that of the primary , and its position is coincident with the cluster radio relic source . we therefore argue that the group has recently merged with the cluster and is located behind the cluster relative to the observer , roughly 0.3 gyr after core passage . under this scenario , the radio relic would owe its existence to the cluster - group merger rather than the ongoing cluster - subcluster merger , which appears to be about 0.2 gyr prior to core passage and entering the primary from behind and to the west @xcite . in the broader picture , if abell 2256 is a representative example then galaxy activity in clusters may be most influenced by the merger history of the parent cluster with outlying groups . these minor mergers would provide a relatively large influx of gas - rich spirals which might undergo starbursts followed by truncation of star formation . mergers of larger partners , while having more profound effects on the icm , may not prove overly important toward understanding star formation in clusters simply because the partners are less likely to include the potential starbursting galaxies . thus , active clusters such as those identified in butcher - oemler type studies would simply correspond to dynamically young clusters in larger - scale structures rich in peripheral groups . nam thanks the national radio astronomy observatory for a predoctoral fellowship , and acknowledges the support of a national research council associateship award at nasa s goddard space flight center . this research was supported in part by nasa through the american astronomical society s small research grant program . we thank keith ashman for providing the original kmm code , and eric greisen for assistance with aips and the multi - resolution clean . markevitch , m. , ponman , t.j . , nulsen , p.e.j . , bautz , m.w . , burke , d.j . , david , l.p . , davis , d. , donnelly , r.h . , forman , w.r . , jones , c. , kaastra , j. , kellogg , e. , kim , d .- w . , kolodziejczak , j. , mazzotta , p. , pagliaro , a. , patel , s. , van speybroeck , l. , vikhlinin , a. , vrtilek , j. , wise , m. , & zhao , p. 2000 , , 541 , 542 l l r r c r r 16:47:57.9 & 78:52:07 & 17784 & 31 & emi & & + 16:50:39.6 & 78:39:07 & 16756 & 42 & abs & & + 16:51:11.5 & 79:06:05 & 17671 & 58 & abs & & + 16:52:32.0 & 78:57:03 & 30106 & 78 & emi & & + 16:52:52.8 & 78:46:29 & 16836 & 19 & emi & & + 16:53:05.9 & 78:30:13 & 16663 & 69 & emi & & + 16:53:13.6 & 78:44:13 & 16397 & 41 & abs & & + 16:53:33.3 & 79:00:00 & 18742 & 37 & emi & & + 16:53:34.3 & 79:06:21 & 18500 & 40 & emi & & + 16:54:06.1 & 78:37:42 & 16674 & 39 & abs & 16710 & 36 + 16:54:16.6 & 79:11:20 & 53488 & 53 & abs & & + 16:54:29.4 & 78:54:34 & 16228 & 39 & abs & & + 16:54:36.1 & 79:14:56 & 39573 & 60 & emi & & + 16:54:58.9 & 79:08:45 & 17433 & 61 & abs & & + 16:55:32.2 & 78:32:02 & 15969 & 57 & abs & 16180 & 35 + 16:55:36.9 & 79:05:10 & 17808 & 30 & abs & & + 16:56:06.9 & 79:23:28 & 26352 & 60 & emi & & + 16:56:09.6 & 78:56:36 & 17815 & 46 & abs & 17851 & 46 + 16:56:16.1 & 78:55:09 & 18288 & 32 & emi & 18309 & 48 + 16:56:18.3 & 79:09:29 & 17598 & 60 & emi & & + 16:57:05.5 & 78:32:48 & 16350 & 56 & abs & 16495 & 31 + 16:57:58.5 & 78:36:15 & 4260 & 22 & emi & 4253 & 45 + 16:58:21.1 & 78:43:25 & 15791 & 67 & abs & 15909 & 52 + 16:58:46.8 & 78:25:21 & 15632 & 62 & abs & 15635 & 49 + 17:00:07.5 & 78:49:52 & 16514 & 68 & abs & & + 17:00:11.1 & 79:00:16 & 17383 & 70 & abs & 17415 & 70 + 17:00:11.5 & 79:03:33 & 17597 & 58 & abs & 17668 & 34 + 17:00:16.0 & 78:34:25 & 18362 & 47 & abs & 18355 & 62 + 17:00:20.4 & 78:33:55 & 16454 & 62 & abs & 16372 & 55 + 17:00:32.6 & 78:56:10 & 15021 & 65 & abs & 15237 & 47 + 17:00:52.3 & 78:41:21 & 17157 & 37 & abs & 17325 & 39 + 17:00:59.2 & 78:44:59 & 15966 & 42 & abs & 16137 & 53 + 17:01:00.2 & 79:05:09 & 18377 & 42 & emi & 18234 & 48 + 17:01:34.6 & 78:50:46 & 19531 & 30 & emi & 19518 & 15 + 17:01:49.5 & 79:02:16 & 19307 & 90 & emi & 19303 & 60 + 17:02:07.5 & 78:17:29 & 17760 & 32 & emi & 18001 & 47 + 17:02:12.3 & 78:51:33 & 19325 & 16 & emi & 19421 & 59 + 17:02:17.8 & 78:45:52 & 20086 & 99 & abs & 19907 & 66 + 17:02:18.6 & 78:46:03 & 19581 & 38 & emi & 19643 & 44 + 17:02:46.8 & 78:51:46 & 16638 & 30 & abs & 16574 & 53 + 17:03:02.9 & 78:35:56 & 16417 & 39 & abs & 16303 & 100 + 17:03:10.9 & 78:14:40 & 18028 & 34 & emi & 17973 & 48 + 17:03:25.4 & 78:20:38 & 18922 & 43 & emi & 18886 & 48 + 17:03:29.5 & 78:37:55 & 17565 & 54 & abs & 17796 & 44 + 17:03:34.5 & 78:50:55 & 18989 & 72 & emi & 18943 & 49 + 17:03:56.5 & 78:44:44 & 17082 & 44 & abs & 17115 & 36 + 17:04:04.2 & 78:33:23 & 17316 & 53 & abs & 17355 & 53 + 17:04:14.5 & 79:23:13 & 26291 & 32 & emi & & + 17:04:15.4 & 78:06:07 & 18107 & 90 & abs & & + 17:04:41.0 & 78:13:49 & 17956 & 58 & abs & 18008 & 56 + 17:04:48.2 & 78:38:29 & 19013 & 38 & abs & 19094 & 49 + 17:04:54.0 & 78:29:54 & 52284 & 60 & abs & 52295 & 114 + 17:05:11.9 & 78:48:14 & 19921 & 32 & emi & 19942 & 48 + 17:05:21.8 & 79:08:05 & 18825 & 42 & emi & 18807 & 48 + 17:05:24.2 & 78:22:00 & 17189 & 55 & abs & 17322 & 36 + 17:05:33.6 & 78:51:16 & 19511 & 46 & emi & 19479 & 54 + 17:06:00.8 & 79:06:43 & 17908 & 68 & abs & 17927 & 47 + 17:06:28.4 & 78:49:15 & 18994 & 22 & emi & 19093 & 45 + 17:07:00.6 & 78:41:23 & 16956 & 62 & abs & 17161 & 79 + 17:07:02.4 & 79:10:19 & 11408 & 31 & emi & 11436 & 49 + 17:07:14.1 & 78:53:18 & 16563 & 36 & abs & 16524 & 42 + 17:07:15.9 & 78:30:08 & 18379 & 30 & emi & 18582 & 48 + 17:07:21.4 & 78:53:38 & 16520 & 56 & abs & 16564 & 35 + 17:08:33.5 & 78:29:54 & 18663 & 39 & abs & 18684 & 46 + 17:08:49.0 & 79:14:31 & 16668 & 44 & abs & & + 17:08:58.2 & 78:50:15 & 16249 & 99 & abs & 16064 & 51 + 17:09:10.0 & 78:52:32 & 18624 & 60 & emi & 18812 & 46 + 17:09:44.3 & 78:06:06 & 18039 & 32 & emi & & + 17:10:14.4 & 78:39:25 & 16922 & 47 & abs & 17043 & 53 + 17:10:15.9 & 78:02:07 & 17801 & 38 & emi & & + 17:10:31.5 & 78:29:10 & 15783 & 65 & abs & 15792 & 66 + 17:10:48.8 & 78:32:18 & 18682 & 36 & abs & 18767 & 31 + 17:13:57.2 & 78:59:33 & 19696 & 30 & abs & 19614 & 62 + l l l r l l r l l c c 16:52:52.8 & 78:46:29 & 16805 & 27 & ke & 16.39 & 0.572 & 0.107 & 2.2 & & sf + 16:53:05.8 & 78:30:13 & 16663 & 69 & me & 16.51 & 0.534 & 0.151 & 2.7 & & sf + 16:54:58.9 & 79:08:45 & 17433 & 61 & mf & 15.38 & 0.338 & 0.127 & 1.3 & & agno + 16:55:36.9 & 79:05:10 & 17808 & 30 & kf & 14.76 & 2.897 & 0.100 & 0.2 & & agno + 16:56:18.3 & 79:09:29 & 17616 & 36 & ke & 15.86 & 0.641 & 0.155 & 0.8 & & sf + 16:58:18.5 & 78:29:34 & 17882 & 47 & b & 14.99 & 3.397 & 0.340 & 0.8 & ad & + 16:58:46.8 & 78:25:21 & 15632 & 62 & mf & 15.23 & 1.000 & 0.218 & 5.0 & & agno + 17:00:25.8 & 78:49:26 & 15376 & 52 & b & 16.58 & 0.225 & 0.071 & 1.5 & & + 17:00:52.3 & 78:41:21 & 17157 & 37 & mf & 15.40 & 7.8 & 0.3 & & i & agno + 17:00:59.2 & 78:44:59 & 15966 & 42 & kf mf & 15.29 & 0.247 & 0.060 & 0.4 & & agno + 17:01:00.2 & 79:05:09 & 18232 & 48 & ke & 16.20 & 0.865 & 0.124 & 0.5 & & sf + 17:01:23.4 & 78:41:16 & 16217 & 67 & b & 15.55 & 0.425 & 0.126 & 2.2 & & + 17:01:34.6 & 78:50:46 & 19531 & 30 & me & 16.64 & 0.549 & 0.083 & 0.4 & & sf + 17:01:49.5 & 79:02:16 & 19342 & 60 & ke & 14.65 & 2.962 & 0.100 & 1.2 & & agnl + 17:02:12.3 & 78:51:33 & 19325 & 16 & ke me & 15.76 & 1.438 & 0.100 & 0.7 & & sf + 17:02:15.0 & 78:35:47 & 18757 & 39 & b & 15.36 & 0.642 & 0.185 & 1.7 & & + 17:02:18.6 & 78:46:03 & 19612 & 23 & ke & 15.52 & 1.687 & 0.104 & 1.2 & k & sf + 17:02:46.8 & 78:51:46 & 16638 & 30 & kf & 15.12 & 1.840 & 0.100 & 0.7 & ac & agno + 17:03:02.9 & 78:35:56 & 16417 & 39 & mf & 14.90 & 50.3 & 0.7 & & b & agno + 17:03:22.4 & 78:46:55 & 19723 & 45 & b & 16.33 & 0.454 & 0.087 & 1.2 & & + 17:03:25.4 & 78:20:38 & 18922 & 43 & me & 16.43 & 0.714 & 0.155 & 0.6 & & sf + 17:03:29.5 & 78:37:55 & 17565 & 54 & mf & 14.73 & 120.6 & 0.3 & & a & agno + 17:03:30.1 & 78:39:55 & 17558 & 36 & b & 15.01 & 40.7 & 0.5 & & c & + 17:03:34.1 & 78:37:48 & 16929 & 47 & b & 14.70 & 1.261 & 0.148 & 2.3 & & + 17:03:34.5 & 78:50:55 & 18989 & 72 & me & 16.02 & 0.697 & 0.115 & 1.4 & & sf + 17:03:56.5 & 78:44:44 & 17082 & 44 & mf & 15.51 & 5.816 & 1.000 & 3.1 & g & agno + 17:04:15.4 & 78:06:07 & 18107 & 90 & kf & 15.74 & 3.583 & 0.200 & 0.4 & & agno + 17:04:48.2 & 78:38:29 & 19103 & 38 & mf & 14.64 & 11.4 & 0.1 & 1.9 & d & agno + 17:05:11.9 & 78:48:14 & 19921 & 32 & me & 16.20 & 0.576 & 0.112 & 0.3 & & sf + 17:05:33.6 & 78:51:16 & 19511 & 46 & me & 15.75 & 4.178 & 0.100 & 0.3 & e & sf + 17:06:43.4 & 78:35:57 & 14910 & 77 & b & 16.17 & 1.025 & 0.076 & 0.9 & & + 17:06:56.4 & 78:41:09 & 16867 & 38 & b & 15.30 & 0.883 & 0.104 & 1.7 & f3 & + 17:07:15.9 & 78:30:08 & 18391 & 36 & ke & 16.41 & 1.329 & 0.086 & 1.2 & & sf + 17:08:33.5 & 78:29:54 & 18663 & 39 & kf mf & 14.78 & 2.895 & 0.100 & 1.0 & aa & agno + 17:08:47.7 & 78:25:56 & 16860 & 7 & b & 16.60 & 0.283 & 0.107 & 2.3 & & + 17:08:58.1 & 78:50:15 & 16249 & 99 & kf & 15.21 & 0.600 & 0.083 & 1.2 & & agno + 17:09:10.0 & 78:52:32 & 18624 & 60 & me & 15.31 & 1.211 & 0.134 & 0.9 & & sf + 17:09:19.1 & 78:30:16 & 16700 & 44 & b & 16.78 & 0.647 & 0.165 & 4.1 & & + 17:09:39.9 & 78:33:07 & 16120 & 33 & b & 15.44 & 0.500 & 0.132 & 2.9 & & + 17:10:15.9 & 78:02:07 & 17801 & 38 & me & 16.58 & 0.405 & 0.105 & 1.7 & & sf + & & & & & & & & & & + & & & & & & & & & & + & & & & & & & & & & + & & & & & & & & & & + & & & & & & & & & & + & & & & & & & & & & + l l r r l l r l l c c c 16:53:19.4 & 78:30:40 & & & & 15.08 & 0.261 & 0.091 & 1.0 & & & + 16:54:16.6 & 79:11:20 & 53488 & 53 & mf & 16.72 & 0.535 & 0.162 & 2.4 & & agno & + 16:54:36.0 & 79:14:56 & 39573 & 60 & ke & 16.72 & 0.520 & 0.145 & 1.4 & & sf & + 16:56:06.9 & 79:23:28 & 26352 & 60 & ke & 15.75 & 2.105 & 0.2 & 1.1 & & sf & + 16:57:58.5 & 78:36:15 & 4260 & 22 & ke me & 15.22 & 1.922 & 0.1 & 0.3 & af & sf & + 17:02:57.9 & 78:03:18 & & & & 16.80 & 0.656 & 0.110 & 0.8 & & & + 17:03:06.8 & 78:11:43 & & & & 16.79 & 0.429 & 0.141 & 3.0 & & & + 17:04:14.5 & 79:23:13 & 26291 & 32 & me & 16.67 & 0.853 & 0.141 & 0.8 & & sf & + 17:04:54.0 & 78:29:54 & 52284 & 60 & kf & 16.64 & 1.898 & 0.154 & 1.5 & & agno & + 17:04:55.4 & 78:51:08 & 19980 & 45 & b & 17.15 & 0.438 & 0.094 & 0.9 & & & + 17:04:56.9 & 78:43:06 & 18682 & 45 & b & 17.05 & 0.419 & 0.096 & 1.2 & & & + 17:06:28.4 & 78:49:15 & 18994 & 22 & me & 16.96 & 1.135 & 0.098 & 1.4 & & sf & + 17:07:02.4 & 79:10:19 & 11408 & 31 & ke me & 16.06 & 0.678 & 0.132 & 0.8 & & sf & + 17:07:40.1 & 78:56:57 & 53835 & 61 & b & 17.57 & 0.276 & 0.078 & 1.6 & & & + 17:09:01.1 & 78:46:18 & 11844 & 26 & b & 15.24 & 2.180 & 0.1 & 0.9 & ae & & + 17:11:20.7 & 78:49:47 & 39493 & 84 & b & 16.64 & 1.231 & 0.1 & 1.0 & & & + 17:14:18.9 & 78:05:20 & & & & 15.10 & 2.390 & 0.270 & 0.8 & & & + 17:18:38.8 & 78:13:38 & & & & 15.90 & 1.299 & 0.100 & 1.0 & & & +

we have selected and analysed the properties of a sample of 2905 galaxies in of the great observatories origins deep survey ( goods ) chandra deep field south ( cdfs ) , to obtain further constraints on the evolution of-selected galaxies with respect to the results already obtained in previous studies . we made use of the public deep multiwavelength imaging from the optical through the infrared ( ir ) bands , in conjunction with available spectroscopic and combo17 data in the cdfs , to construct an optimised redshift catalogue for our galaxy sample . we computed the-band lf and determined that its characteristic magnitude has a substantial brightening and a decreasing total density from to . we also analysed the colours and number density evolution of galaxies with different stellar masses . within our sample , and in contrast to what is observed for less massive systems , the vast majority (% ) of the most massive ( ) local galaxies appear to be in place before redshift . around% of the total assemble between redshifts and and most of them display extremely red colours , suggesting that plausible star formation in these very massive systems should mainly proceed in obscured , short - timescale bursts . the remaining fraction ( up to% ) could be in place at even higher redshifts , pushing the first epoch of formation of massive galaxies beyond the limits of current near - ir surveys . [ firstpage ] galaxies : evolution galaxies : formation galaxies : high - redshift galaxies : luminosity function , mass function

astro-ph0511571

deep multiwavelength surveys are progressively sheding light on the history of galaxy evolution since very high redshifts . from new surveys in the ultraviolet regime ( e.g. martin et al . 2005 ) to the latest ir , sub - millimetre and radio campaigns ( e.g. werner et al . 2004 ; dunlop 2005 ; condon et al . 2003 ) , the advent of numerous datasets is enhancing our understanding of when galaxies were formed and how they evolved through cosmic time . uv / optical observations can unveil the sources of stellar emission but are limited to detect unobscured systems , missing an important fraction of the galaxies with on - going star formation . ir and longer - wavelength surveys provide an unbiased way of studying star - forming systems , but are insensitive to galaxies with little or no dust content . near - ir observations , on the contrary , appear as one of the most suitable methods for making a complete census of galaxy populations , as they are sensitive to the stellar emission from both young and old galaxies , and are also relatively unaffected by the presence of dust . the rest - frame @xmath15 band is the most efficient tracer of the stellar mass content of the universe . recent studies of near - ir - selected galaxies allowed for a rapid progress in setting different constraints on the history of galaxy formation ( cimatti et al . 2002 ; daddi et al . 2003 ; caputi et al . 2004 ; caputi et al . 2005a ; among others ) and , in particular , on the evolution of the stellar mass content of the universe with redshift ( dickinson et al . 2003 ; fontana et al . 2004 ; glazebrook et al . 2004 ; caputi et al . 2005a ; drory et al . 2005 ) . it is now recognised that , although more than a half of the stellar mass has been formed at redshifts @xmath16 $ \sim$}}~}1.5 $ ] , near - ir - selected galaxies are already present before at least redshifts @xmath17 and contain a significant fraction of the stellar mass observed at redshift @xmath5 . however , one of the main problems of existing near - ir surveys is that , when sufficiently deep , they are usually restricted to small areas of the sky , making the results prone to large error bars due to cosmic variance . also , the lack of statistics prevents refined studies of galaxy evolution , where trends for different near - ir sub - populations can be individually investigated . the goods project ( giavalisco et al . 2004 ) is a public program of multiwavelength observations which is now close to completion . the data sets are kindly released to the astronomical community in a fully reduced mode . as part of the goods project , @xmath18 of the cdfs have been observed with the advanced camera for surveys ( acs ) on board the hubble space telescope ( hst ) in four passbands : @xmath3 , @xmath19 , @xmath20 and @xmath21 . in addition , the goods project includes deep near - ir imaging taken with the infrared spectrometer and array camera ( isaac ) on the ` antu ' very large telescope ( antu - vlt ) . a first release of the european southern observatory ( eso ) imaging survey ( goods / eis v0.5 release ) provided data in the @xmath22 , @xmath23 and @xmath24 bands for @xmath25 of the goods / cdfs field . in this sub - region , roche , dunlop & almaini ( 2003 ) selected a deep sample of extremely red galaxies ( ergs ) with @xmath26 ( vega ) , whose redshift distribution and luminosity evolution were studied by caputi et al . a detailed comparison of the properties of ergs with other @xmath2-selected galaxies was presented by caputi et al . ( 2005a ) . over the last year , however , new datasets have been made public within the goods project , increasing the coverage in wavelength and area of the goods / cdfs . the second goods / eis ( v1.0 ) release extends the deep @xmath22 and @xmath24 band imaging to @xmath1 of the goods / cdfs . also , the infrared array camera ( irac ) on board the spitzer space telescope imaged this field at @xmath27 , @xmath28 , @xmath29 and @xmath30 . the availability of these data allows one to map the rest - frame @xmath2-band light up to redshifts @xmath17 . in this work , we take advantage of the further wavelength and area coverage of the goods / cdfs , in conjunction with the information provided by the latest released spectroscopic and combo17 data , to set tighter constraints on the properties and evolution of @xmath2-selected galaxies . the layout of this paper is as follows . first , in section [ sec_sample ] , we present the sample selection and give details on the multiwavelength photometry . in section [ sec_redsh ] , we explain the technique we applied to construct an optimised redshift catalogue for our galaxy sample and show the resulting redshift distribution . in section [ sec_lf ] , we present the rest - frame @xmath2-band luminosity function ( lf ) for our sample of @xmath2-selected galaxies . in section [ sec_mass ] , we discuss several aspects of the evolution of different mass systems . in section [ sec_nearir ] , we explore additional hints from the study of the purely near - ir colours of our galaxies . finally , in section [ sec_conc ] we present some concluding remarks . we adopt throughout a cosmology with @xmath31 , @xmath32 and @xmath33 . a salpeter ( 1955 ) imf over stellar masses @xmath34 is assumed , unless where otherwise stated . we used the public deep isaac @xmath2-band images in the goods / eis v1.0 release to select a catalogue of @xmath0 ( vega ) sources in 131 arcmin@xmath35 of the goods / cdfs . this new release is composed of 21 maps which cover more than twice the area of the goods / eis v0.5 release . in addition , further improvements have been applied to the v1.0 imaging , such as a revised data reduction technique and a uniform photometric calibration of the maps across the entire field . we performed the source extraction using the public code sextractor ( bertin & arnouts , 1996 ) on the @xmath2-band images . we cut the resulting catalogue at a magnitude @xmath36 , a limit half a magnitude brighter than the cut we used to select our @xmath2 sample in caputi et al . ( 2005a ) . several reasons justify our new magnitude limit . first , a few of the additional maps in the v1.0 release have lower exposure times than the average exposure times of the maps composing the v0.5 release . second , this new limit constitutes an adequate balance between having access to galaxies up to very high redshifts and requiring for our catalogue to have a high level of completeness . the completeness levels of our v1.0 sample are very similar to those estimated for the v0.5 sample by caputi et al . ( 2005a ) : 100% , 90% and 86% for @xmath37 , @xmath38 and @xmath39 , respectively . finally , our new sample has been selected having in mind the possibility of a complete quantitative study of galaxy morphology , only possible for high signal - to - noise objects . the results of the latter will be presented in a companion paper ( mclure et al . 2005 , in preparation ) . our total v1.0 @xmath0 sample consists of 3184 sources . we measured multiwavelength photometry for our @xmath2-selected galaxies in order to model their spectral energy distribution ( sed ) , as we shall explain in section [ sec_redsh ] . we ran sextractor in ` double - image mode ' to perform aperture photometry on the goods / eis v1.0 @xmath22-band images , centred at the position of the @xmath2-band extracted sources . we measured magnitudes in an aperture size of 2.83-arcsec diameter for all the objects , to coincide with one of the aperture sizes used to construct the public goods / cdfs acs catalogues . the latter have been constructed running sextractor on the @xmath21-band images , and making ` double - image mode ' runs to perform photometry in the @xmath3 , @xmath19 and @xmath20 bands . we looked for counterparts of our @xmath0 ( vega ) sources in the acs catalogues and adopted the corresponding 2.83-arcsec diameter aperture magnitudes in each of the four acs passbands . a 2.83-arcsec diameter aperture is sufficient to collect all the flux of the vast majority ( @xmath4090% ) of our galaxies in the different optical / near - ir passbands . even in the @xmath2 band , virtually all of our galaxies have 2.83-arcsec - diameter aperture magnitudes in agreement with the corresponding total @xmath2 magnitudes within the error bars . the ( @xmath41 ) versus ( @xmath42 ) colours for the sources in the @xmath0 sample . cross - like symbols correspond to those objects with sextractor stellarity parameter class_star@xmath430.8 . ] deep spitzer / irac imaging for the goods / cdfs is now also publicly available , covering the wavelength range @xmath44 to @xmath45 . at the time of writing , only the first epoch of irac goods / cdfs data had been released . this first epoch includes @xmath44 and @xmath46 images for @xmath47 of the area of the goods / cdfs and @xmath4 and @xmath45 images for another @xmath47 of this area , in such a way that only @xmath48 of the field have data in the four irac passbands . we ran sextractor on the @xmath44 and @xmath4 images and looked for counterparts of the @xmath2-band sources in a matching radius of 1.2 arcsec . we restricted the photometry to only the first two irac channels , as the templates we used for the galaxy sed fitting lose accuracy or do not cover wavelengths beyond @xmath49 . we measured magnitudes in 2.83-arcsec diameter apertures . however , the point - spread function ( psf ) of the irac images is much larger than the characteristic psf of the isaac and acs images : the mean full - width half maximum ( fwhm ) of the irac @xmath44 psf is 1.66 arcsec ( fazio et al . 2004 ) , in comparison with 0.4 arcsec and 0.1 arcsec for the @xmath2 and @xmath21 bands in the isaac and acs instruments , respectively . thus , a 2.83-arcsec diameter aperture is insufficient to encompass the magnitudes measured in the irac bands with the magnitudes obtained in equivalent apertures on the acs and isaac images . we computed aperture corrections for the magnitudes measured on the irac images studying the curve of growth of isolated stars in the field . we determined aperture corrections of 0.50 and 0.55 magnitudes for the 2.83-arcsec - diameter aperture magnitudes in the @xmath44 and the @xmath4 bands , respectively . to clean our catalogues for stars and active galaxies , we applied criteria somewhat more strict than those implemented by caputi et al . ( 2005a ) . we cleaned our catalogues for stars using the sextractor stellarity parameter ` class_star ' , as measured on the acs @xmath21 images . to separate active galaxies , we looked for counterparts of our @xmath0 sources in the public x - ray catalogues available for the cdfs ( szokoly et al . 2004 ) . 222/3184 sources of our initial catalogue had class_star@xmath430.8 . figure [ colstar ] shows the location of these 222 sources in a colour - color diagram , in comparison to the remaining sources in the catalogue . we see that the majority of the objects with class_star@xmath430.8 occupy a clearly separated region in this plot , typically displaying colours @xmath50 $ \sim$}}~}1.1 $ ] ( vega ) . however , we find 6 outliers in the colour - colour diagram , which we studied on a case - by - case basis . 4 out of these 6 outliers are x - ray classified as active galactic nuclei ( agn ) or quasi - stellar objects ( qso ) . the remaining 2 are not detected in the x - rays , but their seds can not be satisfactorily fitted by hyperz ( bolzonella , miralles & pell , 2000 ) with any of the templates in the gissel98 library ( bruzual & charlot 1993 ) , so they are likely to be also stars or agn / qso . as a consequence , we excluded the 222 sources with class_star@xmath430.8 from our sample . on the other hand , we found that 142/3184 of our @xmath2 sources were detected in the x - rays . however , in principle , the fact that a galaxy is classified as active in the x - rays does not necessarily imply that its optical - to - near - ir sed is dominated by an active nucleus . to determine which x - ray - detected @xmath2 sources had to be excluded from our catalogue , we modelled their seds using the templates in the gissel98 library of hyperz . the gissel98 library contains a wide range of synthetic seds based on stellar spectra but no power - law sed template . we determined that the seds of 57/142 x - ray - detected @xmath2 sources could not be satisfactorily fitted by any template in the gissel98 library , so we excluded these 57 objects from our catalogue , as it is likely that their optical / near - ir light is at least partially contaminated by an active nucleus component . 5 additional x - ray - detected @xmath0 sources did not have either a satisfactory fit of their seds , but they already had been excluded on the basis of the stellarity parameter criterion . in summary , our final @xmath0 source catalogue consisted of 2905 galaxies . other no - x - ray - detected active galaxies could still be present in our final galaxy catalog . another characteristic allowing to identify active galaxies is the power - law shape of their seds from the uv / optical through the ir regimes . thus , to investigate the presence of additional contaminating sources within our sample , we cross - correlated our galaxies with the catalogue of @xmath51-selected ir power - law sources in the cdfs constructed by alonso - herrero et al . ( 2005 ) . within our sample , we found only 8 no - x - ray - detected @xmath0 sources included in the alonso - herrero et al . we conclude , then , that the fraction of remaining plausibly - contaminating sources within our sample is basically negligible . a ) photometric versus spectroscopic redshifts for @xmath0 galaxies in the goods / cdfs . b ) distribution of relative errors . the median of this distribution is @xmath52 and the rms is @xmath53 . ] the redshift distribution of @xmath0 normal galaxies in the goods / cdfs . ] we used the public code hyperz ( bolzonella , miralles & pell , 2000 ) with the gissel98 template library ( bruzual & charlot 1993 ) to model the sed and obtain a photometric redshift ( @xmath54 ) for each galaxy in our sample , using photometry in up to eight broad - bands ( @xmath55 and irac @xmath44 and @xmath4 bands ) . only models with solar metallicity have been considered . dust - corrections have been taken into account convolving the template sed with the calzetti et al . ( 2000 ) reddening law . as it was shown in caputi et al . ( 2005a ) , the code hyperz has a tendency to favour high - redshift estimates when degenerate solutions in redshift space exist . to control the hyperz output , we applied the following set of constraints on the obtained photometric redshifts : * we used the @xmath56 relation for radio galaxies ( see figure [ hubble ] ) to determine the maximum possible redshift for each object . radio galaxies delimit the behaviour of the most massive galaxies formed at high redshifts ( e.g. willott et al . 2003 ) and normal galaxies are not usually found below this limit ( caputi et al . 2004 ; caputi et al . * the estimated redshift was constrained to the best - fit solution with a maximum value @xmath57 for all those galaxies with a counterpart in the public shallow u - band catalogues existing for the cdfs - band images exist for the cdfs ( arnouts et al . 2001 ) , they are shallower and poorer in resolution than the acs and isaac images used here and , thus , we decided not to incorporate these data for the input catalogue of the photometric redshift algorithms . ] , as higher redshift sources are unlikely to be bright at such short wavelengths . similarly , the estimated redshift was constrained to a maximum value @xmath58 for those sources detected in the @xmath3 band . * we used the public code bpz ( bentez 2000 ) to obtain a second , independent , set of redshift estimates for all the sources in our catalog . in the cases of sources with hyperz photometric redshifts @xmath59 not confirmed by bpz , we adopted the lower estimates given by the bpz code . * we used public combo17 photometric redshifts ( wolf et al . 2004 ) to replace the estimates of those sources with magnitude @xmath60 ( vega ) and @xmath61 ( the best regime for combo17 redshifts ) . observed @xmath62 magnitudes versus redshifts @xmath63 . the solid line corresponds to the empirical @xmath56 relation for radio galaxies as obtained by willott et al . ( 2003 ) , which approximately coincides with the passive evolution of a @xmath64 starburst formed at redshift @xmath65 . the dotted line is a nominal extrapolation of the same law . ] figure [ zqual]a ) shows the resulting estimated redshifts versus spectroscopic redshifts for all those @xmath0 sources included in different publicly available spectroscopy samples in the goods / cdfs ( e.g. cimatti et al . 2002b ; le fvre et al . 2004 ; vanzella et al . we observe a very good agreement between estimated and real redshifts in most cases . figure [ zqual]b ) shows the corresponding distribution of relative errors : the median is @xmath52 and the standard deviation is @xmath53 . however , in spite of the good accuracy obtained for the photometric redshifts , the incorporation of existing spectroscopic redshifts would be of benefit for the determination of our galaxy redshift distribution . consequently , in order to optimise the quality of our final redshift catalog , we used the available spectroscopic redshifts for the cdfs to replace the photometric estimates of our galaxies whenever possible . the final redshift catalogue for our 2905 @xmath0 galaxies is composed of 686/2905 ( 23.6% ) spectroscopic redshifts , 629/2905 ( 21.7% ) combo17 redshifts and 1590/2905 ( 54.7% ) hyperz / bpz redshifts . figure [ zhisto ] shows the resulting redshift distribution of our sample of 2905 @xmath0 normal galaxies in the goods / cdfs . 81% of these galaxies lie at redshifts @xmath66 , while the remaining 19% is found to be at @xmath67 . the two peaks of the distribution in the redshift bins @xmath68 $ ] and @xmath69 $ ] are produced by known large - scale structure in the cdfs ( e.g. le fvre et al . 2004 ) . the resulting hubble diagram @xmath2 versus @xmath63 is shown in figure [ hubble ] . [ cols="^,^,^,^,^,^,^",options="header " , ] caputi et al . ( 2005a ) computed and analysed the rest - frame @xmath2-band luminosity function ( lf ) from redshifts @xmath70 to @xmath6 . in this work , we take advantage of the larger area covered by the @xmath2-band goods / eis v1.0 release to set tighter constraints on the galaxy rest - frame @xmath2-band lf , and to model its evolution up to @xmath6 . the incorporation of irac @xmath44 and @xmath4 data enables us to compute this lf with no need of magnitude extrapolations up to redshifts @xmath71 and @xmath72 , respectively , and with smaller extrapolations at higher redshifts . anyway , we note that the k - corrections in the @xmath2-band are less dramatic than those in other shorter wavelength bands ( poggianti et al . 1997 ) . the distribution of the differences between the rest - frame absolute @xmath2-band magnitudes @xmath73 obtained from the closest available rest - frame band and the ones obtained extrapolating from the observed @xmath2 magnitudes in all cases . the distribution is centered at @xmath73 ( closest ) @xmath74 and the rms is @xmath75 . ] we obtained the rest - frame absolute magnitude @xmath73 of each galaxy directly from the hyperz output . for those objects with spectroscopic , combo17 or bpz redshifts , we adopted the best sed fitted by hyperz constrained to the predetermined redshift . for the remaining objects , the best - sed fit and redshift estimation were simultaneously obtained . hyperz computes the rest - frame absolute magnitude in the selected filter from the observed passband which is the closest to that filter in the rest - frame . in our case , the rest - frame @xmath73 magnitudes are obtained either from the @xmath2 , @xmath44 or @xmath4 apparent magnitudes , depending on the redshift of the source and the availability of irac data . with this ` closest band ' method , magnitude extrapolations are minimised . it is interesting to compare the rest - frame absolute @xmath2-band magnitudes @xmath73 obtained with this procedure , and those obtained when no irac data is taken into account for the sed fitting , i.e when the rest - frame @xmath2-band magnitude extrapolations are directly made from the observed @xmath2-band in all cases ( as , for instance , in caputi et al . 2005a , where irac data were not available ) . figure [ mksmks ] shows the distribution of the difference between the absolute magnitudes @xmath73 obtained with the two different methods . the statistics has been made over the magnitude range relevant for the computation of our lf ( see below ) , i.e. over all the sources with @xmath76 , as obtained with the ` closest - band ' method used here . we see that , although the resulting distribution has a long tail , the standard deviation @xmath77 is small : the center of the distribution is located at @xmath73 ( closest ) @xmath74 and the rms is @xmath75 . thus , we conclude that , in most cases , the rest - frame absolute magnitudes @xmath73 obtained extrapolating the observed @xmath2-band magnitudes are consistent with those derived from the longer - wavelength data used here . we computed the rest - frame @xmath2-band lf using two different techniques : the @xmath78 method and a maximum likelihood ( ml ) analysis ( marshall et al . 1983 ) . in the @xmath78 method , the lf is directly computed from the data , with no parameter dependence or model assumption . the ml analysis , on the other hand , is a parametric technique which requires an _ a priori _ functional form for the lf and its evolution with redshift . we parametrised the rest - frame @xmath2-band lf with a schechter function ( schechter 1976 ) : @xmath79 d\left(\frac{l}{l^\ast(z)}\right ) , \nonumber\\ \frac{l}{l^\ast(z ) } & = & 10^{-0.4(m - m^\ast(z ) ) } , \end{aligned}\ ] ] and we assumed that the density and luminosity evolved with redshift as @xmath80 , \\ m^\ast(z ) & = & m^\ast(z=0 ) - \left(\frac{z}{z_m}\right)^{k_m}.\end{aligned}\ ] ] the parameters @xmath81 and @xmath82 have been fixed to the local value : @xmath83 and @xmath84 ( kochanek et al . the obtained best - fit values for the remaining ( free ) parameters are given in table [ tab_sch ] . in figure [ kslf ] , we present the resulting rest - frame @xmath2-band lf at redshifts @xmath85 and @xmath86 . the empty symbols indicate the lf obtained with the @xmath78 method . the values shown are strictly above the luminosity completeness limits of each redshift bin . the number of sources in the luminosity - completeness range are 822 , 335 , 145 and 83 for the redshift bins @xmath87 , [ 1.25;1.75 ] , [ 1.75;2.25]$ ] and @xmath88 $ ] , respectively . to correct for the small incompleteness of our sample to @xmath36 , we weighted each source with a multiplicative factor depending on its apparent magnitude : 1.00 , 1.11 and 1.16 for sources with @xmath37 , @xmath38 and @xmath39 , respectively ( see section [ sec_sample ] ) . these correction factors have a minor impact on the resulting lf ( cf . caputi et al . the error bars correspond to the maximum of poissonian errors and the errors due to cosmic variance . we set a fixed value of 15% of the number counts for the cosmic variance error bar at @xmath89 $ \sim$}}~}1 $ ] ( cf . somerville et al . the solid lines show the evolution of the lf with redshift , as obtained through the ml analysis . in the four panels of figure [ kslf ] , we have added for comparison the local k - band lf measured from two micron all sky survey ( 2mass ) data by kochanek et al . ( 2001 ; dotted lines ) . the lfs obtained by the two different methods are in good agreement in the different redshift bins . from the upper - left panel in figure [ kslf ] , we can confirm once more that the bright end ( @xmath90 $ \sim$}}~}-25 $ ] ) of the rest - frame @xmath2-band lf significantly increases from redshifts @xmath5 to @xmath11 . this trend has already been observed in previous works ( drory et al . 2003 ; feulner et al . 2003 ; pozzetti et al . 2003 ; caputi et al . 2004 ; caputi et al . 2005a ) . from our ml analysis we find that , although the total density of objects @xmath91 decreases by a factor @xmath92 between redshifts @xmath5 and @xmath11 , the characteristic magnitude @xmath93 brightens in @xmath94 magnitudes between those redshifts . the overall result is a rise in the density of objects with @xmath90 $ \sim$}}~}-25 $ ] . recently , dahlen et al . ( 2005 ) found that the characteristic magnitude @xmath93 of the rest - frame @xmath22-band lf suffered a dimming from redshifts @xmath95 to @xmath96 . this result was apparently in contradiction with the brightening found by feulner et al . ( 2003 ) for the same lf in a similar redshift range . however , as dahlen et al . ( 2005 ) explain , the one - to - one comparison of the different parameters characterising the schechter function is misleading , as @xmath93 , @xmath81 and @xmath91 are coupled , and different behaviours in the evolution of @xmath93 can still produce consistent lfs . interestingly , on the other hand , dahlen et al . ( 2005 ) infer that , given the dimming they find for @xmath93 in the rest - frame @xmath22-band lf , the rest - frame @xmath2-band lf should present a similar behaviour . they suggested that the brightening observed by different authors in the rest - frame @xmath2-band lf might be due to inaccuracies in the k - corrections used to extrapolate from the observed to the rest - frame @xmath2-band at @xmath72 , rather than being a real effect . the incorporation of irac @xmath44 and @xmath4 data in the present work allows us to determine that this is not the case . once more , one should conclude that a one - to - one comparison of the best - fit parameters obtained for the schechter function ( and even more if they are inferred ) is misleading . the comparison of the densities of objects in each luminosity and redshift bin found by different surveys is the correct way of concluding on the resulting lfs . as it was shown by caputi et al . ( 2004 ; 2005a ) , after the rise of the bright end of the rest - frame @xmath2-band lf from redshifts @xmath5 to @xmath97 , the very bright end ( @xmath90 $ \sim$}}~}-26 $ ] ) of this lf does not show any sign of decline up to at least redshift @xmath98 . our ml analysis suggests that the density of the @xmath90 $ \sim$}}~}-26 $ ] objects might increase up to @xmath99 and then stay constant up to @xmath98 within the error bars ( cf . figure [ kslf ] ) . again , this can be explained by the combined effect of the number and luminosity evolution we obtain through the ml analysis : the total density of objects @xmath91 decreases by a factor @xmath100 between @xmath5 to @xmath98 , but the characteristic magnitude @xmath101 brightens by more than one magnitude between these redshifts . the combined density and luminosity evolution we determined through the ml analysis also implies a particularly strong decrement in the number of intermediate and low luminosity ( @xmath102 ) objects with increasing redshift . we remark , however , that we do not directly see these intermediate and low luminosity objects at high redshifts , given the limits of our survey . we only extrapolate this behaviour from the lf obtained through the ml analysis , which assumes the validity of a schechter function for the description of the lf at all redshifts . thus , two alternative conclusions can be extracted from this result : 1 ) a significant fraction of low and intermediate luminosity objects are created between @xmath98 and @xmath5 , or 2 ) at least as many low luminosity objects are present at high as at low redshifts , and our assumptions for the computation of the lf with the ml analysis are not valid to describe the faint - end of the lf at high redshifts . the direct access to low luminosity objects at high redshifts through deeper surveys is necessary to conclusively prove the behaviour of the faint - end of the lf . in hierarchical models , the creation of new objects is the consequence of the production of galaxy mergers and subsequent triggering of star formation . very recently , conselice ( 2005 ) found that the fraction of mergers in faint and low mass galaxies peaked at lower redshifts than for higher luminosity and more massive galaxies . this result appears to be consistent with the presence of luminous galaxies at high redshifts @xmath103 , as they would be produced from mergers at these or even higher redshifts . the decline in the number of intermediate luminosity objects at high redshifts that we extrapolate from the ml analysis would also be consistent with the fact that a significant fraction of them form through mergers at later times . however , the differential redshift peak observed for mergers of different mass and luminosity still can not decide on the formation epoch of all the lowest luminosity objects , and the two hypotheses discussed above could still be possible . at redshifts @xmath16 $ \sim$}}~}1 $ ] , major mergers become progressively less important ( e.g. patton et al . 2002 ) , so it is expected that the evolution of the @xmath2-band lf is mainly governed by the dimming of the already present old stellar populations . the significant decrease of the bright end ( @xmath104 ) of the lf from redshifts @xmath97 to @xmath5 is probably indicating that the most massive galaxies present at @xmath105 must have finished the construction of the bulk of their stars by this redshift , and mainly fade out all the way through redshift @xmath5 ( cf . section [ sec_masscol ] ) . we used the rest - frame @xmath2-band luminosity of each galaxy to compute its stellar mass . we obtained the rest - frame absolute magnitude @xmath73 ( and corresponding luminosity ) from the hyperz best - fit sed on an individual basis , as explained in section [ sec_lf ] . we used the public code galaxev ( bruzual & charlot 2003 ) to construct grids of mass - to - light ( @xmath106 ) ratios depending on galaxy age , assuming a salpeter imf over stellar masses @xmath34 . the advantage of using luminosities in the rest - fame @xmath2-band is that other parameters , such as the dust - correction or the galaxy star - formation history , have a minor importance in the determination of the @xmath106 ratios . as in caputi ( 2005a ) , we modelled the evolution of these ratios using two galaxy templates , with no dust , corresponding to different exponentially declining star - formation histories with characteristic times @xmath107 and @xmath108 . we determined the @xmath106 of each galaxy using the model most similar to the hyperz best - fit template in each case , and interpolating between the values which corresponded to the closest ages in the grids . our estimated stellar masses are typically accurate within a factor @xmath109 $ \sim$}}~}2 $ ] ( where the larger uncertainties correspond to plausible degeneracies in parameter space , especially in those cases in which the galaxy redshift is simultaneously determined with the best - sed fit ) . on the contrary , the dependence on star - formation history and dust corrections would be much stronger if optical rather than near - ir luminosities were used to derive the stellar masses . as we discussed in section [ sec_lf ] , the rest - frame @xmath2-band luminosity derived as here and the luminosity that would be derived from the extrapolation of the corresponding observed @xmath2-band magnitude are consistent within a rms @xmath75 . this implies that the derived stellar masses will be consistent within @xmath110 , i.e. within a factor @xmath111 . thus , we conclude that the use of the observed b through @xmath2-band data , with no incorporation of irac data , would have introduced an additional uncertainty of @xmath112% on the derived stellar masses . this is a relatively minor contribution to the error budget , which is mainly dominated by plausible degeneracies in parameter space and especially by the adoption of a fixed imf . figure [ imkmass ] shows the observed ( @xmath113 ) colours versus redshift for galaxies with stellar masses within different mass ranges . in all the panels , circles represent those objects with precise ( @xmath113 ) colours , while triangles indicate lower limits to the colours of objects with magnitude @xmath114 ( vega ; equivalent to @xmath109 $ \sim$}}~}2\sigma$ ] detections in the @xmath20 band ) , computed as ( @xmath115 ) . the horizontal dashed line shows the typical colour cut defining extremely red galaxies ( ergs ; @xmath116 , vega ) , while the vertical dotted lines delimit the region of completeness for the lowest masses in each mass bin . the mass completeness limits have been estimated using the median of the k - corrections and the maximal @xmath106 ratios at each redshift ( corresponding to a galaxy formed in a single burst at redshift @xmath117 ) . at redshifts @xmath89 $ \sim$}}~}1 $ ] , typical erg colours indicate the presence of old stellar populations and/or young stellar populations heavily enshrouded by dust . on the contrary , @xmath118 colours at @xmath119 imply the presence of unextincted on - going star - formation . at @xmath16 $ \sim$}}~}1 $ ] , the observed @xmath120 colour cut is merely formal , as only systems with very large dust extinctions would show colours above that limit . inspection of figure [ imkmass ] shows that the @xmath121 colours of galaxies at a given redshift depend on the stellar mass . the fraction of red - to - blue galaxies systematically increases with assembled stellar mass , even considering each mass bin only up to the completeness redshift limit . in particular , the vast majority of galaxies with @xmath8 between redshifts @xmath11 and @xmath122 are ergs . some of these galaxies might be purely passive systems ( cf . e.g. daddi et al . 2005 ) , while in other cases the extremely red colours would only be explained by the superposition of old stellar components and dust ( see caputi et al . 2004 ; papovich et al . 2005 ; and cf . section [ sec_nearir ] ) . the extremely red colours characterising these very massive galaxies reveal important constraints on their star - formation histories . in contrast to what is observed in other less massive systems , our results indicate that unobscured time - extended star - formation at @xmath89 $ \sim$}}~}1 $ ] is rare in galaxies with assembled stellar mass @xmath8 . the main build - up of these systems must be produced in short bursts and/or with considerable amounts of dust . the observed @xmath121 colours versus redshift for galaxies with estimated stellar mass @xmath123 . the symbols are the same as in fig . [ imkmass ] . the solid and dotted - dashed lines show the modelled evolution of the @xmath121 colours for galaxies formed in an instantaneous burst at redshift @xmath124 and @xmath125 , respectively , both with no dust and passive evolution thereafter . ] figure [ imkhighm ] shows an amplified version of the lower - right panel of figure [ imkmass ] . for reference , we added in this figure the modelled evolution of the @xmath121 colours for galaxies formed in an instantaneous burst at redshift @xmath124 and @xmath125 ( solid and dotted - dashed lines , respectively ) , both with no dust and passive evolution thereafter . from figure [ imkhighm ] we can see that , at @xmath119 , galaxies within the erg regime with assembled stellar masses @xmath123 display a wide range of @xmath121 colours . this suggests that they must undergo a variety of star - formation histories , either different formation redshifts or at least different formation redshifts for the youngest stellar population present in them . at redshifts @xmath16 $ \sim$}}~}1 $ ] , on the contrary , the galaxies with stellar mass @xmath123 present in our sample show quite homogeneous @xmath121 colours . this is again in contrast to what is observed in less massive systems . the homogeneous colours displayed by these most massive galaxies suggest that their optical / near - ir light must be dominated by old stellar populations . however , this does not necessarily mean that all of them are following a strictly passive evolution , but that any possible on - going star formation would only produce minor changes to their properties ( e.g. stellar mass ) . in fact , a substantial fraction ( @xmath126 $ \sim$}}~}30\%$ ] ) of these galaxies are detected in the mid - ir , i.e. experience on - going star - formation ( caputi et al . 2005b , 2005c ) . the typical ir - derived star - formation rates of these mid - ir detected galaxies at @xmath127 are @xmath128 ( caputi et al . if we consider that the lifetime of a starburst episode is at most @xmath129 , we infer that these galaxies should be forming @xmath130 per starburst episode at those redshifts , i.e. an additional minor amount to their already existing stellar mass . some of these very massive systems at @xmath16 $ \sim$}}~}1 $ ] could be candidates for cluster e / s0 galaxies , as early - type cluster galaxies show very little dispersion in their colours from @xmath5 ( e.g. bower et al . 1992 ) to @xmath72 ( e.g. ellis et al . 1997 ; stanford et al . 1998 ; blakeslee et al . indeed , the existence of large - scale structure is suggested by the colour - redshift diagram , as a considerable number of our @xmath8 galaxies appear to lie at discrete redshift values . on the contrary , only a morphology study would be able to conclude on the plausible early or late - type classification of these systems ( mclure et al . 2005 , in preparation ) . the comoving number density of galaxies with stellar mass @xmath131 ( circles ) and @xmath132 ( triangles ) ( salpeter 1955 imf , restricted to @xmath133 ) . the filled triangle at @xmath134 denotes the re - computed number density of @xmath132 galaxies after correcting for the effects of large - scale structure in the corresponding redshift bin . the star - like symbols indicate the density values for the same mass thresholds at @xmath5 , obtained integrating the local stellar mass function ( cole et al . 2001 ; bell et al . the dashed lines show the evolution of the number density of haloes with total mass @xmath135 and @xmath136 ( upper and lower curves , respectively ) , as obtained from @xmath137-cdm models of structure formation ( with a code kindly provided by will percival ) . the dotted line corresponds to the predicted number density of galaxies with mass @xmath138 obtained with sph simulations ( nagamine et al . 2005 ) . ] the comoving number density of galaxies with stellar mass @xmath131 ( chabrier 2003 imf ) . the symbols are the same as in figure [ cnumd ] . ] as we did in caputi et al . ( 2005a ) , we analyse the evolution of the number density of different - mass galaxies over cosmic time . in the present work , however , due to the analysis of a considerably larger area , we are able to select a statistically significant sample of very massive galaxies ( @xmath132 ) , which enables us to explore the very - high mass end of the stellar - mass evolution . figure [ cnumd ] shows the comoving number density of galaxies with assembled stellar mass @xmath138 ( circles ) and @xmath132 ( empty triangles ) . the asterisks denote the corresponding local values , as obtained from the integration of the local stellar mass function ( cole et al . 2001 ; bell et al . as in the computation of the galaxy lf ( cf . section [ sec_lf ] ) , the error bars correspond to the maximum of poissonian errors and the errors due to cosmic variance . all the results presented in this section incorporate @xmath139 corrections , but deliberately not sample incompleteness corrections ( which anyway are very small ) . we decided to do so to explicitly show that the contrast betweeen some of the observational and theoretical results discussed below is independent of such corrections . the evolution of the densities of galaxies with stellar mass @xmath138 shows that a minimum of 15% of the local value is in place before redshift @xmath12 ( this value rises to @xmath13% considering the number densities cut at stellar masses @xmath140 , i.e. the regime of strict mass completeness to @xmath12 ) . this results is consistent with caputi et al . ( 2005a ) . however , in contrast to this previous work , we observe here that the number density of @xmath138 systems does increase from redshifts @xmath141 to @xmath142 , although this increment is modest ( @xmath13% of the local value ) . more than 50% of the local @xmath138 galaxies appear to have formed at redshifts @xmath16 $ \sim$}}~}1.5 $ ] . the dashed lines in figure [ cnumd ] trace the evolution of dark - matter haloes with total mass @xmath135 and @xmath136 ( upper and lower curves , respectively ) , as obtained from the extended press - schechter formalism ( bond et al . 1991 ) , with a code kindly provided to us by will percival . the mass thresholds for the halo distributions have been deliberately selected to coincide with the corresponding number densities of galaxies at @xmath5 , assuming a single halo occupation number . the compared observed and predicted density evolutions confirm the trend found in caputi et al . ( 2005a ) : at redshifts @xmath143 $ \sim$}}~}z { \raisebox{-0.13cm}{~\shortstack{$<$ \\[-0.07 cm ] $ \sim$}}~}3 $ ] , the density of haloes with total mass @xmath135 is systematically larger than the density of galaxies with stellar mass @xmath138 . caputi et al . ( 2005a ) also showed that the number density of galaxies and haloes coincide again at higher redshifts @xmath144 , although this can not be re - assessed from the present sample as the @xmath36 cut implies completeness only for stellar masses @xmath132 up to @xmath145 . on the other hand , we investigate the very high mass end ( @xmath132 ) of galaxy number density evolution and also compare with the corresponding theoretical predictions . inspection of figure [ cnumd ] shows that up to @xmath13% of the local massive galaxies with assembled stellar mass @xmath132 is in place before @xmath122 , while the construction of the majority of these very massive systems occurs between redshifts @xmath12 and @xmath146 . within our sample , we find that @xmath147% of the local galaxies with mass @xmath132 is already in place by redshift @xmath148 . this is contrast to what is observed for less massive systems , whose evolution is only halfway complete by that redshift . the comparison with theoretical models shows that , if we assume the same conversion between baryonic and dark - matter halo mass as inferred at @xmath5 , there appear to be too few dark matter haloes of adequate mass in place at high redshift to host the observed number of high mass galaxies . we observe that the deficit is particularly important at @xmath89 $ \sim$}}~}3 $ ] . similar incompatibilites between the presence of very massive systems and the predictions of theoretical models at high redshifts have already been pointed out in previous studies ( e.g. fontana et al . 2003 ; genzel et al . 2003 ; pozzetti et al . 2003 ; cimatti et al . 2004 ; fontana et al . 2004 ; mccarthy et al . 2004 ; saracco et al . our results show that this incompatibility might certainly exist for the most massive galaxies . it could be argued , however , that our excess of very massive galaxies might be biased by cosmic variance effects ( more than what is accounted for within the error bars ) . in effect , as we noted in section [ sec_masscol ] , the existence of large - scale structure within our sample can be deduced from the colour - redshift diagrams . within our sample , we find 16 galaxies with @xmath132 at @xmath149 , yielding a number density @xmath150 . 9/16 of these galaxies have photometric redshifts concentrated in the narrow interval @xmath151 , and there is also a spectroscopically confirmed qso at redshift @xmath152 within our surveyed field ( not included in the final galaxy sample analysed in this work ) , strongly suggesting the presence of large - scale structure for very massive galaxies at these redshifts . to obtain a lower limit on the number density of galaxies with stellar mass @xmath132 at @xmath153 , we re - computed this density excluding the 9 galaxies at @xmath151 . we obtained @xmath154 , i.e. @xmath155 ( filled triangle in figure [ cnumd ] ) , while the density of the corresponding haloes at that redshift is only @xmath156 . thus , it seems that even taking into account the effects of large - scale structure , we can not completely reconcile the theory predictions with the observations at high redshifts . the total stellar mass densities obtained from the present study ( large circles and up - triangle , with error bars , for precise values and lower limit , respectively ) , compared to values previously obtained by other authors : dickinson et al . 2003 ( down - triangles ) ; fontana et al . 2003 ( asterisks ) ; rudnick et al . 2003 ( diamonds ) ; glazebrook et al . 2004 ( right - triangles ) ; fontana et al . 2004 ( squares and double squares for the observed and extrapolated values in the k20 survey , respectively ) ; caputi et al . 2005a ( small circles and double circle for total and @xmath157 mass densities , respectively ) ; drory et al . 2005 ( crosses ) . the circle with a star at @xmath158 show the local value of the total stellar mass density obtained from the integration of the local galaxy stellar mass function ( cole et al . 2001 , bell et al . 2003 ) . the solid line is the evolution of the stellar mass density as obtained from the integration of the star - formation rates derived from sloan digital sky survey ( sdss ) data ( heavens et al . dickinson et al . values correspond to the average of the values obtained with the 1-component and 2-component models with solar metallicity ( see table 3 in dickinson et al . rudnick et al . values are computed only on galaxies with rest - frame v - band luminosity @xmath159 and , thus , miss a significant fraction of the mass . glazebrook et al . values only correspond to galaxies with estimated mass @xmath160 and are at least partially incomplete above redshift @xmath161 . all the values in this figure assume a salpeter imf over stellar masses @xmath133 and a cosmology with @xmath162 , @xmath32 and @xmath33 . ] recently , nagamine et al . ( 2005 ) showed that the properties of high redshift red and massive galaxies could be well reproduced through the use of hydrodynamical simulations with star - formation rate histories that peak at higher redshifts than those used in semianalytic models . it is then interesting to compare also their predictions with our observed evolution of number densities of massive galaxies . the dotted line in figure [ cnumd ] shows the predicted number density of galaxies with stellar mass @xmath138 , as obtained with the nagamine et al . smoothed particle hydrodynamics ( sph ) simulations . we observe that these simulations yield densities below the observed values , and the discrepancies are larger at low than at high redshifts . however , in the nagamine et al . models , galaxies masses are directly obtained from the simulations and are imf independent . thus , for a proper comparison with these model results , we need to test whether the adoption of a different imf to compute our stellar masses could improve the agreement between model and data . figure [ chabrier ] shows the evolution of the comoving number density of galaxies with stellar mass @xmath138 considering a chabrier ( 2003 ) imf . both our observed densities and the local value have been transformed in accordance to this imf . the change from a salpeter ( 1955 ) to a chabrier ( 2003 ) imf produces an overall re - scaling of all our stellar masses by @xmath163 , in such a way that a cut @xmath138 for masses in the chabrier imf corresponds to a cut @xmath164 $ \sim$}}~}2 \times 10^{11 } m_\odot$ ] in the salpeter imf . we repeat the same exercise as in figure [ cnumd ] and compare the observed values with the densities of dark matter haloes with a mass threshold deliberately selected to coincide with the density of galaxies at @xmath5 . as in figure [ cnumd ] , we observe that , at very high redshifts @xmath89 $ \sim$}}~}3 $ ] , the predicted number of haloes appears to be insufficient to encompass the observed densities of @xmath138 ( chabrier ) galaxies . on the contrary , the sph predicted number densities of @xmath138 galaxies by nagamine et al . ( 2005 ) appear to be in quite good agreement with our observed values from redshifts @xmath72 through @xmath165 , although they produce some excess of massive local galaxies . we note that a similar behaviour would be deduced from the densities of dark matter haloes if we did not constrain the comparison to those haloes with densities coinciding with the local densities of massive galaxies . this shows that what we find here is a direct discrepancy with the predictions of hierarchical models , and does not depend on the specific recipes adopted either by semianalytic models or sph simulations . in summary , hierarchical models do not seem to be able to reproduce the observed densities of the most massive galaxies from the local universe to very high redshifts . alternatively , it could also be argued that the use of a single imf is not suitable to convert from luminosities to stellar masses in the whole range of redshifts @xmath5 to @xmath122 ( see e.g. kroupa & weidner 2005 , for a discussion ) . unluckily , the validity of a single imf is an issue extremely hard to test at high redshifts , and it is not possible to overcome the imf uncertainty when doing the comparisons between observations and theory . finally , we computed the evolution of the total stellar mass density with redshift , as derived from our sample of @xmath0 galaxies . figure [ compstm ] shows our obtained stellar mass densities ( large circles and up - triangle , with error bars ) , compared to other values previously published in the literature . from the present sample we find that the total stellar mass densities at @xmath166 and @xmath167 are @xmath168 and @xmath169 , respectively , i.e. @xmath170 and @xmath171 of the local value . these values very well agree with the previous determinations made by caputi et al . ( 2005a ) and other works . at higher redshifts , the completeness limits of our sample only allows us to obtain lower limits on the total stellar mass densities . the up - triangle in figure [ compstm ] represents a lower limit to the total stellar mass density at redshift @xmath98 . as our galaxy sample is only complete for masses @xmath172 up to @xmath141 , we compare this lower limit to the stellar mass density contained in galaxies with stellar mass @xmath173 obtained by caputi et al . ( 2005a ) ( double circle in figure [ compstm ] ) , and observe a very good agreement . the total stellar mass densities at redshifts @xmath89 $ \sim$}}~}2 $ ] have been better constrained by previous deeper surveys . the estimated stellar mass densities at @xmath103 and @xmath144 are ( 20 - 24)% and ( 13 - 18)% of the local values , respectively ( caputi et al . 2005a ; drory et al . the observed versus predicted @xmath174)$ ] and @xmath175)$ ] colours ( upper and lower panels , respectively ) for the @xmath0 ( vega ) galaxies in the restricted area of the goods / cdfs studied by caputi et al . ( 2005a ) . the dashed lines indicate the region of consistency , taking into account the error bars on the observed colours . ] caputi et al . ( 2005a ) predicted the pure near - ir @xmath2/irac colours of @xmath2-selected galaxies based on the best - fit sed of each object . these predictions were made for the galaxies selected in a sub - region of 50.4 arcmin@xmath35 of the area analysed in this work . in the light of the currently available irac data , we can assess the validity of these predictions . figure [ colobspr ] shows the observed versus predicted @xmath174)$ ] and @xmath175)$ ] colours for the @xmath2-band galaxies in the restricted goods / cdfs area ( upper and lower panels , respectively ) . for clarity , we only show those sources with small error bars in their observed irac magnitudes , i.e. @xmath176 and @xmath177 . as in all cases the @xmath2 magnitudes have @xmath178 $ \sim$}}~}0.2 $ ] , the measured colours are typically precise within @xmath179 0.4 mag ( dashed lines in both panels of figure [ colobspr ] ) . from this figure we determine that 89% and 84% of the predicted @xmath174)$ ] and @xmath175)$ ] colours , respectively , are in agreement with the observed colours , within the error bars . we can confirm , then , the majority of the near - ir colours predicted in caputi et al . ( 2005a ) . the observed @xmath174)$ ] and @xmath175)$ ] colours ( upper and lower panels , respectively ) versus redshift for the @xmath0 ( vega ) galaxies in the goods / cdfs . ] figure [ colnearir ] shows the observed @xmath174)$ ] and @xmath175)$ ] colours versus redshift ( upper and lower panels , respectively ) for the @xmath0 ( vega ) galaxies in the larger sample studied in this work . @xmath44 and @xmath4 data are available for alternative @xmath47 of the sources in our sample . once more , for clarity , we only show those sources with magnitude errors @xmath176 and @xmath177 . the variation of these observed colours with redshift confirms the trends predicted by caputi et al . ( 2005a ) within the error bars . inspection of figure [ colnearir ] reveals , however , the existence of a population of low - redshift ( @xmath16 $ \sim$}}~}0.5 $ ] ) galaxies with colours @xmath174)>1.0 $ ] and @xmath175)>1.5 $ ] , which are quite redder than those predicted for low - redshift galaxies in the previous work . the colour excess of these galaxies can not be reconciled with the predictions by simply taking into account the photometry error bars . the analysis of these sources shows that 31 out of the 32 objects satisfying @xmath16 $ \sim$}}~}0.5 $ ] and @xmath174)>1.0 $ ] and all of the 17 objects satisfying @xmath16 $ \sim$}}~}0.5 $ ] and @xmath175)>1.5 $ ] are very low mass systems , with assembled stellar masses @xmath180 . the majority of them are best - fit by young - galaxy templates . the excess of near - ir colour could be explained by the presence of an active nucleus , whose irac magnitudes could not be properly predicted using the gissel98 stellar templates . on the other hand , an alternative and perhaps more likely explanation is that these very red @xmath2/irac colours are produced by the presence of prominent emission lines in the irac bands . the presence of an aromatic feature at @xmath181 is characteristic of star - forming galaxies . these aromatic features are not incorporated in the gissel98 templates , explaining why such very red colours could not have been predicted from the sed models . if this is the case , the colour excess would be the signature of intense star - formation in these young objects . only near - ir spectrocopy would be able to confirm the nature of these sources . figure [ ksm45comp ] shows the observed @xmath175)$ ] colours versus redshift in different stellar mass bins , analogously to figure [ imkmass ] . we observe that the @xmath175)$ ] colours increase up to redshift @xmath182 , and the colour - redshift relation becomes tighter with increasing stellar mass . in contrast , the colour dispersion becomes quite large above that redshift , where the observed near - ir bands start to map the rest - frame optical bands . this colour dispersion is in part due to larger photometry uncertainties , but it is probably also an indication of a wide variety of star - formation histories and dust extinction in these galaxies . figure [ ksm45model ] shows an amplified version of the high - mass panel in figure [ ksm45comp ] . we added for comparison the modelled evolution of the observed @xmath175)$ ] colours for galaxies formed in a single burst at redshifts @xmath125 and @xmath124 , both with no dust and passive evolution thereafter . some interesting conclusions can be extracted from the analysis of this figure . firstly , it is quite evident that the observed near - ir colours of many massive galaxies at high redshifts ( @xmath89 $ \sim$}}~}1.5 $ ] ) can not be explained without considerable amounts of dust ( cf . also caputi et al . 2004 ) . independently of the formation redshift , the colours of a passively - evolving single burst are not sufficient to reproduce the red observer - frame near - ir colours of massive galaxies . we conclude , then , that purely passive galaxies should be a minority at high redshifts . secondly , the colours observed at lower redshifts ( @xmath16 $ \sim$}}~}1.5 $ ] ) indicate that the near - ir light of @xmath8 galaxies must be dominated by evolved stellar populations . a similar conclusion has been obtained from the analysis of the @xmath121 colours of these systems . we note that the tight colour - redshift relation observed for high - mass galaxies at @xmath16 $ \sim$}}~}1.5 - 2.0 $ ] does not imply a common formation redshift or star - formation history . for instance , as we mentioned in section [ sec_masscol ] , some of these galaxies are bright mid - ir sources , while others are not , suggesting a wide variety of on - going star - formation rates ( caputi et al . 2005b , c ) . however , and in contrast to what is progressively observed in less massive systems , all the plausible new star - formation in very massive galaxies at low redshifts could have a minor impact on their overall properties and , in particular , on their sed colours . the evolution of the observed @xmath183)$ ] colours with redshift for galaxies with estimated stellar mass @xmath8 . the solid and dotted - dashed lines show the modelled evolution of the @xmath183)$ ] colours for galaxies formed in an instantaneous burst at redshift @xmath124 and @xmath125 , respectively , both with no dust and passive evolution thereafter . ] with the goods project close to completion , in this work we have been able to set tighter constraints on the evolution of near - ir - selected galaxies in the goods / cdfs field . we took advantage of the latest available very deep and homogeneous multiwavelength photometry , in conjunction with the information provided by complementary spectroscopic programs and the combo17 campaign , to construct an optimised redshift catalogue for the @xmath0 galaxies in the goods - south field . based on this catalogue , we performed a detailed analysis of galaxy luminosity and mass evolution up to very high redshifts @xmath184 . the incorporation of the spitzer / irac data at @xmath185 and @xmath186 allowed us to compute rest - frame @xmath2-band luminosities and stellar masses with considerably smaller extrapolations than those applied in caputi et al . ( 2005a ) . in spite of this important difference , we note that most of the main conclusions obtained by caputi et al . ( 2005a ) have been confirmed by this work . the addition of irac data did not produce any overall change in the derived galaxy luminosities and stellar masses . other authors also have concluded that the use of irac data produced a minor impact on the derived galaxy stellar masses , when other shorter optical / near - ir multiwavelength data are also available ( e.g. shapley et al . 2005 ) . from the modelling of the evolution of the @xmath2-band luminosities , we determined that the characteristic magnitude @xmath93 of the rest - frame @xmath2-band lf significantly brightens from redshifts @xmath5 to @xmath98 . this is simultaneously produced with a decrease in the total density of objects , so we conclude that the @xmath2-band lf has both a substantial luminosity and density evolution with redshift . the characteristics of this evolution reveal some important aspects of the sequence of galaxy formation . for instance , that most of the brightest galaxies are assembled at high redshifts ( @xmath89 $ \sim$}}~}1 $ ] ) , while the creation of a substantial number of intermediate - to - low luminosity objects might continue down to low redshifts ( although the latter will need to be confirmed by future deeper surveys , which have direct access to low - luminosity objects at high redshifts ) . we have also studied the very high stellar - mass end ( @xmath132 ) of galaxy evolution . we found that most of these very massive systems in our sample have been formed in the @xmath187-period of time between redshifts @xmath141 and @xmath148 , although a significant fraction ( up to @xmath13% ) appear to have been in place before redshift @xmath122 . very recently , from the cross - correlation of our @xmath0 galaxy sample with the spitzer/@xmath51 galaxies in the same field , caputi et al . ( 2005b ) found that the most - intense star - forming galaxies at redshifts @xmath188 are mainly massive ( @xmath189 ) galaxies . this is also the main epoch of flourish of submillimetre galaxies ( chapman et al . 2005 ) , which presumably are progenitors of the most massive local galaxies ( e.g. stevens et al . thus , this period appears to be critical to understand the formation and evolution of the bulk of the most massive systems present in the local universe . theoretical models of galaxy formation in their current status are not completely able to account for all of these results . several different clues have been proposed in recent works to try to improve the agreement of the models with the data at high redshifts , including e.g. the effects of qso and supernova feedback ( granato et al . 2004 ; cirasuolo et al . 2005 ) , the adoption of an alternative imf ( baugh et al . 2005 ) or the use of hydrodynamical models ( nagamine et al . however , the discrepancies are not produced by the specific recipes used by theoretical models to account for star formation , but seem to lie at the more fundamental level of hierarchical model predictions . hierarchical models do not seem to be able to reproduce the whole of observational properties from the local universe to high redshifts . a likely alternative to this problem is the invalidity of a universal imf when deriving stellar masses from the observed luminosities . a definitive comprehension of the main epoch of very massive galaxy formation , as well as the determination of the initial distribution of stars in these systems , still constitutes a major challenge from both the observational and theoretical points of view . this paper is based on observations made with the advanced camera for surveys on board the hubble space telescope operated by nasa / esa ; with the infrared spectrometer and array camera on the ` antu ' very large telescope operated by the european southern observatory in cerro paranal , chile ; and the infrared array camera on board the spitzer space telescope , operated by the jet propulsion laboratory , california institute of technology , under a contract with nasa ; and form part of the publicly available goods datasets . we thank the goods teams for providing reduced data products . we are very grateful to kentaro nagamine for kindly providing us with his sph simulation predictions , and to will percival for providing us the code to compute the number densities of dark matter haloes . we thank the anonymous referee for his / her useful comments and suggestions , which improved the discussion of results in this work . kic acknowledges cnes and cnrs funding . rjm acknowledges the support of the royal society . alonso - herrero a. et al . , 2005 , apj , submitted

a mosaic of four uit far - uv ( fuv ) ( = 1620 ) images , with derived stellar and hii region photometry , is presented for most of the bar of the smc . the uv morphology of the smc s bar shows that recent star formation there has left striking features including : a ) four concentrations of uv - bright stars spread from northeast to southwest at nearly equal ( arcmin=0.5 kpc ) spacings ; b ) one of the concentrations , near dem 55 , comprises a well - defined 8-arcmin diameter ring surrounded by a larger h ring , suggestive of sequential star formation . fuv psf photometry is obtained for 11,306 stars in the fuv images , resulting in magnitudes m(162 ) . we present a fuv luminosity function for the smc bar , complete to m(162).5 . detected objects are well correlated with other smc population i material ; of 711 h emission - line stars and small nebulae within the uit fields of view , 520 are identified with fuv sources . the fuv photometry is compared with available ground - based catalogs of supergiants , yielding 191 detections of 195 supergiants with spectral type earlier than f0 in the uit fields . the ( m(162)v ) color for supergiants is a sensitive measure of spectral type . the bluest observed colors for each type agree well with colors computed from unreddened galactic spectral atlas stars for types earlier than about a0 ; for later spectral types the observed smc stars range significantly bluer , as predicted by comparison of low - metallicity and galactic - composition models . redder colors for some stars of all spectral types are attributed to the strong fuv extinction arising from even small amounts of smc dust . internal smc reddenings are determined for all catalog stars . all stars with e(bv).15 are within regions of visible h emission . fuv photometry for 42 h-selected hii regions in the smc bar is obtained for stars and for total emission ( as measured in hii - region - sized apertures ) . the flux - weighted average ratio of total to stellar fuv flux is 2.15 ; consideration of the stellar fuv luminosity function indicates that most of the excess total flux is due to scattered fuv radiation , rather than stars fainter than m(162)=14.5 . both stellar and total emission are well correlated with h fluxes measured by kennicutt and hodge ( 1986 ; hereafter kh ) , yielding fuv / h flux ratios that are consistent with models of smc metallicity , ages from 1 - 5 myr , and moderate ( e(bv)=0.00.1 mag ) internal smc extinction .

astro-ph9612152

uv observations from above the earth s atmosphere are vital for understanding the population i properties of metal - poor galaxies such as the small magellanic cloud ( smc ) . optical - band studies of the smc have shown that this nearby dwarf irregular galaxy provides a unique laboratory for investigating stellar and interstellar evolution at `` primitive '' compositions , compared to the galaxy . many effects of composition differences appear best , or only , in the uv . line blanketing strongly affects uv colors ; similarly , the steep smc extinction curve ( a@xmath5 ) , widely thought to be due to abundances in smc dust , is `` extreme '' only in the uv . moreover , because the energy distributions of hot stars peak at short wavelengths , fuv photometry is more effective than optical - band photometry in determining the temperatures of such stars . here , we present initial results from a uv imaging study of the the smc , based on a mosaic of four far - uv images that were obtained by the ultraviolet imaging telescope ( uit ) during the astro-1 and astro-2 missions in late 1990 and early 1995 . uit observed a total of four 40-arcmin diameter fields nearly completely covering the smc bar at an angular resolution of @xmath13 arcsec better than 75 times the resolution of previous uv imaging studies of the smc ( @xcite ) . a description of individual exposures used in this study is in table [ fuvobstab ] , and a mosaic of the longest exposures for each field is shown as figure [ mosaic ] . details of uit hardware , calibration , operations , and data reduction are in @xcite and @xcite . data discussed here are from images made with the @xmath0=1620 filter , hereafter called b5. astro-1 data also include short exposures of field 1 made in a near - uv ( @xmath0=2490 ) and an additional far - uv ( @xmath0=1520 ) bandpass ( cf @xcite ) which are not discussed here . ( a discussion and analysis of uit imaging of the large magellanic cloud is given in @xcite . ) uit film images are reduced to linearized arrays and absolutely calibrated by comparison to other uv spectrophotometry as described in @xcite . the calibration for astro-2 data has been revised using flight observations of standard fields ( @xcite ) ; estimated absolute uncertainty is 15% for well - exposed pixels . an idl / uit implementation of daophot ( @xcite ; @xcite ) is used to locate stars and perform aperture and psf - fit photometry . both aperture and psf - fit photometry use a 2.9-arcsec ( 2.5-pixel ) radius , which sets the limit for resolution of blended stars and the minimum size for non - stellar sources . psf - fit photometry , using parameters chosen for images individually , is used for subsequent analysis of stellar objects . astrometric solutions are derived for uit images using hst guide stars as standards ( @xcite ) . 11306 stars are measured , and positions , magnitudes , and errors for these objects are provided in the aas cd - rom series . the file includes uit star number , field number , x and y positions in uit pixels , right ascension and declination , m(162 ) derived from psf photometry , and estimated error in m(162 ) . for the smc data , iue observations are used to measure combined aperture corrections and improved single - field absolute calibrations . the corrections are determined by integrating well - exposed and calibrated ( bohlin _ et al . _ 1990 ) iue spectra of isolated field stars in each uit image , and comparing the results with uit psf photometry . the number of iue spectra used for each field ranges from 12 to 31 . statistical uncertainties in mean iue / uit ratios range from .07 to .02 mag , with larger uncertainties for longer uit exposures because of the overexposure of the iue sources on those images . the iue / uit ratio thus derived agrees with the nominal uit absolute calibration within 10% , after application of a directly measured aperture correction . uit magnitudes are defined by m@xmath6 = @xmath32.5 log(f@xmath6)@xmath321.1 , where f@xmath6 is in erg ( @xmath7 s)@xmath8 . data from multiple exposures of each field are combined by selecting data from individual exposures which maximize the signal - to - noise ratio . for stellar images and pixels which are common to two fields , data are used from the field with the longer maximum exposure time , after signal - to - noise selection . in practice signal - to - noise selection is achieved for stellar photometry by restricting each exposure to a selected stellar magnitude range ; this avoids systematic errors at the extremes of the uit dynamic range on single exposures . for photometry of non - stellar objects , an extended - dynamic - range , calibrated composite image is produced for each field by 1 ) boxaveraging all exposures to a @xmath14.5 arcsec pixel size ( to minimize the effects of possible densitometry misalignments of up to @xmath11.12 arcsec ) , and 2 ) selecting and using pixels from the individual boxaveraged frames with exposure values which maximize signal - to - noise ratios . this composite image is subsequently used for all extended source photometry . figure [ mosaic ] shows that fuv emission from the smc originates mostly in hot stellar populations which , while not restricted to clusters , are significantly clumped . no diffuse fuv emission is readily apparent ( but see section 5 ) . the brightest features in the uit images are the clusters ngc 346 and ngc 330 , with additional fuv concentrations approximately centered in fields 1 and 2 . the bright fuv concentrations , spaced along the bar centered at @xmath10.5kpc intervals , have similar clustering and distribution properties to those evident in wide - field fuv images of the lmc ( @xcite ) . comparison of large - scale h@xmath2 and fuv morphologies ( figure [ hapic ] a ) and b ) ) ; @xcite ) can reveal spatial and temporal sequences of star formation over timescales of @xmath120 myr ( @xcite ) . ionized hydrogen emission marks the locations of the earliest - type stars , while detectible fuv flux is emitted by stars as late as a0 . therefore , h@xmath2 is strong for clusters up to a few million years of age , while detectible fuv emission continues for about 20 myr . figures [ hapic]a ) and b ) show that fuv concentrations cover a range of ages , including cases with coincident bright young h@xmath2 sources at the ne and sw ends of the bar , as well as evolved regions without h@xmath2 . uit field 2 , near the bar s center , provides a intriguing instance of what appears to be sequential star formation . a ring of fuv - bright stars with a radius of about 45 arcmin ( @xmath17090 pc at a distance of 60 kpc ; @xcite ) dominates the field . figures [ hapic ] a ) and b ) show that the stellar ring is bounded by , and is apparently interacting with , ionized hydrogen in a concentric , larger ring ( radius 78 arcmin or @xmath1120140pc ) . supernova remnant 0050 - 728 ( @xcite ) , outlined by the circle in figure [ hapic]b , lies at the northern edge of the shell . hi observations ( @xcite , @xcite ) show loops comparable in size to the h@xmath2 ring near these locations at a range of velocities between about 90 and 130 km s@xmath8 ; the ellipse in figure [ hapic]b shows the location of the ring seen at @xmath1110 km s@xmath8 . fabry - perot h@xmath2 measurements ( @xcite ) show linewidths of up to several tens of km s@xmath8 , reinforcing evidence for ties between the distribution of hot stars and the gas dynamics . in a simple model , the hollow - shell morphology of the hii has been driven by winds from now - defunct hot stars and supernova remnants , which have also propagated star formation from the center outward , as evidenced by relatively gas - free stars in the center of the hii ring . uit - based detailed analysis of sequential star formation in this region will be the subject of future work ( @xcite ) . for example , comparing the distribution of uv - bright stars with that of h@xmath2 emission - line stars and small nebulae as catalogued by ( @xcite ) will reveal any systematic stellar age gradients in the 0 - 30 myr range . luminosity functions for the four fields , normalized to the number of stars in each for 14.0 @xmath9 m(162)@xmath9 14.5 , are shown in figure 3 . from turnovers of the luminosity functions we estimate the limiting fuv magnitudes of the fields to be 14.5 , 14.5 , 14.5 , and 15.0 ( @xmath100.5 ) for fields 1 - 4 respectively , consistent with relative exposure times . luminosity functions for fields 1 and 2 show features that may be confirmed from other data . the peak for 11.0 @xmath9 m(162)@xmath9 11.5 in field 1 s luminosity function ( solid line ) is clearly displayed in the field s color - magnitude diagram as a supergiant `` plume '' of stars leaving the main sequence ( @xcite ) . field 2 ( dashed line ) shows a significant overabundance of stars with 10.0 @xmath9m(162)@xmath9 11.0 ( more than 3@xmath11 overabundant compared to each of the other three fields individually ) which may be related to sequential star formation in that field . the high - luminosity ends of these m(162 ) distributions evidently show structure which reflects recent star formation . as shown in previous uit results ( _ e.g. _ @xcite ) fuv sources are generally well - correlated with h@xmath2 emission . we have compared fuv point - source locations with those of the 711 h@xmath2 emission - line stars and small nebulae within the uit fields of view ( @xcite ) , finding 520 matches at a matching radius of 7.2 arcsec ( approximately twice the estimated 1-@xmath11 the uit astrometric uncertainty for these images ) . as described above , a detailed comparison of the spatial distribution of fuv sources with and without associated h@xmath2 emission may provide insights into possible sequential star formation in the smc . we have also correlated our stellar photometry tables with the ground - based stellar photometry compiled by azzopardi & vigneau ( 1982 ; hereafter av ) by positionally matching sources . the precision to which stellar positions are quoted in av ( @xmath10.1 arcmin ) does not permit unambiguous identification of stars fainter than , typically , m(162)@xmath115 because several such stars typically occur in a part of a uit image of that size . for av supergiants , this m(162 ) limit corresponds to a spectral type of @xmath1f0 ; therefore , we do not identify or list m(162 ) values for av stars of spectral type f0 or later . ( where available we have used the spectral types listed in av from @xcite and other sources which employ medium resolution slit spectra as the basis for spectral classification . ) under this spectral type restriction we find 191 sources matched of the 195 within the uit fields of view . results of the photometry of these objects are presented in table [ avtab ] . figure 4 is a color - magnitude diagram of the 191 identified stars common to our data and av , with spectral classes differentiated by different symbols . no reddening corrections have been applied to observed data . smc - composition stellar models are produced by using model atmospheres of @xcite with log(z / z@xmath12 ) = @xmath30.5 ( @xmath1solar/3.2 ) , on stellar evolution models of @xcite with z = 0.004 ( @xmath1solar/5 . ) . these grid points are selected as models with composition parameters nearest to the smc s ( @xmath1solar/4 ; @xcite ) . filled circles outline a 1 myr isochrone , with masses marked ; the tracks to the upper right show the subsequent evolutionary paths of 10 , 15 , 20 , and 40 m@xmath12 stars . model stars are adjusted to distance modulus 18.9 and foreground galactic extinction e(b@xmath3v)=0.02 magnitudes ( @xcite ) . the reddening vector shown , using the smc extinction curve of @xcite and e(b@xmath3v)=0.1 , is a typical large value for the smc ( @xcite ) . the large fuv / v leverage of extinction in the smc ( a@xmath13/e(b - v)=15.75 ) is reflected in the nearly - horizontal orientation of the reddening vector . the stars catalogued by av , the visually brightest non - cluster stars in the smc , are known to be supergiants of a range of spectral types . the evolutionary tracks in figure 4 clearly demonstrate that these stars predominantly have masses between 10 and 20 m@xmath12 , and therefore , ages between 10 and 30 myr . the upper age limit is set by the faintness limit of the av data ; however , the lower limit corresponds to a real absence of supergiant stars brighter than about v=11 ( m@xmath130m@xmath12 . ) while the effects of some reddening ( seen as excesses of ( m(162)@xmath3v ) color of up to 1.5 ) are apparent , most color variation seen in this figure is due to the intrinsic color variations of the stars themselves , as seen from the general segregation of spectral types by color . figure 5 illustrates the strong variation in ( fuv - v ) color with spectral type . the data points are the observed colors , as in figure 4 , plotted against spectral type from av . the solid line is a spline fit to the observed ( m(162)@xmath3v ) colors of galactic supergiants from the atlas of @xcite . the atlas colors have been reddened by the amount of the smc s galactic foreground reddening . the sensitivity of the ( m(162)@xmath3v ) color to spectral type is evident , as are the significant reddenings for some stars , caused by the large fuv leverage of the smc extinction law . the atlas colors are seen to form a blue limit for the observed colors , as expected for unreddened stars . while a true calibration of color _ versus _ spectral type for smc stars is not available , galactic values are probably appropriate for blue stars : ( m(162)@xmath3v ) colors of galactic- and smc - composition model atmospheres differ by less than 0.10 for models that have ( m(162)@xmath3v)@xmath141.0 . however , at galactic ( m(162)@xmath3v)=1.5 , smc models are approximately 0.5 mag bluer , with larger differences for redder stars . therefore , much of the apparent discrepancy between the blue limit of the observed colors and the atlas colors is probably due to the fact that the locus of colors from an `` smc atlas '' would be significantly bluer for stars later than about a0 , than the galactic atlas colors shown . figure 5 provides a straightforward method of computing the reddenings along the lines of sight to the stars in av . we do so by comparing the observed colors to atlas colors for the spectral type , corrected for galactic foreground reddening , and converting the difference to e(b@xmath3v ) using the extinction law of @xcite . extinctions computed in this way are presented in table [ avtab ] for all identified av stars . stars later in spectral type than b8 , which have uncertain intrinsic colors , are noted . figure [ redmap ] shows the resulting e(b@xmath3v ) values plotted at the locations of the stars on the smc , against the h@xmath2 image shown in figure [ hapic]a . all values of e(b@xmath3v ) @xmath150.15 may be seen from figure [ hapic]b to be at least along the line of sight to h@xmath2 emission . the correlation of large reddening values with location implies that these stars are probably associated with the nebulosity . in the future uit photometry of the smc will be supplemented by additional ground - based observations currently underway by the uit team , as well as by the comprehensive photometric surveys of the magellanic clouds ( @xcite ) . h@xmath2 emission from hii regions indirectly traces the ob star content of the regions via the process of nebular ionization by the hot stars and subsequent recombination of the hydrogen gas . the fuv continuum emission ( _ e.g. _ at 1620 ) from hii regions arises from ob stars as well as from the more numerous non - ionizing stars ( mid - b through early a types . ) therefore , the comparison of h@xmath2 and fuv fluxes from hii regions provides a sensitive measure of each hii region s evolutionary state . evolutionary models of single - burst ionizing clusters confirm that the ratio of lyman continuum to fuv flux rapidly decreases with age , with the fuv emission remaining significant for @xmath120 myr ( @xcite ) . we have measured fuv fluxes for 42 h@xmath2-bright hii regions from the catalog of davies , elliott , and meaburn 1976 ( hereafter @xcite ) that were measured at h@xmath2 by kennicutt & hodge 1986 ( hereafter @xcite ) , and have compared the observed flux ratios with predictions from cluster models . we compute fuv fluxes in two ways : by summing the flux from stars in the apertures defined by kh ; and by summing all flux in pixels in the apertures the latter method including diffuse fuv emission . ( our use of the kh aperture definitions causes significant aperture overlaps in a few cases . ) table [ hiitab ] presents fluxes and errors : h@xmath2 from kh ; fuv from stars ( from psf fits to stars in the aperture ) ; and fuv total fluxes from the apertures themselves . for the fuv aperture measurements the `` sky '' flux has been set to zero , and the error term stated does not include a sky contribution . we choose this approximation because the actual sky contribution to any fuv aperture measurement or its uncertainty is very small , and any aperture inside the uit image area will measure a high background due to smc material . we set limits on true sky flux by measuring the average background in astro-2 images made in the uit b5 filter of a nearby field containing part of the globular cluster 47 tuc . these images , long exposures ( 1030 and 1800 sec ) made during orbital day , have average earth limb elevation angles and other observational conditions which are typical of those during uit smc observations , and therefore give representative background measurements ( @xcite ) . as measured in this way the sky contribution to the fuv aperture fluxes is 3.3@xmath16erg ( @xmath7 s arcsec @xmath17)@xmath8 , with an average contribution of 0.75% and a maximum of 1.60% to the individual aperture fluxes . these estimates of `` sky '' contributions are also listed in table [ hiitab ] . figure 7 shows the data from table [ hiitab ] . open diamonds are aperture fluxes , and crosses are stellar photometry . no corrections for galactic foreground reddening have been made to the data . the fuv flux is generally well correlated with its h@xmath2 counterpart ; furthermore , the ratio of aperture to stellar fuv flux is relatively uniform , especially for brighter hii regions . this ratio measures the relative amounts of bright - star and `` diffuse '' light , where `` diffuse '' light here includes contributions from faint stars as well as from dust scattering . the mean value of the ratio of aperture flux to stellar flux , averaged in the log ratio for all hii regions , is 1.79 ; weighted by flux ( the total aperture flux for all hii regions divided by the total stellar flux for all hiiregions ) is 2.15 . two lines of argument point to a dust - scattering origin for most of the additional fuv flux , however . first , @xcite have determined the total - to - stellar flux ratio for the orion nebula , for which the number of undetected stars is small , to be 2.5 for 200 - 300 bandpasses near 1400 , 1823 , 2242 , and 2622 . second , extrapolating our measured luminosity function for the field 4 , which has the deepest exposure , to m(162)=18.0 ( corresponding to an a0 main sequence star , at which spectral type the fuv flux drops rapidly ) predicts an additional stellar flux contribution of 22% and a resulting flux - weighted total - to - stellar ratio of 1.76 significantly smaller than the observed value . scattered - light fractions near the observed smc value are found in the giant hii regions of m33 ( @xcite ) . although the small smc dust abundance ( @xcite ) undoubtedly causes the scattering to be significantly less than for orion , it is nonetheless clear that dust - scattered radiation is a major contributor to the fuv aperture emission from hii regions . figure 7 also displays the fuv / h@xmath2 ratio , known to be a good diagnostic of hii region evolution ( @xcite ) . we have computed f(162)/h@xmath2 ratio values for clusters with a single burst of star formation of smc composition using the idl procedure clustflux ( @xcite ) . the ratio rises essentially monotonically from 0.1 at 1 myr to 4.3 at 10 myr ; ratios for cluster ages of 1myr , 3myr , 6.5 myr , and 8.5 myr , corrected for foreground galactic reddening ( which increases all f(162)/h@xmath2 values by 10% ) are plotted as solid lines in figure 7 . for reference , the arrow in figure 7 shows a typical internal smc extinction of e(b@xmath3v)=0.1 ; @xcite find reddenings local to smc bar hii regions in the range 0.08@xmath9e(b@xmath3v)@xmath90.27 by comparing 6-cm radio , h@xmath2 , h@xmath18 , and nearby stellar measurements . therefore , model and observed ratios agree well for hii - region ages of a few myr and reasonable internal smc extinction ; the observed hii regions are evidently no older than @xmath15 myr for any internal smc reddening , and , for typical assumed smc reddening , are between 1 myr and 3 myr in age . this result is consistent with observations of hii regions in galaxies as disparate as ngc 4449 ( @xcite ) and m81 ( @xcite ) . there is also a clear tendency for h@xmath2-bright hii regions to be either relatively young which is not surprising or unreddened . the absence of significantly older hii regions is most likely a selection effect originating in kh s bias for h@xmath2-bright hiiregions . uit fuv images and derived stellar photometry for most of the bar of the smc are presented . the uv morphology of the smc s bar shows four concentrations of uv - bright stars spread from northeast to southwest at nearly equal ( @xmath130 arcmin ) spacings . one of the concentrations , near dem 55 , appears as a well - defined 8-arcmin diameter ring surrounded by a larger h@xmath2 ring and strongly suggests sequential star formation . fuv psf photometry , resulting in m(162 ) magnitudes , is obtained for 11,306 stars . we present a fuv luminosity function for the smc bar , complete to m(162 ) @xmath114.5 , and compare our photometry with the compiled ground - based data of av . detected objects are well correlated with other smc population i material ; 520 of 711 h@xmath2 emission - line stars and small nebulae within the uit fields of view are identified with fuv sources . for early type stars , the bluest observed ( m(162)@xmath3v ) colors for each spectral type agree well with values computed from unreddened galactic spectral atlas stars for types earlier than about a0 ; for later types , observed smc stars range significantly bluer , as predicted by low - metallicity models . we attribute redder colors for some stars of all spectral types to strong fuv extinction due to even small amounts of smc dust . internal smc reddenings are determined for all catalog stars . all stars with e(b@xmath3v)@xmath40.15 are within regions of visible h@xmath2 emission . fuv photometry is obtained for the resolved stars and for the total emission from dem hii regions in the smc bar . the flux - weighted ratio of total to stellar flux for dem hiiregions using the apertures of kh is 2.15 ; since only 22% more flux is contributed by stars fainter than m(162)=14.5 , most of the excess total flux is due to scattered fuv from dust . stellar and total emission from dem hii regions are well correlated with h@xmath2 fluxes measured by kh . we compute ratios of fuv to h@xmath2 flux for 42 smc hii regions and compare them with model results , finding that the observed ratios for all h@xmath2-selected hii regions are consistent with models of smc metallicity , ages from 1 - 5 myr , and moderate ( e(b - v)=0.0 - 0.1 mag ) extinction . we gratefully acknowledge the contributions made by the many people involved in the _ astro-2 _ mission . we thank robert s. hill , wayne landsman , and michael fanelli for useful discussions , greg bothun for the use of his h@xmath2 image of the smc , joel offenberg for help in preparing publication - quality images , and joan hollis and emily zamkoff for data entry and programming assistance . we also thank lister staveley - smith for kindly supplying a digital version of a part of his hi maps of the complete smc . funding for the uit project has been through the spacelab office at nasa under project number 440 - 51 . rwo gratefully acknowledges nasa support of portions of this research through grants nag5 - 700 and nagw-2596 to the university of virginia . observational data for stars from azzopardi & vigneau 1982 ( av ) which were observed by uit . av stars which are within uit fields but later in spectral type than a9 are omitted , since they are too faint in the fuv for detection by uit . m(162)values are derived from psf photometry as described in the text . e(b@xmath3v ) values are derived by comparing the unreddened ( m(162)@xmath3v ) color for the stellar spectral type with the observed color , using the reddening law of @xcite . fluxes for hii regions from the list of davies , elliott , & meaburn 1976 ( dem ) as measured by kennicutt & hodge 1986 ( kh ) which are completely contained within the uit field of view . `` fuv stellar '' flux is the total flux of stars within the aperture defined by kh ; `` fuv aperture '' flux is the total pixel flux contained within the aperture defined by kh ; and `` sky flux '' is an estimate for the non - smc sky contribution within the aperture , as described in the text . 1 & 00 47 33.7 -73 06 26.8 & 23.9,117.0 & 12/05/90 & 06:16 & 1713 & astro-1 2 & 00 50 58.3 -72 44 56.5 & 25.9,247.0 & 03/16/95 & 23:50 & 2110 & astro-2 3 & 00 56 41.2 -72 28 28.8 & 18.7,459.0 & 03/06/95 & 06:14 & 2517 & astro-2 4 & 00 59 24.2 -72 11 08.4 & 36.2,179.0,898 . & 03/09/95 & 04:04 & 4966 & astro-2 1 & 2b & 1686 & 00 43 51.2 -73 08 54 & b0 & 13.49 & 10.51 & 0.10 & 0.05 & 2 & 3 & 70 & 00 44 29.6 -72 58 00 & b0 & 14.12 & 12.65 & 0.07 & 0.17 & 3 & 6 & 1699 & 00 45 19.8 -73 15 18 & b0 & 13.46 & 10.25 & 0.08 & 0.03 & 4 & 7 & 1417 & 00 45 33.7 -73 04 31 & b0 & 14.57 & 10.85 & 0.06 & -0.01 & 5 & 8 & 1505 & 00 45 36.2 -72 59 12 & b8 & 13.53 & 12.07 & 0.05 & -0.02 & 6 & 9 & 1694 & 00 45 37.1 -73 14 07 & b2 & 13.05 & 11.25 & 0.07 & 0.08 & 7 & 11 & 1622 & 00 45 54.3 -73 16 13 & b2 & 13.56 & 11.48 & 0.05 & 0.05 & 8 & 12 & 1429 & 00 46 2.8 -73 06 19 & o9 & 13.21 & 10.11 & 0.07 & 0.06 & 9 & 14 & 321 & 00 46 33.2 -73 05 55 & o5 & 13.77 & 12.73 & 0.07 & 0.28 & 10 & 16 & 1558 & 00 46 55.8 -73 08 25 & b1 & 13.19 & 12.03 & 0.10 & 0.16 & 11 & 17 & 1522 & 00 47 3.6 -73 05 55 & b1 & 13.76 & 11.49 & 0.12 & 0.08 & 12 & 17a & 1679 & 00 47 4.8 -73 06 19 & b2 & 13.37 & 11.42 & 0.10 & 0.07 & 13 & 18 & 1433 & 00 47 13.1 -73 06 25 & b1 & 12.48 & 10.98 & 0.04 & 0.14 & 14 & 19 & 2766 & 00 47 15.8 -72 49 55 & b6 & 13.00 & 13.38 & 0.08 & 0.16 & 15 & 20 & 158 & 00 47 29.9 -73 01 25 & b8 : & 12.14 & 14.76 & 0.08 & 0.31 & 16 & 21 & 1690 & 00 47 32.6 -73 10 56 & b1 & 14.14 & 11.05 & 0.05 & 0.01 & 17 & 22 & 1444 & 00 47 39.8 -73 07 38 & b2 & 12.25 & 10.98 & 0.06 & 0.12 & 18 & 23 & 1670 & 00 47 40.1 -73 22 43 & b3 & 12.25 & 12.03 & 0.08 & 0.17 & 19 & 24 & 1675 & 00 47 42.9 -73 02 25 & o9 & 13.78 & 10.96 & 0.07 & 0.08 & 20 & 25 & 1597 & 00 47 47.5 -73 12 19 & b5 & 13.19 & 11.35 & 0.10 & -0.01 & 21 & 26 & 514 & 00 47 50.5 -73 08 14 & b0 & 12.55 & 12.56 & 0.07 & 0.28 & 22 & 27 & 947 & 00 47 51.6 -73 14 14 & a2 & 12.17 & 14.12 & 0.05 & 0.10 & 23 & 28 & 1662 & 00 47 55.0 -73 21 14 & b0 & 13.42 & 11.34 & 0.08 & 0.12 & 24 & 31 & 361 & 00 48 7.2 -73 06 32 & b8 : & 12.52 & 13.20 & 0.06 & 0.15 & 25 & 32 & 1939 & 00 48 10.2 -72 43 44 & b1 & 14.20 & 10.95 & 0.03 & 0.00 & 26 & 34 & -1 & 00 48 18.7 -73 23 38 & b2 : & 13.81 & 0.00 & 0.00 & -9.99 & 27 & 35 & -1 & 00 48 21.3 -72 51 02 & b0 : & 14.13 & 0.00 & 0.00 & -9.99 & 28 & 36 & 1934 & 00 48 23.9 -72 43 38 & b1 & 14.02 & 10.82 & 0.04 & 0.00 & 29 & 37 & 1676 & 00 48 27.0 -73 03 20 & b5 & 12.88 & 11.43 & 0.07 & 0.03 & 30 & 38 & 142 & 00 48 28.2 -73 00 32 & a0 & 12.85 & 13.89 & 0.05 & 0.11 & 31 & 39a & 1476 & 00 48 31.4 -73 15 32 & w6 & 99.99 & 10.94 & 0.06 & -9.99 & 32 & 41 & 1767 & 00 48 36.1 -72 52 50 & b2 & 14.56 & 11.05 & 0.06 & -0.06 & 33 & 42 & 2265 & 00 48 40.2 -72 58 14 & b5 & 13.49 & 12.16 & 0.09 & 0.03 & 34 & 43 & 1875 & 00 48 48.8 -72 46 14 & b1 & 14.08 & 10.34 & 0.04 & -0.04 & 35 & 45 & -1 & 00 48 50.3 -73 22 02 & b8 & 14.15 & 0.00 & 0.00 & -9.99 & 36 & 48 & 1709 & 00 49 3.4 -73 21 32 & b3 & 11.03 & 9.45 & 0.12 & 0.06 & 37 & 50 & 1765 & 00 49 16.7 -72 52 32 & b1 & 13.11 & 10.15 & 0.05 & 0.02 & 38 & 51 & 1777 & 00 49 32.8 -72 51 08 & b0 & 14.05 & 10.81 & 0.03 & 0.03 & 39 & 53 & 2495 & 00 49 45.8 -72 52 38 & a0 & 12.96 & 14.41 & 0.05 & 0.14 & 40 & 55 & 1197 & 00 49 48.2 -73 17 45 & b5 & 13.40 & 13.66 & 0.06 & 0.16 & 41 & 56 & 1751 & 00 49 51.7 -72 55 39 & b5 & 11.17 & 9.14 & 0.04 & -0.02 & 42 & 58 & 1769 & 00 49 58.3 -72 51 45 & b2 & 14.38 & 10.83 & 0.04 & -0.06 & 43 & 59 & 788 & 00 49 58.4 -73 11 33 & a0 & 13.40 & 13.54 & 0.06 & 0.04 & 44 & 62 & 2361 & 00 50 1.2 -72 55 03 & b3 & 14.34 & 11.64 & 0.03 & -0.03 & 45 & 63 & 672 & 00 50 0.9 -73 10 08 & a0 & 13.48 & 13.12 & 0.06 & 0.00 & 46 & 65 & 1448 & 00 50 6.9 -73 07 39 & b6 & 11.13 & 11.18 & 0.08 & 0.14 & 47 & 66 & 1642 & 00 50 6.7 -73 16 21 & b0 & 13.48 & 11.27 & 0.09 & 0.11 & 48 & 67 & 2039 & 00 50 11.9 -72 32 27 & b0 & 13.66 & 10.22 & 0.07 & 0.01 & 49 & 69 & 2442 & 00 50 18.0 -72 53 21 & b0 & 13.35 & 10.01 & 0.21 & 0.02 & 50 & 70 & 2007 & 00 50 18.9 -72 38 03 & b0 & 12.38 & 8.86 & 0.04 & 0.00 & 51 & 73 & 1677 & 00 50 28.4 -73 03 09 & b0 & 14.08 & 10.95 & 0.07 & 0.04 & 52 & 75 & 1757 & 00 50 32.9 -72 52 27 & o9 & 12.79 & 9.30 & 0.04 & 0.03 & 53 & 77 & 1827 & 00 50 34.2 -72 47 39 & b0 & 13.91 & 10.41 & 0.03 & 0.01 & 54 & 80 & 1824 & 00 50 44.3 -72 47 33 & b1 & 13.33 & 9.83 & 0.04 & -0.02 & 55 & 82 & 1925 & 00 50 49.8 -72 42 33 & b2 & 14.13 & 10.09 & 0.06 & -0.10 & 56 & 83 & 1930 & 00 50 52.8 -72 42 03 & b1 & 13.38 & 9.80 & 0.04 & -0.03 & 57 & 85 & 2440 & 00 51 0.8 -72 52 57 & b0 & 13.75 & 11.33 & 0.05 & 0.09 & 58 & 87 & 3400 & 00 51 8.3 -72 40 03 & b0 & 13.90 & 12.76 & 0.11 & 0.19 & 59 & 89 & 1121 & 00 51 10.4 -73 15 45 & b0 & 14.47 & 14.03 & 0.09 & 0.25 & 60 & 90 & -1 & 00 51 12.1 -72 28 09 & a5 & 12.58 & 0.00 & 0.00 & -9.99 & 61 & 91 & 2325 & 00 51 11.5 -72 55 09 & b8 & 12.61 & 12.36 & 0.04 & 0.08 & 62 & 93 & 1940 & 00 51 20.8 -72 41 21 & b0 : & 14.13 & 11.07 & 0.06 & 0.04 & 63 & 94 & 1783 & 00 51 20.6 -72 49 33 & b1 & 13.99 & 10.15 & 0.04 & -0.05 & 64 & 95 & 1885 & 00 51 21.9 -72 44 03 & o9 & 13.91 & 9.90 & 0.04 & -0.01 & 65 & 99 & 1809 & 00 51 25.4 -72 47 45 & b2 & 13.01 & 10.59 & 0.04 & 0.03 & 66 & 100 & 1743 & 00 51 26.9 -72 57 45 & b3 & 14.29 & 11.44 & 0.05 & -0.04 & 67 & 102 & 2036 & 00 51 36.5 -72 32 09 & b2 & 14.29 & 11.11 & 0.07 & -0.03 & 68 & 103 & 1847 & 00 51 36.7 -72 45 57 & b1 & 13.36 & 10.27 & 0.03 & 0.01 & 69 & 104 & 1800 & 00 51 39.0 -72 47 57 & b1 & 13.13 & 9.93 & 0.04 & 0.00 & 70 & 105 & 3885 & 00 51 41.9 -72 27 57 & a0 & 12.24 & 12.61 & 0.07 & 0.05 & 71 & 106 & 2005 & 00 51 44.1 -72 37 15 & b1 & 14.32 & 11.28 & 0.05 & 0.01 & 72 & 109 & 1976 & 00 51 50.6 -72 39 15 & b1 & 13.73 & 10.43 & 0.03 & -0.01 & 73 & 110 & 1877 & 00 51 52.8 -72 44 03 & a0 & 12.17 & 12.06 & 0.06 & 0.02 & 74 & 111 & 2035 & 00 51 56.7 -72 31 57 & b1 & 13.84 & 10.52 & 0.06 & -0.01 & 75 & 112 & 2031 & 00 51 58.5 -72 33 15 & b0 & 14.15 & 10.27 & 0.05 & -0.02 & 76 & 114 & 1971 & 00 52 3.1 -72 39 22 & b0 & 14.93 & 11.10 & 0.05 & -0.02 & 77 & 122 & 1932 & 00 52 25.7 -72 40 52 & b6 & 12.79 & 11.44 & 0.05 & 0.02 & 78 & 123 & 1758 & 00 52 27.8 -72 50 58 & b8 & 13.22 & 11.59 & 0.06 & -0.03 & 79 & 126 & 1961 & 00 52 31.7 -72 39 22 & b0 & 13.47 & 10.26 & 0.04 & 0.03 & 80 & 131 & 1841 & 00 52 41.0 -72 45 10 & b8 & 12.61 & 11.94 & 0.07 & 0.05 & 81 & 133 & 2003 & 00 52 44.2 -72 36 46 & b0 & 13.91 & 9.87 & 0.04 & -0.04 & 82 & 137 & 1865 & 00 52 53.0 -72 44 04 & b4 & 12.33 & 10.46 & 0.04 & 0.01 & 83 & 138 & 1784 & 00 52 52.9 -72 48 22 & b0 & 14.28 & 10.94 & 0.05 & 0.02 & 84 & 141 & 3723 & 00 53 10.5 -72 33 40 & b3 : & 14.50 & 12.79 & 0.04 & 0.05 & 85 & 143 & 1970 & 00 53 27.1 -72 38 22 & b0 & 14.12 & 10.32 & 0.04 & -0.02 & 86 & 144 & 1742 & 00 53 33.7 -72 56 16 & b0 & 14.06 & 10.13 & 0.08 & -0.03 & 87 & 145 & 1983 & 00 53 36.0 -72 37 40 & b1 & 13.35 & 10.10 & 0.05 & 0.00 & 88 & 148 & 1882 & 00 53 42.5 -72 42 28 & b0 & 14.28 & 10.40 & 0.04 & -0.02 & 89 & 149 & 1774 & 00 53 53.6 -72 48 22 & b2 & 13.96 & 10.79 & 0.07 & -0.03 & 90 & 150 & 4034 & 00 53 58.3 -72 28 35 & b6 & 12.72 & 11.44 & 0.05 & 0.03 & 91 & 151 & 1807 & 00 53 59.6 -72 45 53 & b5 & 12.30 & 10.27 & 0.06 & -0.02 & 92 & 152 & 5913 & 00 54 4.2 -72 31 40 & a3 & 11.87 & 12.23 & 0.03 & -0.06 & 93 & 153 & 3451 & 00 54 4.0 -72 37 17 & b8 & 13.58 & 12.51 & 0.06 & 0.01 & 94 & 154 & 3095 & 00 54 9.9 -72 41 34 & b0 : & 13.55 & 13.30 & 0.05 & 0.26 & 95 & 155 & 2975 & 00 54 15.2 -72 42 59 & b0 : & 14.34 & 14.54 & 0.10 & 0.30 & 96 & 156 & 4384 & 00 54 19.4 -72 17 47 & a0 & 14.17 & 13.99 & 0.06 & 0.01 & 97 & 157 & 3938 & 00 54 23.0 -72 17 05 & b0 & 14.33 & 11.07 & 0.05 & 0.03 & 98 & 158 & 3946 & 00 54 23.5 -72 18 59 & b1 & 14.06 & 10.22 & 0.04 & -0.05 & 99 & 160 & 2340 & 00 54 43.5 -72 51 53 & b0 & 13.99 & 11.34 & 0.07 & 0.07 & 100 & 164 & 4070 & 00 55 16.0 -72 41 17 & b0 & 14.16 & 10.71 & 0.04 & 0.01 & 101 & 165 & 3994 & 00 55 19.9 -72 25 59 & b7 & 12.79 & 11.65 & 0.06 & 0.03 & 102 & 169 & 4077 & 00 55 40.3 -72 44 54 & b1 & 13.90 & 11.07 & 0.05 & 0.03 & 103 & 171 & 6141 & 00 55 48.2 -72 39 23 & b6 & 13.34 & 12.77 & 0.05 & 0.09 & 104 & 172 & 10309 & 00 55 54.9 -72 08 54 & b8 & 13.37 & 13.37 & 0.06 & 0.10 & 105 & 174 & 8802 & 00 56 38.6 -72 01 36 & a7 & 12.43 & 14.52 & 0.07 & -0.06 & 106 & 175 & 4036 & 00 56 38.9 -72 36 30 & b1 & 13.65 & 11.12 & 0.05 & 0.05 & 107 & 176 & 3947 & 00 56 39.7 -72 25 06 & b1 & 13.93 & 11.10 & 0.04 & 0.03 & 108 & 177 & 6454 & 00 56 45.1 -72 03 30 & o5 & 14.60 & 10.50 & 0.03 & 0.03 & 109 & 178 & 3970 & 00 56 48.6 -72 28 42 & b2 & 14.38 & 11.29 & 0.06 & -0.03 & 110 & 180 & 7161 & 00 56 55.8 -72 24 12 & b5 & 13.19 & 12.27 & 0.09 & 0.07 & 111 & 181 & 11001 & 00 56 57.8 -72 17 36 & a0 & 13.83 & 13.15 & 0.04 & -0.03 & 112 & 182 & 6480 & 00 57 1.6 -72 08 06 & b0 & 14.33 & 10.81 & 0.03 & 0.00 & 113 & 184 & 7143 & 00 57 14.4 -72 22 24 & b8 & 14.18 & 12.57 & 0.07 & -0.03 & 114 & 185 & 6777 & 00 57 24.4 -72 01 30 & b8 : & 13.28 & 11.97 & 0.06 & 0.00 & 115 & 186 & 4000 & 00 57 27.8 -72 33 06 & o9 & 14.10 & 10.63 & 0.05 & 0.03 & 116 & 189 & 3958 & 00 57 33.3 -72 28 48 & b1 & 14.51 & 11.21 & 0.05 & -0.01 & 117 & 190 & 11396 & 00 57 35.1 -72 26 48 & b6 & 13.52 & 13.19 & 0.08 & 0.11 & 118 & 191 & 6482 & 00 57 37.8 -72 13 00 & b1 & 13.63 & 10.42 & 0.06 & 0.00 & 119 & 192 & 6490 & 00 57 38.2 -72 21 49 & o9 & 14.58 & 11.06 & 0.05 & 0.03 & 120 & 193 & 7152 & 00 57 38.1 -72 24 31 & b1 & 15.40 & 12.54 & 0.09 & 0.03 & 121 & 194 & 4008 & 00 57 45.6 -72 35 30 & b0 & 13.95 & 10.73 & 0.04 & 0.03 & 122 & 195 & 5818 & 00 57 45.5 -72 40 12 & b5 & 13.53 & 14.47 & 0.04 & 0.21 & 123 & 196 & 7163 & 00 57 54.7 -72 27 37 & b1 & 13.93 & 11.57 & 0.09 & 0.07 & 124 & 199 & 5274 & 00 58 2.3 -72 35 37 & a3 & 13.41 & 13.45 & 0.03 & -0.09 & 125 & 200 & 4025 & 00 58 8.1 -72 38 25 & a0 & 12.10 & 11.37 & 0.05 & -0.03 & 126 & 201 & 6479 & 00 58 10.7 -72 10 55 & b0 & 14.04 & 10.76 & 0.06 & 0.02 & 127 & 202 & 6855 & 00 58 15.5 -72 07 25 & b0 & 14.33 & 11.20 & 0.05 & 0.04 & 128 & 204 & 3998 & 00 58 22.5 -72 35 13 & b1 : & 14.38 & 11.65 & 0.05 & 0.04 & 129 & 205 & 11006 & 00 58 23.5 -72 21 31 & a2 & 12.30 & 13.41 & 0.04 & 0.03 & 130 & 207 & 6429 & 00 58 33.7 -71 55 43 & o7 & 14.37 & 10.70 & 0.06 & 0.05 & 131 & 208 & 5549 & 00 58 33.7 -72 39 25 & o9 & 14.10 & 11.49 & 0.10 & 0.10 & 132 & 209 & 7135 & 00 58 35.9 -72 24 55 & b0 & 14.46 & 11.45 & 0.04 & 0.05 & 133 & 210 & 6484 & 00 58 36.7 -72 16 19 & b3 & 12.67 & 10.73 & 0.05 & 0.03 & 134 & 211 & 7145 & 00 58 41.8 -72 26 13 & b8 & 11.50 & 11.67 & 0.07 & 0.11 & 135 & 213 & 6607 & 00 58 55.1 -71 56 31 & a2 & 12.11 & 12.14 & 0.08 & -0.05 & 136 & 214 & 6964 & 00 58 55.2 -72 13 13 & b3 & 13.34 & 12.15 & 0.03 & 0.09 & 137 & 215 & 3955 & 00 58 55.9 -72 32 01 & b1 & 12.76 & 10.15 & 0.06 & 0.05 & 138 & 217 & 7050 & 00 59 2.3 -72 18 49 & b2 : & 14.59 & 11.48 & 0.03 & -0.03 & 139 & 218 & 7064 & 00 59 5.2 -72 19 37 & b1 & 13.80 & 11.29 & 0.05 & 0.06 & 140 & 219 & 7010 & 00 59 7.7 -72 16 31 & b2 : & 14.50 & 11.23 & 0.06 & -0.04 & 141 & 220 & 6446 & 00 59 10.9 -72 05 43 & o9 & 14.50 & 10.52 & 0.06 & -0.01 & 142 & 221 & 3945 & 00 59 10.2 -72 31 31 & b0 & 13.46 & 11.34 & 0.05 & 0.12 & 143 & 222 & 7080 & 00 59 13.5 -72 20 55 & b3 & 13.20 & 11.42 & 0.06 & 0.04 & 144 & 223 & 4007 & 00 59 13.6 -72 38 55 & b0 & 13.67 & 10.64 & 0.02 & 0.04 & 145 & 224 & 6732 & 00 59 16.9 -72 04 37 & b0 & 14.22 & 11.23 & 0.05 & 0.05 & 146 & 225 & 4313 & 00 59 16.8 -72 29 43 & b8 & 14.05 & 13.07 & 0.03 & 0.02 & 147 & 226 & 6481 & 00 59 21.4 -72 17 07 & b0 & 14.42 & 10.72 & 0.05 & -0.01 & 148 & 229 & 8932 & 00 59 27.5 -72 09 49 & o9 & 11.86 & 10.98 & 0.16 & 0.24 & 149 & 230 & 6435 & 00 59 30.1 -72 01 02 & b2 & 12.75 & 10.64 & 0.05 & 0.05 & 150 & 231 & 10562 & 00 59 29.6 -72 20 14 & a0 & 14.20 & 13.54 & 0.04 & -0.03 & 151 & 232 & -1 & 00 59 32.9 -72 10 43 & o9 & 12.36 & 0.00 & 0.00 & -9.99 & 152 & 234 & 6440 & 00 59 44.3 -72 04 13 & b2 & 12.98 & 10.75 & 0.04 & 0.04 & 153 & 237 & 7019 & 00 59 53.5 -72 19 02 & b8 & 12.59 & 12.26 & 0.03 & 0.07 & 154 & 238 & 6475 & 00 59 56.0 -72 13 32 & o9 & 13.77 & 10.28 & 0.06 & 0.03 & 155 & 240 & 3921 & 01 00 1.1 -72 23 50 & b3 & 14.03 & 11.42 & 0.05 & -0.02 & 156 & 242 & 9303 & 01 00 7.9 -72 13 56 & b1 & 12.11 & 12.61 & 0.07 & 0.29 & 157 & 245 & 6543 & 01 00 16.7 -71 57 02 & b5 & 13.34 & 11.74 & 0.06 & 0.01 & 158 & 250 & 6589 & 01 00 23.8 -71 59 26 & b9 & 12.82 & 12.71 & 0.08 & 0.06 & 159 & 251 & 3932 & 01 00 22.9 -72 30 44 & o7 & 14.75 & 11.09 & 0.06 & 0.05 & 160 & 252 & 6450 & 01 00 30.6 -72 10 56 & b2 : & 13.07 & 10.63 & 0.05 & 0.03 & 161 & 257 & 6486 & 01 00 45.1 -72 23 50 & b3 & 12.79 & 10.87 & 0.05 & 0.03 & 162 & 260 & 6453 & 01 00 52.0 -72 13 32 & b1 & 13.28 & 10.67 & 0.05 & 0.05 & 163 & 263 & 6813 & 01 01 6.9 -72 12 57 & b6 & 12.85 & 12.07 & 0.04 & 0.07 & 164 & 264 & 6422 & 01 01 8.4 -71 59 56 & b0 & 12.36 & 9.50 & 0.12 & 0.06 & 165 & 266 & 6489 & 01 01 10.1 -72 27 26 & b1 & 12.63 & 9.90 & 0.06 & 0.04 & 166 & 267 & 6439 & 01 01 16.0 -72 06 32 & o8 & 14.92 & 10.72 & 0.03 & -0.01 & 167 & 268 & 6448 & 01 01 16.4 -72 12 39 & b0 : & 13.14 & 10.42 & 0.06 & 0.07 & 168 & 270 & 6477 & 01 01 17.5 -72 17 27 & b9 & 11.42 & 10.60 & 0.05 & 0.00 & 169 & 271 & 6476 & 01 01 20.5 -72 17 15 & b0 & 13.46 & 10.59 & 0.06 & 0.06 & 170 & 272 & 9881 & 01 01 23.4 -72 20 09 & b2 & 14.52 & 12.41 & 0.05 & 0.05 & 171 & 273 & 7855 & 01 01 27.9 -72 07 03 & a1 & 12.18 & 13.32 & 0.04 & 0.07 & 172 & 274 & 6483 & 01 01 29.8 -72 23 15 & b0 & 14.02 & 10.60 & 0.06 & 0.01 & 173 & 277 & 6824 & 01 01 33.0 -72 14 39 & b2 & 14.07 & 11.20 & 0.05 & -0.01 & 174 & 279 & 6428 & 01 01 35.2 -72 03 03 & o9 & 14.20 & 10.46 & 0.04 & 0.01 & 175 & 280 & 6580 & 01 01 39.9 -72 02 27 & b0 & 14.66 & 11.74 & 0.05 & 0.05 & 176 & 282 & 6447 & 01 01 50.4 -72 13 09 & o7 & 14.83 & 10.81 & 0.04 & 0.02 & 177 & 286 & 6605 & 01 01 57.8 -72 04 15 & a5 & 12.35 & 13.64 & 0.07 & -0.06 & 178 & 287 & 6444 & 01 01 57.5 -72 12 39 & b0 & 12.86 & 9.65 & 0.05 & 0.03 & 179 & 290 & 6561 & 01 02 0.8 -72 02 21 & b1 & 13.93 & 11.49 & 0.07 & 0.06 & 180 & 291 & 6816 & 01 02 4.6 -72 15 27 & b1 & 14.78 & 12.61 & 0.04 & 0.08 & 181 & 292 & 6923 & 01 02 4.5 -72 19 03 & b3 & 13.10 & 11.52 & 0.05 & 0.06 & 182 & 296 & 6445 & 01 02 8.8 -72 13 15 & o8 & 14.38 & 10.77 & 0.03 & 0.04 & 183 & 297 & 6419 & 01 02 10.4 -72 00 21 & b7 & 12.18 & 10.63 & 0.06 & 0.00 & 184 & 298 & 6560 & 01 02 12.7 -72 02 51 & a0 & 12.52 & 12.58 & 0.09 & 0.03 & 185 & 299 & 6980 & 01 02 13.9 -72 22 09 & o9 & 14.61 & 12.34 & 0.07 & 0.13 & 186 & 300 & 6442 & 01 02 14.8 -72 11 15 & b0 & 14.46 & 10.69 & 0.05 & -0.01 & 187 & 301 & 6512 & 01 02 15.7 -71 59 45 & b0 & 14.24 & 11.24 & 0.08 & 0.05 & 188 & 302 & 9841 & 01 02 19.3 -72 22 03 & b0 & 14.35 & 14.54 & 0.11 & 0.30 & 189 & 303 & 6418 & 01 02 22.3 -72 00 15 & b1 & 12.81 & 10.16 & 0.07 & 0.05 & 190 & 312 & 6441 & 01 02 45.8 -72 12 03 & b0 & 13.62 & 9.93 & 0.02 & -0.01 & 191 & 314 & 8760 & 01 02 48.6 -72 16 39 & b2 & 12.90 & 10.57 & 0.12 & 0.04 & 192 & 315 & 6438 & 01 02 50.0 -72 10 09 & b8 & 10.92 & 10.82 & 0.07 & 0.09 & 193 & 318 & 6436 & 01 02 54.8 -72 09 51 & o9 & 13.59 & 10.53 & 0.05 & 0.06 & 194 & 321 & 6433 & 01 02 57.8 -72 08 03 & o9 & 13.88 & 10.42 & 0.06 & 0.03 & 1 & 14 & 5.00 & 1.00 & 1.047 & 0.081 & 2.397 & 0.074 & 0.022 2 & 15 & 0.70 & 0.30 & 1.067 & 0.101 & 1.665 & 0.061 & 0.015 3 & 17 & 1.70 & 0.20 & 0.190 & 0.016 & 0.559 & 0.036 & 0.006 4 & 18 & 4.70 & 0.70 & 1.158 & 0.105 & 1.781 & 0.061 & 0.015 5 & 20 & 2.90 & 0.50 & 0.755 & 0.048 & 1.987 & 0.063 & 0.015 6 & 22 & 2.00 & 0.20 & 0.498 & 0.032 & 1.059 & 0.036 & 0.006 7 & 23 & 8.40 & 1.00 & 0.253 & 0.015 & 1.017 & 0.041 & 0.006 8 & 24 & 2.80 & 0.30 & 0.146 & 0.010 & 0.747 & 0.036 & 0.006 9 & 30 & 2.90 & 0.30 & 0.405 & 0.033 & 1.217 & 0.037 & 0.006 10 & 31 & 2.50 & 0.30 & 0.201 & 0.013 & 0.378 & 0.025 & 0.002 11 & 32 & 40.00 & 4.00 & 6.202 & 0.463 & 11.814 & 0.107 & 0.040 12 & 35 & 1.30 & 0.20 & 1.362 & 0.085 & 4.246 & 0.084 & 0.030 13 & 36 & 0.90 & 0.20 & 1.793 & 0.108 & 4.655 & 0.085 & 0.030 14 & 37 & 10.00 & 1.50 & 1.978 & 0.116 & 4.825 & 0.085 & 0.030 15 & 40 & 4.20 & 0.50 & 0.779 & 0.050 & 1.446 & 0.037 & 0.006 16 & 42 & 3.20 & 0.40 & 0.364 & 0.024 & 0.943 & 0.036 & 0.006 17 & 43 & 6.50 & 0.70 & 0.545 & 0.032 & 1.177 & 0.037 & 0.006 18 & 45 & 11.50 & 1.20 & 0.589 & 0.049 & 1.203 & 0.050 & 0.010 19 & 46 & 2.30 & 0.30 & 0.720 & 0.033 & 2.779 & 0.034 & 0.022 20 & 47 & 1.70 & 0.20 & 0.556 & 0.026 & 1.069 & 0.018 & 0.006 21 & 49 & 5.00 & 1.30 & 1.669 & 0.114 & 3.286 & 0.061 & 0.015 22 & 54 & 28.00 & 2.80 & 3.557 & 0.212 & 7.581 & 0.052 & 0.040 23 & 55 & 25.00 & 2.50 & 4.585 & 0.217 & 8.356 & 0.047 & 0.040 24 & 56 & 4.00 & 0.40 & 3.464 & 0.151 & 5.338 & 0.051 & 0.030 25 & 57 & 0.70 & 0.20 & 0.312 & 0.020 & 0.745 & 0.018 & 0.006 26 & 63 & 40.00 & 4.00 & 22.478 & 1.077 & 33.105 & 0.097 & 0.139 27 & 69 & 19.50 & 4.00 & 4.353 & 0.217 & 13.957 & 0.059 & 0.075 28 & 77 & 1.30 & 0.20 & 2.616 & 0.139 & 2.750 & 0.024 & 0.010 29 & 80 & 13.50 & 1.50 & 5.744 & 0.326 & 8.266 & 0.044 & 0.089 30 & 83 & 0.40 & 0.20 & 0.670 & 0.039 & 0.550 & 0.009 & 0.006 31 & 84 & 0.20 & 0.20 & 0.379 & 0.024 & 0.276 & 0.006 & 0.002 32 & 85 & 1.30 & 0.30 & 2.562 & 0.156 & 2.369 & 0.022 & 0.030 33 & 86 & 1.30 & 0.30 & 2.145 & 0.126 & 2.465 & 0.022 & 0.030 34 & 90 & 1.70 & 0.50 & 2.534 & 0.129 & 2.916 & 0.015 & 0.040 35 & 93 & 0.80 & 0.20 & 1.575 & 0.091 & 1.552 & 0.016 & 0.015 36 & 94 & 1.30 & 0.20 & 1.603 & 0.080 & 1.653 & 0.016 & 0.015 37 & 98 & 2.00 & 0.20 & 0.387 & 0.014 & 0.383 & 0.004 & 0.002 38 & 100 & 0.70 & 0.20 & 0.299 & 0.017 & 0.353 & 0.005 & 0.006 39 & 101 & 0.60 & 0.20 & 0.363 & 0.023 & 0.387 & 0.005 & 0.006 40 & 102 & 3.40 & 0.40 & 1.410 & 0.070 & 1.251 & 0.009 & 0.010 41 & 103 & 155.00 & 16.00 & 29.339 & 1.747 & 93.790 & 32.747 & 0.158 42 & 111 & 1.20 & 0.20 & 0.521 & 0.037 & 0.491 & 0.006 & 0.006

we describe new radial velocity and x - ray observations of extremely low - mass white dwarfs ( elm wds , ) in the sloan digital sky survey data release 4 and the mmt hypervelocity star survey . we identify four new short period binaries , including two merger systems . these observations bring the total number of short period binary systems identified in our survey to 20 . no main - sequence or neutron star companions are visible in the available optical photometry , radio , and x - ray data . thus , the companions are most likely wds . twelve of these systems will merge within a hubble time due to gravitational wave radiation . we have now tripled the number of known merging wd systems . we discuss the characteristics of this merger sample and potential links to underluminous supernovae , extreme helium stars , am cvn systems , and other merger products . we provide new observational tests of the wd mass - period distribution and cooling models for elm wds . we also find evidence for a new formation channel for single low - mass wds through binary mergers of two lower mass objects .

1011.4073

before the recent discovery of the six short period binary wd systems containing @xmath1 elm wds @xcite , there were only six double wd systems known to have short enough orbital periods to merge within a hubble time ( see * ? ? ? * and references therein ) . even though the supernovae ia progenitor survey ( spy ) identified @xmath2100 new double degenerate systems from a sample of 1000 wds observed using the very large telescope , @xcite identify only eight merger candidates ( including both wd+wd and sdb+wd systems ) from the spy and the literature . this fraction ( 8 out of 1000 ) is relatively small compared to the fraction of merger systems ( 6 out of 8) among the elm wds in the literature @xcite . low - mass ( @xmath3 ) he - core wds are formed when a companion strips the outer envelope from a post main - sequence star before the star reaches the tip of the red giant branch and ignites helium . they are usually found in close binaries , mostly double degenerate systems @xcite . however , about half of the low - mass wds in the field do not show any radial velocity variations , indicating that they are single @xcite . @xcite argue that these single low - mass wds may come from old metal - rich stars that truncate their evolution prior to the helium flash from severe mass loss . they estimate a binary fraction of 50% for @xmath4 m@xmath5 wds . however , they predict that the binary fraction rises to 100% for @xmath0 elm wds , since such extreme mass loss rates are not expected even for the most metal - rich stars in the galaxy . to study the binary fraction of elm wds and to find future merger systems , we have undertaken an mmt survey of all previously identified elm wds from the sdss dr4 area @xcite and the hypervelocity star survey of @xcite . the latter sample is described in a companion paper by @xcite . part of the sdss dr4 sample was discussed earlier in @xcite . here , we present radial velocity observations of 7 more elm wd candidates from the sdss sample . the combined sample contains 23 wds . we also report on _ chandra _ x - ray observations of three elm wds to search for x - ray evidence of milli - second pulsar ( msp ) companions . while almost all sdss elm wds show radial velocity variations , none show evidence of main - sequence or neutron star companions in the optical , radio @xcite , or x - ray ( this paper and * ? ? ? hence , they are almost certainly binary wd systems . out of the 20 short period binary systems in our sample , 12 have merger times shorter than a hubble time . our elm survey has now tripled the number of known merging wd systems . based on our sample and previously known short period binary wd systems , we study the mass - period distribution and future evolutionary prospects , including the connections to unusual stellar populations and underluminous supernovae ( sne ) . our paper is organized as follows . in section 2 we discuss our radial velocity and x - ray observations . in section 3 we present an analysis of the optical spectroscopy and x - ray data and the nature of the companions . in section 4 we discuss the sample characteristics of elm wds , including the binary fraction , mass - period distribution , merging systems , and the implications for underluminous sne and other unusual merger products . we conclude in section 5 . @xcite identified 13 low - mass wd candidates with @xmath6 . @xcite discuss follow - up observations of four of these targets . we selected 6 more targets for follow - up mmt spectroscopy . these targets are sdss j002228.45 + 003115.5 , j002207.65@xmath7101423.5 , j123410.36@xmath7022802.8 , j162542.10 + 363219.1 , j204949.78 + 000547.3 , and j225242.25@xmath7005626.6 . we also observed sdss j234536.48@xmath7010204.8 , another elm wd candidate identified by @xcite . there are three more targets in the @xcite elm wd sample that lack follow - up spectroscopy . these targets ( sdss j1056 + 6536 , j1426 + 0100 , and j1630 + 4233 ) have @xmath8 based on the model fits by @xcite and @xmath9 based on the fits by @xcite . radial velocity observations of these objects will be useful for the identification of more short period binary wd systems . we used the 6.5 m mmt equipped with the blue channel spectrograph to obtain moderate resolution spectroscopy of seven elm wd candidates over 17 different nights between 2008 september and 2010 july . we operate the spectrograph with the 832 line mm@xmath10 grating in second order , providing wavelength coverage 3600 to 4500 and a spectral resolution of 1.0 - 1.2 , depending on the slit size used . we obtain all observations at the parallactic angle , with a comparison lamp exposure paired with every observation . we flux - calibrate using blue spectrophotometric standards @xcite , and we measure radial velocities using the cross - correlation package rvsao @xcite . our radial velocity measurement procedures are described in @xcite . x - ray emission from radio msps is expected to be less highly beamed than radio emission , as a major component of the x - ray emission comes from the hot neutron star surface @xcite . blackbody emission from the surface of a possible pulsar companion to the elm wds will be gravitationally bent , allowing observation of @xmath1175% of the neutron star surface in x - rays even if the radio pulsar beam ( produced well above the surface ) misses our line of sight @xcite . all 15 radio msps with precise positions in unconfused regions of the globular cluster 47 tuc have been clearly detected in x - rays @xcite , with no observed correlation between their radio and x - ray luminosities ( expected , as they are produced via different mechanisms ) . there is also no evidence for differences in the x - ray properties of globular cluster vs. galactic msps @xcite . this result allows us to use the 47 tuc msp sample ( with accurate @xmath12 values due to its well - known distance ) to predict that other msps should have x - ray luminosities above @xmath12(0.5 - 6 kev)@xmath13 erg / s , the minimum @xmath12 of msps in 47 tuc . thus , deep x - ray upper limits can rule out the presence of an msp , as in the elm wd sdss j0917 + 4638 @xcite . the msps in 47 tuc show x - ray spectra dominated by thermal blackbody - like emission from the neutron star polar cap surfaces , with temperature @xmath14 k @xcite . some msps also show additional ( sometimes much higher @xmath12 ) components due to magnetospheric emission or pulsar wind shocks , but there is no evidence for a msp lacking the thermal component . we obtained _ chandra _ observations of three elm wds ( sdss j105353.89 + 520031.0 , j123410.36@xmath7022802.8 , and j092345.59 + 302805.0 ) to search for msp x - ray emission . none of these wds had been previously observed in the x - rays since the _ rosat _ all - sky survey @xcite , where the short exposure time was insufficient to place useful limits . we observed the three systems with _ chandra _ s acis - s instrument in very faint mode , for the exposure times noted in table 1 . we used ciao 4.2 ( with caldb 4.2.1 ) to reprocess the observations with current calibrations and reduce the backgrounds using very faint mode cleaning . we constructed images in the 0.3 - 6 kev band , and found no x - ray photons within the 1 error circles around each source . we use poisson statistics @xcite to place 99% confidence upper limits on the count rate . we compute distances to the elm wds using the models of @xcite and the sdss @xmath15 magnitudes , and @xmath16 values using the colden tool @xcite . we use pimms and the x - ray spectrum of the faintest msp in 47 tuc ( 47 tuc - t , 134 ev blackbody ) to produce 0.5 - 6 kev @xmath12 upper limits , which we list in table 1 . the 99% confidence @xmath12 upper limits are all at least a factor of 10 fainter than the faintest msp identified in 47 tuc , and 21 times fainter than the median @xmath12 of these msps @xcite . thus the lack of x - ray emission from these wds is strong evidence that the unseen companions are not msps . four of our seven targets exhibit significant radial velocity variations , with peak - to - peak velocity amplitudes between 124 and 232 km s@xmath10 . we weight each velocity measurement by its associated error and solve for the best - fit orbit using the code of @xcite . figures 1 - 6 show the observed radial velocities and the best fit orbits for our targets . the heliocentric radial velocities are best fit with circular orbits with periods 1.9 - 11.8 hr . we present the best - fit orbital period ( @xmath17 ) , semi - amplitude ( @xmath18 ) of the radial velocity variations , systemic velocity ( which includes a small gravitational redshift term ) , the time of spectroscopic conjunction , and the mass function in table 2 . j2049 + 0005 is excluded from this list due to its low surface gravity , which implies that it is not a wd ( see below ) . two of the stars in our sample , j2252@xmath70056 and j2345@xmath70102 , do not show significant velocity variations . the best - fit orbital solutions have semi - amplitudes 25 - 43 km s@xmath10 . these two objects may be low inclination systems or just single stars . we perform model fits to each individual spectrum and also to the average composite spectra using synthetic da wd spectra kindly provided by d. koester . this model grid covers temperatures up to 30,000 k. we use an nlte grid computed by i. hubeny ( see * ? ? ? * ) to fit the spectrum of j2345@xmath70102 , which requires a solution slightly above 30,000 k. we use the individual spectra to obtain a robust estimate of the errors in our analysis . figure 7 shows the composite spectra and our fits using the entire wavelength range . the best - fit @xmath19 and @xmath20 values are given in table 3 . we obtain best - fit solutions of @xmath21 k and @xmath22 for six of our targets , confirming that they are low - mass wds . figure 8 shows the effective temperatures and surface gravities for our targets ( red circles ) plus the previously identified elm wds in the literature @xcite . filled triangles show the wd companions to milli - second pulsars psr j1012 + 5307 and j1911 - 5958a ( * ? ? ? * ; * ? ? ? * see 4.4 ) . solid lines show the constant mass tracks for low mass wds from @xcite models ( updated by * ? ? ? this figure shows that our six wd targets have masses ranging from 0.20 to 0.42 @xmath23 . the best - fit orbital period for j0022 + 0031 is 11.8 hr . however , the limited number of follow - up observations allow for an alias at 7.9 hr period . j0022 + 0031 is a binary system containing a 120 myr old 0.38 @xmath23 wd and a @xmath24 companion star . based on the updated @xcite models , it has an absolute magnitude of @xmath25 mag and a distance of 790 pc . for all six wds in our sample , we combine the spectra near maximum blue - shifted and red - shifted velocities into two composite spectra . if there is a contribution from a companion object , it may be visible as an asymmetry in the line profiles . we do not see any obvious asymmetries in the line profiles and our optical spectroscopy does not reveal any spectral features from companion objects . a @xmath26 main - sequence star companion to j0022 + 0031 would have @xmath27 mag @xcite , brighter than the low - mass wd ( @xmath28 mag ) and detectable in the @xmath29-band . such a main - sequence star companion is ruled out based on the sdss photometry . using the mean inclination angle for a random stellar sample , @xmath30 , the companion has a mass of 0.26 @xmath23 . there is a 2.5% probability that the companion is a 1.4 - 3 @xmath23 neutron star . given the small probability of a neutron star companion and the unsuccesful searches for neutron stars around other elm wds ( section 2.2 , see also * ? ? ? * ; * ? ? ? * ) , the companion is most likely another wd . short period binaries may lose angular momentum through gravitational wave radiation and merge within a hubble time . the merger time for the j0022 + 0031 system is longer than a hubble time . j0022@xmath71014 is a 1.9 hr period binary system with a 0.33 @xmath23 wd primary and a @xmath31 secondary . the primary wd is 70 myr old , it has @xmath32 = 9.2 mag , and is located at a distance of 1.3 kpc . a @xmath31 main - sequence star companion would have @xmath33 mag , 30% fainter than the elm wd . such a companion is ruled out based on the sdss photometry . the companion to j0022@xmath71014 is most likely another wd , and specifically a low - mass wd . there is a 83% probability that the companion is less massive than 0.45 @xmath23 , a he - core low - mass wd . this system will merge in less than 730 myr . j0022@xmath71014 has a proper motion of @xmath34 and @xmath35 mas yr@xmath10 @xcite . based on the mass and radius estimates , the gravitational redshift of the wd is 8.1 km s@xmath10 . therefore , the true systemic velocity is @xmath36 km s@xmath10 . the velocity components with respect to the local standard of rest as defined by @xcite are @xmath37 , and @xmath38 km s@xmath10 . j0022@xmath71014 is a disk star . j1234@xmath70228 was originally identified as a very low mass wd by @xcite and is included in the dr4 wd catalog of @xcite . based on our mmt observations over 5 different nights separated by about a year , j1234@xmath70228 is a binary system with an orbital period of 2.2 hr . the visible component of the binary is a 70 myr old wd with @xmath32 = 8.4 mag at a distance of 780 pc . for an inclination angle of 60@xmath39 , the mass function requires a 0.11 @xmath23 companion . a main - sequence companion of this mass would contribute 7% excess flux in the @xmath40band . no excess flux is observed in the sdss photometry and spectrum of this object , therefore the companion is most likely another wd ( see also * ? ? ? there is a 94% chance that the companion is a low - mass ( @xmath41 ) wd , and more interestingly there is a 85% probability that the combined mass of the two stars in this system is lower than 0.45 @xmath23 . the merger time for the j1234@xmath70228 system is shorter than 2.7 gyr . therefore , within the next few gyr , this system will merge and likely form a single low - mass wd . j1234@xmath70228 has a proper motion of @xmath42 and @xmath43 mas yr@xmath10 @xcite . based on the mass and radius estimates , the gravitational redshift of the wd is 3.7 km s@xmath10 . therefore , the true systemic velocity is + 46.6 km s@xmath10 . the velocity components with respect to the local standard of rest as defined by @xcite are @xmath44 , and @xmath45 km s@xmath10 . j1234@xmath70228 is also a disk star . j1625 + 3632 has a best - fit orbital period of 5.6 hr . the mmt spectrum reveals a weak line at 4471 . our best - fit da wd model has @xmath46 23570 k and @xmath47 6.12 , similar to the sdb star hd 188112 @xcite . based on these model fits , the primary is a @xmath2160 myr old 0.20 @xmath23 wd at a distance of 2.8 kpc . j1625 + 3632 does not display a significant proper motion , but given its distance and systemic velocity , it is likely a member of the thick disk or halo . the relatively small semi - amplitude ( @xmath48 km s@xmath10 ) of the radial velocity variations imply that the lower limit on the mass of the companion is 0.07 @xmath23 . a main - sequence companion of this mass would contribute less than 1% in the optical and would not be detected in our observations or the sdss photometry . given the accuracy of the sdss photometry , any main - sequence companion more massive than about 0.17 @xmath23 would have been detected . hence , only 0.07 - 0.17 @xmath23 main - sequence companions are allowed based on the mass function and optical photometry . near - infrared observations will be useful to rule out such companions . j1625 + 3632 may be a low inclination system . however , the probability of a neutron star companion is only 0.6% . given the narrow range of possible main - sequence companions and the unlikelihood of a neutron star companion , the companion is likely another wd . there is a 96% chance that the companion is less massive than 0.45 @xmath23 , i.e. , a low - mass he - core wd . figure 9 shows the best - fit wd model to our mmt spectrum of j2049 + 0005 . this star was identified as a @xmath19 = 8660 k and @xmath20 = 5.48 wd by @xcite . further analysis of the same sdss spectrum by @xcite suggested that j2049 + 0005 has @xmath49 , and therefore is not a wd . similarly , our model fits to the higher resolution and higher signal - to - noise ratio mmt spectrum find a best - fit solution that is exactly our lowest gravity model ( @xmath20 = 5 ) . this model poorly matches the high order balmer lines , indicating that the surface gravity is lower than @xmath20 = 5 and that j2049 + 0005 is most likely an a star . based on 30 different spectra , the best - fit orbital period is 10.3 hr with @xmath50 km s@xmath10 . however , the radial velocity variations are not significant . the best - fit orbit has a @xmath51 slightly better than a constant - velocity fit . if j2252@xmath70056 is a binary , it is likely a low inclination or a pole - on system . given the 25 km s@xmath10 upper limit to the velocity semi - amplitude , j2252@xmath70056 could be a binary with either month - long ( assuming an 0.5 @xmath23 companion and @xmath52 ) orbital period or relatively pole - on inclination ( @xmath53 for 2 - 12 hr periods ) . there is no evidence of a companion in our mmt spectrum or the sdss observations . j2345@xmath70102 is another object for which we detect no significant velocity variations . the best - fit orbital period is 7.2 hr with @xmath54 km s@xmath10 , but the best - fit orbit is a slightly better fit to the data than a constant - velocity fit . again , if j2345@xmath70102 is a binary , it may be a low inclination system . the observed 43 km s@xmath10 upper limit to the velocity semi - amplitude corresponds to a binary with either month - long orbital period or relatively pole - on inclination ( @xmath55 for 2 - 12 hr periods ) . the mmt spectrum of this object is best - explained with a 23 myr old 0.42 @xmath23 wd at a distance of 2.0 kpc . based on 5 epochs from the usno - b+sdss , @xcite measure a proper motion of @xmath56 and @xmath57 mas yr@xmath10 . given its relatively large systemic velocity , proper motion , and distance , j2345@xmath70102 is clearly a halo star . about half of the known 0.4 @xmath23 wds seem to be single ( see the discussion in * ? ? ? hence , the lack of significant radial velocity variations is not very surprising for this system . @xcite , @xcite , and this paper discuss radial velocity observations of 11 elm wd candidates identified in the sdss dr4 @xcite and 12 elm wds identified in the hypervelocity star survey of @xcite . one of the targets is common to both samples . in addition , one of the dr4 targets ( j2049 + 0005 ) is most likely an a - star . @xcite , @xcite , and @xcite describe radial velocity observations of two more elm wds identified in the literature . thus , there are 23 ( candidate ) elm wd systems with comprehensive radial velocity measurements . all but three of these systems ( j1448 + 1342 , j2252@xmath70056 , and j2345@xmath70102 ) are in binary sytems . limiting this sample to 19 wds with @xmath58 , 18 are in short period binary systems . thus , close binary evolution is required for the formation of elm wds . the only system that does not display significant radial velocity variations , j1448 + 1342 , is at the upper mass limit of this selection ; all 18 wds with @xmath59 are indeed binaries . the average velocity semi - amplitude of these 18 binaries is 236 km s@xmath10 , whereas the upper limit for the velocity semi - amplitude of j1448 + 1342 is 35 km s@xmath10 . an average system viewed at @xmath60 would be consistent with the observations of j1448 + 1342 . for a randomly distributed sample of orbital inclinations , there is a 1.1% chance that @xmath60 . thus , there is a 21% likelihood of finding one of the 19 systems with @xmath60 . it is possible that j1448 + 1342 is a pole - on binary system ( see also * ? ? ? * ) . hence , the binary fraction of @xmath58 wds is at least 95% and it may be as high as 100% . this result is in contrast to the binary fraction of slightly more massive wds . @xcite , @xcite , and @xcite demonstrate that about 30 - 50% of @xmath61 wds are single ( see also * ? ? ? single low - mass wds may form through enhanced mass loss from a metal - rich progenitor star @xcite , but we do not expect even the most metal - rich stars to loose enough mass to create @xmath1 wds . here we confirm this expectation that the lowest mass wds all form in close binary systems . double degenerate ( dd ) merger systems are one of the two main accepted formation channels for type ia supernovae ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * and references therein ) . however , only a handful of dd systems are known to have combined mass above the chandrasekhar limit and short enough orbital periods to merge through gravitational wave radiation within a hubble time . @xcite started the heroic spy survey for short period dd systems . they used the very large telescope to observe 1014 stars brighter than 16.5 mag , including 75% of the known wds accessible from the southern hemisphere . @xcite report that the spy survey discovered about 100 new dd systems , one of which is a potential sne ia progenitor ( wd 2020@xmath7425 , * ? ? ? @xcite describe the discovery of another massive dd merger system and also a summary of the current discoveries from the spy survey ( their figure 4 ) . figure 10 shows the total masses and periods for the short period dd systems ( including sdb+wd ) discovered by the spy survey ( kindly made available to us by s. geier ) and our elm wd sample . to explain the observed sne ia rate , only one out of 1000 wds is expected to be a sne ia progenitor @xcite . hence , the relatively few number of potential sne ia progenitors in the spy survey is not surprising . what is striking is that the number of merger systems found in the spy survey is significantly smaller than our elm wd survey . we discovered 12 binary wd merger systems ( @xmath62 ) in a sample of 23 stars , whereas the spy survey discovered a handful of such systems in a sample of 1000 stars . the reason for this is the binary common envelope evolution required to form elm wds . figure 11 shows the total masses and periods for previously identified wd+wd binary systems in the literature @xcite . there were only six double wd merger systems known prior to our work . with the additional 12 merger systems presented in table 4 , we now have tripled the number of wd merger systems known . @xcite studied the period and mass distribution of double wd systems using population synthesis models . the details of these models including selection effects due to orbital evolution of the systems ( disappearence of the shortest period systems due to mergers ) , target selection biases in the current surveys , and different cooling models are discussed in their paper . @xcite find a correlation between the masses and periods ; more massive wd primaries have longer periods . this is not surprising ; shorter period systems would start interacting earlier in their evolution compared to longer period systems and experience enhanced mass - loss during the red giant phase , hence end up as lower mass wds . they also find that the mass ratio distribution peaks at @xmath63 , favoring equal mass binary components . this distribution is consistent with the observed mass ratio distribution of binary wds known prior to our work . figure 12 shows the masses and orbital periods for our sample of elm wds and the previously known binary wd systems in the literature compared to the population synthesis models of ( * ? ? ? * ; * ? ? ? * kindly made available to us by g. nelemans ) . depending on which cooling model is used , the model predictions change somewhat ( see * ? ? ? * for a discussion of the cooling models used here ) . however , the main trend in the mass versus period distribution is evident ; lower mass primaries are in shorter period systems . this result is consistent with the observed distribution of masses and periods for previously known binary wd systems and the elm wd systems . all but one of the @xmath64 systems have periods shorter than a day , whereas the majority of the more massive wds have periods longer than a day . the overall number distribution of the population synthesis models and the observations can not be directly compared without taking into account the obvious biases in the target selection . the sample of previously known binary wds are selected from a magnitude limited sample down to 16 - 17 mag , whereas our sample mainly comes from fainter targets selected from the sdss , which has its own spectroscopic target selection bias . nevertheless , an overabundance of 0.17 @xmath23 wds is evident in figure 12 . a comparison between the population synthesis models using ( * ? ? ? * top panel ) and the modified ( * ? ? ? * bottom panel ) cooling models shows that the expected number of the lowest mass wds changes depending on the cooling model used . the latter models predict a significantly larger number of 0.17 @xmath23 wds . based on the @xcite models , a 0.25 @xmath23 wd cools 4@xmath65 faster than a 0.17 @xmath23 wd down to an absolute magnitude of @xmath66 mag . hence , the observed mass distribution of elm wds in a magnitude - limited survey is skewed toward lower mass objects @xcite . population synthesis models and a large , unbiased , magnitude - limited survey of low - mass and elm wds like that of @xcite will be useful to constrain the elm wd cooling models . another important prediction of the population synthesis models is that there should not be many systems with periods less than an hour . we have three elm wd systems with periods close to 1 hr . the shortest period system , j1053 + 5200 has a period of 61 min @xcite . the shortest period hydrogen - rich cataclysmic variable ( cv ) v485 cen has an orbital period of 59 min @xcite . both short period binary wds and cvs are expected to contribute to the am cvn population , which contains interacting wd binary systems with orbital periods ranging from a few minutes to about an hour . the two longest period am cvn systems have periods of 56 min and 66 min , respectively @xcite . it seems that hydrogen - deficient am cvn interacting binary systems have periods less than about an hour and the hydrogen - rich detached wd systems have orbital periods about an hour or longer . before the sdss , elm wds were traditionally found as companions to msps . the masses for the msp companions are usually measured from the pulsar timing measurements since the wds are usually too faint for follow - up optical spectroscopy . @xcite and @xcite obtained keck and vlt spectroscopy of companions to two msp systems . @xcite derive @xmath46 8550 @xmath67 25 k and @xmath68 g = 6.75 @xmath67 0.07 for the companion to psr j1012 + 5307 , whereas @xcite obtain @xmath46 10,090 @xmath67 150 k and @xmath68 g = 6.44 @xmath67 0.20 for the companion to psr j1911 - 5958a . these two wds are included in figure 8 . @xcite discovered the lowest surface gravity wd known at the time , sdss j0917 + 4638 , with @xmath69 . @xcite discovered two wds with even lower surface gravities , sdss j1233 + 1602 and j2119@xmath70018 . these wds are included in our elm wd sample and in figure 8 . based on the @xcite and @xcite models , all of these wds , including the two msp companions , have @xmath70 . there are about a dozen wds in figure 8 with @xmath71 , but none below that mass limit . however , this result is most likely due to observational biases present in the sdss and the hypervelocity star survey . the former is , of course , an incomplete spectroscopic survey and the latter has a color selection that excludes @xmath72 10000 k objects with @xmath73 ( see figure 1 in * ? ? ? based on the @xcite models , a 0.16 @xmath23 wd takes 15 gyr to cool down from 9200 k to about 7000 k with an increase in surface gravity from @xmath74 5.7 to 6.4 and a decrease in luminosity from @xmath75 8.1 to 11.1 mag . such wds are not included in the hypervelocity star survey color selection . based on radio timing measurements , 45 out of 99 msps with periods less than 20 ms have companions less massive than @xmath76 ( b. kiziltan 2010 , private communication ) . the spectral types of the companions are unknown . however , a large fraction of these objects are expected to be low - mass wds ; nature produces @xmath77 wds in msp systems . lower mass elm wds should be produced in binary wd systems as well . population synthesis calculations for double wds do not predict many wds below 0.17 @xmath23 , but such wds may eventually be found in double wd systems in the sdss or other spectroscopic surveys . the end result of the future evolution of the 12 elm wd merger systems in our sample depends on the mass ratios of the two components . the systems with extreme mass ratios ( @xmath78 ) should have stable mass transfer at initial roche lobe filling . stable mass transfer leads to the formation of am cvn systems that transfer angular momentum back to the orbit , which increases the orbital period . if the higher mass wds in such systems have masses close to the chandrasekhar limit , accretion may result in a type ia sne explosion . however , the probability of such an event is small ( @xmath79 7% ) for our systems . this probability would be higher if sub - chandrasekhar mass wd explosions also contribute to the observed type ia sne population ( * ? ? ? * ; * ? ? ? * and references therein ) . however , our elm survey is biased against sne ia progenitors due to the selection of very low - mass primaries . figure 13 shows the mass ratios and periods for our elm wd sample plus the previously known binary wd systems . single - lined systems are shown as filled triangles , whereas the double - lined systems and the eclipsing wd system nltt 11748 are shown as filled circles . open circles mark the objects with merger times shorter than a hubble time . out of the 18 merger systems , six have mass ratios smaller than 0.25 ( assuming an average inclination of 60@xmath39 ) , and are likely to form am cvn systems . these six systems also have the highest total mass ( @xmath80 ) in our sample . j1233 + 1602 is the most extreme case with @xmath81 for @xmath30 . it is striking that the number of extreme mass ratio systems in our sample ( six ) is the same as the number of am cvn systems found in the sdss data relase 5 area @xcite . these systems are potential sne .ia progenitors @xcite . based on a magnitude - limited sample of six elm wd mergers , @xcite find that half of their sample has mass ratios of @xmath20.15 and they estimate a contribution factor of @xmath823% to the am cvn formation rate from the observed population of elm wds . the fraction of extreme mass ratio systems in our survey is also 50% ( six out of 12 elm wds ) , supporting the conclusions reached by @xcite . the other six elm wd systems in our sample have mass ratios larger than 0.4 . @xcite discuss various evolutionary scenarios for these objects . these systems will merge within a hubble time and form a variety of exotic objects including underluminous sne , extreme helium stars ( rcrb ) , or single helium - enriched subdwarf o stars . the merger rate of elm wds is roughly the same as the observed rate of underluminous sne @xcite . sdss j1234@xmath70228 is a unique elm wd merger system . there is a 85% chance that the combined mass of the two stars in this system is lower than 0.45 @xmath23 . therefore , within the next few billion years , this system will merge and likely form a a single low - mass wd . there are several scenarios to explain the presence of single low - mass ( @xmath83 ) wds , including enhanced mass - loss from metal - rich progenitors @xcite and mass loss through common envelope evolution with a low - mass brown dwarf or planetary mass companion @xcite . single low - mass wds can also form through mergers of two lower - mass wds @xcite , though no such progenitors have been found until now . here we have uncovered a potential progenitor system for a future single low - mass wd , j1234@xmath70228 . about 10% of da wds are low - mass @xcite and half of these low - mass wds are single @xcite , corresponding to a space density of 2.5 @xmath84 kpc@xmath85 @xcite . the sdss dr4 wd catalog covers 4783 square degrees with @xmath86 mag @xcite . we estimate the local space density using the modified @xmath87 method and the disk model of @xcite . based on one system , j1234@xmath70228 , the space density of merging wd systems that can form single low - mass wds is 1.1 kpc@xmath85 . this number is likely uncertain by a factor of 10 due to incompleteness of the sdss spectroscopic target selection . nevertheless , it is several orders of magnitude smaller than the estimated space density of single low - mass wds . hence , this formation channel does not significantly contribute to the single low - mass wd population in the solar neighborhood . our mmt elm survey has now discovered 20 short period binary wd systems , including 12 merger systems . only six binary wd merger systems were known prior to our work . the elm survey has now tripled the number of such systems . these merger systems provide potential formation channels for a variety of interesting objects including underluminous sne , am cvn , extreme helium stars , single subdwarf stars , and now single low - mass wds . all but one of the 19 stars with @xmath58 are in binaries , strongly arguing for a binary formation channel for elm wds . the combined sample of more than 40 short period wd systems found in our elm survey and the literature demonstrate that the shortest period detached wd systems have @xmath21 hr orbital periods and they also have the lowest mass primaries . in addition , the lowest mass wds in double wd systems have @xmath88 . lower - mass objects may be produced in such systems but are currently undetected in the sdss and other spectroscopic surveys . the sample of elm wd merger systems will be significantly enlarged by our continuing elm survey based on sdss photometry . such a magnitude - limited survey will not suffer from the biases present in our current sample , and will provide invaluable information on the merger rates of elm wd systems and their contribution to underluminous sne and am cvn systems . furthermore , a large sample of elm wds and population synthesis models will improve our understanding of the cooling models for elm wds . we thank g. nelemans for providing his binary wd population synthesis model results , s. geier for the latest results from the spy survey , and d. koester for wd model spectra . support for this work was provided by nasa through the _ spitzer space telescope _ fellowship program , under an award from caltech . _ facilities : _ augusteijn , t. , van der hooft , f. , de jong , j. a. , & van paradijs , j. 1996 , , 311 , 889 ageros , m. a. , camilo , f. , silvestri , n. m. , kleinman , s. j. , anderson , s. f. , & liebert , j. w. 2009a , , 697 , 283 ageros , m. a. , et al . 2009 , , 700 , l123 allende prieto , c. , hubeny , i. , & smith , j. a. 2009 , , 396 , 759 badenes , c. , mullally , f. , thompson , s. e. , & lupton , r. h. 2009 , , 707 , 971 bassa , c. g. , van kerkwijk , m. h. , koester , d. , & verbunt , f. 2006 , , 456 , 295 beloborodov , a. m. 2002 , , 566 , l85 bildsten , l. , shen , k. j. , weinberg , n. n. , & nelemans , g. 2007 , , 662 , l95 bogdanov , s. , grindlay , j. e. , & rybicki , g. b. 2006 , , 648 , l55 branch , d. , livio , m. , yungelson , l. r. , boffi , f. r. , & baron , e. 1995 , , 107 , 1019 brown , j. m. , kilic , m. , brown , w. r. , & kenyon , s. j. 2010 , , submitted brown , w. r. , geller , m. j. , kenyon , s. j. , & kurtz , m. j. 2006 , , 647 , 303 brown , w. r. , kilic , m. , allende prieto , c. , & kenyon , s. j. 2010a , , 723 , 1072 brown , w. r. , kilic , m. , allende prieto , c. , & kenyon , s. j. 2010b , , in press dickey , j. m. , & lockman , f. j. 1990 , , 28 , 215 di stefano , r. 2010 , , 719 , 474 driebe , t. , schoenberner , d. , bloecker , t. , & herwig , f. 1998 , , 339 , 123 eisenstein , d. j. , et al . 2006 , , 167 , 40 gehrels , n. 1986 , , 303 , 336 geier , s. , heber , u. , kupfer , t. , & napiwotzki , r. 2010 , , 515 , a37 hansen , b. m. s. 1999 , , 520 , 680 heber , u. , edelmann , h. , lisker , t. , & napiwotzki , r. 2003 , , 411 , l477 heinke , c. o. , grindlay , j. e. , edmonds , p. d. , cohn , h. n. , lugger , p. m. , camilo , f. , bogdanov , s. , & freire , p. c. 2005 , , 625 , 796 hogg , d. w. , blanton , m. r. , roweis , s. t. , & johnston , k. v. 2005 , , 629 , 268 iben , i. , jr . , & tutukov , a. v. 1984 , , 54 , 335 iben , i. , jr . , tutukov , a. v. , & yungelson , l. r. 1997 , , 475 , 291 kawka , a. , vennes , s. , & vaccaro , t. r. 2010 , , 516 , l7 kenyon , s. j. & garcia , m. r. 1986 , , 91 , 125 kilic , m. , allende prieto , c. , brown , w. r. , & koester , d. 2007 , , 660 , 1451 kilic , m. , brown , w. r. , allende prieto , c. , pinsonneault , m. h. , & kenyon , s. j. 2007b , , 664 , 1088 kilic , m. , stanek , k. z. , & pinsonneault , m. h. 2007 , , 671 , 761 kilic , m. , brown , w. r. , allende prieto , c. , swift , b. , kenyon , s. j. , liebert , j. , & ageros , m. a. 2009 , , 695 , l92 kilic , m. , brown , w. r. , & mcleod , b. 2010 , , 708 , 411 kilic , m. , brown , w. r. , allende prieto , c. , and kenyon , s. j. 2010 , , 716 , 122 kilic , m. , allende prieto , c. , brown , w. r. , ageros , m. a. , kenyon , s. j. , and camilo , f. 2010 , , 721 , l158 kotak , r. 2008 , asp conf . 401 , rs ophiuchi ( 2006 ) and the recurrent nova phenomenon , ed . a. evans , m. f. bode , t. j. obrien , & m. j. darnley ( san francisco , ca : asp ) , 150 kromer , m. , sim , s. a. , fink , m. , rpke , f. k. , seitenzahl , i. r. , & hillebrandt , w. 2010 , , 719 , 1067 kroupa , p. , & tout , c. a. 1997 , , 287 , 402 kulkarni , s. r. & van kerkwijk , m. h. 2010 , , 719 , 1123 kurtz , m. j. , & mink , d. j. 1998 , , 110 , 934 liebert , j. , bergeron , p. , eisenstein , d. , harris , h. c. , kleinman , s. j. , nitta , a. , & krzesinski , j. 2004 , , 606 , l147 liebert , j. , bergeron , p. , & holberg , j. b. 2005 , , 156 , 47 marsh , t. r. 1995 , , 275 , l1 marsh , t. r. , gaensicke , b. t. , steeghs , d. , southworth , j. , koester , d. , harris , v. , & merry , l. 2010 , , submitted massey , p. , strobel , k. , barnes , j. v. , & anderson , e. 1988 , , 328 , 315 maxted , p. f. l. , marsh , t. r. , & moran , c. k. j. 2000 , , 319 , 305 mullally , f. , badenes , c. , thompson , s. e. , & lupton , r. 2009 , , 707 , l51 munn , j. a. , et al . 2004 , , 127 , 3034 napiwotzki , r. , edelmann , h. , heber , u. , karl , c. , drechsel , h. , pauli , e .- m . , & christlieb , n. 2001 , , 378 , l17 napiwotzki , r. , et al . 2004a , spectroscopically and spatially resolving the components of the close binary stars , 318 , 402 napiwotzki , r. , et al . 2007 , 15th european workshop on white dwarfs , 372 , 387 nelemans , g. , & tauris , t. m. 1998 , , 335 , l85 nelemans , g. , yungelson , l. r. , portegies zwart , s. f. , & verbunt , f. 2001 , , 365 , 491 nelemans , g. , yungelson , l. r. , & portegies zwart , s. f. 2004 , , 349 , 181 nelemans , g. , et al . 2005 , , 440 , 1087 panei , j. a. , althaus , l. g. , chen , x. , & han , z. 2007 , , 382 , 779 roelofs , g. h. a. , groot , p. j. , steeghs , d. , marsh , t. r. , & nelemans , g. 2007 , , 382 , 1643 saio , h. , & jeffery , c. s. 2000 , , 313 , 671 serenelli , a. m. , althaus , l. g. , rohrmann , r. d. , & benvenuto , o. g. 2001 , , 325 , 607 serenelli , a. m. , althaus , l. g. , rohrmann , r. d. , & benvenuto , o. g. 2002 , , 337 , 1091 sim , s. a. , rpke , f. k. , hillebrandt , w. , kromer , m. , pakmor , r. , fink , m. , ruiter , a. j. , & seitenzahl , i. r. 2010 , , 714 , l52 sion , e. m. , holberg , j. b. , oswalt , t. d. , mccook , g. p. , & wasatonic , r. 2009 , , 138 , 1681 steinfadt , j. d. r. , kaplan , d. l. , shporer , a. , bildsten , l. , & howell , s. b. 2010 , , 716 , l146 van kerkwijk , m. h. , bergeron , p. , & kulkarni , s. r. 1996 , , 467 , l89 vennes , s. , kawka , a. , vaccaro , t. r. , & silvestri , n. m. 2009 , , 507 , 1613 voges , w. , et al 1999 , , 349 , 389 zavlin , v. e. , pavlov , g. g. , sanwal , d. , manchester , r. n. , trmper , j. , halpern , j. p. , & becker , w. 2002 , , 569 , 894 zavlin , v. e. 2006 , , 638 , 951 lcccccc j105353.89 + 520031.0 & 9963 & 20.50 & 1.08 & @xmath89 & @xmath90 & @xmath91 + j123410.36@xmath7022802.8 & 9964 & 5.50 & 0.78 & @xmath92 & @xmath93 & @xmath94 + j092345.59 + 302805.0 & 9965 & 1.03 & 0.27 & @xmath95 & @xmath96 & @xmath97 + ccrrcc j0022@xmath980031 & 0.4914 @xmath67 0.0254 & 80.8 @xmath67 1.3 & @xmath720.3 @xmath67 0.8 & 732.71818 & 0.02681 @xmath67 0.00190 + j0022@xmath71014 & 0.0799 @xmath67 0.0030 & 145.6 @xmath67 5.6 & @xmath738.5 @xmath67 3.7 & 732.76530 & 0.02553 @xmath67 0.00310 + j1234@xmath70228 & 0.0914 @xmath67 0.0040 & 94.0 @xmath67 2.3 & @xmath9850.3 @xmath67 1.8 & 917.75366 & 0.00787 @xmath67 0.00067 + j1625@xmath983632 & 0.2324 @xmath67 0.0396 & 58.4 @xmath67 2.7 & @xmath795.0 @xmath67 2.1 & 922.74700 & 0.00480 @xmath67 0.00105 + j2252@xmath70056 & & & @xmath723.2 @xmath67 0.8 & & + j2345@xmath70102 & & & @xmath7161.2 @xmath67 0.4 & & ccccccccc j0022@xmath980031 & 17890 @xmath67 110 & 7.38 @xmath67 0.02 & 0.38 & @xmath99 & 0.26 & 2.5% & 1.9% & @xmath100 + j0022@xmath71014 & 18980 @xmath67 380 & 7.15 @xmath67 0.04 & 0.33 & @xmath101 & 0.23 & 2.3% & 1.4% & @xmath102 + j1234@xmath70228 & 18000 @xmath67 170 & 6.64 @xmath67 0.03 & 0.23 & @xmath103 & 0.11 & 0.9% & 0.3% & @xmath104 + j1625@xmath983632 & 23570 @xmath67 440 & 6.12 @xmath67 0.03 & 0.20 & @xmath105 & 0.08 & 0.6% & 0.2% & @xmath106 + j2252@xmath70056 & 19450 @xmath67 270 & 7.00 @xmath67 0.02 & 0.31 & & & & & + j2345@xmath70102 & 33130 @xmath67 450 & 7.20 @xmath67 0.04 & 0.42 & & & & & + crccrcccc j0022@xmath71014 & 18980 & 7.15 & 0.07989 & 145.6 & 0.33 & @xmath101 & 0.23 & @xmath102 + j0755 + 4906 & 13160 & 5.84 & 0.06302 & 438.0 & 0.17 & @xmath107 & 1.12 & @xmath108 + j0818 + 3536 & 10620 & 5.69 & 0.18315 & 170.0 & 0.17 & @xmath109 & 0.33 & @xmath110 + j0822 + 2753 & 8880 & 6.44 & 0.24400 & 271.1 & 0.17 & @xmath111 & 1.05 & @xmath112 + j0849 + 0445 & 10290 & 6.23 & 0.07870 & 366.9 & 0.17 & @xmath113 & 0.88 & @xmath114 + j0923 + 3028 & 18350 & 6.63 & 0.04495 & 296.0 & 0.23 & @xmath115 & 0.44 & @xmath116 + j1053 + 5200 & 15180 & 6.55 & 0.04256 & 264.0 & 0.20 & @xmath109 & 0.33 & @xmath117 + j1233 + 1602 & 10920 & 5.12 & 0.15090 & 336.0 & 0.17 & @xmath118 & 1.20 & @xmath119 + j1234@xmath70228 & 18000 & 6.64 & 0.09143 & 94.0 & 0.23 & @xmath103 & 0.11 & @xmath104 + j1436 + 5010 & 16550 & 6.69 & 0.04580 & 347.4 & 0.24 & @xmath120 & 0.60 & @xmath121 + j2119@xmath70018 & 10360 & 5.36 & 0.08677 & 383.0 & 0.17 & @xmath122 & 1.04 & @xmath123 + nltt 11748 & 8690 & 6.54 & 0.23503 & 273.4 & 0.18 & 0.76 & & 7.20 + crr j0022 + 0031 & 732.722684 & @xmath772.2 @xmath67 30.3 + & 732.858519 & 22.4 @xmath67 18.8 + & 732.939955 & 21.2 @xmath67 36.0 + & 732.957466 & @xmath719.3 @xmath67 50.4 + & 733.728259 & @xmath745.4 @xmath67 19.5 + & 733.850239 & 15.3 @xmath67 11.6 + & 733.951397 & @xmath7102.0 @xmath67 47.9 + & 1094.846527 & @xmath712.4 @xmath67 12.4 + & 1100.810977 & 59.7 @xmath67 15.6 + & 1100.827123 & 39.7 @xmath67 10.6 + & 1100.841278 & 45.5 @xmath67 26.9 + & 1100.858385 & 80.6 @xmath67 14.8 + & 1100.872447 & 74.6 @xmath67 19.8 + & 1100.890352 & 93.5 @xmath67 15.6 + & 1100.904427 & 31.7 @xmath67 16.6 + & 1100.919566 & 51.9 @xmath67 23.7 + & 1101.895459 & 19.9 @xmath67 22.9 + & 1101.906050 & 36.9 @xmath67 21.6 + & 1101.917786 & 17.5 @xmath67 21.0 + & 1102.904928 & 38.2 @xmath67 13.9 + & 1102.915530 & 33.1 @xmath67 26.3 + & 1384.944307 & 100.6 @xmath67 29.7 + & 1385.929157 & @xmath775.1 @xmath67 40.6 + & 1387.925074 & @xmath714.8 @xmath67 72.2 + & 1395.943892 & @xmath796.8 @xmath67 19.8 + & 1396.936261 & @xmath787.0 @xmath67 17.1 + & 1396.976913 & @xmath711.9 @xmath67 49.8 + j0022@xmath71014 & 732.748459 & @xmath7156.9 @xmath67 27.6 + & 732.759199 & @xmath759.1 @xmath67 41.4 + & 732.832753 & @xmath7232.5 @xmath67 32.4 + & 732.843424 & @xmath777.3 @xmath67 32.2 + & 732.913470 & @xmath7199.3 @xmath67 24.1 + & 732.924142 & 13.5 @xmath67 20.7 + & 733.781989 & @xmath7198.3 @xmath67 44.7 + & 733.872718 & @xmath7125.5 @xmath67 47.9 + & 1094.862244 & 25.5 @xmath67 14.1 + & 1102.756314 & 89.0 @xmath67 31.1 + & 1102.770423 & 67.9 @xmath67 31.6 + & 1102.787923 & @xmath7148.8 @xmath67 20.6 + & 1102.801997 & @xmath7214.7 @xmath67 22.9 + & 1102.818895 & 33.5 @xmath67 18.9 + & 1102.832969 & 59.5 @xmath67 15.6 + & 1102.848212 & 46.9 @xmath67 18.2 + & 1102.862274 & @xmath7121.9 @xmath67 19.0 + & 1102.876371 & @xmath7148.9 @xmath67 33.8 + j1224@xmath70228 & 917.805724 & @xmath71.6 @xmath67 6.2 + & 922.728604 & 98.1 @xmath67 6.4 + & 922.815478 & 111.6 @xmath67 5.8 + & 922.904031 & 132.2 @xmath67 14.9 + & 923.754131 & @xmath716.3 @xmath67 5.2 + & 923.867625 & @xmath722.5 @xmath67 4.6 + & 923.931236 & 6.6 @xmath67 7.9 + & 924.850731 & 52.6 @xmath67 13.4 + & 924.857965 & @xmath761.9 @xmath67 17.8 + & 1275.829077 & 136.3 @xmath67 10.1 + & 1275.834841 & 73.9 @xmath67 11.9 + & 1275.841936 & 17.8 @xmath67 10.5 + & 1275.847700 & @xmath720.6 @xmath67 7.8 + & 1275.853441 & @xmath715.4 @xmath67 9.3 + & 1275.860270 & @xmath713.8 @xmath67 11.7 + & 1275.866022 & @xmath721.6 @xmath67 11.9 + & 1275.871763 & @xmath727.3 @xmath67 8.4 + & 1275.919252 & 146.8 @xmath67 11.0 + & 1275.925560 & 103.6 @xmath67 12.5 + j1625 + 3632 & 922.872132 & @xmath790.0 @xmath67 12.8 + & 922.957182 & @xmath7146.1 @xmath67 9.7 + & 923.002704 & @xmath739.0 @xmath67 9.6 + & 923.837051 & @xmath7160.2 @xmath67 10.4 + & 923.912980 & @xmath790.8 @xmath67 8.4 + & 923.956569 & @xmath754.9 @xmath67 7.6 + & 924.004812 & @xmath773.7 @xmath67 16.0 + & 925.000912 & @xmath7134.7 @xmath67 11.0 + & 925.012835 & @xmath7140.7 @xmath67 21.9 + & 1276.936718 & @xmath712.2 @xmath67 14.2 + & 1276.943038 & @xmath725.0 @xmath67 17.5 + & 1276.951372 & @xmath711.3 @xmath67 17.6 + & 1276.957668 & @xmath757.6 @xmath67 17.7 + & 1276.964011 & @xmath760.2 @xmath67 13.3 + & 1276.972808 & @xmath77.3 @xmath67 24.7 + & 1276.979128 & @xmath734.2 @xmath67 27.0 + & 1276.985447 & @xmath750.1 @xmath67 26.9 + & 1384.783927 & @xmath751.8 @xmath67 20.6 + & 1385.738969 & @xmath759.6 @xmath67 11.9 + & 1396.823036 & @xmath723.7 @xmath67 42.7 + j2252@xmath70056 & 732.661721 & @xmath733.0 @xmath67 8.4 + & 732.706500 & @xmath752.3 @xmath67 13.5 + & 732.786660 & @xmath740.0 @xmath67 11.8 + & 732.872943 & @xmath717.0 @xmath67 16.2 + & 733.653315 & @xmath724.9 @xmath67 16.9 + & 733.752698 & 4.8 @xmath67 13.1 + & 733.761692 & @xmath739.8 @xmath67 13.9 + & 733.829687 & @xmath718.2 @xmath67 18.7 + & 733.892232 & @xmath711.5 @xmath67 11.7 + & 1094.801506 & @xmath722.9 @xmath67 7.3 + & 1100.666435 & @xmath718.0 @xmath67 16.4 + & 1100.681967 & 4.7 @xmath67 9.2 + & 1100.696261 & @xmath714.5 @xmath67 8.9 + & 1100.712232 & 34.4 @xmath67 8.3 + & 1100.726399 & @xmath712.9 @xmath67 9.5 + & 1100.742289 & 1.4 @xmath67 8.7 + & 1100.756386 & 26.4 @xmath67 10.8 + & 1100.770645 & @xmath729.4 @xmath67 15.4 + & 1100.781270 & @xmath729.9 @xmath67 11.5 + & 1101.650617 & @xmath735.3 @xmath67 15.1 + & 1101.659471 & 20.6 @xmath67 13.9 + & 1101.672399 & @xmath7118.1 @xmath67 16.3 + & 1101.682989 & 27.1 @xmath67 16.7 + & 1101.698706 & @xmath735.8 @xmath67 13.6 + & 1101.709307 & @xmath72.8 @xmath67 16.8 + & 1101.719897 & @xmath733.6 @xmath67 17.3 + & 1102.645214 & @xmath786.1 @xmath67 19.9 + & 1102.655804 & 28.9 @xmath67 20.7 + & 1102.667655 & @xmath765.7 @xmath67 23.2 + & 1102.678245 & @xmath763.8 @xmath67 15.3 + j2345@xmath70102 & 732.689649 & @xmath7219.3 @xmath67 27.1 + & 732.799579 & @xmath7130.2 @xmath67 38.3 + & 732.810239 & @xmath794.4 @xmath67 35.0 + & 732.889312 & @xmath7114.6 @xmath67 30.3 + & 733.697952 & @xmath7129.3 @xmath67 31.9 + & 733.711030 & @xmath7159.5 @xmath67 53.8 + & 733.809108 & @xmath7142.7 @xmath67 24.4 + & 733.928533 & @xmath7153.0 @xmath67 47.3 + & 1094.830335 & @xmath7175.5 @xmath67 28.8 + & 1101.735791 & @xmath7139.1 @xmath67 59.8 + & 1101.751161 & @xmath7242.8 @xmath67 64.4 + & 1101.765224 & @xmath7243.8 @xmath67 41.6 + & 1101.786080 & @xmath7109.0 @xmath67 31.3 + & 1101.800154 & @xmath797.8 @xmath67 23.5 + & 1101.816149 & @xmath7199.4 @xmath67 40.1 + & 1101.830211 & @xmath7138.7 @xmath67 25.3 + & 1101.846137 & @xmath7109.9 @xmath67 26.4 + & 1101.860199 & @xmath7138.8 @xmath67 22.3 + & 1102.701680 & @xmath7136.8 @xmath67 29.2 + & 1102.719307 & @xmath795.9 @xmath67 65.7 + & 1102.736402 & @xmath7135.1 @xmath67 46.7 +

a geometry - based density functional theory is presented for mixtures of hard spheres , hard needles and hard platelets ; both the needles and the platelets are taken to be of vanishing thickness . geometrical weight functions that are characteristic for each species are given and it is shown how convolutions of pairs of weight functions recover each mayer bond of the ternary mixture and hence ensure the correct second virial expansion of the excess free energy functional . the case of sphere - platelet overlap relies on the same approximation as does rosenfeld s functional for strictly two - dimensional hard disks . we explicitly control contributions to the excess free energy that are of third order in density . analytic expressions relevant for the application of the theory to states with planar translational and cylindrical rotational symmetry , e.g. to describe behavior at planar smooth walls , are given . for binary sphere - platelet mixtures , in the appropriate limit of small platelet densities , the theory differs from that used in a recent treatment [ l. harnau and s. dietrich , phys . rev . e * 71 * , 011504 ( 2004 ) ] . as a test case of our approach we consider the isotropic - nematic bulk transition of pure hard platelets , which we find to be weakly first order , with values for the coexistence densities and the nematic order parameter that compare well with simulation results .

cond-mat0511472

dispersions of non - spherical colloidal particles are model systems to study a broad range of phenomena in condensed matter ranging from fluid phase separation to the emergence of liquid crystalline ordering . examples of such systems are clay suspensions @xcite , dispersed gibbsite platelets @xcite , mixtures of silica spheres and silica - coated boehmite rods @xcite , wax disks @xcite , or nonaqueous suspensions of laponite and montmorillonite @xcite . while already pure systems often possess complex liquid crystal phase behavior , binary mixtures additionally may demix into bulk phases with different chemical compositions , e.g. like in dispersions of disks and spheres @xcite . such mixtures are often viewed as composed of a primary species ( here the spheres ) that interact with an effective depletion potential that is generated by the second component referred to as the depletion agent ( here the disks ) . the depletion interaction is primarily attractive with a range of attraction similar to the size of the depletant agent and with a strength that is ruled by the concentration of the depletion agent . under appropriate conditions the depletion interaction may be sufficiently strong in order to drive a phase transition that is anaologous to the gas - liquid phase separation in simple substances , and that manifests fluid - fluid demixing when regarding the full mixture . recent theoretical work has been devoted to fluids of platelike particles near a hard wall @xcite , an interaction site model for lamellar colloids @xcite , wetting and capillary nematization of binary hard - platelet and hard - rod fluids @xcite , the lamellar zwanzig model @xcite , and colloidal hard - rod fluids near geometrically structured substrates @xcite . the phase diagram of mixtures of hard colloidal spheres and discs was obtained within a free - volume scaled particle approach @xcite . the depletion potential between two spheres immersed in a sea of platelets was studied in detail @xcite , and found to compare well with predictions from the derjaguin approximation @xcite . sedimentation was found to influence liquid crystal phase transitions of colloidal platelets @xcite , as well as multi - phase equilibria in mixtures of platelets and ideal polymer @xcite . the fundamental - measure theory ( fmt ) is an ( approximate ) density - functional theory ( dft ) @xcite , originally proposed by rosenfeld for additive hard sphere mixtures @xcite . an early extension to convex non - spherical particles has been given in refs . when applied to homogeneous and isotropic fluid states , this theory yields the correct second virial coefficient of the equation of state , but fails to recover the exact density functional up to second order in density . to remedy the latter problem , an interpolation between the hard sphere rosenfeld functional and the onsager functional for elongated rods was proposed @xcite and applied to the bulk isotropic - nematic transition . an fmt that originated from the treatment of parellel hard cubes @xcite was also used very successfully for the zwanzig model , see e.g.@xcite . see also refs.@xcite for the treatment of various hard body system mixtures . recently attention was paid to binary mixtures of hard rods and polymers @xcite . also notable is the treatment of hard body fluids based on two - point measures @xcite . the bolhuis - frenkel model of hard spheres and vanishingly thin hard needles @xcite can be considered as the simplest model hard core mixture of spheres and rods . previous work was concerned with the formulation of a dft for this model @xcite , and an extension to include rod - rod interaction on the onsager ( second virial ) level @xcite . predictions for the orientation ordering of the rods at a free interface between ( isotropic ) sphere - rich and sphere - poor phases @xcite were successfully confirmed by simulations @xcite . adding a third component to this system , the fluid demixing phase behavior of ternary mixtures of spheres , rods , and model polymers was found to be rich @xcite . quenching one of the components led to investigations of hard spheres immersed in random rod networks @xcite , and of the isotropic - nematic transition of rods in matrices of immobilized spheres @xcite . the present paper is concerned with the introduction of hard platelets to the fmt framework . central to fmt is the so - called deconvolution of the mayer bond @xmath0 between particles of species @xmath1 and @xmath2 into `` weight functions '' @xmath3 that are characteristic of the shape of particles of each species @xmath1 ; the label @xmath4 indicates the type of weight function . recall that the mayer bond is related to the pair interaction potential @xmath5 between species @xmath1 and @xmath2 via @xmath6 , where @xmath7 , @xmath8 is the boltzmann constant and @xmath9 is temperature . for hard bodies @xmath10 if the two particles overlap and @xmath11 otherwise . in the fmt framework a convolution integral is used to express @xmath0 as a sum of terms each of which is bilinear in the weight functions , i.e. possesses the structure @xmath12 , where @xmath13 label the species , @xmath14 label the type of weight function , and the asterisk indicates the ( three - dimensional ) convolution . the weight functions are characteristic of the geometry of single particles , i.e. vanish ( already ) beyond the physical extent of _ one _ particle . note that this is different from the behavior of @xmath0 , which extends to larger separations and vanishes provided _ two _ particles do not overlap . however the convolution of pairs of ( shorter ranged ) weight functions restores the correct ( longer ranged ) non - locality of the mayer bond . this concept was originally introduced for hard sphere mixtures @xcite . an extension to arbitrarily shaped convex bodies was proposed in refs.@xcite , which yields the correct second virial coefficients , but gives only an approximation for the mayer bonds . this is insufficient to describe e.g. nematic ordering ( see e.g. ref . @xcite for a discussion ) . however , obtaining exact deconvolutions is possible in principle , as the cases of the mayer bond between a sphere and a vanishingly thin needle @xcite and between a pair of thin rods @xcite demonstrate . to give an overview of the general structure of fmt , the weight functions @xmath15 are used to to build so - called weighted densities via convolution with the bare one - body density distribution ( of each species in the case of a mixture ) . the weighted densities are then input as arguments to an ( analytically given ) excess free energy density , @xmath16 . integrating @xmath16 over space ( and director space in case of anisotropic particles ) yields the excess free energy of the system , which is the only unknown contribution to the grand potential functional being the central quantity in ( classical ) dft . for any external potential ( like describing gravity or a wall ) minimizing the grand potential with respect to the one - body density distribution(s ) gives both the value of the grand potential and the structural information contained in the density profile(s ) . here we show how to treat platelet - shaped particles and mixtures with both spheres and needles within the same framework . the relevant weight functions are given and it is shown explicitly how all mayer bonds for the ternary mixture are obtained through convolutions . for two - dimensional hard disks the rosenfeld functional only yields an approximation for the exact mayer bond . despite this deficieny the two - dimensional rosenfeld functional is considered to be a useful tool to study inhomogeneous situations , see e.g. ref . @xcite for an investigation of laser - induced freezing and melting of confined colloidal particles . the treatment of the sphere - platelet case inevitably leads to the same defect . we do achieve , however , exact deconvolutions of the mayer bonds for platelet - platelet and platelet - needle interactions . furthermore we treat explicitly contributions to the excess free energy functional that are of third order in densities . while the treatment of second order terms was a crucial input for the original construction of fmt for hard spheres @xcite , this route has been only recently pursued in detail @xcite , and the authors conclude that they come `` close to the edge of fmt '' in attempting to incorporate properties of the exact expansion of the free energy functional in density on the third virial level into the approximate framework of fmt . in particular these delicate cases appear when particle touch , i.e. configurations that are close to the edge of the particles . hence we approximate the exact third virial level ( the triangle in the diagrammatic mayer @xmath0-expansion ) by terms that are non - vanishing only for cases with common triple intersection of the three particles involved . global prefactors are used to compensate for the `` lost cases '' @xcite in order to yield reasonable third virial coefficients . as a test case for the accuracy of the theory we consider the pure system of hard platelets . this is known to undergo a weak first order isotropic - nematic phase transition @xcite . in contrast to the application to thin rods , the onsager ( second virial ) functional for platelets yields only a qualitatively correct account of the phase transition : the transition density , density jump , and value of the nematic order parameter in the coexisting nematic are significantly overestimated . our theory has the same ( exact ) second order contribution to the free energy , but also sports a further third - order term . higher than third - order terms are absent due to the scaled - particle roots @xcite of the approach ; the vanishing thickness and hence vanishing volume of the platelets effectively truncates the series . our results show that fmt considerably improves on the onsager treatment , and yields a quantitatively correct picture , albeit with slightly too small transitition densities . harnau and dietrich recently considered bulk and wetting phenomena in a binary mixture of colloidal hard spheres and hard platelets @xcite , an interesting limiting case of the present ternary mixture . we postpone a detailed discussion of the relationship of the current work to ref . @xcite to sec . [ secplateconclusions ] . the paper is organized as follows . in sec . [ secplatemodel ] we define the model of a ternary hard body mixture of spheres , needles , and platelets . in sec . [ secdeconvolution ] the mayer bonds are represented as convolutions of ( weight ) functions characteristic for single particles . in sec . [ sectriangle ] we proceed to control the third virial level . readers who are interested in the structure of the resulting theory rather than its derivation might wish to skip secs.[secdeconvolution ] and [ sectriangle ] and directly go to sec.[secweighteddensities ] where the weighted densities are defined as convolutions of the weight functions with the bare density profiles . [ secdft ] presents the excess free energy functional . explicit expressions for the relevant quantities in planar and uniaxial geometry are given in sec.[secplanargeometry ] . we apply the theory to the isotropic - nematic transition of hard platelets in sec . [ secresults ] , and conclude in sec . [ secplateconclusions ] . we consider a mixture of hard spheres ( species @xmath17 ) with diameter @xmath18 , where @xmath19 is the radius , hard platelets ( species @xmath20 ) of diameter @xmath21 and vanishing thickness , and hard needle - like rods ( species @xmath22 ) of length @xmath23 and vanishing thickness . the pair interaction potential @xmath5 between any two particles of species @xmath24 is infinite if their geometrical shapes overlap and zero otherwise . the one - body density distribution of species @xmath25 is denoted by @xmath26 , @xmath27 , and @xmath28 , respectively , where @xmath29 is the position of the particle center , and @xmath30 is a unit vector pointing along ( normal to ) the shape of the needle ( platelet ) describing the particle orientation in space . for completeness we first summarize results from the literature for hard spheres and their mixtures with needles . rosenfeld s hard sphere weight functions @xcite are @xmath31 where @xmath32 is the unit step ( heaviside ) function and @xmath33 is the dirac distribution . here and in the following the @xmath34 are quantities with the dimension of @xmath35 ; the subscript @xmath36 indicates a vectorial quantity . further , linearly dependent , weight functions are @xmath37 , @xmath38 , and @xmath39 . for pure hard spheres the mayer bond is obtained through @xmath40 where the ( three - dimensional ) convolution is defined as @xmath41 and also implies a scalar product between vectors , as appears in the last term on the right hand side of eq . ( [ eqfss ] ) . a pair of vanishingly thin needles does not experience excluded volume interactions , but needles do interact with hard spheres . for such a binary mixture the needle weight functions used in ref.@xcite were obtained from the prescription of refs.@xcite , and are given by @xmath42 , \label{eqw0n}\\ w_1^{n}({{\bf r}},{{\bf \omega } } ) & = \frac{1}{4}\int_{-l/2}^{l/2 } { \mathrm{d}}l\,\delta({{\bf r}}- l{{\bf \omega } } ) . \label{eqw1n}\end{aligned}\ ] ] introducing a `` mixed '' weight function that is non - vanishing on the surface of a sphere , but carries a dependence on orientation , @xmath43 allows one to obtain the mayer bond between sphere and needle through convolutions , @xmath44 see ref . @xcite for the explicit calculation . for the case of residual rod - rod interactions in the onsager limit , @xmath45 can also be deconvolved into weight functions , see ref . @xcite for the details . in the following , however , we will restrict ourselves to the case of vanishingly thin needles . as a prerequisite for our subsequent treatment of platelets in three dimensions , we give an overview of rosenfeld s functional for the model of strictly two - dimensional disks . this constitues a prerequisite for considering the overlap between a hard sphere and a hard platelet in three dimensions ( sec . [ secmayersd ] below ) . here we deal with a multicomponent two - dimensional mixture of disks with radii @xmath46 of species @xmath1 , and characterized by the pair potential @xmath47 if @xmath48 and zero otherwise , where @xmath49 is the center - center distance between the disks of species @xmath1 and @xmath2 . ( when regarded from the viewpoint of the three - dimensional model , all disks possess the same orienation @xmath30 , perpendicular to the two - dimensional plane of position coordinates ; we will not delve into this delicate dimensional crossover in the following ) . rosenfeld s weight functions for this model are @xmath50 and there is an additional linearly dependent weight function , @xmath51 . the exact mayer bond @xmath0 between species @xmath1 and @xmath2 is then approximated through @xmath52 , with @xmath53 where here ( and only here ) @xmath54 denotes the two - dimensional convolution , @xmath55 and again the convolution implies a scalar product between vectors . as our subsequent treatment of the three - dimensional sphere - platelet overlap relies heavily on eq . ( [ eqmayertwod ] ) , we calculate the explicit functional dependence of @xmath56 on @xmath49 and discuss some of its properties . we choose two - dimensional cartesian coordinates @xmath29 and @xmath57 , such that @xmath58 and @xmath59 , which allows us to write the first term on the r.h.s . of eq . ( [ eqmayertwod ] ) as @xmath60 which is , by the cosine theorem , equal ( up to a factor of @xmath61 ) to the length of the arc which the rim of the disc @xmath1 traces across the interior of disc @xmath2 ( see fig . [ fig:2darc ] ) . by symmetry , the second term on the r.h.s . of eq . ( [ eqmayertwod ] ) gives eq.([eq:2d : arc ] ) with @xmath46 and @xmath62 interchanged . to calculate the remaining third term on the r.h.s . ( [ eqmayertwod ] ) , we use the same setup as above and obtain @xmath63 in the special case that both platelets have the same radius , @xmath64 , as relevant for describing the pure hard disk system , eq.([eq:2d : cusp ] ) further simplifies to @xmath65 we plot @xmath56 in fig . [ fig2d ] for three typical values of the size ratio @xmath66 . for small separations @xmath49 , such that the ( small ) disk @xmath1 lies completely inside the ( big ) disk @xmath2 , the result is exact , @xmath67 for @xmath68 . as @xmath49 increases , @xmath69 first decreases and then increases , exhibiting a divergence at contact , @xmath70 . beyond this threshold , for @xmath71 , @xmath72 is ( correctly ) indentically zero . note that the deviations around the exact value inside the core , @xmath10 , balance such that the second virial coefficient is exact , @xmath73 . and @xmath62 and center - center distance @xmath49 . shown are the @xmath74 and @xmath75 axis of the coordinate system used , the angle @xmath76 between the y - axis and @xmath46 , as well as half of the part ( bold ) of the circumference of disk @xmath1 that lies inside disk @xmath2 . ] , of two - dimensional hard disks of radii @xmath46 and @xmath62 , for size ratios @xmath77 ( from left to right ) as a function of the ( scaled ) center - to - center distance @xmath78 , where @xmath79 . in sec . [ secmayersd ] the same result is obtained for the mayer bond in three dimensions between a sphere of radius @xmath19 and a platelet of radius @xmath80 as a function of the ( scaled ) distance @xmath78 between the center of the platelet and the axis of the sphere ( parallel to the orientation @xmath30 of the platelet ) , where @xmath81 . the ( scaled ) distance between the center of the sphere and the plane of the platelet is @xmath82 ( right to left ) , and @xmath83 . the result for the mayer bond obtained is exact in cases ( not shown ) where the platelet i ) fully cuts through the sphere or ii ) lies completely inside the sphere ( provided @xmath84 ) . ] although the particle shapes appear to be equivalent , overlapping pair configurations are very different for freely rotating platelets in three dimensions as compared to the strictly two - dimensional case . note that in three dimensions we only need to consider cases of differing orientations of both platelets , @xmath85 , as configurations with strictly equal orientations , @xmath86 , carry vanishing statistical weight . the intersection volume differs markedly from the two - dimensional case ( area of intersection ) , requiring very different weight functions to express the mayer bond . this task can be accomplished exactly , in contrast to the two - dimensional case above . the direct application of rosenfeld s recipe @xcite to platelets yields the scalar weight functions @xmath87 corresponding to the surface , @xmath88 , integral mean curvature , @xmath89 , and euler characteristic , @xmath90 , of the platelets , respectively . the weight function that describes the particle volume vanishes , @xmath91 , due to the vanishing thickness , and hence vanishing volume of the platelets , @xmath92 . in contrast to rosenfeld s treatment @xcite , we here aim at an exact deconvolution of @xmath93 , where @xmath29 is the center - center distance vector between both particles and @xmath30 and @xmath94 are their orientations . we introduce @xmath95 where @xmath96 , given through eq . ( [ eqw1drosenfeld ] ) , describes the platelet rim , and @xmath94 is the orientation of the second platelet . keeping rosenfeld s surface weight function , @xmath97 as given in eq . ( [ eqw2drosenfeld ] ) , we recover the mayer bond between platelets via convolution , @xmath98 as the r.h.s . of eq . ( [ eqfdd ] ) consists of two symmetric terms , it is sufficient to consider the first one ; the second one gives an equivalent contribution to @xmath99 . without loss of generality we place the particles such that one platelet is located at the origin with its orientation vector pointing up ( into the positive @xmath100-direction ) , and place the center of the other platelet in the @xmath75@xmath100 plane ; see fig . [ fig : dd ] for an illustration . the chosen coordinates are : @xmath101 , @xmath102 , @xmath103 , @xmath104 , and the integration variable is @xmath105 . is located in the @xmath74@xmath75 plane with its center at the origin . platelet @xmath106 has its center in the @xmath75@xmath100 plane and an arbitrary orientation . the calculation checks for the intersection of the rim of @xmath20 and the interior of @xmath106 ( bold dot ) . the one - dimensional intersection volume of the two platelets is shown as a bold line . ] using ( [ eqw1drosenfeld ] ) and ( [ eqw1dd ] ) we insert the explicit form @xmath107 into the first term on the r.h.s . of ( [ eqfdd ] ) to obtain @xmath108\right|}\\ & \qquad\times{\delta\left({{\smash{r_{d}^{}}}}-\sqrt{r'^2+z'^2}\right)}{\delta\left(z'\right ) } { \,\theta\!\left({{\smash{r_{d}^{}}}}-|(r'\sin{\varphi}',r'\cos{\varphi}'-r , z'-z)|\right)}\\ & \qquad\times { \delta\left({\bigl(\begin{smallmatrix}r'\sin{\varphi}'\\r'\cos{\varphi}'-r\\z'-z\end{smallmatrix}\bigr)}\cdot { \bigl(\begin{smallmatrix}\sqrt{1-{\bar z}^2}\sin{\bar{\varphi}}\\ \sqrt{1-{\bar z}^2}\cos{\bar{\varphi}}\\ { \bar z}\end{smallmatrix}\bigr)}\right ) } \end{split}\label{eq : mayer : dd : simpleint}\\ \begin{split } & = \frac{{{\smash{r_{d}^{}}}}}{2}\int_0^{2\pi}{\mathrm{d}}{\varphi}'{\left| { \bigl(\begin{smallmatrix}\sin{\varphi}'\\ \cos{\varphi}'\\ 0\end{smallmatrix}\bigr ) } \cdot{\bigl(\begin{smallmatrix}-\sqrt{1-{\bar z}^2}\cos{\bar{\varphi}}\\ \sqrt{1-{\bar z}^2}\sin{\bar{\varphi}}\\0\end{smallmatrix}\bigr)}\right|}\\ & \qquad\times{\,\theta\!\left({{\smash{r_{d}^{}}}}-|({{\smash{r_{d}^{}}}}\sin{\varphi}',{{\smash{r_{d}^{}}}}\cos{\varphi}'-r ,- z)|\right)}\\ & \qquad\times{\delta\left({\bigl(\begin{smallmatrix}{{\smash{r_{d}^{}}}}\sin{\varphi}'\\ { { \smash{r_{d}^{}}}}\cos{\varphi}'-r\\-z\end{smallmatrix}\bigr ) } \cdot{\bigl(\begin{smallmatrix}\sqrt{1-{\bar z}^2}\sin{\bar{\varphi}}\\ \sqrt{1-{\bar z}^2}\cos{\bar{\varphi}}\\ { \bar z}\end{smallmatrix}\bigr)}\right ) } \end{split}\label{eq : mayer : dd : manyphi}\\ \begin{split } & = \frac{{{\smash{r_{d}^{}}}}}{2}\int_0^{2\pi}{\mathrm{d}}{\varphi}'{\left|\sqrt{1-{\bar z}^2}\sin({\bar{\varphi}}-{\varphi}')\right| } { \,\theta\!\left(2r{{\smash{r_{d}^{}}}}\cos{\varphi}'-r^2-z^2\right)}\\ & \qquad\times { \delta\left ( r\sqrt{1-{\bar z}^2}\cos{\bar{\varphi}}-{{\smash{r_{d}^{}}}}\sqrt{1-{\bar z}^2}\cos({\varphi}'-{\bar{\varphi } } ) + z{\bar z}\right ) } \end{split}\label{eq : mayer : dd : singlephi}\\ \begin{split } & = \frac{1}{2}\sum_\pm{\,\theta\!\left(2r{{\smash{r_{d}^{}}}}\cos\left({\bar{\varphi}}\pm\arccos \frac{r\sqrt{1-{\bar z}^2}\cos{\bar{\varphi}}+z{\bar z}}{{{\smash{r_{d}^{}}}}\sqrt{1-{\bar z}^2}}\right)-r^2-z^2\right)}\\ & \qquad\times{\,\theta\!\left({{\smash{r_{d}^{}}}}\sqrt{1-{\bar z}^2}-{\left|r\sqrt{1-{\bar z}^2}\cos{\bar{\varphi}}+z{\bar z}\right|}\right ) } , \end{split}\label{eq : dd : result}\end{aligned}\ ] ] where the integrations over @xmath109 and @xmath110 in eq . ( [ eq : mayer : dd : simpleint ] ) are straightforward . from ( [ eq : mayer : dd : manyphi ] ) to ( [ eq : mayer : dd : singlephi ] ) we rewrote the argument of the dirac delta distribution such that @xmath111 appears only once , using the identity @xmath112 . given the complexity of the overlap condition between two arbitrarily oriented platelets in three dimensions , it is not surprising that eq . ( [ eq : dd : result ] ) is an involved expression . it can be viewed as counting the number of intersections between the platelet @xmath106 and the rim of platelet @xmath20 . and @xmath106 to refer both to the platelets and the points of their origins . consider the triangle @xmath113 between the center of platelet @xmath20 , the intersection @xmath114 of platelet @xmath106 with the rim of @xmath20 , and the center of @xmath106 projected onto the @xmath74@xmath75 plane ( call this point @xmath115 ) . this triangle has by construction two known sides , @xmath116 and @xmath49 . the length of the third side of @xmath113 can be calculated by the cosine theorem from the angle @xmath117 . using the pythagorean theorem on the triangle between @xmath114 and the real and projected centers of @xmath106 , the length @xmath118 can be calculated and compared to the radius of @xmath106 , @xmath116 . this is accomplished by the first step function . in order to calculate the angle needed for the cosine theorem , consider @xmath113 and the right - angled triangle @xmath119 that is obtained by continuing @xmath120 over @xmath114 to form a right angle at the new point @xmath121 . then , @xmath122 and @xmath123 . therefore , the @xmath124 equals the angle @xmath125 , and we obtain the angle @xmath117 by adding @xmath126 . the second step function in eq . ( [ eq : dd : result ] ) checks if the plane defined by @xmath106 intersects with the rim of @xmath20 . ] the second term in eq . ( [ eqfdd ] ) counts the number of intersections between platelet @xmath20 and the rim of platelet @xmath106 . hence in total there are two ( zero ) intersections for an overlapping ( nonoverlapping ) configuration , and indeed the mayer bond between platelets , eq . ( [ eqfdd ] ) , is recovered through convolution of ( orientation - dependent ) single - particle measures . we next consider the platelet - needle mayer bond , @xmath127 , where @xmath29 is the difference vector from disk center to needle center , and @xmath128 and @xmath129 are the orientations of the disk and needle , respectively . we define a `` mixed '' weight function for the platelets that is non - vanishing on the platelet surface , but carries an additional dependence on the rod orientation , @xmath130 where @xmath97 , as given in eq . ( [ eqw2drosenfeld ] ) , describes the platelet surface . this allows us to obtain the mayer bond between platelet and needle via @xmath131 the validity of eq . ( [ eqfdn ] ) can be seen by employing cylindrical coordinates and putting the platelet on the @xmath75-axis , its orientation being along the @xmath100-axis . position the needle in such a way that its intersection with the @xmath74@xmath75 plane is the origin ( cf.fig . [ fig : dn ] ) . in such a way that it points at the origin . the platelet is located at @xmath132 . ] the needle is described by @xmath133 and @xmath134 . the coordinates of the platelet are @xmath135 and @xmath136 ; the integration variable is @xmath137 . from ( [ eqw2dn ] ) and ( [ eqw2drosenfeld ] ) we obtain the explicit expression @xmath138 , that we insert together with @xmath139 , as given in eq . ( [ eqw1n ] ) , into eq . ( [ eqfdn ] ) to obtain @xmath140 where the integrals over @xmath109 and @xmath141 are straightforward to calculate . in eq . , the integral over @xmath111 is performed first , rendering the @xmath142-integral trivial . the result , eq . ( [ eq : dn : result ] ) , is indeed the mayer bond between platelet and needle : due to the set - up , an overlap can occur only at the origin and the step functions in eq . ( [ eq : dn : result ] ) provide a means to assess whether or not this is the case . the first step function checks whether the origin is inside the platelet while the second step function checks whether the needle intersects the origin . treating the sphere - platelet mayer bond leads to the most involved geometry of pairwise particle overlap in the ternary mixture under consideration . it turns out that this case is intimately related to the mayer bond of the strictly two - dimensional hard disk mixtures . as detailed in sec . [ sectwod ] the fmt gives only an approximate representation in this case . still the rosenfeld functional for two - dimensional hard disks is a reasonably accurate theory , see e.g.@xcite for a recent study . here we deal with this problem on the same level of approximation as the two - dimensional rosenfeld case . note that cutting the sphere with the plane of the platelet yields a disk as the shape of intersection . we need to consider its overlap with a platelet , and introduce the weight functions @xmath143 where @xmath144 , as given in eq . ( [ eqw2s ] ) , describes the sphere surface , and @xmath96 , as given in eq.([eqw1drosenfeld ] ) , describes the platelet rim . eq.([eqwv2sd ] ) defines a vector field tied to the surface of the sphere ( with radius @xmath19 ) . in contrast to the `` radial hedgehog '' of the ( classic ) vector weight function @xmath145 , as given in eq . ( [ eqwv2s ] ) , the direction of the current vector field is radial with respect to the ( platelet ) @xmath30 direction and its magnitude decreases towards either pole a configuration one could refer to as a `` cylindrical hedgehog '' . note further that @xmath146 , in accordance with @xmath147 . the vector - field of eq . ( [ eqwv1d ] ) resembles a corona and is the straightforward generalization of the corresponding two - dimensional hard disk weight function , @xmath148 , given in eq . ( [ eqwv1twod ] ) . we use these functions to approximate @xmath149 with @xmath150 where again the convolution implies a scalar product between vectors . the three terms on the r.h.s . of eq . ( [ eqfsdapproximation ] ) represent the arc that the rim of the platelet traces inside the sphere , the arc that the sphere traces on the platelet ( as illustrated in figs . [ fig : sd : sphere ] and [ fig : sd : platelet ] , respectively ) , and an additional contribution of the cusps where the two arcs meet . we put the platelet into the @xmath74@xmath75 plane with its center at the origin . the center of the sphere is located in the @xmath75@xmath100 plane , see fig [ fig : sd : sphere ] ; the coordinates chosen are @xmath101 , @xmath102 , @xmath105 . is equal to the angle under which the arc shown in bold appears . ] is equal to the angle under which the arc shown in bold appears . ] we obtain for the first term on the r.h.s . ( [ eqfsdapproximation ] ) @xmath151 by the cosine theorem , this is equal ( up to a factor of @xmath61 ) to the length of the arc which the rim of the disc traces inside the sphere , see eq . . note also that @xmath152 is the squared radius of the circle that the platelet cuts out of the sphere . for the second term on the r.h.s . of eq . ( [ eqfsdapproximation ] ) we obtain @xmath153 which is indeed the length of the arc that the sphere traces on the platelet divided by @xmath61 , see eq . the relevant weight functions to treat the third and the fourth term on the r.h.s.of eq . ( [ eqfsdapproximation ] ) are explicitly @xmath154/(\pi r)$ ] and @xmath155 , hence @xmath156 \nonumber\\ \begin{split } & = \frac{1}{2\pi } \int_0^{2\pi}{\mathrm{d}}{\varphi}'{\int_0^\infty}{\mathrm{d}}r'\,r'{\int_{-\infty}^\infty}{\mathrm{d}}z ' { \delta\left(r-\sqrt{r'^2+z'^2}\right)}\\ & \qquad\times{\delta\left({{\smash{r_{d}^{}}}}-\sqrt{r^2+r'^2 - 2rr'\cos{\varphi}'+(z'-z)^2}\right)}\\ & \qquad\times{\delta\left(z'-z\right ) } \left(\frac{r'}{\sqrt{r'^2+z'^2 } } -\frac{{\bigl(\begin{smallmatrix}r'\sin{\varphi}'\\r'\cos{\varphi}'\\0\end{smallmatrix}\bigr ) } \cdot{\bigl(\begin{smallmatrix}r'\sin{\varphi}'\\ r'\cos{\varphi}'-r\\z'-z\end{smallmatrix}\bigr ) } } { { { \smash{r_{d}^{}}}}\sqrt{r'^2+z'^2}}\right ) \end{split}\\ \begin{split } & = \frac{1}{2\pi } \int_0^{2\pi}{\mathrm{d}}{\varphi}'{\int_0^\infty}\,r'\,{\mathrm{d}}r ' { \delta\left(r-\sqrt{r'^2+z^2}\right)}\\ & \qquad\times{\delta\left({{\smash{r_{d}^{}}}}-\sqrt{r^2+r'^2 - 2rr'\cos{\varphi}'}\right ) } \left(\frac{r'}{\sqrt{r'^2+z^2}}-\frac{r'^2-rr'\cos{\varphi } ' } { { { \smash{r_{d}^{}}}}\sqrt{r'^2+z^2}}\right ) \end{split}\\ \begin{split } & = \frac{r}{2\pi}\int_0^{2\pi}{\mathrm{d}}{\varphi } ' { \delta\left({{\smash{r_{d}^{}}}}-\sqrt{r^2+r^2-z^2 - 2r\sqrt{r^2-z^2}\cos{\varphi}'}\right)}\\ & \qquad\times \frac{\sqrt{r^2-z^2}}{r } \left(1-\frac{\sqrt{r^2-z^2}-r\cos{\varphi } ' } { { { \smash{r_{d}^{}}}}}\right ) { \,\theta\!\left(r-|z|\right ) } \end{split}\\ \begin{split } & = \frac{1}{\pi } { \,\theta\!\left(2r\sqrt{r^2-z^2}-|r^2+r^2-z^2-{\smash{r_{d}^{2}}}|\right ) } { \,\theta\!\left(r-|z|\right)}\\ & \qquad\times\frac{r^2-(\sqrt{r^2-z^2}-{{\smash{r_{d}^{}}}})^2 } { \sqrt{4r^2(r^2-z^2)-(r^2+r^2-z^2-{\smash{r_{d}^{2}}})^2 } } \end{split}\\ & = \frac{1}{\pi } { \,\theta\!\left({{\smash{r_{d}^{}}}}-|{\smash{r_{d}^{\prime}}}-r|\right ) } \frac{r^2-({\smash{r_{d}^{\prime}}}-{{\smash{r_{d}^{}}}})^2 } { \sqrt{4r^2{\smash{r_{d}^{\prime2}}}-(r^2+{\smash{r_{d}^{\prime2}}}-{\smash{r_{d}^{2}}})^2}}.\end{aligned}\ ] ] considering again that @xmath157 is the squared radius of the circle that the plane of the platelet cuts out of the sphere , we recover rosenfeld s approximation for the mayer bond of hard disks in two dimensions , see eq . . hence the same defect , as displayed in fig . [ fig2d ] for two - dimensional hard disks , is encountered for the overlap between sphere and platelet . based on the very reasonable quality of the two - dimensional rosenfeld functional , we expect this to do little harm in typical applications of the theory . the contribution of third order in densities to the excess free energy of a mixture has the generic exact form of a sum of triangle diagrams , @xmath158 where @xmath159 label the species , the sum runs over all species , @xmath160 is the mayer bond between particles of species @xmath1 and @xmath2 with coordinates @xmath161 and @xmath162 , respectively , and @xmath161 ( and its primed versions ) is a composite variable appropriate for the species under consideration , i.e. for the case of ( uniaxial ) rotators like platelets @xmath163 , @xmath164 , and for spheres simply @xmath165 , @xmath166 . even for one - component hard spheres fmt yields only an approximation to ( [ eqfexcthirdvirialmixture ] ) due to the occurrence of `` lost cases '' @xcite : the fmt approximation for the triple product of mayer bonds in ( [ eqfexcthirdvirialmixture ] ) is non - vanishing only for configurations @xmath29 , @xmath167 , and @xmath168 , such that three spheres with ( hard sphere ) radius @xmath19 centered at these positions share a non - vanishing common volume of intersection . note that the existence of such triple intersection implies pairwise intersections , but not vice versa : an illustrative example is that constituted by three touching spheres ; see e.g. @xcite for an in - depth discussion . the lost cases are compensated by overcounting the cases with triple intersection , such that a reasonable ( exact in the pure hard sphere case ) third virial coefficient for the bulk fluid results . the upshot is that new weight functions result , that can be , via weighted densities , incorporated to improve the accuracy of the density functional approximation . here we follow a similar strategy and aim at controlling the dependence on positions _ and _ orientations through appropriately constructed weight functions . for triplets of platelets in configurations with common triple intersection we can express @xmath169 where @xmath170 is a constant , @xmath97 is the surface weight function defined in eq . ( [ eqw2drosenfeld ] ) , and we introduce an additional surface weight that carries a dependence on _ two _ further ( platelet ) orientations , @xmath171 with normalization such that @xmath172 . setting @xmath173 yields equality in eq . ( [ eqtripleddd ] ) for particle configurations with common triple intersection of the three particles shapes . considering the bulk equation of state for pure platelets from spt , and in particular its ( approximative ) third virial coefficient , the compensation for the lost cases ( those without common triple intersection , but with pairwise intersections ) results in @xmath174 . in the following we will rather use the value from fmt , @xmath175 , which is numerically similar to the spt result . the advantage of the fmt scheme is that it consistently applies to mixtures , see ref.@xcite for a detailed discussion . for the platelet - sphere mixture the triangle diagram stemming from a pair of platelets and a single sphere , provided that there is a a common intersection of all three particle shapes , can be expressed as @xmath176 , \label{eqtriplesdd}\end{aligned}\ ] ] where @xmath177 is a constant , and we have introduced a sphere weight function that depends on two platelet orientations , @xmath178 with the normalization @xmath179 . again @xmath180 yields equality in eq.([eqtriplesdd ] ) provided that there is common triple intersection . we will use the fmt compensation , @xmath181 . the remaining case is that of triplets consisting of a single platelet and a pair of spheres . this case involves similar geometry as the platelet - sphere overlap , and hence we do not aim at an exact treatment . again in the spirit of the two - dimensional hard disk case ( cf . sec . [ sectwod ] ) , we use an approximation , @xmath182,\end{aligned}\ ] ] where @xmath183 generates the best possible approximation and @xmath184 compensates for the lost cases in the fmt approximation . we first summarize the novel weight functions that describe the platelet degrees of freedom in the ternary mixture . from the original rosenfeld prescription we have obtained [ see eqs.([eqw2drosenfeld])-([eqw0drosenfeld ] ) in sec.[secplateletplateletmayerbond ] ] @xmath185 in order to treat platelet - platelet and platelet - needle overlap we have introduced [ see eq . ( [ eqw1dd ] ) in sec.[secplateletplateletmayerbond ] and eq . ( [ eqw2dn ] ) in sec.[secplateletneedlemayerbond ] ] @xmath186 respectively . to treat sphere - platelet overlap we use [ see eq.([eqw2sd])-([eqwv1d ] ) in sec . [ secmayersd ] ] @xmath187 the consideration of the third order contributions to the excess free energy has led to [ see eqs . ( [ eqw2ddd ] ) and ( [ eqw2sdd ] ) in sec . [ sectriangle ] ] @xmath188 all weight functions are used to build weighted densities via convolutions with the bare density profiles . for hard spheres the scalar and vectorial weighted densities are respectively given by @xmath189 where @xmath190 and @xmath191 as given in and below eqs . ( [ eqw3s])-([eqwv2s ] ) are the classic rosenfeld hard sphere functions . adding non - spherical particles requires convolving the sphere density profile with orientation - dependent weight functions . the coupling to needles is accomplished via @xmath192 where @xmath193 as given in eq . ( [ eqw2sn ] ) is a mixed sphere - needle function . treating the platelets follows a similar outline , @xmath194 with @xmath195 and @xmath196 given in eqs.([eqw2sd ] ) and ( [ eqwv2sd ] ) , respectively , and @xmath197 as given in eq . ( [ eqw2sdd ] ) . for needles the pure weight functions are @xmath198 where @xmath199 and @xmath200 are given in eqs.([eqw0n ] ) and ( [ eqw1n ] ) , respectively . in case one considers needle - needle interactions , additional weighted densities need to be taken into account @xcite . the degrees of freedom of the platelets are modelled through the weighted densities @xmath201 with the weight functions @xmath97 , @xmath96 , and @xmath202 , as given in eqs . ( [ eqw2drosenfeld ] ) , ( [ eqw1drosenfeld ] ) , and ( [ eqw0drosenfeld ] ) , respectively , and @xmath203 as given in eq . ( [ eqwv1d ] ) . the double orientation - dependent weight function @xmath204 is given in eq . ( [ eqw1dd ] ) . the weight function that we have introduced to control the triangle diagrams , @xmath205 , is given in eq . ( [ eqw2ddd ] ) . the remaining coupling of the platelets to the needles is achieved via @xmath206 where we have used @xmath207 as given in eq . ( [ eqw2dn ] ) to build a mixed weighted density . in ref . @xcite the excess helmholtz free energy functional for particles with rotational degrees of freedom was expressed not only as a spatial integral ( as was proposed in refs.@xcite ) , but also as an integral over the director space . in our present explicit consideration of three particle correlations an additional integration over director space is required , @xmath208 where the ( reduced ) free energy density , @xmath16 , is a function of the set of weighted densities , @xmath209 , where @xmath1 labels the species and @xmath4 labels the type of weight function . for hard spheres @xmath210 with @xmath211 where we have suppressed here and in the following the dependence of the weighted densities on the space coordinate @xmath29 . eq.([eqphis ] ) is the original rosenfeld form @xcite that yields , for bulk fluids , the ( reduced ) free energy density as obtained by the percus - yevick compressibility ( or scaled - particle ) route . improved versions exist that feature exact dimensional crossover @xcite , and the carnahan - starling equation for bulk fluid states @xcite . for binary mixtures of spheres and needles the reduced free energy density is @xmath212 , where @xmath213 is given in ( [ eqphis ] ) , and @xmath214 describes the effect of needle - sphere interactions @xcite , given as @xmath215 where we suppress the dependence of the weighted densities on @xmath29 , but make the dependence on orientation explicit in the notation . the free energy for bulk fluid states obtained from ( [ eqphisn ] ) is the same as that obtained from a thermodynamic ( free - volume - like ) perturbation theory @xcite , that results in a fluid - fluid demixing binodal that compares well with results from simulations . for a system of pure platelets @xmath216 , with @xmath217 where the prefactor of the third order term in density is according to rosenfeld @xmath218 . scaled - particle theory @xcite gives a slightly different value , namely @xmath219 ; see ref.@xcite for a detailed comparison of the predictions of these approaches for bulk fluid properties . eq.([eqphid ] ) is the simplest form of a free energy density that features the correct second virial level and a similar free energy for bulk isotropic fluid states as scaled - particle theory . for binary sphere - platelet mixtures we obtain @xmath220 , where the hard sphere contribution @xmath221 is given in eq . ( [ eqphis ] ) , and @xmath222 for a binary platelet needle mixture the free energy density is @xmath223 , where the pure platelet contribution @xmath224 is given in eq . ( [ eqphid ] ) , and a simple second - order coupling between both species is introduced via @xmath225 due to the vanishing thickness of particles of both species , any higher order terms vanish according to spt . finally , for ternary sphere - platelet - needle mixtures the free energy density is @xmath226 where the first three terms on the r.h.s . are given through eqs.([eqphis ] ) , ( [ eqphisn ] ) and ( [ eqphisd ] ) , respectively . in many practical situations one is faced with inhomogeneities that depend only on a single ( cartesian ) coordinate , while being translationally invariant in the two remaining directions . a smooth planar wall , where the fluid density profile(s ) only depend on the perpendicular distance , @xmath100 , from the wall is a primary example . for rotators an additional simplification arises from ( cylindrical ) rotational symmetry around the @xmath100-axis . hence such problems are characterized solely by @xmath100 , and the tilt angle @xmath227 between the particle orientation and the @xmath100-axis . we give in the following explicit expressions for ( reduced ) weight functions appropriate for efficient numerical treatment of such situations . for completeness we also give the results for spheres and needles @xcite . let us start by defining the full set of reduced densities ; we use the parametrizations @xmath228 and @xmath229 . for the spheres @xmath230 further mixed weight functions couple spheres to needles , @xmath231 and spheres to platelets @xmath232 for the needles @xmath233 for the platelets the effective weight functions for planar geometry are obtained by integrating over the lateral coordinates @xmath234 and @xmath235 @xmath236 using the effective weight functions the weighted densities can then be written as @xmath237 @xmath238 where in eqs . - ( and only here ) @xmath54 denotes the one - dimensional convolution , @xmath239 . in contrast to the needle case the integrands depend non - trivially on @xmath240 . we finally give all results for the effective weight functions in planar geometry . for hard spheres the well - known results are @xmath241 where @xmath242 is the unit vector pointing along the @xmath100-axis . the linearly dependent weight functions are @xmath243 , @xmath244 , @xmath245 . the mixed sphere - needle weight function is given through @xmath246 further mixed sphere weight functions are @xmath247 ^ 2},\\ { { \bf w}_{{{v}2}}^{sd}}(z,\vartheta ) & = \frac{8\pi}{{{\smash{r_{d}^{}}}}}z\sin^2\vartheta\sqrt{{\smash{r_{d}^{2}}}-z^2}\theta(r-|z|){{\bf e}}_z,\\ w_2^{sdd}(z,\vartheta,\vartheta ' ) & = { \,\theta\!\left(r-|z|\right)}\int_0^{2\pi}{\mathrm{d}}{\varphi } '' \begin{cases } \frac{8}{\pi}|t| & \text{if } t^2 > s^2\\ \frac{16}{\pi^2}\bigl(\sqrt{s^2-t^2 } + t \arcsin\frac{t}{s}\bigr ) & \text{else , } \end{cases}\end{aligned}\ ] ] where @xmath248 is a unit vector in the @xmath100-direction , @xmath249 , and @xmath250 . for the needles the effective weight functions are @xmath251 carrying out the integrations in eq . ( [ eqwdiskeffective ] ) yields @xmath252 comparing eqs . ( [ eqw1dplanar ] ) and ( [ eqw0dplanar ] ) to the corresponding expressions for needles , eqs . ( [ eqw1nplanar ] ) and ( [ eqw0nplanar ] ) , reveals that the different particle geometries imprint marked differences in the functional dependence on @xmath100 and @xmath227 . clearly this can be traced back to the differences in the full weight functions for needles [ eqs . ( [ eqw0n ] ) and ( [ eqw1n ] ) ] and platelets [ eqs . ( [ eqw1drosenfeld ] ) and ( [ eqw0drosenfeld ] ) ] . note that for the present case of vanishing particle thicknesses , @xmath253 has no needle counterpart to compare with , as the needle surface vanishes . @xmath254 @xmath255 where @xmath256 . @xmath257 where @xmath258 ^ 2 \\ \qquad + \sin(\vartheta'+\vartheta)\sin(\vartheta'-\vartheta ) \sin^2\vartheta''\sin^2{\varphi } '' \biggr\}^{1/2 } \\ \quad + \sin\vartheta\sin\vartheta''\cos\vartheta'\sin{\varphi } '' \arcsin k(\vartheta,\vartheta',\vartheta'',{\varphi}'')\biggr ) & \text{else , } \end{cases}\ ] ] and @xmath259}}\ ] ] for the remaining mixed weight functions we obtain @xmath260 already the binary subsystems with platelets ( leaving aside the full ternary mixture ) , i.e. the sphere - platelet and rod - platelet mixtures , are expected to display rich bulk and interfacial properties due to the competition between depletion and orientational order . in order to find a simple yet demanding test case for the present theory , we restrict ourselves further to the pure system of platelets . in bulk this undergoes an isotropic - nematic phase transition , where @xmath261 in the isotropic phase , and @xmath262 in the nematic phase being peaked around the nematic director that indicates the preferred direction of alignment of the particles ; @xmath263 is the angle between @xmath264 and the nematic director . the strength of the nematic order is conveniently measured via the nematic order parameter , @xmath265 where @xmath266 is the second legendre polynomial . we use the generic form of the free energy , eq.([eqfexcgeneric ] ) , as a multiple integral over a free energy density appropriate for pure platelets , @xmath224 as given in eq.([eqphid ] ) . this function depends on the weighted densities @xmath267 , @xmath268 , and @xmath269 , as defined in eqs.([eqnnud ] ) ( with @xmath270 ) , ( [ eqn1dd ] ) , and ( [ eqn2ddd ] ) , respectively . the weighted densities in turn are built as convolutions of the bare platelet one - body density , @xmath271 , with the weight functions @xmath272 , @xmath273 , and @xmath274 , as given in ( [ eqw2drosenfeld ] ) , ( [ eqw1dd ] ) , and ( [ eqw2ddd ] ) [ and summarized in eqs . ( [ eqsumw2drosenfeld ] ) , ( [ eqsumw1dd ] ) , and ( [ eqsumw2ddd ] ) ] , respectively . to carry out the actual calculations , we have started from the corresponding expressions in planar geometry ( see sec . [ secplanargeometry ] ) , that we have further ( numerically ) reduced to @xmath100-independent quantities , as appropriate for the spatially homogeneous isotropic and nematic phases . alternatively , our fmt excess free energy functional for hard platelets can be viewed as being composed of the exact second - order virial contribution and a further ( approximate ) third order term that i ) vanishes for configurations without common triple intersection of three platelets , and ii ) is constant for configurations with common triple intersection ; the constant is adjusted to give a reasonable value for the third virial coefficient . note that the theory for the full ternary mixture has a similar structure , but features higher order ( in density ) contributions that arise from the finite volume , and hence finite packing fraction , of the hard spheres . we minimize the grand potential functional @xmath275 ) = f_{\rm id}[\rho_d ] + f_{\rm exc}[\rho_d ] -\mu_d \int d{{\bf r}}\frac{d{{\bf \omega}}}{4\pi } \rho_d({{\bf r}},{{\bf \omega } } ) , \label{eqgrandpotential}\ ] ] where the ideal gas contribution is given by @xmath276 = \int d{{\bf r}}\frac{d{{\bf \omega}}}{4\pi } \rho_d({{\bf r}},{{\bf \omega}})[\ln(\rho_d({{\bf r}},{{\bf \omega}})\lambda_d^3)-1],\ ] ] where @xmath277 is the chemical potential , the dependence on temperature and volume has been suppressed in the notation , and we formally set the thermal wavelength equal to the platelet radius , @xmath278 , to fix an arbitrary additive constant to the chemical potential . for a given value of the ( scaled ) chemical potential @xmath279 , we minimized @xmath280 with respect to @xmath281 in either phase . inserting the resulting distribution into eq.([eqgrandpotential ] ) yields the grand potential of the system , and the condition for equality of the grand potential in both phases locates the phase transition . the numerical implementation uses free minimization , i.e. no parametrized form of @xmath282 is assumed a priori . we use an equidistant grid in @xmath263-space with 100 grid points in the interval @xmath283 $ ] . the integration over @xmath284 in eq.([eqw2dddplanar ] ) to obtain @xmath274 was done with 200 grid points in @xmath285 $ ] ; this only needs to be performed once at the start of the calculation . we have used alternatively fixed - point iteration or simulated annealing to minimize the density profile ; results from both approaches were consistent . up to 1000 steps were sufficient to obtain convergence . .results for the bulk isotropic - nematic phase transition from different approaches . given are values for the ( scaled ) coexisting densities in the isotropic and nematic phase , @xmath286 and @xmath287 , respectively , the ( scaled ) density jump , @xmath288 , and the value of the nematic order parameter in the coexisting nematic phase , @xmath289 , as obtained from fmt , and the simulation results by frenkel and eppenga @xcite , as well as theoretical results obtained from the onsager second virial functional for hard platelets . [ cols="<,<,<,<",options="header " , ] the resulting values of the densities in the coexisting isotropic and nematic phases , @xmath290 and @xmath291 , respectively , as well as the value of the order parameter in the coexisting nematic phase , @xmath289 , are in good agreement with frenkel and eppenga s classic simulation results @xcite , see tab . [ tabtransitionin ] . the fmt values somewhat underestimate the coexistence densities and give a too high value for @xmath289 . the onsager second virial functional , on the other hand , overestimates significantly the coexistence densities as well as @xmath289 . hence the actual phase transition , as characterized by the jump in coexistence densities and in order parameter , is weaker than predicted by the onsager treatment . to further illustrate our findings , we plot in fig . [ fig : sofmu ] the variation of @xmath17 with @xmath292 and in fig . [ fig : sofrho ] the variation of @xmath17 with @xmath293 . we also plot the behavior of the metastable nematic branch , i.e. for statepoints where the nematic phase remains locally stable against small fluctuations , although globally the grand potential for the isotropic phase is the lower one . it is apparent that the value of @xmath289 very sensitively depends on the precise location of the phase transition . taking this into account we find the overall the agreement of the fmt results with simulation data to be very reasonable , and to give confidence both into applications to interfacial phenomena of platelets , as well as into the accuracy of the fmt for the full ternary mixture . for a bulk system of pure hard platelets as a function of the ( scaled ) chemical potential , @xmath292 . shown are results from fmt in the stable ( solid line ) and metastable ( dashed line ) nematic . values in the coexisting nematic and isotropic phases , @xmath289 and @xmath294 , are shown as obtained from fmt ( squares ) and the onsager - type second virial treatment ( dots ) . ] , but as a function of the ( scaled ) density of platelets @xmath295 . note the considerable reduction of the coexistence density gap obtained from fmt as compared to the onsager - like treatment . ] in conclusion we have derived a geometry - based density functional theory for hard body mixtures of spheres , platelets and needles . both the needles and the ( circular ) platelets possess vanishing thickness and hence constitute the simplest examples of prolate and oblate model particles , respectively . our treatment of the mixture is based on the so - called deconvolution of the mayer bond into single - particle functions which vanish beyond the extent of the particle . the mayer bonds are recovered upon convolution of the single - particle functions . the construction of the full functional relies further on controlling the third virial level and on rosenfeld s scaled - particle treatment . in order to facilitate future applications , like wetting of planar walls or capillary phenomena in planar slits , we have given explicit reduced expressions for the relevant quantities in planar geometry . in a recent contribution harnau and dietrich propose and apply a dft for binary platelet - sphere mixtures @xcite . they obtain a platelet - sphere functional by starting from the rod - sphere functional of refs . @xcite , and in particular from the explicit expressions for the needle weight functions in planar and uniaxial geometry modifying the definition of the relevant angle between the particle orientation and the @xmath100-axis the rod weight functions are taken to play the role of platelet weight functions . the resulting excess free energy functional is linear in the platelet density , limiting the theory to small densities of platelets . recall that for binary mixtures where one component ( the depletant ) is ideal , the absence of higher than linear order terms in the density distribution of this component is a good approximation . examples are the above rod - sphere mixture and the asakura - oosawa model of colloid - polymer mixtures , where the polymers are described as non - interacting spheres . while in these cases the pure depletant system is an ideal gas , pure platelets constitute an interacting system . hence higher than linear order terms in the platelet density should only be irrelevant at low platelet densities . the current theory is in accordance with scaled - particle theory @xcite and yields second and third - order contributions to the excess free energy , but higher order terms are absent . this behavior can be traced back to the vanishing volume of the platelets . in the present work we have shown explicitly how our dft for the sphere - platelet mixture reduces to the correct low - density limit . in contrast to the procedure in ref . @xcite , we have treated the full three - dimensional problem and have obtained , up to a defect already present for two - dimensional hard disks , the deconvolution of the sphere - platelet mayer bond and hence the appropriate platelet weight functions . projecting those to planar and uniaxial symmetry ( appropriate for fluid states at a planar smooth wall like investigated in ref . @xcite ) reveals that the expresssions differ markedly from those for sphere - rod mixtures . this might come as no surprise given the fact that the genuine shapes of the particles are one of the building blocks of the geometry - based dft . an immediate consequence is that the platelet weight functions , and hence the form of the excess free energy functional , also differ from the expressions used in ref.@xcite . our successful description of the bulk isotropic - nematic transition of hard platelets can be viewed as the first genuine fmt treatment of liquid crystalline ordering in a continuum model ( without any interpolation as e.g. in ref . possible future applications of our theory include capillary and wetting phenomena , influence of gravity or other external fields , and the study of free interfaces between demixed ( and possibly liquid crystalline ) phases . further testing the accuracy of the theory , i.e.for the predictions of the bulk direct correlation functions , see e.g. @xcite , against results from computer simulations is clearly desirable . professor siegfried dietrich and dr . ludger harnau are acknowledged for useful discussions . we thank dr . marjolein dijkstra and dr . ren van roij for stimulating exchange and for communicating their unpublished results to us . andrew archer is acknowledged for valuable comments on the manuscript . this work is supported by the sfb - tr6 `` colloidal dispersions in external fields '' of the german science foundation ( deutsche forschungsgemeinschaft ) .

the recent _ square root law _ ( srl ) for covert communication demonstrates that alice can reliably transmit bits to bob in uses of an additive white gaussian noise ( awgn ) channel while keeping ineffective any detector employed by the adversary ; conversely , exceeding this limit either results in detection by the adversary with high probability or non - zero decoding error probability at bob . this srl is under the assumption that the adversary knows _ when _ alice transmits ( if she transmits ) ; however , in many operational scenarios he does not know this . hence , here we study the impact of the adversary s ignorance of the time of the communication attempt . we employ a slotted awgn channel model with slots each containing symbol periods , where alice may use a single slot out of . provided that alice s slot selection is secret , the adversary needs to monitor all slots for possible transmission . we show that this allows alice to reliably transmit bits to bob ( but no more ) while keeping the adversary s detector ineffective . to achieve this gain over srl , bob does not have to know the time of transmission provided , .

1403.1013

recent revelations of massive surveillance programs @xcite have emphasized the necessity of secure communication systems that do not just protect the content of the user s message from being decoded , but prevent the detection of its transmission in the first place . indeed , encrypted data or even just the transmission of a signal can arouse suspicion , and even the most theoretically robust cryptographic security scheme can be defeated by a determined adversary using non - computational methods such as side - channel analysis . covert , or _ low probability of detection / intercept _ ( lpd / lpi ) communication capability is thus very important in extremely sensitive situations ( e.g. , during military operations or for organization of civil unrest ) . in the covert communication scenario , alice transmits a message to bob over a noisy channel while the adversary , warden willie , attempts to detect her transmission . the channel from alice to willie is also subject to noise . thus , while alice transmits low - power covert signals to bob , willie attempts to classify these signals as either noise on his channel or signals from alice . we recently showed that the _ square root law _ ( srl ) governs covert communication : provided that alice and bob pre - share a secret of sufficient length and alice can tolerate a small probability of being detected by willie , she can reliably transmit @xmath0 bits to bob in @xmath1 uses of an additive white gaussian noise ( awgn ) channel while keeping ineffective any detector employed by willie . conversely , attempting to transmit more that @xmath0 either results in detection by willie with probability one or non - zero decoding error probability at bob as @xmath6 @xcite . follow - on work has addressed the size of the pre - shared secret @xcite , the optimal constant hidden by the asymptotic ( big-@xmath7 ) notation @xcite , network aspects of covert communication @xcite , extensions of srl to bosonic channels where the adversary is only limited by the laws of quantum mechanics @xcite , and , finally , practical limitations on the adversary that improve the covert throughput to beyond the srl limit @xcite . here we explore another limitation on the adversary s capabilities . studies of covert communication up to now assume that willie knows when alice may start to transmit ( if she does ) . however , there are many practical scenarios where this assumption can be relaxed and alice s communication start time is unknown to willie . for example , a possible transmission time can be pre - arranged in advance . alice s message may also be much shorter than the total time available to transmit it ( e.g. , a few seconds out of the day when both alice and bob are available ) . thus , since willie does not know when alice may transmit , he has to monitor a much longer time period than the duration of alice s transmission . here we show how alice can leverage willie s ignorance of her transmission time to transmit significant additional information to bob . ( 85,13 ) ( 0,5.5)(14.8,4 ) ( 0,10)(14.8,5)@xmath8 ( 15,5.5)(14.85,4 ) ( 15,10)(14.8,5)@xmath8 ( 30,5.5)(12.35,4 ) ( 42.5,5.5)(14.85,4 ) ( 42.5,10)(14.8,5)@xmath8 ( 57.5,5.5)(12.3,4 ) ( 70,5.5)(14.8,4 ) ( 70,10)(15,5)@xmath8 ( 0,0)(85,5)@xmath9 ( 57.5,14)(-1,-2)3 ( 58,14)(20,3)slot used by alice and bob in our scenario , alice communicates to bob over an awgn channel . willie also has an awgn channel from alice . unlike the setting in @xcite , the channel is slotted , as described in figure [ fig : timing ] . each of @xmath2 slots contains @xmath1 symbol periods , where @xmath2 is an increasing function of @xmath1 . if alice used all @xmath10 symbol periods for transmission , then , by the square root law in @xcite , she could reliably transmit @xmath11 covert bits denotes an asymptotic upper bound on @xmath12 ( i.e. there exist constants @xmath13 such that @xmath14 for all @xmath15 ) , @xmath16 denotes an upper bound on @xmath12 that is not asymptotically tight ( i.e. for any constant @xmath17 , there exists constant @xmath18 such that @xmath19 for all @xmath15 ) , and @xmath20 denotes a lower bound on @xmath12 that is not asymptotically tight ( i.e. for any constant @xmath17 , there exists constant @xmath18 such that @xmath21 for all @xmath15 ) . ] to bob . however , alice uses only a single slot @xmath22 , which she keeps secret from willie , who is thus forced to monitor all @xmath2 slots . a nave application of the square root law @xcite allows alice to reliably transmit @xmath0 covert bits in this scenario . we demonstrate that alice can transmit @xmath23 bits reliably on this channel while maintaining arbitrarily low probability of detection by willie . conversely , we show that the transmission of @xmath24 bits either results in alice being detected with high probability or unreliable communication . the cost of covert communication on the awgn channel is the secret that alice and bob share before the transmission . remarkably , we demonstrate that the _ multiplicative _ increase ( by a factor of @xmath25 ) in the number of covert bits that alice can transmit reliably to bob does not require bob to know the timing of the transmission if @xmath4 , where @xmath26 is a constant ; to realize the @xmath25 gain when @xmath27 only an additive expense of an extra @xmath28 secret bits is needed to indicate to bob the slot employed by alice . timing is thus a very useful resource for covert communication . it also necessitates a vastly different analysis than that in @xcite . specifically , the relative entropy based bounds on the probability of detection error employed in @xcite are too loose to yield our achievability results , and we therefore have to apply other techniques from mathematical statistics . the main result of this paper is the following : suppose the channel between alice and each of bob and willie experiences independent additive white gaussian noise ( awgn ) with constant power @xmath29 and @xmath30 , respectively . also suppose that , if alice chooses to transmit , she uses one of the @xmath2 slots chosen randomly . each slot contains @xmath1 symbol periods , where @xmath31 . then , for any @xmath32 , alice can reliably transmit @xmath33 bits to bob in a selected slot while maintaining a probability of detection error by willie greater than @xmath34 . conversely , if alice tries to transmit @xmath35 bits using @xmath1 consecutive symbol periods , either willie detects with arbitrarily low probability of error or bob can not decode her message with arbitrary low probability of decoding error as @xmath6 . after introducing our slotted channel model in section [ sec : prerequisites ] , we prove the achievability and the converse in sections [ sec : achievability ] and [ sec : converse ] , respectively . we discuss the relationship of our paper to other work in covert communication in section [ sec : discussion ] and conclude in section [ sec : conclusion ] . ( 85,42 ) ( 33,38)(14,4 ) ( 7,40)(0,-1)3 ( 73,40)(0,-1)3 ( 7,40)(1,0)26 ( 47,40)(1,0)26 ( 0,31)(15,6)alice ( 15,34)(1,0)38.25 ( 16,26.5)(25,5)@xmath36 ( 17,21.5)(25,5)@xmath37 ( 47,34 ) ( 47,34)(0,-1)16.5 ( 47,15 ) ( 5,5 ) ( 0,0 ) ( 0,-2 ) ( 0,1 ) 4 ( -2,0)(1,0 ) 4 ( 49.5,15)(1,0)12.5 ( 62,12)(15,6)willie ( 45,2)(58,12 ) ( 32,15)(1,0)12.5 ( 25,12)(7,6)@xmath38 ( 52.25,21.5)(7,6)@xmath39 ( 55.75,34 ) ( 5,5 ) ( 0,0 ) ( 0,-2 ) ( 0,1 ) 4 ( -2,0)(1,0 ) 4 ( 55.75,27.5)(0,1)4 ( 58.25,34)(1,0)6.75 ( 65,31)(15,6)bob ( 63,25)(25,5)decode @xmath40 we use the discrete - time slotted awgn channel model with real - valued symbols depicted in figures [ fig : timing ] and [ fig : framework ] . the channel has @xmath2 slots , each containing @xmath1 symbol periods . alice selects slot @xmath22 uniformly at random prior to transmission . if alice chooses to transmit , she sends a vector of @xmath1 real - valued symbols @xmath41 during slot @xmath22 . the awgn on bob s channel is described by an independent and identically distributed ( i.i.d . ) sequence @xmath42 of @xmath10 zero - mean gaussian random variables with variance @xmath43 ( i.e. , @xmath44 ) . bob receives @xmath45 where @xmath46 $ ] is a vector of observations collected during slot @xmath47 . if alice transmits during slot @xmath22 , @xmath48 . for any slot that is not used for transmission , @xmath49 ( this includes all slots @xmath50 , and slot @xmath22 when alice does not transmit ) . similarly , willie observes @xmath51 where @xmath52 $ ] is a vector of observations collected during slot @xmath47 . the awgn on willie s channel is described by an i.i.d . sequence @xmath53 of @xmath10 zero - mean gaussian random variables with variance @xmath54 ( i.e. , @xmath55 ) . if alice transmits during slot @xmath22 , @xmath56 . for any slot that is not used for transmission , @xmath57 ( again , this includes all slots @xmath50 , and slot @xmath22 when alice does not transmit ) . willie does not know @xmath22 and has to perform a statistical hypothesis test on his entire set of observations @xmath58 to determine whether alice is transmitting . willie performs a statistical hypothesis test @xcite on @xmath58 , where the null hypothesis @xmath59 is that alice does not transmit and each sample is an i.i.d . realization of awgn @xmath60 . the alternate hypothesis @xmath61 is that alice transmits , and the samples from one of the slots come from a different distribution . the rejection of @xmath59 when it is true is a false alarm ( fa ) and the acceptance of @xmath59 when it is false is a missed detection ( md ) . the probability of error @xmath62 is the sum of these error probabilities weighted by the prior probabilities of the hypotheses . we assume that the prior probabilities @xmath63 , and note that unequal priors do not affect our scaling results ( * ? ? ? thus , @xmath64 . as in @xcite , alice desires to ensure that any detector willie employs can only be slightly better than a random guess of her transmission state . since a random guess results in @xmath65 , formally , alice has to maintain @xmath66 for an arbitrary choice of @xmath32 . this requirement also has a natural signal processing interpretation via the receiver operating characteristic ( roc ) curve ( * ? ? ? 2.2.2 ) , which plots probability of detection @xmath67 versus @xmath68 . since @xmath69 , @xmath66 implies that the roc curve lies arbitrarily close to the line of no - discrimination ( the diagonal line where @xmath70 ) . we discuss how to achieve this next . we proved the achievability theorems in @xcite by upper - bounding the relative entropy between the distributions of willie s sequence of observations @xmath58 corresponding to hypotheses @xmath59 and @xmath61 . however , these bounds lack the necessary tightness in this context . therefore , we take a different approach by explicitly analyzing willie s optimal detector assuming that his only unknowns are : a. slot @xmath22 used by alice for transmission ( if she transmits ) , as depicted in figures [ fig : timing ] and [ fig : framework ] ; and b. a secret shared between alice and bob prior to the potential transmission . thus , willie is given alice s channel input distribution , the distribution of the awgn on the channel between alice and willie , and the slot boundaries depicted on figure [ fig : timing ] . this effectively provides willie with the complete statistical model of observations @xmath58 under both hypotheses @xmath59 and @xmath61 , allowing him to construct an optimal detector . here we determine what it takes for alice to ensure that willie s optimal detector performs only slightly better than a random guess of her transmission state , and how much data she can reliably transmit to bob in this manner . the optimal detector for willie is the likelihood ratio test ( lrt ) . this follows from the neyman pearson lemma ( * ? ? ? * chapters 3.2 and 13.1 ) , and the fact that alice s transmission state is binary ( implying that @xmath59 and @xmath61 are simple ) . the lrt compares the likelihood ratio @xmath71 to a threshold @xmath72 , where @xmath73 and @xmath74 are the likelihood functions under @xmath59 and @xmath61 , respectively . @xmath59 or @xmath61 is chosen based on whether @xmath75 is smaller or larger than @xmath72 ( if it is equal , a random decision is made ) : @xmath76 the lrt statistic @xmath75 is a function of the sequence of observations @xmath58 , and , as such , is a random variable . per its definition in section [ sec : prerequisites ] , @xmath58 is parameterized by the slot length @xmath1 and which hypothesis is true ( that is , alice s transmission state ) . let @xmath77 where @xmath78 indicates the true hypothesis ( @xmath59 or @xmath61 ) . since one - to - one transformations of both sides in do not affect the performance of the test , we analyze a detector that is equivalent to the one defined in but employs the following test statistic : @xmath79 where @xmath80 is a one - to - one function defined later . denote by @xmath81 and @xmath82 convergence of random variable @xmath83 to random variable @xmath84 in probability and in distribution , respectively . the following lemma establishes sufficient conditions for covertness of alice s transmission : [ lemma : lrt_bound ] if lrt statistic is described by random variables @xmath85 where @xmath86 and @xmath87 , as well as @xmath88 with @xmath89 , then @xmath66 for any @xmath32 and a sufficiently large @xmath1 . consider any @xmath32 . suppose willie chooses threshold @xmath72 arbitrarily . when @xmath90 , the false alarm probability is lower - bounded as follows : @xmath91 similarly , when @xmath92 , the probability of missed detection is lower - bounded as follows : @xmath93 denote by @xmath94 the event that either @xmath95 when alice is quiet or @xmath96 when alice transmits . since we assume equiprobable priors , @xmath97 while @xmath98 converges in distribution to @xmath99 , this only ensures pointwise convergence in the argument of the distribution function ( that is , the distribution function @xmath100 of @xmath98 converges pointwise rather than uniformly in @xmath101 to the distribution @xmath102 of @xmath99 ) . however , since @xmath72 is the @xmath103 value in an arbitrary sequence , in appendix [ app : convergence ] we exploit the uniform convergence on any finite number of points and the monotonicity of the distribution function to show that setting @xmath104 yields @xmath105 such that for all @xmath15 and any @xmath72 , @xmath106 since @xmath86 and @xmath87 , there exists @xmath107 such that for all @xmath108 , @xmath109 and @xmath109 . the intersection of these events and the event @xmath110 yields a detection error event . by combining their probabilities using demorgan s law and the union bound , we lower - bound @xmath66 for all @xmath111 . _ remark : _ lemma [ lemma : lrt_bound ] holds when @xmath98 converges to any distribution provided it has a continuous density , however , here we do not need such generality . in order to employ lemma [ lemma : lrt_bound ] , we require a one - to - one function @xmath80 that re - scales @xmath112 so that the convergence conditions hold . in the proofs that follow , @xmath113 , where @xmath114 is a sequence of @xmath2 independent random variables , each corresponding to a slot of length @xmath1 . since alice is limited to slot @xmath22 for a potential transmission , the only random variable in @xmath114 that is distributed differently under each hypothesis is @xmath115 . we thus denote by @xmath116 and @xmath117 the random variable corresponding to slot @xmath22 under @xmath59 and @xmath61 , respectively . regardless of alice s transmission state , @xmath118 is an i.i.d . sequence , with @xmath116 distributed identically to the elements of @xmath118 . therefore , @xmath119 let s denote by @xmath120 and @xmath121 respectively the mean and variance of @xmath116 and @xmath122 , @xmath123 . we define @xmath80 as follows : @xmath124 thus , the re - scaled test statistic @xmath125 , @xmath78 , is expressed as follows : @xmath126 by the central limit theorem ( clt ) for triangular arrays ( * ? ? ? * th . 27.2 ) , @xmath127 where @xmath89 . thus , the weighted sum in and corresponds to @xmath98 in lemma [ lemma : lrt_bound ] . now consider the term corresponding to slot @xmath22 , @xmath128 . it effectively offsets @xmath129 s mean away from zero and its distribution depends on alice s transmission state @xmath130 . thus , depending on which hypothesis is true , it maps to either @xmath131 or @xmath132 in lemma [ lemma : lrt_bound ] . to prove achievability of covert communication , we show that there exists a coding scheme for alice such that the random variable describing the lrt statistic has the form given in @xmath133 and @xmath134 , with the terms in the sums satisfying the regularity conditions required by the clt and the term corresponding to slot @xmath22 converging to zero in probability . this allows us to establish the covertness of alice s transmission by applying lemma [ lemma : lrt_bound ] . we prove reliability by extending the random coding arguments from @xcite . first , we state an achievability theorem under an average power constraint . [ th : achievability ] suppose alice has a slotted awgn channel to bob with @xmath31 slots , each containing @xmath1 symbol periods . then , provided that alice and bob have a secret of sufficient length , if alice chooses to , she can transmit @xmath33 bits in a single slot while @xmath135 and @xmath136 for arbitrary @xmath32 and @xmath137 . * construction : * alice secretly selects slot @xmath22 uniformly at random out of the @xmath2 slots . alice s channel encoder takes as input blocks of length @xmath138 bits and encodes them into codewords of length @xmath1 symbols . we employ random coding arguments and independently generate @xmath139 codewords @xmath140 from @xmath141 for messages @xmath142 , each according to @xmath143 , where @xmath144 and symbol power @xmath145 is defined later . the codebook codebooks , where the @xmath146 codebook has a block of non - zero symbols in the @xmath146 slot . selection of the @xmath22-th slot is equivalent to selection of the @xmath22-th codebook and the message is encoded by choosing a codeword from the selected codebook . ] is used only to send a single message and , along with @xmath22 , is the secret not revealed to willie , though he knows how it is constructed , including the value @xmath147 . * analysis ( willie ) : * denote by @xmath148 the power in slot @xmath47 . since willie s channel from alice is corrupted by awgn with power @xmath54 , the likelihood function of the observations @xmath58 under @xmath59 is : @xmath149.\end{aligned}\ ] ] since willie does not know which of the @xmath2 slots alice randomly selects for communication , nor the codebook alice and bob use , but knows that alice s signal is gaussian , the likelihood function of the observations @xmath58 under @xmath61 is : @xmath150.\end{aligned}\ ] ] the lrt statistic @xmath112 is the ratio between and . re - arranging terms yields : @xmath151.\end{aligned}\ ] ] when alice does not transmit in the @xmath146 symbol period , @xmath152 since willie observes awgn ; when alice transmits , @xmath153 by construction . let @xmath154 , @xmath155 , @xmath156 be a sequence of i.i.d . chi - squared random variables with @xmath1 degrees of freedom . then @xmath157 for all @xmath158 under @xmath59 and @xmath159 under @xmath61 , while @xmath160 under @xmath61 . let @xmath161 $ ] for all @xmath159 , @xmath162 $ ] , and @xmath163 $ ] . application of @xmath80 in to yields the expression for @xmath125 in the form defined in and . using the moment generating function ( mgf ) @xmath164 of a chi - squared random variable , we have : @xmath165=1\\ \sigma^2_u(n)&=\left(\frac{\sigma_{\rm w}^2}{\sigma_{\rm w}^2+p_{\rm f}}\right)^{n}\mathbb{e}\left[\exp\left(\frac{p_{\rm f}x_t}{\sigma_{\rm w}^2+p_{\rm f}}\right)\right]-1=\left(\frac{\sigma_{\rm w}^4}{\sigma_{\rm w}^4-p_{\rm f}^2}\right)^{\frac{n}{2}}-1.\end{aligned}\ ] ] since @xmath145 , @xmath120 and @xmath166 satisfy conditions for the clt ( * ? ? ? 27.2 ) and @xmath127 , @xmath89 . also , when alice does not transmit , by chebyshev s inequality : @xmath167 since @xmath31 and @xmath145 , @xmath168 as @xmath6 . when alice transmits , @xmath169}&=\left(\frac{\sigma_{\rm w}^2}{\sigma_{\rm w}^2+p_{\rm f}}\right)^{\frac{n}{2}}\mathbb{e}\left[\exp\left(\frac{p_{\rm f}x_t}{2\sigma_{\rm w}^2}\right)\right]=\left(\frac{\sigma_{\rm w}^4}{\sigma_{\rm w}^4-p_{\rm f}^2}\right)^{\frac{n}{2}}\\ { \mathbb{e}\left[\left(u_{t_{\rm a}}^{(n,1)}\right)^2\right]}&=\left(\frac{\sigma_{\rm w}^2}{\sigma_{\rm w}^2+p_{\rm f}}\right)^{n}\mathbb{e}\left[\exp\left(\frac{p_{\rm f}x_t}{\sigma_{\rm w}^2}\right)\right]=\left(\frac{\sigma_{\rm w}^2}{\sigma_{\rm w}^2+p_{\rm f}}\right)^{n}\left(\frac{\sigma_{\rm w}^2}{\sigma_{\rm w}^2 - 2p_{\rm f}}\right)^{\frac{n}{2}}.\end{aligned}\ ] ] by chebyshev s inequality : @xmath170}{\left(\delta\sigma_u(n)\sqrt{t(n)-1}-{\mathbb{e}\left[u_{t_{\rm a}}^{(n,1)}\right]}\right)^{2}}.\end{aligned}\ ] ] let s divide the numerator and the denominator in the rhs of by @xmath171 $ ] . first , note that : @xmath172}}{\sqrt{\operatorname{var}\left[u_{t_{\rm a}}^{(n,1)}\right]}}&=\frac{1}{\sqrt{\left(1+\frac{p_{\rm f}^2}{\sigma_{\rm w}^4(1 - 2p_{\rm f}/\sigma_{\rm w}^2)}\right)^{\frac{n}{2}}-1}}\\ & \label{eq : ineq1}\leq \frac{1}{\left(1+\frac{p_{\rm f}^2}{\sigma_{\rm w}^4(1 - 2p_{\rm f}/\sigma_{\rm w}^2)}\right)^{\frac{n}{4}}}\leq 1,\end{aligned}\ ] ] where the second inequality in follows from @xmath30 and @xmath147 satisfying @xmath173 . also , @xmath174}\geq \frac{\sigma^2_u(n)}{{\mathbb{e}\left[\left(u_{t_{\rm a}}^{(n,1)}\right)^2\right]}}$ ] where @xmath175}}=\left(1-\frac{2p_{\rm f}^2}{\sigma_{\rm w}^4(1-p_{\rm f}/\sigma_{\rm w}^2)}\right)^{\frac{n}{2}}-o(1)$ ] . when @xmath176 , setting @xmath177 for a constant @xmath178 ensures @xmath179 as @xmath6 . convergence follows from noting that @xmath180 $ ] , and using the taylor series expansion of @xmath181 at @xmath182 . when @xmath183 , convergence is obtained by setting @xmath184 . therefore , by lemma [ lemma : lrt_bound ] , setting @xmath185 ensures @xmath66 for any @xmath32 . * analysis ( bob ) : * let bob employ the maximum likelihood ( ml ) decoder ( i.e. , minimum distance decoder ) . if bob knows the value of @xmath22 , since @xmath186 , alice can reliably ( i.e. , with bob s decoding error probability , averaged over all the codebooks , decaying to zero as @xmath6 ) transmit @xmath187 covert bits , where @xmath188 is a constant @xcite . however , knowledge of @xmath22 is unnecessary for bob if @xmath4 , where @xmath26 is a constant . let s augment alice and bob s gaussian codebook with the origin @xmath189 ( indicating `` no transmission '' ) and have bob attempt to decode each of the @xmath2 slots . the squared distance between a codeword @xmath190 and @xmath191 is @xmath192 , where @xmath193 . repeating the analysis of bob s detection error probability from @xcite using the distance between @xmath190 and @xmath191 instead of @xmath194 yields a looser upper bound on the probability of the decoding error in each slot . by the union bound over all @xmath2 slots , the overall probability of error is @xmath195 . if @xmath176 , then clearly bob s decoding error probability decays to zero if alice attempts to transmit @xmath196 bits in a randomly selected @xmath1-symbol slot @xmath22 . if @xmath183 , then , @xmath197 , and @xmath4 where @xmath198 ensures that bob s decoding error probability decays to zero if alice attempts to transmit @xmath199 bits in a randomly selected @xmath1-symbol slot @xmath22 . therefore , @xmath3 covert bits can be transmitted reliably using slot @xmath22 . unfortunately , representing real - valued codewords requires unbounded storage , which means that the length of the secret pre - shared by alice and bob is infinite . to address this , we consider finite alphabet input , which also satisfies a peak power constraint on the transmitter . the remark in ( * ? ? ? * section iii ) allows both improvement of bob s decoding performance and reduction of the size of the pre - shared secret to @xmath200 bits ( provided @xmath4 ) . bits using a protocol described in ( * ? ? ? * appendix ) and , possibly , to @xmath0 bits using resolvability - based approach in @xcite . however , this is outside of the scope of this work . ] [ th : achievability_peak_power ] suppose alice has a slotted awgn channel to bob with @xmath31 slots , each containing @xmath1 symbol periods , and that her transmitter is subject to the peak power constraint @xmath201 . then , provided that alice and bob have a secret of sufficient length , if alice chooses to , she can transmit @xmath33 bits in a single slot while @xmath135 and @xmath136 for any @xmath32 and @xmath137 . * construction : * alice secretly selects slot @xmath22 uniformly at random out of the @xmath2 slots in which to communicate . she encodes the input in blocks of length @xmath138 bits into codewords of length @xmath1 bits / symbol with the symbols drawn from alphabet @xmath202 , where @xmath203 satisfies the peak power constraint @xmath204 and is defined later . alice independently generates @xmath139 codewords @xmath205 for messages @xmath206 from @xmath207 according to @xmath143 , where @xmath208 . as in the proof of theorem [ th : achievability ] , this single - use codebook is not revealed to willie , though he knows how it is constructed , including the value of @xmath203 . while in this proof the entire codebook is secretly shared between alice and bob , the amount of shared secret information can be reduced using the remark in ( * ? ? ? * section iii ) . * analysis ( willie ) : * the model for the awgn channel from alice to willie did not change from theorem [ th : achievability ] , implying that the likelihood function of the observations @xmath58 under @xmath59 is given by . since willie does not know which of the @xmath2 slots alice randomly selects for communication , nor the codebook alice and bob use , but knows how the codebook is constructed , the likelihood function of the observations @xmath58 under @xmath61 is : @xmath209.\end{aligned}\ ] ] the lrt statistic @xmath112 is the ratio between and . re - arranging terms yields : @xmath210}{2^n}\sum_{\mathbf{b}\in\{-1,1\}^n}\exp\left[{\frac{a}{\sigma_{\rm w}^2}\sum_{y_i\in\mathbf{y}_{\rm w}(t)}y_ib_i}\right].\end{aligned}\ ] ] let @xmath211}{2^n}\sum_{\mathbf{b}\in\{-1,1\}^n}\exp\left[{\frac{a}{\sigma_{\rm w}^2}\sum_{y_i\in\mathbf{y}_{\rm w}(t)}y_ib_i}\right]$ ] . application of @xmath80 in to yields the expression for @xmath125 in the form defined in and . when alice does not transmit , @xmath212\sim\log\mathcal{n}\left(0,\frac{na^2}{\sigma_{\rm w}^2}\right)$ ] , where @xmath213 denotes the log - normal distribution with location @xmath214 and scale @xmath215 . thus , @xmath216}{2^n}\sum_{\mathbf{b}\in\{-1,1\}^n}{\mathbb{e}\left[\exp\left[{\frac{a}{\sigma_{\rm w}^2}\sum_{y_i\in\mathbf{y}_{\rm w}(t)}y_ib_i}\right]\right]}=1\end{aligned}\ ] ] to obtain @xmath217 , we calculate the second moment of @xmath218 and @xmath219 , @xmath123 : @xmath220}&=\frac{\exp\left[{-\frac{na^2}{\sigma_{\rm w}^2}}\right]}{2^{2n}}\sum_{\mathbf{b}\in\{-1,1\}^n}\sum_{\mathbf{d}\in\{-1,1\}^n}{\mathbb{e}\left[\exp\left[{\frac{a}{\sigma_{\rm w}^2}\sum_{y_i\in\mathbf{y}_{\rm w}(t)}y_i(b_i+d_i)}\right]\right]}\\ \label{eq : calc_su1}&=\frac{1}{2^{2n}}\sum_{\mathbf{b}\in\{-1,1\}^n}\sum_{\mathbf{d}\in\{-1,1\}^n}\exp\left[\frac{a^2}{\sigma_{\rm w}^2}\sum_{i=1}^nb_id_i\right]\\ \label{eq : calc_su2}&=\frac{1}{2^{2n}}\sum_{\mathbf{b}\in\{-1,1\}^n}\sum_{\mathbf{z}\in\{-1,1\}^n}\exp\left[\frac{a^2}{\sigma_{\rm w}^2}\sum_{i=1}^nz_i\right]\\ \label{eq : calc_su3}&=\cosh^n\left(\frac{a^2}{\sigma_{\rm w}^2}\right),\end{aligned}\ ] ] where follows from @xmath221\sim\log\mathcal{n}\left(0,\frac{a^2}{\sigma_{\rm w}^2}\sum_{i=1}^n(b_i+d_i)^2\right)$ ] ; is because for a given @xmath222 and any @xmath223 , @xmath224 $ ] is a unique vector in @xmath225 ; and , follows from appendix [ app : derive_cosh ] . thus , @xmath127 , @xmath89 . also , when alice does not transmit , by chebyshev s inequality : @xmath226 since @xmath31 , @xmath168 as @xmath6 . when alice transmits , by construction , willie observes @xmath227 , @xmath228 , where @xmath229 $ ] is drawn equiprobably from @xmath225 . thus , @xmath169}&=\frac{\exp\left[{-\frac{na^2}{2\sigma_{\rm w}^2}}\right]}{2^{2n}}\sum_{\mathbf{c}\in\{-1,1\}^n}\sum_{\mathbf{b}\in\{-1,1\}^n}{\mathbb{e}\left[\exp\left[{\frac{a}{\sigma_{\rm w}^2}\sum_{y_i\in\mathbf{y}_{\rm w}(t)}y_ib_i}\right]\right]}\\ \label{eq : calc_muta}&=\frac{1}{2^{2n}}\sum_{\mathbf{c}\in\{-1,1\}^n}\sum_{\mathbf{b}\in\{-1,1\}^n}\exp\left[\frac{a^2}{\sigma_{\rm w}^2}\sum_{i=1}^nb_ic_i\right]=\cosh^n\left(\frac{a^2}{\sigma_{\rm w}^2}\right),\end{aligned}\ ] ] where follows from @xmath212\sim\log\mathcal{n}\left(\frac{a^2}{\sigma_{\rm w}^2}\sum_{i=1}^nc_ib_i,\frac{na^2}{\sigma_{\rm w}^2}\right)$ ] , the argument for above , and appendix [ app : derive_cosh ] . by the definition of variance and the law of total expectation , @xmath230\leq\frac{1}{2^n}\sum_{\mathbf{c}\in\{-1,1\}^n}{\mathbb{e}\left[\left(u_{t_{\rm a}}^{(n,1)}|\mathbf{c}\text{~sent}\right)^2\right]}$ ] , where @xmath231}&=\frac{\exp\left[{-\frac{na^2}{\sigma_{\rm w}^2}}\right]}{2^{2n}}\sum_{\mathbf{b}\in\{-1,1\}^n}\sum_{\mathbf{d}\in\{-1,1\}^n}{\mathbb{e}\left[\exp\left[{\frac{a}{\sigma_{\rm w}^2}\sum_{y_i\in\mathbf{y}_{\rm w}(t)}y_i(b_i+d_i)}\right]\right]}\\ & = \frac{\exp\left[{-\frac{na^2}{\sigma_{\rm w}^2}}\right]}{2^{2n}}\sum_{\mathbf{b}\in\{-1,1\}^n}\sum_{\mathbf{d}\in\{-1,1\}^n}\exp\left[\frac{a^2}{\sigma_{\rm w}^2}\sum_{i=1}^nc_i(b_i+d_i)+\frac{(b_i+d_i)^2}{2}\right]\\ & = \frac{1}{2^{2n}}\sum_{\mathbf{b}\in\{-1,1\}^n}\sum_{\mathbf{d}\in\{-1,1\}^n}\exp\left[\frac{a^2}{\sigma_{\rm w}^2}\sum_{i=1}^nc_ib_i+(c_i+b_i)d_i\right]\\ \label{eq : calc_suta}&\leq\cosh^n\left(\frac{a^2}{\sigma_{\rm w}^2}\right)\cosh^n\left(\frac{2a^2}{\sigma_{\rm w}^2}\right),\end{aligned}\ ] ] with following from @xmath232 and arguments for above . by chebyshev s inequality : @xmath233}\right)^{2}}.\end{aligned}\ ] ] dividing both numerator and denominator of by @xmath234 , we note that @xmath235}}{\cosh^{\frac{n}{2}}\left(\frac{a^2}{\sigma_{\rm w}^2}\right)\cosh^{\frac{n}{2}}\left(\frac{2a^2}{\sigma_{\rm w}^2}\right)}=\left(\frac{\cosh(a^2/\sigma_{\rm w}^2)}{\cosh(2a^2/\sigma_{\rm w}^2)}\right)^{\frac{n}{2}}\leq 1 $ ] , since @xmath236 for @xmath237 . also , @xmath238 . when @xmath176 , setting @xmath239 for a constant @xmath178 ensures @xmath179 as @xmath6 . convergence follows from noting that @xmath240\geq\exp\left[-\frac{2na^4}{\sigma_{\rm w}^4}\right]$ ] . when @xmath183 , convergence is obtained by setting @xmath241 . therefore , by lemma [ lemma : lrt_bound ] , setting @xmath242 ensures @xmath66 for any @xmath32 . * analysis ( bob ) : * suppose alice transmits @xmath190 . as in the proof of theorem [ th : achievability ] , let bob employ the ml decoder that suffers an error event @xmath243 when the received vector @xmath244 is closer in euclidean distance to codeword @xmath194 , @xmath245 . knowledge of @xmath22 allows the application of appendix [ app : corr ] : since @xmath242 , alice can ensure reliable decoding ( i.e. , bob s error probability , averaged over all the codebooks , decaying to zero as @xmath6 ) for a covert transmission containing @xmath246\right)$ ] bits , where @xmath188 is a constant . however , as in theorem [ th : achievability ] , knowledge of @xmath22 is unnecessary for bob if @xmath4 where @xmath247 is a constant . again , let s augment alice and bob s codebook with the origin @xmath189 ( indicating `` no transmission '' ) and have bob attempt to decode each of the @xmath2 slots . denoting the decoding error probability in slot @xmath47 by @xmath248 , we employ the union bound over all slots to upper - bound the overall decoding error probability : @xmath249 the decoding error probability for one of the @xmath250 slots that alice does not use is the probability that the received vector is closer to some codeword than the origin @xmath191 : @xmath251 where the inequality is the union bound . since the euclidean distance between each codeword and @xmath191 is @xmath252 , by ( * ? ? ? * eq . ( 3.44 ) ) : @xmath253 where @xmath254 and the inequality is because of the upper bound @xmath255 ( * ? ? ? * eq . ( 5 ) ) . substituting into yields @xmath256 to upper - bound the decoding error probability for the slot that alice uses to transmit , we combine the bounds in and as follows : @xmath257,\frac{a^2\log_2e}{8\sigma_{\rm b}^2}\right\}}.\end{aligned}\ ] ] combining , , and yields @xmath258,\frac{a^2\log_2e}{8\sigma_{\rm b}^2}\right\}}.\end{aligned}\ ] ] if @xmath176 , then alice sets @xmath259 . thus , for large enough @xmath1 , @xmath260 $ ] and bob s decoding error probability decays to zero if alice attempts to transmit @xmath261 bits in a randomly selected @xmath1-symbol slot @xmath22 , where @xmath188 . if @xmath183 , then , @xmath262 , and @xmath4 where @xmath263 ensures that bob s decoding error probability decays to zero if alice attempts to transmit @xmath264,\frac{c_{\rm p}\sigma_{\rm w}^2\log_2e}{16\sigma_{\rm b}^2}\right\}$ ] bits in slot @xmath22 . here we show that alice can not transmit @xmath24 bits both reliably and covertly using one of @xmath2 @xmath1-symbol slots . alice attempts to send one of @xmath265 ( equally likely ) @xmath138-bit messages reliably to bob using a sequence of @xmath1 consecutive symbol periods out of @xmath10 , where @xmath266 , and each message is encoded arbitrarily into @xmath1 symbols . unlike in the previous section , here willie is oblivious to the locations of the slot boundaries , alice s codebook construction scheme and other properties of her signal . nevertheless , by dividing his sequence of @xmath10 observations into a set of @xmath2 non - overlapping subsequences , and employing a simple threshold detector on the maximum subsequence power , willie can detect alice if she attempts to transmit @xmath24 covert bits reliably . we assume that alice is power - constrained , precluding transmission of @xmath267 bits in @xmath1 awgn channel uses ( see standard arguments in ( * ? ? ? * chapter 9 ) ) . we therefore limit our analysis to @xmath176 . [ th : converse ] if alice attempts to transmit @xmath35 bits using a sequence of @xmath1 consecutive symbol periods that are arbitrarily located inside a sequence of @xmath10 symbol periods , then , as @xmath6 , either willie can detect her with high probability , or bob can not decode with arbitrarily low probability of error . willie divides the sequence @xmath58 of @xmath10 observations into a set of @xmath2 non - overlapping subsequences @xmath268 , with each @xmath269 containing @xmath1 consecutive observations . denote by @xmath148 the observed power in each subsequence and @xmath270 . for a threshold @xmath271 , willie accuses alice of transmitting if @xmath272 . suppose alice does not transmit . willie s probability of false alarm is @xmath273 . let @xmath274 . to find @xmath275 so that willie s detector has an arbitrary probability of false alarm @xmath276 as @xmath6 , note that each @xmath157 where @xmath154 , @xmath155 , @xmath156 is a sequence of i.i.d . chi - squared random variables each with @xmath1 degrees of freedom . we have @xmath277 where @xmath278 . for the desired @xmath276 , @xmath279 . using a chernoff bound for the tail of a chi - squared distribution ( * lemma 2.2 ) , we obtain : @xmath280 where results from a taylor series expansion of @xmath281 at @xmath182 . discarding low order terms and solving for @xmath275 yields @xmath282 . taylor series expansion of @xmath283 at @xmath182 yields @xmath284 . thus , setting @xmath285 with some constant @xmath286 yields the desired probability of false alarm @xmath276 . now suppose alice uses an arbitrary codebook @xmath287 and transmits codeword @xmath190 using @xmath1 consecutive symbol periods . denote the average symbol power of @xmath190 by @xmath288 . since alice uses @xmath1 consecutive symbols , her transmission overlaps at most two of willie s subsequences , which we denote @xmath22 and @xmath289 . denote by @xmath290 and @xmath291 the power from alice s transmission in subsequences @xmath22 and @xmath289 , respectively , with @xmath292 . willie s probability of missing alice s transmission is @xmath293 where the factorization in is because alice s codeword and the noise in other subsequences are independent . @xmath294 does not depend on alice s codeword . however , since the codeword is an unknown deterministic signal that is added to awgn on willie s channel to alice , @xmath295 and @xmath296 are non - central chi - squared random variables with @xmath1 degrees of freedom and respective non - centrality parameters @xmath297 and @xmath298 . without loss of generality , assume that @xmath299 . thus , @xmath290 satisfies @xmath300 and the expected value and variance of @xmath301 are bounded as follows ( * ? ? ? d.1 ) : @xmath302&\geq\sigma_{\rm w}^2n+\frac{np_{\rm f}}{2}\\ \label{eq : p1_var}\operatorname{var}\left[y_{t_{\rm a}}\right]&\leq2n\sigma_{\rm w}^4 + 4n\sigma_{\rm w}^2p_{\rm f}\end{aligned}\ ] ] since @xmath303 , chebyshev s inequality with and yields @xmath304\right|>\mathbb{e}[y_{t_{\rm a}}]-\sigma_{\rm w}^2(n+c\sqrt{n\log t(n)})\right)\leq\frac{2\sigma_{\rm w}^4 + 4\sigma_{\rm w}^2p_{\rm f}}{\left(\frac{\sqrt{n}p_{\rm f}}{2}-c\sigma_{\rm w}^2\sqrt{\log t(n)}\right)^2}.\end{aligned}\ ] ] therefore , if @xmath305 , as @xmath306 , @xmath62 can be made arbitrarily small . the proof of the non - zero lower bound on the decoding error probability @xmath307 if alice tries to transmit @xmath24 bits in a single slot using average symbol power @xmath308 follows from the arguments in the proof of theorem 2 in @xcite . attempting to transmit @xmath24 bits in a single slot while maintaining covertness by using average symbol power @xmath308 results in the non - zero lower bound on the decoding error probability @xmath307 . however , a natural question is whether it is possible to transmit @xmath24 covert bits _ in expectation_. if this were possible , then alice and bob could employ error correction coding to achieve throughput that exceeds the limit imposed by the codebook construction in the proofs of theorems [ th : achievability ] and [ th : achievability_peak_power ] . however , we can use the arguments in the remark that follows the proof of theorem 2 in @xcite to demonstrate that the expected number of covert bits transmissible in a single slot can not exceed @xmath33 , obviating any additional layers of error control coding . here we relate our work to other studies of covert communication . an overview of the area can be found in @xcite . covert communication is an ancient discipline @xcite of hiding messages in innocuous objects . modern steganographic systems @xcite hide information by altering the properties of fixed - size , finite - alphabet covertext objects ( e.g. images ) , and are subject to a similar srl as covert communication : @xmath0 symbols in covertext of size @xmath1 may safely be modified to hide an @xmath309-bit message @xcite . the similarity between the srls in these disciplines is from the mathematics of statistical hypothesis testing , as discussed in @xcite . however , in steganography , the transmission to bob is noiseless , which allows the extra @xmath310 factor . batch steganography uses multiple covertext objects to hide a message and is subject to the steganographic srl described above @xcite . the batch steganography interpretation of covert communication using the timing - based degree - of - freedom that is described here is equivalent to using only one of @xmath2 covertext objects of size @xmath1 to embed a message . willie , who knows that one covertext object is used but not which one , has to examine all of them . we are not aware of any work on this particular problem , but it is likely that one could extend our result to it . we also note that more recent work on steganography shows that an empirical model of the covertext suffices to break the steganographic srl , allowing the embedding of @xmath200 bits in an @xmath1-symbol covertext @xcite . however , this technique relies on embedding messages in covertext by _ replacing _ part of it something that can not be done in standard communication systems unless alice controls willie s noise source . the emergence of radio - frequency ( rf ) communication systems necessitated the development of means to protect them from jamming , detection , and eavesdropping . spread - spectrum techniques @xcite address these issues by transmitting a signal that requires bandwidth @xmath311 on a much wider bandwidth @xmath312 , thus , effectively suppressing the power spectral density of the signal below the noise floor . this provides both covertness as well as the resistance to jamming , fading , and other interference . however , while the spread - spectrum architectures are well - developed , the fundamental srl for covert communication has been derived only recently @xcite . this resulted in the revival of the field , with follow - on work focusing on reducing the size of the pre - shared secret @xcite , fully characterizing the optimal constant hidden by the big-@xmath7 notation of the srl @xcite , and extending the srl to bosonic channels with willie limited only by the laws of quantum mechanics @xcite . finally , while here we improve on the srl by exploiting willie s ignorance of transmission timing , other studies explore even stronger assumptions on his limitations . in particular , authors in @xcite examine the impact of the noise in willie s channel estimates , which allows @xmath200 covert bits to be transmitted in @xmath1 uses of the channel even when willie has the upper and lower bounds on the estimates . thus , positive - rate rather than srl - governed covert communication is possible when willie s knowledge of the channel is incomplete . we have shown that secretly pre - arranging a choice of a single @xmath1-symbol period slot out of @xmath2 allows alice to reliably transmit @xmath313 bits on an awgn channel to bob while rendering willie s detector arbitrarily close to ineffective . surprisingly , the multiplicative increase in transmitted information over the result in @xcite is obtained without the need for bob to know which slot holds the transmission if @xmath4 , where @xmath247 is a constant , and , when @xmath27 only an additive expense of an extra @xmath28 pre - shared secret bits is needed . in the future we plan on combining this work with our recent results on jammer - assisted covert communication @xcite to enable covert networks . since @xmath98 is independent of which hypothesis is true , can be re - written : @xmath314 denote the standard gaussian distribution function by @xmath315 where @xmath316 is the standard gaussian density function . the convergence of @xmath100 to @xmath102 is pointwise in @xmath101 , and , since @xmath72 is the @xmath103 value in an arbitrary sequence , we can not use this fact directly . ( 85,20 ) ( 66,12)(1,0)19 ( 25,12)(-1,0)25 ( 60,12)(-1,0)29 ( 61,9.90)(4,4)@xmath317 ( 26,9.90)(4,4)@xmath317 ( 26.2,7)(4,4)@xmath318 ( 0,2.5)(10,4)@xmath319 region 1 ( 13,4)(0,1)3 ( 72,4)(0,1)3 ( 0,7)(1,0)85 ( 75,2.5)(10,4)region 3 @xmath320 ( 37,2.5)(10,4)region 2 ( 13,10.5)(0,1)3 ( 72,10.5)(0,1)3 ( 11,7)(4,4)@xmath321 ( 70,7)(4,4)@xmath322 ( 8,11)(0,1)2 ( 5,7)(5,4)@xmath323 ( 18,11)(0,1)2 ( 15,7)(5,4)@xmath324 ( 23,11)(0,1)2 ( 67,11)(0,1)2 ( 64,7)(5,4)@xmath325 ( 77,11)(0,1)2 ( 74,7)(5,4)@xmath326 ( 54,15.5)(8,4)@xmath327 ( 46,15.5)(8,4)@xmath72 ( 38,15.5)(8,4)@xmath328 ( 45,15)(1,0)10 ( 45,14.5)(0,1)1 ( 50,14)(0,1)2 ( 55,14.5)(0,1)1 ( 32,11)(0,1)2 ( 37,11)(0,1)2 ( 42,11)(0,1)2 ( 40,7)(4,4)@xmath329 ( 47,11)(0,1)2 ( 52,11)(0,1)2 ( 57,11)(0,1)2 ( 54.7,7)(4,4)@xmath330 however , let s choose finite constants @xmath331 and @xmath332 , and partition the real number line into three regions as shown in figure [ fig : regions ] . clearly , for any @xmath1 , @xmath72 is in one of these regions . next we demonstrate that holds for an arbitrary @xmath72 by appropriately selecting @xmath321 , @xmath322 , and @xmath275 . consider @xmath333 , or region 1 in figure [ fig : regions ] : @xmath334 because the convergence of @xmath100 to @xmath102 is pointwise , given @xmath275 , @xmath335 , and @xmath336 , there exists @xmath337 such that , for all @xmath338 , @xmath339 when @xmath333 . similarly for @xmath340 , or region 3 in figure [ fig : regions ] : @xmath341 again , because the convergence of @xmath100 to @xmath102 is pointwise , given @xmath275 , @xmath335 , and @xmath342 , there exists @xmath343 such that , for @xmath340 and all @xmath344 , @xmath345 finally , consider @xmath72 satisfying @xmath346 , or region 2 in figure [ fig : regions ] . let s assume that @xmath322 and @xmath321 are selected so that @xmath347 is an integer multiple of @xmath275 ( e.g. , using larger @xmath322 than necessary , which results in the rhs of being smaller ) . consider a sequence @xmath348 where @xmath349 , @xmath350 , @xmath351 , @xmath352 , @xmath353 , @xmath354 . sequence @xmath348 partitions region 2 into @xmath355 subregions , and , for any @xmath72 satisfying @xmath346 , there exists @xmath356 such that @xmath357 , as illustrated in figure [ fig : regions ] . since @xmath100 is monotonic , @xmath358 since the convergence of @xmath100 to @xmath102 is pointwise , for a given @xmath329 , @xmath275 , and @xmath335 , there exists @xmath359 such that for all @xmath360 , @xmath361 where follows from @xmath362 . setting @xmath363 and @xmath364 yields the desired lower bound for all @xmath365 when @xmath72 satisfies @xmath346 . thus , for any @xmath72 , when @xmath15 where @xmath366 , @xmath367 . we argue by induction on @xmath1 . the base case @xmath369 is trivial . assume the claim holds for @xmath1 . now , @xmath370&=\frac{1}{2\cdot2^{n}}\sum_{{\genfrac{}{}{0pt}{}{\mathbf{x}\in\{-1,1\}^{n}}{x_{n+1}\in\{-1,1\}}}}\exp[ax_{n+1}]\exp\left[a\sum_{i=1}^{n } x_i\right]=\cosh(a)\cosh^n(a).\end{aligned}\ ] ] here we analyze bob s decoding error probability @xmath307 when the transmission time @xmath22 is known and alice uses binary modulation @xmath202 that satisfies the peak power constraint and ensures that willie s detection error probability @xmath66 . this appendix provides an alternative to the analysis of @xmath307 in the proof of theorem 1.2 in @xcite ( analysis of @xmath307 in the proof of theorem 1.2 in @xcite contains minor technical errors which do _ not _ change the main results ) . the construct presented here is adapted for the proof of theorem [ th : achievability_peak_power ] in section [ sec : achievability ] . suppose alice transmits @xmath190 . recall that @xmath190 is drawn from a codebook @xmath371 containing codewords that are independently generated according to @xmath372 , where @xmath208 . bob uses an ml decoder which suffers an error event @xmath243 when the received vector @xmath373 is closer in euclidean distance to codeword @xmath194 , @xmath245 . the decoding error probability , averaged over all the codebooks , is then : @xmath374\leq\sum_{{\genfrac{}{}{0pt}{}{i=1}{i\neq k}}}^{2^m}\mathbb{e}_{\mathbf{c}(w_k)}\mathbb{p}(e_{k\rightarrow i})\label{eq : av_bincodebook_err},\end{aligned}\ ] ] where @xmath375 $ ] denotes the expectation over random variable @xmath376 , and the inequality is the union bound . let @xmath377 denote the euclidean ( @xmath378 ) distance between two codewords . then , by ( * ? ? ? * eq . ( 3.44 ) ) : @xmath379\leq\mathbb{e}_{\mathbf{d}}\left[\frac{1}{2}\exp\left(-\frac{\|\mathbf{d}\|_2 ^ 2}{8\sigma_{\rm b}^2}\right)\right],\end{aligned}\ ] ] where @xmath254 and the inequality is because of the upper bound @xmath255 ( * ? ? ? * eq . ( 5 ) ) . euclidean distance @xmath380 depends on the number of locations @xmath381 where @xmath190 and @xmath194 differ , which is binomially - distributed by construction , where each location is different with probability @xmath382 . since when the codewords are different in @xmath381 locations , @xmath383 , we have @xmath384&=\frac{1}{2}\sum_{j=0}^n \exp\left(-\frac{ja^2}{2\sigma_{\rm b}^2}\right)\binom{n}{j}\frac{1}{2^n}=\frac{1}{2^{n+1}}\left[1+\exp\left(-\frac{a^2}{2\sigma_{\rm b}^2}\right)\right]^n\label{eq : exp_bincodeword_err},\end{aligned}\ ] ] where the last equality is an application of binomial theorem . substitution of into yields @xmath385\right)}.\end{aligned}\ ] ] thus , if alice attempts to transmit @xmath386\right)$ ] bits , where @xmath188 is a constant , bob s decoding error probability decays to zero as @xmath6 . if @xmath387 , then clearly @xmath388 bits . if @xmath389 , then , application of the bounds @xmath390 and @xmath391 yields @xmath392 . in particular , if @xmath54 is known to alice and @xmath393 as prescribed by the analysis of willie s detector in ( * ? ? ? 1.2 ) , @xmath394 . b. a. bash , d. goeckel , and d. towsley , `` square root law for communication with low probability of detection on awgn channels , '' in _ proc . of ieee international symposium on information theory ( isit ) _ , cambridge , ma , jul . 2012 . h. che , m. bakshi , and s. jaggi , `` reliable deniable communication : hiding messages in noise , '' in _ proc . of ieee international symposium on information theory ( isit ) _ , istanbul , turkey , jul . 2013 , arxiv:1304.6693 . m. bloch , `` covert communication over noisy memoryless channels : a resolvability perspective , '' in _ proc . of ieee international symposium on information theory ( isit ) _ , hong kong , china , 2015 , arxiv:1503.08778 . l. wang , g. w. wornell , and l. zhang , `` limits of low - probability - of - detection communication over a discrete memoryless channel , '' in _ proc . of ieee international symposium on information theory ( isit ) _ , hong kong , china , 2015 , arxiv:1506:03236 . s. kadhe , s. jaggi , m. bakshi , and a. sprintson , `` reliable , deniable , and hidable communication over multipath networks , '' in _ proc . of ieee international symposium on information theory ( isit ) _ , honolulu , hi , jul . 2014 , arxiv:1401.4451 . r. soltani , b. a. bash , d. goeckel , s. guha , and d. towsley , `` covert single - hop communication in a wireless network with distributed artificial noise generation , '' in _ proc . of conf . on commun . , control , comp ( allerton ) _ , monticello , il , 2014 . b. a. bash , s. guha , d. goeckel , and d. towsley , `` quantum noise limited communication with low probability of detection , '' in _ proc . of ieee international symposium on information theory ( isit ) _ , istanbul , turkey , jul . b. a. bash , a. h. gheorghe , m. patel , j. l. habif , d. goeckel , d. towsley , and s. guha , `` covert optical communication , '' arxiv:1404:7347 , university of massachusetts technical report um - cs-2014 - 004 , 2014 . m. chiani , d. dardari , and m. k. simon , `` new exponential bounds and approximations for the computation of error probability in fading channels , '' _ ieee transactions on wireless communications _ , vol . 2 , no . 4 , pp . 840845 , jul . 2003 . , `` batch steganography and pooled steganalysis , '' in _ information hiding _ , ser . lecture notes in computer science , j. l. camenisch , c. s. collberg , n. f. johnson , and p. sallee , eds.1em plus 0.5em minus 0.4emspringer berlin heidelberg , 2007 , vol . 4437 , pp .

marine viruses shape the structure of the microbial community . they are , thus , a key determinant of the most important biogeochemical cycles in the planet . therefore , a correct description of the ecological and evolutionary behavior of these viruses is essential to make reliable predictions about their role in marine ecosystems . the infection cycle , for example , is indistinctly modeled in two very different ways . in one representation , the process is described including explicitly a fixed delay between infection and offspring release . in the other , the offspring are released at exponentially distributed times according to a fixed release rate . by considering obvious quantitative differences pointed out in the past , the latter description is widely used as a simplification of the former . however , it is still unclear how the dichotomy `` delay versus rate description '' affects long - term predictions of host - virus interaction models . here , we study the ecological and evolutionary implications of using one or the other approaches , applied to marine microbes . to this end , we use mathematical and eco - evolutionary computational analysis . we show that the rate model exhibits improved competitive abilities from both ecological and evolutionary perspectives in steady environments . however , rate - based descriptions can fail to describe properly long - term microbe - virus interactions . moreover , additional information about trade - offs between life - history traits is needed in order to choose the most reliable representation for oceanic bacteriophage dynamics . this result affects deeply most of the marine ecosystem models that include viruses , especially when used to answer evolutionary questions . = -1 bacteriophage ; phytoplankton ; latency period ; burst size ; evolutionarily stable strategy ; eco - evolutionary dynamics .

1312.5528

viruses are the most numerous organisms on earth . they play diverse roles in the biotic component of practically any ecosystem . especially remarkable is the case of marine ecosystems . marine viruses are important sources of mortality at every trophic level . potential hosts range from whales and commercial fish species to zooplankton , heterotrophic bacteria and microbial autotrophs @xcite . viruses are key components of the microbial loop and , therefore , the biogeochemical cycle of elements such as nitrogen or phosphorus @xcite . they are responsible for more than @xmath0 of marine bacterial mortality @xcite , contributing importantly to shaping the community @xcite . the relevance of _ virioplankton _ not only stems from the `` predatory '' pressure they exert , but also from the subsequent release of organic nutrients ( able to supply a considerable amount of the nutrient demand of , e.g. heterotrophic bacterioplankton @xcite ) ; or their contribution to microbial genetic diversity in the ocean through horizontal gene transfer @xcite . the vast majority of these roles are assumed by marine viruses that eventually kill the host cell @xcite . the standard _ lytic infection _ can be summarized in the following steps @xcite : _ i ) _ free viruses diffusing in the medium encounter and attach to cells at a certain _ adsorption rate _ ; _ ii ) _ after injecting its nucleic acid into the host cell , the virus takes control of the host synthesis machinery in order to replicate its genetic material ( dna or rna , depending on the type of virus @xcite ) and produce the proteins that will form the components of the viral offspring ( _ eclipse period _ ) ; _ iii ) _ during the _ maturation _ stage ( or _ rise period _ ) , the new _ virions _ are assembled ; _ iv ) _ finally , the virus synthesizes the _ holin _ protein , which perforates the plasma membrane allowing viral endolysins ( lysoenzymes ) to reach and lyse the cell wall , thereby releasing offspring and cellular organic compounds to the medium . the _ latent period _ ( steps _ ii_-_iv _ above ) , controlled by the so - called _ gene t _ ( or holin gene ) @xcite , is one of the most important viral life - history traits . so are the _ burst size _ ( offspring number , intimately related to the duration of the infection ) , and the adsorption rate . the latent period is studied intensively in the viral literature not only due to its ecological importance , but also owing to the small pleiotropic effect that its evolutionary change has on other phenotypic traits @xcite . on the other hand , the latent period links ecological and evolutionary change , as mutations in this trait influence the demography of the population and the environment influences which latent periods are favored by selection @xcite , closing in this way an eco - evolutionary feedback loop @xcite . furthermore , the short generation times and numerous offspring of viruses facilitate rapid evolution @xcite , and a possible overlap between ecological and evolutionary timescales . all these factors provide evidence for the importance of using a proper description of the ecological interactions between virus and host in order to make reliable evolutionary predictions . in the theoretical literature for marine viruses , mostly centered on viruses that infect bacteria ( _ bacteriophages _ ) , host - virus interactions are represented in two different ways . one approach explicitly considers the latent period imposing a fixed delay between the adsorption and the release of the offspring @xcite . in the other approach , new viruses are continuously released at a certain _ lytic rate _ , with cells that are simultaneously infected bursting at different post - infection times , exponentially distributed @xcite . thus , in the delay model the survival of each and every infected cell is ensured up to an infection age that equals the fixed latent period , whereas survival responds to a probabilistic rule in the rate model . the latter can actually be seen as a simplification of the former that facilitates mathematical and computational analysis of the interactions . indeed , the ecological outcome of the two approaches seems to be , a priori , qualitatively similar in spite of the obvious difference in the timing of the infection @xcite . while in the delay model progeny show periods of no release ( e.g. initial stages of viral culture experiments ) , in the rate model viral offspring are liberated at all times . however , little attention has been paid to quantifying thoroughly how these differences affect the long - term predictions by the two kinds of models . here , we aim to fill this gap . in this paper , we focus on the eco - evolutionary differences between the two approaches to the description of the lytic infection cycle . this comparison may prove very useful to assess the evolutionary consequences of the simplifying assumptions in these models , and therefore the long - term reliability of a whole group of different models for host - virus dynamics available in the literature . the rate - based approach is used to model not only diverse aspects of host - lytic virus interactions @xcite , but also other types of viral infection cycles such as lysogeny @xcite or shedding @xcite . in the latter , viruses continuously produce and release virions during the entire infection period . some examples include filamentous phages , and viruses of an enormous importance for humans such as ebola , sars , smallpox , varicella - zoster virus , and hiv @xcite . in some retroviruses such as hiv , both burst and continuous production modes have actually been suggested @xcite . thus , this question transcends purely technical matters such as model selection . indeed , this study can potentially serve to compare the evolutionary strategies of a wide selection of viruses with very different infection cycles . as a model case , we use bacteriophages , due to their importance for biogeochemical cycles ; it also allows us to resort to the extensive modeling bibliography available , in which the two approaches to the infection cycle are used . on the other hand , we consider mutations only in the holin gene , in order to isolate the effects of evolution on the key differentiating trait for the two strategies : the latent period ( or , equivalently , lysis rate ) . thus , we first present the two models for lytic infection . after briefly comparing them from an ecological perspective , we turn our attention to their evolutionary divergences . under this framework , we discuss the ecological and evolutionary contrast between the two forms for the life - history _ trade - off _ between latent period and burst size that have been proposed in the literature . finally , we comment on the implications of all the above for the descriptions of host - virus interactions in general , and marine bacteriophages in particular . this study will contribute to the reliability of long - term predictions regarding the interaction between a wide variety of viruses and their hosts . in order to compare the two approaches to the infection cycle , we first set common idealized environmental conditions by using two - stage chemostats @xcite . two - stage chemostats are basically composed of a continuous culture for bacterial hosts , coupled to a continuous culture of co - existing bacteria and viruses . a flow of nutrients from a fresh medium to the first chemostat facilitates bacterial growth , and a flow of `` fresh '' hosts from the first chemostat to the second chemostat allows for the development of the viral population . finally , both virus and bacterial cells are washed out from the second chemostat at a certain rate . the described flows , which can loosely resemble e.g. the continuous passage or migratory events occurring in the mammalian gastrointestinal tract @xcite , enable a steady state for the overall system . from the perspective of marine bacteriophages , _ quasi_-stationary conditions may be found in stratified waters where cyanobacteria , among the most common targets for virioplankton , dominate . such a steady state is very convenient from the mathematical standpoint , as is the continuous source of hosts , which helps alleviate the oscillations that are frequently observed in standard predator - prey models @xcite ( see below ) . in addition , the continuous flow of uninfected hosts constitutes a relief for the bacterial population from the evolutionary pressure of the virus and , therefore , prevents bacteria from embarking on an otherwise expected co - evolutionary arms race @xcite . this allows us to focus on viral evolution only . thus , two - stage chemostats provide a controlled environment whose conditions are easily reproducible in the laboratory ; they also offer general results that can be adapted to other environments , as discussed below . lastly , the environmental parameters are chosen to avoid multiple infections ( see table [ table ] in appendix [ appa ] ) , preventing in this way any kind of intra - cellular competition among viruses . let us study first the approach to lysis in which the individuals of the viral population release their offspring after exactly the same latent period . the model describing this cycle implements explicitly the delay between adsorption and burst . if @xmath1 $ ] represents the concentration of uninfected bacterial cells , @xmath2 $ ] the concentration of free viruses , and @xmath3 $ ] the concentration of infected bacteria , the dynamics of the interactions between host and virus can be modeled using the equations @xcite : @xmath4(t)}{dt}&=&\mu\;[c]-k[c][v]-w[c]+w[c_{0}]\label{delayed_c_eq}\\ \dfrac{d[i](t)}{dt}&=&k[c][v]-\left(k[c]_{t - l}[v]_{t - l}\right)e^{-wl}-w[i]\label{delayed_i_eq}\\ \dfrac{d[v](t)}{dt}&=&\left(b\;k[c]_{t - l}[v]_{t - l}\right)e^{-wl}-k[c][v]-m[v]-w[v],\label{delayed_v_eq}\end{aligned}\ ] ] where @xmath5 represents the growth rate of uninfected hosts ; @xmath6 , the adsorption rate ; @xmath7 , the viral mortality or decay rate ; @xmath8 , the washout rate ; @xmath9 and @xmath10 , the viral latent period and burst size , respectively ; and the subscript @xmath11 indicates that the term is evaluated a lytic cycle ( latent period ) in the past . we initially consider a monomorphic viral population ( i.e. all individuals share the same phenotype or trait values ) . thus , the first equation describes the dynamics of the host population as a result of growth ( first term ) , infection events ( second term ) , washout process ( third term ) and inflow of uninfected cells ( fourth term ) . the second equation considers the dynamics of infected cells , whose number grows due to adsorption events ( first term ) , and decreases due to dilution ( last term ) , and lysis of cells ( second term ) ; the latter term is the result of correcting the number of cells that were infected @xmath9 time steps before , @xmath12_{t - l}[v]_{t - l}$ ] , using the probability for those cells to survive dilution during that time ( @xmath13 term ) @xcite . likewise , the free virus population grows owing to those lysed cells ( first term , number of lysis events times the burst size ) , and is reduced by adsorption ( second term ) , natural mortality ( third term ) or dilution ( last term ) . we assume a simple monod formulation for the growth rate of bacteria , given by : @xmath14)=\dfrac{\mu_{max}[n]}{[n]+k_{n } } , \label{monod_equation}\ ] ] in which @xmath15 is the maximum growth rate for the cell and @xmath16 its half - saturation constant ( defined as the concentration at which the growth rate of the cell equals half its maximum ) . see table [ table ] for units . note that we use here the standard assumption that infected cells effectively allocate all their resources to viral production ( i.e. @xmath17 ) . to these equations , we must add the dynamics of the nutrient : @xmath18}{dt}=w\left([n_{0}]-[n]\right)-\mu([n])[c]/y , \label{nutrient_equation}\ ] ] where @xmath19 $ ] is the inflow of nutrient feeding uninfected host cells and @xmath20 is a yield or efficiency parameter accounting for the efficiency for bacterial cells to transform uptake into growth . the equations above can be solved for the stationary state , expected under chemostat conditions . writing , for simplicity , @xmath21_{st})$ ] , we obtain a trivial solution for the virus - free configuration @xmath2_{st}=[i]_{st}=0 $ ] , @xmath1_{st}=w[c_{0}]/(w-\overline{\mu})$ ] . this solution is always feasible ( see appendix [ appb ] ) . on the other hand , the non - trivial steady state is given by the expressions : @xmath22_{st}&=&\dfrac{w+m}{k\left(be^{-wl}-1\right)}\label{stat_delayed_c}\\ \left[i\right]_{st}&=&\dfrac{(\overline{\mu}-w)(w+m)}{k\;w}\dfrac{\left(1-e^{-wl}\right)}{\left(be^{-wl}-1\right)}+\left(1-e^{-wl}\right)[c_{0}],\label{stat_delayed_i}\\ \left[v\right]_{st}&=&\dfrac{\overline{\mu}-w}{k}+\dfrac{w[c_{0}]\left(be^{-wl}-1\right)}{w+m}\label{stat_delayed_v},\end{aligned}\ ] ] feasible as long as @xmath23 and @xmath24/[c]_{st}\right)$ ] ( see appendix [ appb ] ) . the stability of the trivial solution is determined by the basic reproductive number , @xmath25 . this observable is the expected number of secondary infections arising from a single individual in an equilibrium susceptible population @xcite . thus , @xmath26 indicates an eventual fall into the phage - free state , whereas @xmath27 points to the instability of this trivial state . from eq.([delayed_v_eq ] ) , it is easy to find that @xmath28_{st}e^{-wl}}{k[c]_{st}+m+w}$ ] . thus , the trivial equilibrium changes stability for @xmath29 or , in other words , for a @xmath1_{st}$ ] given by eq.([stat_delayed_c ] ) . therefore , the latter condition opens the possibility of a _ transcritical bifurcation _ @xcite . however , the deduction of general stability conditions for the non - trivial state is a highly nontrivial task , beyond the scope of this article . we refer the reader to the extensive mathematical literature devoted to study of the local and global stability of host - virus systems similar to the one presented here @xcite . on the other hand , oscillations are a common outcome of predator - prey interactions , and frequently seen in bacteriophage models @xcite . for the sake of mathematical tractability ( especially for evolutionary matters ) , we focus our analysis on the region of the parameter space where stationarity can be found ( but see discussion ) . with these words of caution , we assume hereafter that the generic feasible steady state above fulfills those stability conditions and proceed with the rest of the analysis . indeed , eqs.([stat_delayed_c])-([stat_delayed_v ] ) prove to be stable for the realistic range of parameters used in this study , as shown in our simulations below . the chosen parametrization represents generically marine lytic t bacteriophages and a bacterial species ( see table [ table ] ) . finally , the stationary growth rate for the viral population ( per - capita change in the concentration of free virus ) is given by : @xmath30_{st}. \label{dm_fitness}\ ] ] because the average number of surviving offspring per cell is given by : @xmath31 eq.([dm_fitness ] ) indicates that stationary co - existence is possible ( i.e. @xmath32 ) only when @xmath33}+1 $ ] ( i.e. @xmath29 ) , which ensures that the average number of offspring per cell is larger than one . following the same notation above , infections in which the offspring are released at a certain lysis rate @xmath34 can be described by the equations : @xmath4(t)}{dt}&=&\mu\;[c]-k[c][v]-w[c]+w[c_{0}]\label{rate_c_eq}\\ \dfrac{d[i](t)}{dt}&=&k[c][v]-k_{l}[i]-w[i]\label{rate_i_eq}\\ \dfrac{d[v](t)}{dt}&=&b\;k_{l}[i]-k[c][v]-m[v]-w[v],\label{rate_v_eq}\end{aligned}\ ] ] where the delay terms have been replaced by instantaneous terms ( i.e. evaluated at time @xmath35 ) . in this way , there is unceasingly new virions joining the free virus population , with host cells being lysed at a rate @xmath36 . eqs.([monod_equation])-([nutrient_equation ] ) complete the description of the dynamics of the system . the stationary states are described by the trivial configuration @xmath2_{st}=[i]_{st}=0 $ ] , @xmath1_{st}=w[c_{0}]/(w-\overline{\mu})$ ] , always feasible ( see appendix [ appb ] ) , and the non - trivial steady state : @xmath22_{st}&=&\dfrac{\left(w+m\right)\left(k_{l}+w\right)}{k\left[k_{l}\left(b-1\right)-w\right]}\label{stat_rate_c}\\ \left[i\right]_{st}&=&\dfrac{(\overline{\mu}-w)(w+m)}{k\left[k_{l}\left(b-1\right)-w\right]}+\dfrac{w[c_{0}]}{k_{l}+w}\label{stat_rate_i}\\ \left[v\right]_{st}&=&\dfrac{\overline{\mu}-w}{k}+w[c_{0}]\dfrac{k_{l}\left(b-1\right)-w}{\left(w+m\right)\left(k_{l}+w\right)}\label{stat_rate_v},\end{aligned}\ ] ] feasible for @xmath37 and @xmath24/[c]_{st}\right)$ ] . the definition of the basic reproductive number , @xmath38_{st}}{\left(k[c]+m+w\right)\left[v\right]_{st}}$ ] , leads once more to a potential transcritical bifurcation for @xmath1_{st}$ ] given by eq.([stat_rate_c ] ) , which results from the condition @xmath29 . also similarly to the previous model , our realistic parametrization is able to yield stable solutions . thus , we assume that the stationary state fulfills the stability conditions and refer the reader to the existing literature for a detailed study of these @xcite . lastly , we can deduce the viral growth rate as in the case of the dm : @xmath39_{st } , \label{rm_fitness}\ ] ] for which , by realizing that : @xmath40 we can conclude again that the nontrivial stationary state is maintained thanks to the condition @xmath33}+1 $ ] , again equivalent to @xmath29 . it is advisable to note that latent period and burst size are not independent . the number of offspring is determined by the timing of the lysis . moreover , the time spent in producing new virions increases the generation time ( sum of extra- and intra - cellular viral lifetime ) . this sets an obvious life - history trade - off realized in the dichotomy `` immediate but low reproduction '' and `` delayed but larger offspring '' that shapes the evolution of @xmath10 and @xmath9 . little information about life - history trade - offs is available for marine viruses . however , two mathematical forms have been suggested for this specific trade - off in the general bacteriophage literature . one takes into account that the parental virus is utilizing limited host resources to synthesize the virions @xcite : @xmath41 where @xmath42 is the maturation rate , @xmath43 represents the eclipse period , and @xmath44 is the decay rate for the bacterial resources . the other form assumes that the time needed to deplete host resources is much larger than the latent period , therefore simplifying the exponential relationship above to a linear function @xcite : @xmath45 although the first option seems more mechanistic , most experimental evidence points to the linear relationship as the more frequently observed form for the trade - off @xcite . in spite of the lack of information , and for the sake of concreteness , we assume here that marine phages can potentially show any of these two trade - offs . thus , we perform all the calculations keeping in mind that @xmath10 is an increasing function of @xmath9 , @xmath46 , replacing later such a function by each of the two forms mentioned above . aiming to gain some knowledge on the evolutionary consequences linked to one or the other lytic descriptions , we now focus our attention on invasion experiments . invasion analysis provides a unified framework with which we can reach some classic results , together with novel ones ( see appendix [ appb ] ) . as explained above , we consider only alterations on the gene _ t _ , controlling the duration of the latent period . thus , mutants and residents differ only in @xmath9 ( and , therefore , in @xmath10 as well ) . we consider that the form of the trade - off , @xmath46 , is the same for both viral populations . if we assume that the invading mutant ( subscript @xmath42 ) perturbs the otherwise stable state of the resident population ( subscript @xmath47 ) , the possibility for invasion is decided by the sign of the invasion - matrix eigenvalue ( see appendix [ appb ] ) : @xmath48_{st}b_{m}l_{m}e^{-(k[c_{r}]_{st}+m)l_{m}}\right)-\left(k[c_{r}]_{st}+w+m\right ) , \label{dm_eigen_sol_text}\ ] ] where @xmath49 is the so - called lambert function , defined as the solution to @xmath50 @xcite . the analysis of the sign of @xmath51 provides the condition for strategy @xmath52 to resist any invasion : @xmath53 as shown in appendix [ appb ] , the solution to this equation minimizes @xmath1_{st}$ ] ( eq.([stat_delayed_c ] ) ) and maximizes @xmath54 ( eq.([dm_fitness ] ) ) . thus , @xmath52 is an ess . this result is also reached after defining the invasion fitness function , @xmath55 , and analyzing its derivatives . the result of solving eq.([dm_der_sol_text ] ) for the exponential trade - off , eq.([exp_tradeoff ] ) , and linear trade - off , eq.([lin_tradeoff ] ) , is summarized in table [ table0 ] . these results can be graphically obtained representing the pairwise invasibility plots ( pip ) @xcite for the two forms of @xmath46 . fig.[pips ] ( left ) portrays the case of the exponential trade - off . the sign of the invasion fitness depicted in the pip provides essential information . because @xmath56 for any @xmath57 when @xmath58 , these solutions for the dm ( table [ table0 ] ) resist any invasion . @xmath52 also maximizes @xmath54 , and therefore is an ess . on the other hand , @xmath59 for @xmath60 and @xmath61 , ensuring that phenotypes closer to @xmath52 can invade populations with phenotypes further from that strategy . therefore , @xmath52 is also a convergence - stable strategy ( css ) @xcite . dm & cexponential ' '' '' ' '' '' + linear ' '' '' ' '' '' & p7cm|p5@xmath62&@xmath63 ' '' '' ' '' '' + @xmath64&@xmath65 ' '' '' ' '' '' + rm & cexponential ' '' '' ' '' '' + linear ' '' '' ' '' '' & p7cm|p5@xmath66&@xmath67 ' '' '' ' '' '' + @xmath68&@xmath68 ' '' '' ' '' '' + after following similar steps to those of the previous section , the condition to be fulfilled for @xmath52 to be an ess candidate is ( see appendix [ appb ] ) : @xmath69 we can now combine eq.([rm_der_sol_text ] ) with eqs.([exp_tradeoff ] ) and ( [ lin_tradeoff ] ) to obtain @xmath52 for the exponential and linear trade - offs , respectively . in the case of the former , the evolutionarily stable strategy can be found in table [ table0 ] . this singularity maximizes fitness ( see above ) and , therefore , is an ess . furthermore , @xmath52 is , at least for the chosen parametrization , a css . these results can be easily confirmed , as in the previous case , with the numerical analysis of the derivatives of @xmath70 or plotting the corresponding pip , qualitatively similar to the one shown in fig.[pips ] ( left ) . interestingly , combining eq.([lin_tradeoff ] ) with eq.([rm_der_sol_text ] ) ( or @xmath71 ) does not offer any feasible solution . moreover , the only line providing a change of sign for the invasion fitness is the diagonal @xmath72 ( see fig.[pips ] , right panel ) . invasion analysis focuses on natural selection . the little pleiotropy expected for mutations affecting @xmath9 , and the steady environment reached under chemostat conditions , allow invasion analysis to provide reliable evolutionary predictions if there are infinitesimal differences between mutant and parent phenotypes @xcite . these limitations are shared by most of the available theoretical frameworks , which may also require ecological equilibrium as necessary condition for mutation / immigration events to occur @xcite . we present now numerical simulations aimed to check if the esss calculated above are indeed reached in the absence of these constraints . to this end , we use an eco - evolutionary framework in which new mutants are created in the system at random times not necessarily coinciding with ecological stationary states . similarly to the framework employed in @xcite , new phenotypes can be introduced periodically in the system ( invasion by migration ) , or by creating mutants through a genetic algorithm used at random times . these times are in part determined by each phenotype s population size and a common , fixed mutation rate . new phenotypes are identical to the mutating phenotype except for the latent period , which changes in an _ arbitrarily _ large amount . the simulation scheme is basically as follows : the model equations ( ecological interactions describing either the dm or the rm ) are numerically integrated ; at times calculated as specified above , a new ( invading / mutant ) virus phenotype is introduced into the system . the ecological dynamics then resumes , now with a new viral population in the medium competing against the existing ones ( namely the dominant resident and contenders ) for the host , which is their only available resource . this competition drives some phenotypes to extinction , which may change which species dominates , and thus the @xmath73 of the total viral population . these steps are repeated until , eventually , one phenotype comes along that is able to resist any invasion . this phenotype will remain as the dominant strain regardless of the other existing or incoming phenotypes . if this phenotype is indeed an ess , its fitness will be larger than that of any other strain . thus , due to its competitive advantage , the population size of the dominant phenotype will eventually be much larger than that of any other species in the system , and the average @xmath73 of the population will converge in the long term to this species trait pair , @xmath74 . the evolutionary succession described above can be observed in fig.[sims ] ( left ) . this figure portrays the average latent period of the population , as well as the latent period of the instantaneous dominant phenotype , over time . we keep in these simulations the parametrization given by table [ table ] . after a long transient , the population reaches an evolutionarily stable state , ess@xmath75 . as depicted in fig.[sims ] ( right ) , the evolutionary steady states obtained in different realizations of the framework fluctuate around the analytical ess , with a very small variance . this result is robust , for it is observed either using random mutation times or periodic immigration events . moreover , this result is also reached when an `` everything - is - everywhere '' ( eie ) approach is used @xcite . in eie approaches , a large number of fixed phenotypes , that is , with no possibility for evolutionary change , is used to initialize the system ; these phenotypes , intended to represent all possible genetic variability , compete for the available resource until only one strain remains . as we observe in fig.[sims ] ( right ) , the phenotype able to out - compete the rest is consistently close to the ess predicted analytically for the lytic strategy and trade - off used in the simulation . these results are observed for any combination of model / trade - off studied above except for the linear case of the rm for which , as deduced earlier , no ess is expected . in any case studied , an initial test simulation with monomorphic viral and host populations showed stationary values for @xmath1 $ ] , @xmath2 $ ] and @xmath3 $ ] that match the analytical solutions , confirming the assumed stability of the stationary state . the steady - state value of the observables deduced above , and how they change with environmental ( chemostat ) conditions , can give us some initial insight on the ecological behavior of the two lytic models . as stated before , the ecological outcome of the two modes is qualitatively similar . for instance , for both release models , @xmath1_{st}$ ] and @xmath54 are positively correlated with @xmath8 ( fig.[all_vs_w ] , left ) . @xmath2_{st}$ ] , on the other hand , shows non - monotonicity ( fig.[all_vs_w ] , right ) . these results remain valid for any of the two trade - off functions above . this positive correlation of the amount of resources needed ( host cells ) and viral growth rate with the dilution rate is not trivial , attending to eqs.([stat_delayed_c ] ) and ( [ dm_fitness ] ) , and eqs.([stat_rate_c ] ) and ( [ rm_fitness ] ) . for the host population , increasing the dilution rate increases the nutrient input rate , fostering host growth ; in consequence , the viral population can grow faster as well . on the other hand , we can measure the relative difference @xmath76 in the stationary concentrations of host , virus and infected cells between the two models ( defined as @xmath77_{st_{rm}}/\left[x\right]_{st_{dm}}$ ] , with @xmath78 , @xmath79 , or @xmath80 , respectively ) , as a function of @xmath10 and @xmath9 . in fig.[ecol_comparison ] , we can observe that the stationary @xmath1 $ ] is lower in the case of the rm for any feasible pair of @xmath9 and @xmath10 , while the stationary @xmath2 $ ] is larger . in this case , the exponential trade - off has been used , but a qualitatively similar pattern is observed for the linear form of the trade - off ( results not shown ) . this improved ecological performance for the rm is a consequence of the different timing between the two lytic cycles @xcite : while in the dm there is a fixed delay between the infection of any cell and its subsequent effects ( i.e. viral reproduction and disease propagation ) , in the rm distributed bursts have immediate consequences on the population . accordingly , the final viral population for the rm is larger and grows faster ; it also needs less resources ( fig.[ecol_comparison ] , and fig.[dm_v_and_muv ] ) . according to classic competition theory @xcite , this would indicate that viral populations releasing continuously their offspring ( e.g. shedding viruses ) are better competitors for any feasible combination of @xmath10 and @xmath9 . however , the fact that this strategy performs better in isolation does not necessarily mean that it can out - compete the delayed strategy when present in the same environment @xcite . moreover , the fact that delayed lysis and not shedding dominates in , e.g. marine environments highlights the limitation of these simplified models to produce reliable ecological and evolutionary predictions without the proper modifications @xcite . we can also define the rate of infection ( roi ) as @xmath81 $ ] @xcite , whose behavior parallels that of the viral population size as we have assumed @xmath6 to be constant . for the parametrization in table [ table ] , the frequency of infection in the population is never beyond @xmath82 infections per day . this number is reduced to @xmath83 infections per day when @xmath84=0 $ ] ( one - stage chemostat ) . such low values confirm the suitability of the single - infection assumption used here . moreover , the difference in roi , as @xmath85 , is negative for any realistic value of @xmath8 , @xmath9 or @xmath10 , due to the ( obvious ) higher speed of infection spread by the rate - based model . lastly , we can compare the nontrivial stationary states of the two models once dilution and other mortality events are discounted from the viral offspring . it is easy to see that , when the steady - state solutions , eq.([stat_delayed_c])-([dm_fitness ] ) and eq.([stat_rate_c])-([rm_fitness ] ) , are expressed as functions of each model s @xmath86 , eq.([av_b_delay ] ) and eq.([av_b_rate ] ) respectively , both models offer identical values for @xmath1_{st}$ ] , @xmath2_{st}$ ] , and @xmath54 , while @xmath3_{st}$ ] is larger for the dm ( fig.[newfig ] , left ) . in other words , dm viruses need more infections to maintain growth , population density and resource requirements similar to those of the rm . however , for the same values of @xmath87 , the rm shows a much larger amount of surviving offspring than the dm ( fig.[newfig ] , right ) , explaining why the former out - performs the latter in figs.[all_vs_w]-[dm_v_and_muv ] . note that these relative differences depend only on the latent period , and therefore are not influenced by the particular trade - off assumed . focusing now on the ess , for both delay and rate models not only does the ess minimize the amount of resources needed by the virus ( i.e. @xmath1_{st}$ ] , eqs.([dm_der_sol_text ] ) and ( [ rm_der_sol_text ] ) ) , but it also maximizes the viral population size and its fitness ( fig.[dm_v_and_muv ] ) . these results have been obtained theoretically in the past for the dm under low roi and optimality conditions ( e.g. @xcite ) . this explains why the ess coincides with the dominant phenotype in eie experiments : the ess makes the best use of the available resources and , thus , out - competes any other phenotype . in addition , the fitness value at the ess increases with @xmath1_{st}$ ] for both cycles ( eqs.([dm_fitness ] ) and ( [ rm_fitness ] ) ) . the hump shape shown in fig.[dm_v_and_muv ] ( right ) has been experimentally observed @xcite , pointing to the possibility of singular strategies in controlled environments . on the other hand , the dm results in a smaller fitness than the rate - based release for any value of the latent period , including the ess ( fig.[dm_v_and_muv ] , right ) . importantly , the selection gradient ( slope of the fitness function ) close to the ess is much larger for the delay model than for the rate model . this enhanced selection strength accelerates evolution , because it enhances differences between phenotypes @xcite . it is noteworthy to mention that the maturation period at the ess , @xmath88 ( step _ iii ) _ of the cycle description above ) is phage - independent for the dm . in this model , the time needed to assembly the new virions depends exclusively on the dilution rate and/or the host physiological state ( a function of @xmath44 ) . in the rm , however , maturation also depends on @xmath43 . this is due to the fact that for this model the end of the eclipse period immediately leads to a possible start of the infectious stage of the population ( release of offspring ) . this dependence on @xmath43 provides the phage with control over its entire reproductive cycle . let us delve now into the reasons for the lack of ess in the case of the rm and linear trade - off . mathematically , the condition for the mutant to invade , @xmath89_{st}<[c_{r}]_{st}$ ] , is translated into @xmath90 using eqs.([stat_rate_c ] ) and ( [ lin_tradeoff ] ) ; the phenotype with the larger burst size always invades . furthermore , there is no limit to this alternation , as there is no extrema to the viral fitness ( fig.[l_quantity_quality ] , left ) . this is owing to the fact that , for a linear relationship between burst size and latent period , the fitness cost of increasing @xmath10 and @xmath9 in the rm is always smaller than the benefit . in consequence , there is no change in sign for the invasion fitness other than that expected when the roles of mutant and resident are exchanged ( see fig.[pips ] , right ) . from an evolutionary point of view , the key is again the timing of the offspring release . in the rate - based model , the trade - off between latent period and burst size influences the average number of virions liberated per unit time in the population , but offspring start to be released instantaneously . thus , unless the resources in the host set a limit to @xmath10 ( case of the exponential trade - off , eq.([exp_tradeoff ] ) and fig.[l_quantity_quality ] , left ) , the impact on viral fitness of increasing @xmath10 at the expense of increasing @xmath9 is always positive . note that the fitness associated with the linear trade - off is always larger than that of the exponential one ( fig.[l_quantity_quality ] , left ) . on the other hand , increasing @xmath73 in the case of the delay description has a much stronger effect on viral fitness , due to the increase in the time needed to release _ any _ of the new virions @xcite . this enhanced differentiation between approaches is eventually translated into the larger selection strength observed in fig.[dm_v_and_muv ] ( right ) . finally , fig.[l_quantity_quality ] ( right ) shows the behavior of @xmath52 when @xmath8 , positively correlated with host quantity ( see fig.[all_vs_w ] ) , varies . for both models , a larger host availability or quality select for shorter latent periods ( fig.[l_quantity_quality ] , right ) . thus , improved growth conditions favor shorter generation times . this result has been observed experimentally @xcite and obtained theoretically by other means ( see e.g. @xcite ) . for the rm , however , @xmath52 is always larger than that of the dm . nonetheless , the continuous release allows the fitness in rm populations to be larger than that of dm ones to the smaller impact of varying the latent period for the former ( fig.[dm_v_and_muv ] , right ) . as commented above , host - virus models are prone to oscillatory equilibria . indeed , the models presented here can show such behavior by , e.g. increasing @xmath6 beyond @xmath91 . oscillations prevent stationary solutions from being realized , for example due to sudden population collapse or strong fluctuations around stationary values . thus , parametrizations leading to oscillations prevent the ecological and evolutionary analysis presented here from being applicable . the presence of a second chemostat ( i.e. @xmath84\ne0 $ ] ) has helped us find a realistic region of the parameter space where oscillations are not present . developing an analytical framework able to predict ecological and evolutionary behavior in the presence of such oscillations remains elusive to this date . the numerical value for the ess and the qualitative behavior described here have been observed in phage experiments ( tables [ table0 ] and [ table ] ) . however , this is not to say that real bacteriophages are close to their ( optimal ) evolutionarily stationary state . the intra- and inter - specific variability observed in measured latent periods indicates that there are other factors , not considered in this idealized study , that contribute to the exact value of the latent period shown by real viruses . for instance , constantly changing environmental conditions that include hosts potentially co - evolving with the virus will most certainly change the latent period . ] selected for in each case . this is especially true for marine environments , for which strict stationary conditions are hardly found even in stratified waters . keeping the same formalism used here , we can implement easily more realistic environments changing eq.([nutrient_equation ] ) ( for instance , incorporating remineralization of nutrients by bacteria , or periodic pulses ) , and playing with the sources of `` fresh '' host or dilution . nonetheless , the study under chemostat conditions presented here provides valuable information on the qualitative behavior of different lytic strategies . continuous cultures are very extended in the experimental literature . as shown here , in these environments populations with distributed latent periods prove to have many ecological and evolutionary advantages . everything being equal , rm populations start releasing offspring ( i.e. become infectious ) earlier , providing the rm virus with competitive advantage over the dm one . moreover , dynamic evolution of the latent period under the same environmental conditions leads the rate - based population toward an ess in which the virus utilizes less resources to produce larger population sizes and larger viral fitness with smaller generation times . the strength of selection for this model is , however , much smaller than for the case of dm viruses . smaller fitness gradients slow down evolutionary succession and , thus , may prevent the ess for rm viruses from being eventually realized . moreover , it can in principle put , e.g. shedding strategies in competitive disadvantage if the bacterial host co - evolves , because a burst lytic virus may adapt more quickly to changes in the host . even in the absence of dm viruses , rm viruses may not adapt quickly enough to host co - evolution , leading to a much slower ( or even eventually vanishing ) red - queen dynamics . however , the expected changing fitness landscape will influence non - trivially the adaptation rate in both descriptions . for viruses that can show either infection cycle , shedding provides quick invasion whereas burst lysis provides quick adaptation . on the other hand , we have mathematically proved that the linear form for the trade - off between burst size and latent period yields an endless evolutionary succession in the case of the rm . linear trade - offs have been experimentally observed for bacteriophages , and maxima for the viral fitness have been measured in the same experiments . therefore , the standard formulation of the rm presented here may lead to unrealistic predictions about bacteriophage evolution in steady environments . thus , if used to describe long - term behavior of this and other phages such as phages with non - linear virion assemblage or , e.g. shedding , more appropriate environments , biological constraints , and trade - offs need to be used . possible improvements to capture a shedding cycle correctly should include an explicit eclipse period in the exponential release distribution , and de - couple offspring release from cell mortality . on the other hand , further research is needed in order to find which functional forms for the relation between life - history traits correspond specifically to marine phages . thus , the classic modeling trade - off between simplicity and realism materializes once again . the study presented here shows that , with the trade - offs analyzed , even though applying the simpler rm model to bacteriophages provides qualitatively similar ecological results to those of the dm , the former may not be reliable for evolutionary matters , leaving the latter as the only available alternative . on the other hand , the delay terms complicate the analysis of the dm . this prevents its inclusion in bigger modeling frameworks such as models for oceanic biogeochemistry , which lack a realistic explicit representation of marine viruses . thus , a new theoretical description is needed , able to capture the essential ecological _ and _ evolutionary aspects of the marine host - virus dynamics without the use of delay terms . finding new paths to the regularization of the delay terms fulfilling these features remains as an open question . another open issue relates to finding a theoretical framework able to tackle the eco - evolutionary interactions in host - virus systems . in this paper , we focused on the analytical description of stationary states in both ecological and evolutionary timescales , ensured by the chosen environmental conditions and parametrization . however , such a framework could describe analytically more realistic situations such as the evolutionary succession ( i.e. transients ) observed in our simulations . this theoretical framework would be able to capture the feedback loop between ecology and evolution provided by rapid evolutionary events , non - vanishing evolutionary jumps , and overlapping generations . these three features break in one way or another the simplifying assumptions of the available theoretical frameworks such as adaptive dynamics . this study evidences the importance of taking into account _ both _ ecological and evolutionary aspects of the dynamics between host and virus , subject to rapid evolution . this is especially relevant if we are interested in reliable long - term predictions for the system under scrutiny . the inclusion of viruses in the description of biogeochemical cycles is one important example , as reliable estimates of viral dynamics are crucial to understand any future climate change scenario . another sound example is phage therapy , which is re - emerging as an alternative to antibiotics . the design of efficient treatments requires a reliable estimate not only of instantaneous population sizes but also of possible co - evolutionary events between phages and bacteria . the new theoretical alternatives suggested here will prove to be essential to this end . we want to thank anne maria eikeset , yoh iwasa , roger kouyos , duncan menge , alex washburne , and joshua weitz for helpful discussions in early stages of this research . we thank the editor and the two anonymous reviewers for their feedback , which has improved enormously this manuscript . we are also grateful for the support from nsf under grant oce-1046001 and by the cooperative institute for climate science ( cics ) of princeton university and the national oceanographic and atmospheric administration s ( noaa ) geophysical fluid dynamics laboratory ( gfdl ) . for both models , a trivial stationary solution is given by the phage - free state @xmath2_{st}=[i]_{st}=0 $ ] , @xmath1_{st}=w[c_{0}]/(w-\overline{\mu})$ ] , with @xmath21_{st})$ ] . stationary conditions also require that : @xmath92}{dt}=0&\longleftrightarrow\left([n_{0}]-[n]_{st}\right)=&\overline{\mu}\dfrac{[c]_{st}}{y}\\ & & = \overline{\mu}\dfrac{w[c_{0}]}{y\left(w-\overline{\mu}\right ) } , \label{nstat1}\end{aligned}\ ] ] or , rearranging terms : @xmath93-[n]_{st}\right)\left(w-\overline{\mu}\right)=\overline{\mu}\dfrac{[c_{0}]}{y } \label{nstat2}\ ] ] because @xmath21_{st})$ ] increases with @xmath94_{st}$ ] and vanishes at zero , and the two terms on the left hand side decrease with @xmath94_{st}$ ] and are positive at zero , the intermediate value theorem ensures a solution to the equation above for which both sides are positive ( i.e. @xmath95_{st})<w$ ] and @xmath96_{st}<[n_{0}]$ ] . thus , the trivial stationary state is always feasible . for both models , an analysis similar to the one above ensures that @xmath95_{st})$ ] and @xmath96_{st}<[n_{0}]$ ] . on the other hand , the necessary positivity condition for @xmath1_{st}$ ] , @xmath2_{st}$ ] , and @xmath3_{st}$ ] imposes : @xmath14_{st})\ge w\left(1-\dfrac{[c_{0}]}{[c]_{st}}\right).\ ] ] and , @xmath97 for the dm , whereas for the rm the latter condition becomes @xmath98 . if we assume that the invader perturbs the otherwise stable state of the resident population ( subindex @xmath47 ) , the dynamic equations for the mutant ( subindex @xmath42 ) can be written as : @xmath99(t)}{dt}&=&\left(b_{m}\;k[c_{r}]_{st}[v_{m}]_{t - l_{m}}\right)e^{-wl_{m}}-k[c_{r}]_{st}[v_{m}]-m[v_{m}]-w[v_{m}],\label{delayed_vm_eq}\\ \dfrac{d[i_{m}](t)}{dt}&=&k[c_{r}]_{st}[v_{m}]-\left(k[c_{r}]_{st}[v_{m}]_{t - l_{m}}\right)e^{-wl_{m}}-w[i_{m}]\label{delayed_im_eq},\end{aligned}\ ] ] where @xmath100_{st}$ ] ( @xmath89_{st}$ ] ) represents eq.([stat_delayed_c ] ) calculated using the resident ( mutant ) traits . by definition , the evolutionarily stable strategy ( ess ) can not be invaded by any mutant or immigrant phenotype . thus , the sign of the invasion eigenvalue ( i.e. eigenvalue associated with the equations above ) will provide the conditions for the phenotype ( @xmath52 , @xmath101 ) to be uninvadable . the eigenvalues @xmath51 are the result of solving @xmath102 @xcite , where @xmath103 is the jacobian matrix associated with the instantaneous terms of the equations , @xmath104 that of the delayed terms , and @xmath80 is the identity matrix . the condition above can be translated into : @xmath105_{st}\left(b_{m}\;e^{-(w+\lambda)l_{m}}-1\right)-w - m-\lambda&0\\ -k[c_{r}]_{st}\left(e^{-(w+\lambda)l_{m}}-1\right)&-w-\lambda \end{vmatrix}=\left(a(\lambda)-\lambda\right)\left(b-\lambda\right)=0 . \label{delayed_j_condition2}\ ] ] one eigenvalue is , trivially , given by @xmath106 . thus , if the other eigenvalue , resulting from solving the implicit equation @xmath107_{st}\left(b_{m}\;e^{-(w+\lambda)l_{m}}-1\right)-w - m$ ] is positive , the mutant can invade , whereas a negative value will ensure unbeatability for the resident . this remaining eigenvalue is given by : @xmath48_{st}b_{m}l_{m}e^{-(k[c_{r}]_{st}+m)l_{m}}\right)-\left(k[c_{r}]_{st}+w+m\right ) , \label{dm_eigen_sol}\ ] ] where @xmath49 is the so - called lambert function , defined as the solution to @xmath50 @xcite . the condition @xmath108 provides the marginal case : @xmath109_{st}\;\;\;\;\;\;\longleftrightarrow\;\;\;\;\;\;[c_{m}]_{st}=[c_{r}]_{st } , \label{dm_c_condition}\ ] ] for the resident to resist invasion ( i.e. @xmath110 ) , @xmath100_{st}<[c_{m}]_{st}$ ] . alternatively , for the mutant to invade ( @xmath111 ) , @xmath89_{st}<[c_{r}]_{st}$ ] . therefore , the phenotype that minimizes @xmath1_{st}$ ] will be an ess candidate . thus : @xmath4_{st}}{dl } = 0 & \longleftrightarrow&\dfrac{db}{dl}=w\;b;\nonumber\\\nonumber\\&\longleftrightarrow & f'\left(l^{*}\right)=w\;f\left(l^{*}\right ) . \label{dm_der_sol}\end{aligned}\ ] ] this condition , deduced in @xcite by other means , can also be deduced looking for the invading strategy that maximizes @xmath54 , eq.([dm_fitness ] ) . from eq.([dm_der_sol ] ) , it easily follows that @xmath112_{st}/dl^{2}$ ] is positive at @xmath52 . thus , the solution to eq.([dm_der_sol ] ) indeed provides an uninvadable strategy that maximizes fitness , i.e. the ess . equivalently , the condition given by eq.([dm_der_sol ] ) can be found by realizing that eq.([dm_eigen_sol ] ) can be used to define the invasion fitness function , @xmath55 . thus , evolutionary singularities are given by the points @xmath52 such that the derivative of @xmath113 with respect to @xmath9 vanishes , @xmath114 . it is also possible to show that both the second derivative of @xmath70 and @xmath115 are , at least for the parametrization in table [ table ] , negative when evaluated at @xmath116 . in this way , the analytical conditions for @xmath52 to be an ess and a css , respectively , are fulfilled . we now follow similar steps to deduce the expressions for the ess in the case of the rate model . the equations for the mutant are , in this case : @xmath99(t)}{dt}&=&b_{m}\;k_{l , m}[i_{m}]-k[c_{r}]_{st}[v_{m}]-m[v_{m}]-w[v_{m}]\label{rate_vm_eq}\\ \dfrac{d[i_{m}](t)}{dt}&=&k[c_{r}]_{st}[v_{m}]-k_{l , m}[i_{m}]-w[i_{m}]\label{rate_im_eq},\end{aligned}\ ] ] and the characteristic equation for the invasion eigenvalue is given by : @xmath117_{st}\\ k_{l , m}b_{m}&-k[c_{r}]_{st}-w - m-\lambda \end{vmatrix}= \begin{vmatrix } a-\lambda&b\\ c&d-\lambda \end{vmatrix}=0 . \label{rate_j_condition}\ ] ] @xmath118 and @xmath119 are by definition negative in any feasible scenario . thus , the only remaining condition to be fulfilled for the resident state to be uninvadable is , following the routh - hurwitz criteria , @xmath120 . after some algebra , this condition is translated again into @xmath100_{st}<[c_{m}]_{st}$ ] . therefore , the phenotype minimizing @xmath1_{st}$ ] will be a possible ess . thus , using eq.([stat_rate_c ] ) : @xmath4_{st}}{dl } = 0 & \longleftrightarrow&\dfrac{db}{dl}=\dfrac{w\;b}{1+w\;l};\nonumber\\\nonumber\\&\longleftrightarrow & f'\left(l^{*}\right)=\dfrac{w\;f\left(l^{*}\right)}{1+w\;l^{*}}. \label{rm_der_sol}\end{aligned}\ ] ] this same condition can be reached by defining the invasion fitness function @xmath121 and using @xmath114 . 99 c. a. suttle , marine viruses major players in the global ecosystem , nat . microbiol . 5 ( 2007 ) 801 - 812 . j. a. fuhrman , marine viruses and their biogeochemical and ecological effects , nature 399 ( 1999 ) 541 - 548 . c. a. suttle , the significance of viruses to mortality in aquatic microbial communities . ( 1994 ) 237 - 243 . c. winter , t. bouvier , m. g. weinbauer , and t. f. thingstad , trade - offs between competition and defense specialists among unicellular planktonic organisms : the `` killing the winner '' hypothesis revisited . microbiol . and mol . , 74 ( 2010 ) 42 - 57 . k. e. wommack and r. r. colwell , virioplankton : viruses in aquatic ecosystems . microbiol . and mol . 64 ( 2000 ) 69 - 114 . s. w. wilhelm and c.a . suttle , viruses as regulators of nutrient cycles in aquatic environments . in _ microbial biosystems : new frontiers _ , proceedings of the 8th international symposium on microbial ecology . bell c.r . , brylinsky m. , johnson - green p. ( ed ) , atlantic canada society for microbial ecology , halifax , canada , 1999 . m. f. marston , f. j. pierciey , jr . , a. shepard , g. gearin , j. qi , c. yandava , s. c. schuster , m. r. henn , and j. b. h. martiny , rapid diversification of coevolving marine _ synechococcus _ and a virus . natl . acad . of sci . usa 109 ( 2012 ) 4544 - 4549 . s. t. abedon , phage evolution and ecology . microbiol . 67 ( 2009 ) 1 - 45 . r. m. wilcox and j. a. fuhrman , bacterial viruses in coastal seawater : lytic rather than lysogenic production . mar . ( 1994 ) 35 - 45 . m. g. weinbauer , ecology of prokaryotic viruses . fems microbiol . 28 ( 2004 ) 127 - 181 . s. t. abedon , p. hyman , and c. thomas , experimental examination of bacteriophage latent - period evolution as a response to bacterial availability . 69 ( 2003 ) 7499 - 7506 . j. j. bull , optimality models of phage life history and parallels in disease evolution . j. of theor . 241 ( 2006 ) 928 - 938 . j. j. bull , j. millstein , j. orcutt , and h. a. wichman , evolutionary feedback mediated through population density , illustrated with viruses in chemostats . 167 ( 2006 ) e39-e51 . f. pelletier , d. garant , and a.p . hendry , eco - evolutionary dynamics . 364 ( 2009 ) 1483 - 1489 . lennon and j. b. h. martiny , rapid evolution buffers ecosystem impacts of viruses in a microbial food web . lett . 11 ( 2008 ) 1178 - 1188 . b. r. levin , f. m. stewart , and l. chao , resource - limited growth , competition , and predation : a model and experimental studies with bacteria and bacteriophage . 111 ( 1977 ) 3 - 24 . e. beretta and y. kuang , modeling and analysis of a marine bacteriophage infection . biosci . 149 ( 1998 ) 57 - 76 . m. middelboe , bacterial growth rate and marine virus - host dynamics . 40 ( 2000 ) 114 - 124 . r. j. weld , c. butts , and j. a. heinemann , models of phage growth and their applicability to phage therapy . ( 2004 ) 1 - 11 . j. s. weitz , h. hartman , and s. a. levin , coevolutionary arms races between bacteria and bacteriophage . usa 102 ( 2005 ) 9535 - 9540 . t. evans , r. g. bowers , and m. mortimer , adaptive dynamics of temperate phages . evol res . 12 ( 2010 ) 413 - 434 . j. e. pearson , p. krapivsky , and a. s. perelson , stochastic theory of early viral infection : continuous versus burst production of virions . plos comp . ( 2011 ) e1001058 . m. a. nowak and c. r. m. bangham , population dynamics of immune responses to persistent viruses . science 272 ( 1996 ) 74 - 79 . y. husimi , k. nishigaki , y. kinoshita , and t. tanaka , cellstat : a continuous culture system of a bacteriophage for the study of the mutation rate and the selection process at the dna level . . instrum . 53 ( 1982 ) 517522 . s. t. abedon , selection for bacteriophage latent period length by bacterial density : a theoretical examination . 18 ( 1989 ) 79 - 88 . h. l. smith and h. r. thieme , persistence of bacteria and phages in a chemostat . j. math . 64 ( 2012 ) 951 - 979 . f. dercole and s. rinaldi , analysis of evolutionary processes : the adaptive dynamics approach and its applications . princeton university press , princeton , nj , 2008 . e. beretta and y. kuang , modeling and analysis of a marine bacteriophage infection with latency period . nonlinear anal . real world appl . 2 ( 2001 ) 35:74 . j. s. weitz and j. dushoff , alternative stable states in host phage dynamics , theor . ecol . 1 ( 2008 ) 13 - 19 . i.- n. wang , d. e. dykhuizen , and l. b. s lobodkin , the evolution of phage lysis timing . evol . ( 1996 ) 545 - 558 . i.- n. wang , lysis timing and bacteriophage fitness . genetics 172 ( 2006 ) 17 - 26 . r. m. corless , g. h. gonnet , d. e. g. hare , d. j. jeffrey , and d. e. knuth , on the lambert @xmath122 function . adv . 5 ( 1996 ) 329 - 359 . s. a. h. geritz , e. kisdi , g. meszna , and j. a. j. metz , evolutionarily singular strategies and the adaptive growth and branching of the evolutionary tree . 12 ( 1998 ) 35 - 57 . j. j. bull , d. w. pfennig , and i.- n. wang , genetic details , optimization and phage life histories . trends ecol . ( 2004 ) 76 - 82 . d. n. l. menge and j. s. weitz , dangerous nutrients : evolution of phytoplankton resource uptake subject to virus attack . 257 ( 2009 ) 104 - 115 . j. bruggeman and s. a. l. m. kooijman , a biodiversity - inspired approach to aquatic ecosystem modeling . 52 ( 2007 ) 1533 - 1544 . m. j. follows , s. dutkiewicz , s. grant , s. w. chisholm , emergent biogeography of microbial communities in a model ocean . science 315 ( 2007 ) 1843 - 1846 . d. tilman , resource competition and community structure . princeton university press , princeton , nj , 1982 . n. l. komarova , viral reproductive strategies : how can lytic viruses be evolutionarily competitive ? , j. theor . ( 2007 ) 766 - 784 . l. m. proctor , a. okubo , and j. a. fuhrman . calibrating estimates of phage - induced mortality in marine bacteria : ultrastructural studies of marine bacteriophage development from one - step growth experiments . 25 ( 1993 ) 161 - 182 . e. l. litchman , c. a. klausmeier , o. m. schofield , and p. falkowski , the role of functional traits and trade - offs in structuring phytoplankton communities : scaling from cellular to ecosystem level . ( 2007 ) 1170 - 1181 . s. t. abedon , t. d. herschler , and d. stopar , bacteriophage latent - period evolution as a response to resource availability . microbiol . 67 ( 2001 ) 4233 - 4241 .

the magnitude - limited catalog of the southern sky redshift survey ( ssrs2 ) , is used to characterize the properties of galaxies hosting active galactic nuclei . using emission - line ratios , we identify a total of 162 ( 3% ) seyfert galaxies out of the parent sample with 5399 galaxies . the sample contains 121 seyfert 2 galaxies and 41 seyfert 1 . the ssrs2 seyfert galaxies are predominantly in spirals of types sb and earlier , or in galaxies with perturbed appearance as the result of strong interactions or mergers . seyfert galaxies in this sample are twice as common in barred hosts than the non - seyferts . by assigning galaxies to groups using a percolation algorithm we find that the seyfert galaxies in the ssrs2 are more likely to be found in binary systems , when compared to galaxies in the ssrs2 parent sample . however , there is no statistically significant difference between the seyfert and ssrs2 parent sample when systems with more than 2 galaxies are considered . the analysis of the present sample suggests that there is a stronger correlation between the presence of the agn phenomenon with internal properties of galaxies ( morphology , presence of bar , luminosity ) than with environmental effects ( local galaxy density , group velocity dispersion , nearest neighbor distance ) .

astro-ph0307180

since bright agns are such a rare phenomenon , the way a sample is selected can affect the interpretation of the observational results . as an example , the largely accepted unified model for agns is sometimes challenged by the results of some analyses that may be dependent on the sample selection ( e.g. , * ? ? ? seyferts selected on the basis of x - ray or uv excess are biased in favor of the type 1 . on the other hand , samples selected in the ir are prone to contain galaxies undergoing strong star formation activity , and also biased in favor of luminous seyfert nuclei . the fir correlates weakly with the agn emission and is very likely related to a concurrent starburst . the identification of large numbers of galaxies hosting active galactic nuclei has been possible thanks to the use of spectroscopic surveys using objective prisms ( e.g. , @xcite , kiss ; @xcite and references therein ) . however , since these surveys are specifically designed to detect galaxies with emission - lines , using their catalogues to compare the properties of agn host galaxies with the general field galaxies is not straightforward . the need of an isotropic property to assure a fair selection is therefore very important . the use of magnitude - limited surveys is an effective means of selecting galaxies in a relatively unbiased way , such that detailed comparisons between the properties of agns and the entire sample can be done . the first analysis of this kind was by @xcite , who used the cfa1 spectroscopic database to identify agn hosts , and calculated their luminosity function to determine their density relative to the entire sample of galaxies . @xcite found that agns dominate the bright end of the cfa1 survey luminosity function . a similar conclusion was reached by @xcite who used the hamburg - eso prism survey , and found that the spatial densities of seyfert 1 galaxies at @xmath00.07 are consistent with those of @xcite . a detailed analysis using a smaller sample than the cfa1 was done by @xcite and @xcite who selected galaxies brighter than @xmath1 = 12.5 , for which high resolution ( and high signal - to - noise ) spectroscopy was obtained . @xcite and @xcite showed that @xmath2 40% of all galaxies present agn - like spectra , about 2/3 of the sample being liners , while a proportion of 50% to 75% of all agn are found in bulge - dominated galaxies . the largest sample of agns to date is those compiled by @xcite and @xcite , both works using the sloan digital sky survey to detect agns . using the h@xmath3 flux as an estimator of the energy output from the agn , @xcite find that the ratio between seyfert 1 and seyfert 2 galaxies is roughly the same at low luminosities , while at higher luminosities , seyfert 1 galaxies dominate by a large margin . in that work they find that there is no correlation between the nuclear luminosity and the host galaxy luminosity . in the present work we examine the properties of agn hosts relative to the population of field galaxies . for this we use a sample of agns identified in the ssrs2 survey @xcite . this survey reaches about 1 magnitude deeper than the cfa1 and contains more than twice the number of galaxies . in this work , the ssrs2 is used as a `` control sample '' against which several properties of the agn hosts are compared , therefore allowing a systematic characterization of the properties of galaxies relative to the general field galaxy sample . this paper is divided as follows : in section 2 we describe the sample of galaxies , followed in section 3 several analyses of the properties of agn hosts as a function of the internal properties of galaxies . in section 4 we examine how the agn phenomenon correlates with the environment where the host galaxy is located . a summary of our results is presented in section 5 . in the present paper we use the following cosmological parameters : @xmath4 = 70 kms@xmath5 mpc@xmath5 , @xmath6 = 0.3 , @xmath7 = 0.7 to characterize the agn properties in the nearby universe , we used an updated version of the southern sky redshift survey ( ssrs2 , * ? ? ? * ) . this catalog contains 5399 galaxies with @xmath8 covering 1.69 sr of the southern sky , limited in the southern galactic hemisphere by @xmath9 and @xmath10 , and in the northern galactic hemisphere by @xmath11 and @xmath12 . this updated version of ssrs2 contains more precise positional information , new or improved radial velocities , revised morphological classifications , and has had removed galaxies with magnitudes above the nominal limit as well as false detections in the hst guide star catalogue . it also contains galaxies that where not included previously because of misclassification . this catalog is currently 99.99% complete in redshift , and in our database we have optical spectra available for @xmath2 68% of the ssrs2 galaxies , while the remaining radial velocities were obtained from the literature . the spectra were inspected and agns identified using the diagnostic diagrams of line intensity ratios proposed by @xcite and @xcite . we classify as seyfert galaxies objects that satisfy : @xmath13\lambda 6584 \enskip / \enskip h\alpha } \enskip > \enskip 0.6\ ] ] and @xmath14\lambda 5007 \enskip / \enskip h\beta } \enskip > \enskip 3\ ] ] in dubious cases , we also used the following ratios whenever available : @xmath15\lambda 3727 \enskip / \enskip [ oiii]\lambda 5007 } \enskip < \enskip 1\ ] ] and/or @xmath16\lambda 6300 \enskip / \enskip h\alpha } \enskip < \enskip 0.3\ ] ] we classify as seyfert 1 those galaxies with broad @xmath17 lines in their spectra . using the prescriptions described above , we found 98 seyfert galaxies . some of them were detected in the ssrs2 survey and were reported in the papers by @xcite and @xcite . in cases where the seyfert nature was dubious due to low signal to noise or small wavelength coverage of the spectra , new spectroscopic data were obtained . additional seyfert galaxies in the region covered by the ssrs2 were included after searching the ned database and the literature . the literature sometimes presents conflicting classifications , or does not state clearly how galaxies were classified . a very common confusion is to classify liners as seyferts . for all candidate objects recovered from the literature , we obtained spectra , and measured the line ratios so their seyfert nature could be confirmed . this ensures that the present catalog has a homogeneous classification procedure . a total of 64 additional galaxies were added to our list . the catalog of ssrs2 seyfert galaxies ( hereafter , agnc ) is presented in table 1 , which contains the object name , equatorial coordinates , apparent magnitude , morphological type , axial ratio @xmath18 , the heliocentric radial velocity and the seyfert type ( 1 or 2 ) . the apparent magnitudes in table 1 were derived from the hst guide star catalogue @xcite , and refer to the galaxy as a whole , there being no decomposition between the contribution from the host galaxy and the agn . the tabulated magnitudes are shown by @xcite to be isophotal magnitudes with a limiting isophote close to 26 mag arc sec@xmath19 ; these magnitudes are also very close to the @xmath20-zwicky magnitude system used in the cfa1 survey . the conversion between @xmath20-zwicky and total magnitudes is given by @xmath20-zwicky = @xmath1 + 0.26 @xcite . in the subsequent analyses , the observed magnitudes are corrected for galactic extinction using the dirbe / iras maps @xcite , and using @xmath21 = 4.035 , where we assume that the bandpass of the plates used for the gsc is the same as that used for the apm survey @xcite . in the analyses , the radial velocities of galaxies are corrected to the local group barycenter following @xcite . the absolute magnitudes are calculated using the expressions presented by @xcite . for the k - corrections we use morphological - type dependent expressions due to @xcite . in figure 1 we show the distribution of apparent magnitudes of galaxies in the agnc and the ssrs2 . this figure suggests there might be a lack of agns for magnitude bins fainter than m@xmath22=15.0 . this possible incompleteness is also suggested by the @xmath23 statistic @xcite , which for the ssrs2 sample is 0.47 , while for the agns , it is 0.38 . an independent assessment of the catalogue completeness can be made using the completeness test ( @xmath24 ) proposed by @xcite , shown in figure 2 . in this test , the completeness is estimated by calculating for each galaxy @xmath25 the ratio between the number of galaxies with absolute magnitudes @xmath26 ( @xmath27 ) and the number of galaxies ( @xmath28 ) visible to survey limit ( @xmath29 ) in both cases within the volume out to @xmath30 . from these one calculates the estimator @xmath31 which is related to the ratio between the cumulative luminosity function @xmath32 the expectation value and variance of @xmath33 are respectively : @xmath34 and @xmath35 @xmath24 is then defined as the ratio between the fluctuations of the cumulative luminosity function and its variance : @xmath36 ( see @xcite for the full derivation ) . the underlying assumption that is made in this test is that there is no dependence between an object s luminosity and its location . this test shows that the sample completeness can not be rejected at a confidence level greater than 95% , suggesting that the incompleteness of the present sample may not be severe at the adopted limiting magnitude ( @xmath37 = 15.5 ) . in figure 3 we show the radial velocity distribution of agnc and ssrs2 galaxies . for radial velocities @xmath38 13,000 km s@xmath5 we see that both samples present few galaxies , this velocity range being populated by galaxies about 1.5 magnitudes brighter than @xmath39 . the kolmogorov - smirnov test ( ks ) , applied to the ssrs2 and agnc samples shows that their velocity distributions differ at the 80% confidence level . this may be reflecting the fact that bright agns will contribute significantly to the total emission of galaxies , therefore enhancing the malmquist bias , such that brighter and more distant galaxies will be included in the sample . the average distance , @xmath40 , of the galaxies in the samples are @xmath41 mpc and @xmath42 mpc . we show in figure 4 the fraction of available spectra as a function of galaxy morphology for galaxies in the ssrs2 , where the galaxy morphologies are coded following @xcite . the ssrs2 , in general , samples spectroscopically all morphologies at a similar rate , with exception of t=-1 ( s0a ) and very late types , such as sds and irregular galaxies . this figure can also be used to estimate the fraction of agns that may have been missed in the ssrs2 because no spectra are available . by considering the fraction of galaxies with spectra in the ssrs2 ( 68% ) database , and the fraction of agns detected in each morphological type , the expected number of active galaxies in the 32% without previous spectra in our database should be of about 42 . this is less than the 64 galaxies recovered from the inspection of the literature and may be an indication that the agnc is not drastically affected by the lack of spectra for 32% of ssrs2 galaxies . in this section we use the agnc to characterize the properties of seyfert galaxies , and compare them to a representative ensemble of galaxies of the local universe the total number of seyfert galaxies identified in the ssrs2 is 162 . this comprises about 3% of the total number of galaxies in the sample ( 5399 ) . this number is consistent with the proportion of 2% found by @xcite for the 48 seyferts of the cfa1 , the 4.7% found by @xcite for the sdss and the @xmath2 4% found for kiss by @xcite . we should note that @xcite find that the revised shapley - ames catalog of bright galaxies ( rsa * ? ? ? * ) , contains 5% ( 91 galaxies ) of agns , although they claim , using completeness arguments , that agns may comprise up to 16% of the sample . this is supported also by @xcite who find that about 11% of the galaxies in a subsample of 486 nearby galaxies selected from the rsa catalog are seyferts . in a mid - ir selected sample of 891 galaxies , @xcite find a fraction of 9% of agns ( 34 type 1 and 44 type 2 ) . part of these differences may be explained by the criteria used in different works to classify a galaxy as an agn . in particular , iras samples are strongly biased towards luminous , dusty , early - type spiral galaxies . we should note that the present sample of agns identified from the ssrs2 may be deficient in low - luminosity agns , particularly when compared to works like @xcite who used a different strategy that allowed uncovering lower luminosity agns , through the use of high - dispersion spectroscopy @xcite . in the agnc catalogue , which contains 162 galaxies , there are 121 seyfert type 2 galaxies and 41 type 1 , corresponding to a ratio of @xmath2 3:1 . for the kiss , the proportion is @xmath2 2:1 @xcite , while the ratio between narrow - line and broad - line agns in the sdss is also 2:1 @xcite . other authors that measured this proportion found ratios of 4:1 for the rsa @xcite , 1:1 for the cfa1 @xcite , and 1:1 found by @xcite for their 12 @xmath43 m flux - limited sample of galaxies . we display the fraction of galaxies as a function of morphological type in figure 5 . the agnc ( solid line ) and ssrs2 ( dashed line ) distributions differ at a confidence level of 99.99% . the seyferts are distributed preferentially among early - type spirals , confirming previous results ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , with 70% of the agns between the morphological types s0a and sb , there being @xmath2 5% of agns in ellipticals and very few hosted by spirals later than sb . similar results are found by @xcite in their analysis of sdss data . although sdss has no morphologies , @xcite were able to show that most of the agn hosts in their data are found in centrally concentrated galaxies , similar to early type galaxies . there is a high proportion of seyferts in galaxies with type 15 , which contain about 10% of the total sample of agns . this morphological type is used in the ssrs2 to denote galaxies presenting evidence of advanced merger event . evidence of segregation in terms of morphologies was presented by @xcite , who claimed that seyfert 1s are more often found in earlier spirals compared to seyfert 2s . by analysing the kiss galaxies @xcite find that seyfert 2 ( and liners ) are preferentially found in redder hosts ( @xmath44 = 0.92 ) , while seyfert 1 hosts typically have bluer colors ( @xmath44=0.70 ) . this apparent contradiction between the two works can be explained by the use of integrated colors of galaxies i.e. , the contribution of the agn is very hard to be measured separately , added to the fact that the correlation between galaxy morphology and color has a large dispersion . yet both works suggest that the seyfert types are preferentially found in different types of host . the agnc was used to investigate whether this segregation in morphologies between the seyfert 1 and 2 hosts is found in the ssrs2 ( figure 6 ) . by using a ks test we find that both samples may be considered identical at the 84 % confidence level , showing that for the ssrs2 there is no strong correlation between the host galaxy morphology and the seyfert type . the distribution of absolute magnitudes , @xmath45 for the agnc and ssrs2 catalogs is displayed in figure 7 panel ( a ) , and between the seyfert types 1 and 2 in panel ( b ) . panel ( a ) clearly shows that galaxies hosting agns have a peaked distribution , while the non - actives present a tail at lower luminosities . the mean values of @xmath45 are : -20.06 @xmath46 1.32 for the ssrs2 , -20.59 @xmath46 0.87 for agns , -20.85 @xmath46 0.77 for type 1 and -20.50 @xmath46 0.89 for type 2 . the evidence here is that agns reside preferentially in high luminosity hosts , as already suggested by @xcite and confirmed by @xcite , and @xcite . since we can not isolate the light from the agn from that of the host with the present data , it is unclear how much the observed distribution is due to the host galaxy and how much is due to the agn itself . as shown by @xcite , the correlation between the @xmath47 and typical @xmath48 luminosities of galaxies with agns as measured from an hst - selected sample , is fairly tight for high - luminosity agns ( rms @xmath21 mag ) , but shows a broad dispersion once lower - luminosity agns are considered ( rms @xmath491 mag ) . this may be explained by the difficulty in isolating the contribution from the agn even in the case of hst data . this interpretation is supported by @xcite who show using sdss data , that there is no correlation between the agn luminosity and that of the host galaxy . notwithstanding , the tendency we have found may be attributed , at least in part , to the relation between the black hole mass and that of the galaxy bulge . since seyfert hosts are frequently found in spiral galaxies , it is possible that optically - selected samples may be biased against identifying agns in edge - on systems by projection effects . this is because the gas which is ionized by the nucleus is not detected along the line of sight , being obscured by the intervening gas and dust in the plane of the galaxy . the distribution of the axial ratio @xmath18 for the non - interacting spirals of the agnc is similar to that of a magnitude - limited sample of thin circular disks @xcite . the frequency distribution of axial ratios for all seyferts is presented in figure 8 panel ( a ) . the lack of objects with very small axial ratios @xmath18 is expected because of the disk thickness and also to the less numerous edge - on systems in magnitude - limited samples . the small decrease in frequencies for @xmath18 above 0.8 may be attributed to the existence of noncircular disks . in panel ( b ) of figure 8 we show the distribution of axial ratios for seyfert 2 galaxies ( solid line ) and seyfert 1 ( dashed line ) . seyfert 2s tend to be more abundant at smaller values of @xmath18 while seyfert 1s are commoner at @xmath18 closer to 1 . a ks test shows that both samples differ at the 97% confidence level . this behavior is consistent with the interpretation that part of seyfert 2 galaxies are seyfert 1s heavily obscured in more inclined systems . bars are claimed to be detected in up to 50% of normal galaxies ( e.g. , * ? ? ? * ; * ? ? ? * ) , although some authors claim that the fraction may be as high as @xmath5070% ( e.g. , * ? ? ? the presence of this non - axisymmetric gravitational perturbation can induce radial gas inflow , fueling the seyfert activity as described by @xcite . the incidence of bars as a function of morphological type was reported by @xcite , although @xcite do not find any trend . @xcite using ir hst data , find that bars are more abundant in seyfert hosts ( 73% ) than in non - seyferts ( 50% ) . on the other hand , @xcite , @xcite and @xcite find similar occurrence of barred systems in seyferts and normal galaxies . in order to detect the presence of bars , dss images of ssrs2 galaxies were visually inspected ( by magm ) . we do not differentiate between strong and weak bars . we are not able to identify the totality of bars due to the saturation of the central parts of galaxies in dss plates or the inability to detect bars in edge - on systems and for galaxies with small apparent sizes . nevertheless , we are able to detect , at least , the relative differences between the ssrs2 and agnc , since the procedure used to identify bars for both samples was identical . the fraction of barred galaxies as a function of morphological type , for the ssrs2 and the agnc , is displayed in figure 9 . we find that the ssrs2 shows no dependence between the fraction of galaxies with bars and morphology . on the other hand , the higher frequency of barred galaxies in the agnc sample is noticeable . the barred fraction in the ssrs2 is 14.1% while for the agnc is 28.4% . these distributions are different at the 99.97% confidence level . if we consider only the interval of morphological types between s0a and sbc , where we have the majority of seyferts , the result is still the same . the percentage of barred systems in the ssrs2 is smaller than that found in the rsa catalog by @xcite . this discrepancy can be partly explained by our inability to detect bars visually in galaxies with apparent diameters smaller than 0.5@xmath51 . the distribution of the fraction of galaxies with bars as a function of morphology , but now discriminating between seyfert 1 and seyfert 2 galaxies is shown in figure 10 . the distribution is fairly noisy , and by applying a ks test we find that the difference between both samples is not significant . to determine the spatial density of seyfert galaxies in the ssrs2 , we use the the non - parametric step - wise maximum likelihood method of @xcite to measure the luminosity function of agn hosts . although in principle we could calculate the parametric @xcite function , we have chosen not to do so because of the small sample size and the strong correlation between the different schechter parameters . by calculating the step - wise maximum likelihood for both the ssrs2 sample and the seyfert galaxy subsample we can compare the spatial density of agn hosts relative to the parent sample as a function of absolute magnitude . prior to the calculation of the luminosity function , galaxies with corrected radial velocities smaller than 500 kms@xmath5 or greater than 15000 kms@xmath5 are removed from the sample . the swml is calculated for galaxies in the range -22.5 @xmath52 -16.0 , and is shown in figure 11 where the agn sample is represented as solid squares . the error bars were obtained following @xcite . the histogram showing the number of galaxies per absolute magnitude bin is shown in the lower part of the figure . in figure 11 , in addition to the agnc non - parametric luminosity function , we show previous determinations by other authors converted into the cosmological model and magnitude system adopted in this paper : @xcite represented as crosses ; @xcite , open circles ; @xcite , open triangles ; @xcite , shown as solid triangles , and @xcite as stars . the figure also shows the swml calculated for the ssrs2 parent sample ( solid triangles ) . the agnc luminosity function is in very good agreement with @xcite and @xcite . the agreement with both @xcite and @xcite is not as good , which could be explained by the rather smaller samples of those two works . the agnc number density ( as well as those of * ? ? ? * ; * ? ? ? * ) are systematically lower than measured by @xcite , probably reflecting the presence of liners in the agn sample of the latter work . in figure 12 we show the ratio between the number density of agn hosts relative to the that of the parent sample as a function of absolute magnitude . in the figure we show this ratio as a differential distribution ( open triangles ) where we calculate the ratio at each magnitude bin . we also show the cumulative distribution ( solid circles ) , where all galaxies in each subsample in a given magnitude bin and brighter are added . the figure , also suggests that on average slightly more that 1% of galaxies down to @xmath45=-16 are agn hosts . however , as brighter absolute magnitudes are probed , the proportion of agn hosts increases , a trend already noted by huchra & sargent ( 1973 ) and @xcite . at @xmath39 , corresponding to @xmath53 -20.5 , agns are found in @xmath2 3% of galaxies . in this section , we examine how the agn phenomenon correlates with the environment where the host galaxy is located . an unresolved issue in the study of agns is fueling the `` central motor '' of the agn , whereby mass in the form of gas ( or stars ) is fed into the mass concentration generating the agn . among the mechanisms proposed to explain the transportation of gas from the disk of a spiral galaxy to its nucleus , is the interaction with a close galaxy ( e.g. ; * ? ? ? * ; * ? ? ? several observational studies claim that agn hosts show an excess of companions relative to normal galaxies ( e.g. ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , while other works find no evidence of such an excess of companions ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? part of this controversy may be explained by the way the agn and control samples are generated . the influence of the environment over agn phenomenon can be examined from scales ranging from the typical separation of close companions ( @xmath2100 kpc ) to that of groups and clusters ( @xmath2 1 mpc ) . in particular , at large scales , it is possible that even competing mechanisms could play a complex role . in such places the agn may be fueled from gas captured from close companion , during an interaction with a close companion . on the other hand , the central monster could be `` starved '' by the removal of gas from the galaxy through the mechanism of ram pressure stripping of the intracluster medium , as has already been proposed to explain anemic galaxies . one of the first works showing that seyfert galaxies follow the general trend of clustering in the same way as normal galaxies do , but avoiding rich clusters , was by @xcite . this trend was subsequently confirmed by @xcite , who found that the fraction of agns in rich clusters is @xmath501% , similar to the proportion of seyfert galaxies in the field galaxy population . in a sample of galaxies selected in regions of high density of galaxies , typical of the central parts of clusters and compact groups of galaxies , @xcite find an excess of seyferts when compared to a control sample of galaxies from regions with very low density of objects . this result is confirmed by @xcite , using the uzc - cg catalogue to identify compact groups of galaxies . the proportion of galaxies hosting an agn in compact groups is still not well known , and estimates range from 70% @xcite to 50% @xcite . in fact , the latter authors conclude , that after correcting for the different morphological distributions of those environments , the dense galaxy environment in the compact groups triggers neither the agn activity nor the nuclear starburst . the proportion of agns in lower density ( `` loose '' ) groups is even less well known . @xcite compare the emission - line properties of galaxies in a small sample of groups with and without seyferts . they conclude that groups containing seyferts contain more frequently objects with strong emission lines indicative of intense star formation , relative to groups without seyfert galaxies . in another analysis of loose groups of galaxies , @xcite find no significant differences in the dynamical properties ( velocity dispersions ) of groups harboring seyferts as compared to those without seyferts . we re - examine the correlation between the presence of the agn phenomenon with local density by identifying groups in the ssrs2 . this is done using a percolation algorithm applied to the ssrs2 catalog . we generated a catalog of groups of galaxies similarly to @xcite , for which we examined the distribution of seyfert galaxies . the algorithm is that described by @xcite and @xcite . since we are interested in `` real physical systems '' , the adopted friends - of - friends algorithm , searches for possible group member galaxies , keeping a fixed surrounding density contrast ( @xmath54 ) relative to the mean density of galaxies of the entire sample . groups were selected so they correspond to a density enhancement @xmath55 . this level was adopted following @xcite who showed that this density contrast level gives the best compromise between identifying as many physical loose groups as possible , including systems with high velocity dispersion , but minimizing contamination of the catalog by pseudo - groups as well as groups with interlopers . the search for companions around galaxies is carried out taking into account projected separations satisfying @xmath56 and with line - of - sight velocity differences , @xmath57 in the above expressions @xmath58 , @xmath59 and @xmath60 are the radial velocities of the galaxies , and @xmath61 their angular separation . the quantities @xmath62 and @xmath63 are search parameters scaled according to the expressions below in order to take into account the variation in the sampling of the galaxy luminosity function , @xmath64 , with distance @xmath65 where @xmath66^{-1/3}\ ] ] @xmath67 @xmath68\ ] ] @xmath69 is the selection parameter at a fixed fiducial radial velocity , @xmath70 . @xmath63 is scaled in the same way as @xmath62 . the adopted values for @xmath69 and @xmath71 are 0.352 h@xmath5mpc and 350 km s@xmath5 respectively . the apparent magnitude , @xmath72 , while the luminosity function parameters are @xmath73 galaxies mag@xmath5 mpc@xmath74 , @xmath75 and @xmath76 @xcite . the resulting group catalog contains systems with at least 4 members and mean radial velocities @xmath77 12,000 km s@xmath5 ( see @xcite for more details ) . the algorithm also outputs lists of galaxies for which no companions are found ( `` isolated '' ) , as well as binary and triple systems . the distribution of agnc and the entire ssrs2 in each of these multiplicity classes is presented in table 2 . an inspection of the table suggests that agns tend to be more common in binary systems , showing no preference for being located in groups with three or more members when compared to the control sample . the catalogue of groups was also used to examine the distributions of velocity dispersions for groups containing seyferts against the entire group sample . if galaxy interactions are supposed to enhance / trigger the nuclear activity , we should expect that the systems presenting the lowest velocity dispersions are more likely to contain seyfert hosts . the median values are 166 kms@xmath5 for groups with seyferts and 170 km s@xmath5 for the entire sample , which are essentially identical . the same conclusion is found by applying the ks - test to the distribution of velocity dispersions , therefore confirming the results of @xcite . for binaries we find a very weak tendency ( at the 50% confidence level ) for seyfert galaxies to be present in systems with higher median velocity dispersions ( 102 km s@xmath5 ) , relative to the entire sample ( 78 km s@xmath5 ) . the average group density , @xmath78 , can be calculated by means of the expression @xmath79 where , @xmath80 is the number of galaxies and @xmath81 the mean pairwise separation of a given group . the distributions of @xmath78 , show no significant difference between seyfert and non - seyfert groups for binary , triple or even higher multiplicity orders . in figure 13 we display the median values and upper and lower quartiles for @xmath78 for several group multiplicities , and there is no evidence of the agn phenomenon being correlated with the group density . one of the suggestions of @xcite is that close companion galaxies enhance the star formation rather than directly enhancing seyfert - like activity . on the other hand @xcite conclude that seyferts have on average about twice the number of companions than other galaxies in their control sample . @xcite also find a significant excess of large companions for seyfert 2 galaxies but no such evidence for seyfert 1 . on the other hand , @xcite report no difference between the presence of close companions for seyfert galaxies relative to non - active galaxies . we address this issue estimating the separation of the nearest neighbor , @xmath82 , for galaxies in our sample . in this analysis we use the ssrs2 catalogue restricting it to contain objects with velocities smaller than 12000 km s@xmath5 , therefore minimizing the problem of decreasing completeness of the sample with distance . we also neglect galaxies which are located at projected distances @xmath83 1 mpc from the survey boundaries . we further restrict the sample to galaxies with morphological types between s0a and sbc , the interval which contains approximately 80% of the seyfert galaxy sample . this is done to minimize the bias induced by the morphology - density relation , since we have a negligible fraction of seyferts hosted by elliptical galaxies , which are objects generally found in high density environments . the distributions of the nearest neighbor , @xmath82 , for agnc and ssrs2 samples were examined using the ks test which shows that both samples are part of the same parent population at the 98% confidence level . if we compare the distribution of the second nearest neighbor , a similar result is found . we find a trend of seyfert 1s presenting , on average , closer companions than the type 2s , but the result is not statistically significant . the nearest neighbor distance is also used to compute the maximum tidal influence , @xmath84 , that a companion may exert on the agn host galaxy . this tidal influence is proportional to the companion s mass divided by the cube of the separation , @xmath82 . assuming that light traces mass , the tidal parameter for each host is @xmath85 . the distribution of @xmath84 values for the agnc and ssrs2 are identical at the 86% confidence level . the seyfert types 1 and 2 are considered the same only at the 20% confidence level , with type 1s presenting higher values of @xmath84 , compared to type 2s . in the present work we have used the magnitude - limited ssrs2 to identify galaxies with active galactic nuclei , and to investigate how the properties of agn host galaxies relate to general population of galaxies in the parent sample . the main results of this work are summarized below . * we find that 162/5339 ( 3% ) of the galaxies in the ssrs2 are seyfert galaxies . this number is comparable to other works studying the local universe ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , as well as with the number of agns detected in a high redshift sample ( e.g. , * ? ? ? the ratio of seyfert 2s to 1s is about 3:1 . * the majority of the agn host galaxies in the ssrs2 are in systems classified morphologically betwen s0a and sb , which are galaxies containing significant bulge components . this result is consistent with findings of @xcite . we find that @xmath2 10% of the seyfert galaxies are found in hosts with indication of an ongoing merger event . there is no significant difference between the morphological distributions of seyfert 1 relative to seyfert 2 hosts . * the agns are preferentially detected in high luminosity hosts ( @xmath86 ) . this result is similar to previous findings ( e.g. , * ? ? ? * ) , though in conflict with that found by @xcite using the sdss database . this discrepancy can be explained by the fact that the present sample does not allow separating the contribution to the total luminosity due to the agn from that of the host galaxy . * seyfert galaxies are twice as frequent in barred hosts than in galaxies without nuclear activity . * using a percolation algorithm we find that seyfert galaxies are more commonly found in binary systems than in groups with higher multiplicities or as `` isolated '' galaxies , when compared to galaxies in the parent sample . * we do not find strong correlation between the presence of an agn with indicators of the strength of gravitational perturbation by companions , such as the projected local density , the group velocity dispersion , the nearest neighbor distance and the maximum tidal influence . * we find marginal evidence that seyfert 1 galaxies have closer companions and are more susceptible to tidal effects than seyfert 2 galaxies are . * the results above suggest that the large - scale environment does not seem to play an important role in enhancing the agn activity . on the other hand , internal characteristics such as morphology , luminosities , existence of bars and other asymmetries are more strongly correlated with the presence of the agn . the authors thank the referee , dr . john huchra , whose suggestions allowed us to improve the text . we also thank alberto rodrguez - ardilla and lei hao for useful discussions and caryl gronwall , vicki sarajedini and john salzer for providing the kiss non - parametric luminosity function . this research has made use of nasa / ipac extragalactic database ( ned ) . rsm acknowledges financial support from pibic - cnpq scholarship , magm to cnpq grant 301366/86 - 1 and faperj e-26/171.619/2001 , and cnaw to nsf ast 95 - 29028 and ast 00 - 71198 . da costa , l.n . , willmer , c.n.a , pellegrini , p.s . , chaves , o.l . , rite , c. , maia , m.a.g . , geller , m.j . , latham , d.w . , kurtz , m.j . , huchra , j.p . , ramella , m. , fairall , a.p . , smith , c. , & lipari , s. 1998 , aj 116 , 1 gronwall , c. , sarajedini , v.l . , & salzer , j.j . 2001 , in `` agn surveys '' , i.a.u . colloquium 184 , eds . green , e.ye . khachikian , & d.b . sanders asp conference series ( san francisco : asp ) , astro - ph/0111519 kauffmann , g. , heckman , t.m . , tremonti , c. , brinchmann , j. , charlot , s. , white , s.d.m . , ridgway , s. , brinkmann , j. , fukugita , m. , hall , p. , ivezic , z. , richards , g. , & schneider , d. 2003 , astro - ph/0304239 sarajedini , v.l . 2002 , in `` agn surveys , asp conference proceedings , vol . 284 . edited by r.f . green , e.ye . khachikian , and d.b . sanders . '' , ( san francisco : astronomical society of the pacific , in press ) lllrrrcrr 00 03 32.1 & -10 44 41 & ngc 7808 & 14.47 & 8922 & -2 & n & 1.00 & s1 + 00 53 29.8 & -08 46 04 & ngc 0291 & 14.64 & 5695 & 2 & y & 0.45 & s2 + 00 10 01.0 & -04 42 38 & mrk 0937 & 14.70 & 8851 & 3 & n & 1.00 & s1 + 00 10 54.3 & -21 04 03 & eso 538- g 025 & 15.02 & 7782 & 2 & n & 0.22 & s2 + 00 18 35.9 & -07 02 56 & gsc 4670 00946 & 14.36 & 5620 & 0 & n & 1.00 & s2 + 00 19 44.0 & -14 07 18 & ic 0009 & 15.42 & 12622 & 0 & n & 0.75 & s2 + 00 26 00.0 & -02 55 07 & mcg -01 - 02 - 013 & 15.49 & 20282 & 15 & n & 0.72 & s2 + 00 30 53.8 & -09 12 07 & mcg -02 - 02 - 035 & 15.28 & 5939 & -2 & n & 0.50 & s2 + 00 34 13.8 & -21 26 21 & eso 540- g 001 & 13.96 & 8030 & 4 & y & 0.69 & s1 + 00 35 48.8 & -13 36 38 & ngc 0166 & 15.18 & 6020 & 0 & n & 0.33 & s2 + 00 38 18.4 & -14 50 07 & mcg -03 - 02 - 027 & 15.40 & 10971 & 3 & n & 0.67 & s2 + 00 41 11.7 & -21 07 54 & eso 540- g 014 & 15.38 & 1653 & -5 & n & 0.50 & s2 + 00 42 52.8 & -23 32 29 & ngc 0235a & 14.18 & 6589 & 15 & n & 0.44 & s2 + 00 54 54.5 & -32 01 54 & eso 411- g 029 & 15.04 & 9622 & 0 & n & 0.38 & s2 + 00 55 02.6 & -19 00 18 & eso 541- g 001 & 13.61 & 6366 & 3 & y & 0.62 & s2 + 00 58 22.2 & -36 39 37 & eso 351- g 025 & 15.12 & 10415 & 2 & n & 0.58 & s2 + 01 02 17.4 & -19 40 09 & eso 541-ig 012 ned01 & 15.22 & 16905 & 15 & n & 0.85 & s2 + 01 08 47.6 & -15 50 34 & ic 0078 & 14.46 & 12066 & 1 & n & 0.41 & s2 + 01 11 27.5 & -38 05 01 & ngc 0424 & 13.90 & 3496 & 0 & n & 0.50 & s1 + 01 12 19.2 & -32 03 43 & ngc 0427 & 14.87 & 10011 & -2 & y & 0.62 & s1 + 01 14 07.0 & -32 39 03 & ic 1657 & 13.12 & 3564 & 3 & n & 0.25 & s2 + 01 19 25.0 & -15 41 07 & mcg -03 - 04 - 046 & 15.26 & 15234 & 4 & y & 1.00 & s2 + 01 23 54.4 & -35 03 56 & ngc 0526a & 14.66 & 5725 & 15 & n & 0.53 & s2 + 01 31 50.4 & -33 08 10 & eso 353- g 009 & 14.07 & 4970 & 2 & y & 1.00 & s2 + 01 39 24.8 & -09 24 04 & ngc 0640 & 15.45 & 7490 & 2 & n & 0.67 & s2 + 01 43 37.6 & -33 42 20 & eso 353- g 038 & 14.85 & 8884 & 5 & n & 0.36 & s2 + 01 51 41.7 & -36 11 16 & eso 354- g 004 & 15.08 & 10045 & 3 & n & 0.93 & s1 + 01 59 51.3 & -06 50 25 & ic 0184 & 14.87 & 5382 & 1 & n & 0.50 & s2 + 02 01 06.5 & -06 48 56 & ngc 0788 & 13.50 & 4078 & -5 & n & 0.77 & s2 + 02 06 53.1 & -36 27 08 & ngc 0824 & 14.05 & 5836 & 2 & y & 0.95 & s2 + 02 09 20.9 & -10 08 00 & ngc 0833 & 13.64 & 3867 & 3 & n & 0.47 & s2 + 02 10 11.4 & -09 03 36 & gsc 5281 00378 & 15.23 & 12491 & 3 & n & 0.50 & s2 + 02 24 40.6 & -19 08 30 & eso 545- g 013 & 13.59 & 10253 & 3 & n & 0.73 & s1 + 02 30 05.5 & -08 59 53 & mcg -02 - 07 - 024 & 15.25 & 4859 & -2 & n & 0.86 & s1 + 02 31 51.0 & -36 40 16 & ic 1816 & 13.66 & 5215 & 3 & n & 0.93 & s2 + 02 34 37.8 & -08 47 15 & ngc 0985 & 15.00 & 19150 & 15 & n & 1.00 & s1 + 02 35 13.5 & -29 36 17 & eso 416- g 002 & 15.06 & 17700 & 1 & n & 0.50 & s2 + 02 38 45.2 & -30 48 24 & eso 416- g 005 & 15.34 & 18587 & 5 & n & 0.80 & s1 + 02 41 04.8 & -08 15 21 & ngc 1052 & 12.08 & 1519 & -3 & n & 0.70 & s2 + 02 41 38.7 & -28 10 17 & ic 1833 & 14.28 & 4952 & -2 & n & 0.58 & s2 + 02 43 07.8 & -08 46 26 & ngc 1071 & 15.41 & 11302 & 3 & n & 0.45 & s2 + 02 44 02.9 & -26 11 11 & eso 479- g 030 & 15.29 & 10504 & 2 & n & 0.24 & s2 + 02 46 19.0 & -30 16 29 & ngc 1097 & 10.23 & 1195 & 3 & n & 0.62 & s1 + 02 49 33.8 & -38 46 12 & eso 299- g 020 & 13.96 & 5008 & 1 & n & 0.45 & s2 + 02 51 40.3 & -16 39 04 & ngc 1125 & 13.87 & 3303 & 0 & y & 0.50 & s2 + 02 56 09.8 & -13 41 07 & mcg -02 - 08 - 031 & 15.38 & 6935 & 3 & n & 0.50 & s2 + 02 56 21.5 & -32 11 06 & eso 417- g 006 & 14.34 & 4947 & -2 & n & 0.83 & s2 + 02 57 49.6 & -10 10 07 & mcg -02 - 08 - 033 & 14.02 & 4516 & 3 & n & 0.30 & s2 + 03 00 04.3 & -10 49 28 & mcg -02 - 08 - 038 & 15.27 & 9770 & 0 & y & 0.50 & s1 + 03 00 30.7 & -11 24 07 & mcg -02 - 08 - 039 & 14.57 & 8989 & 3 & y & 0.69 & s2 + 03 02 13.2 & -23 35 20 & gsc 6438 00439 & 14.93 & 10514 & -2 & n & 0.39 & s1 + 03 08 10.8 & -22 57 39 & ngc 1229 & 14.82 & 10676 & 1 & n & 0.69 & s2 + 03 11 14.7 & -08 55 20 & ngc 1241 & 13.26 & 4036 & 3 & y & 0.61 & s2 + 03 24 48.7 & -03 02 32 & ngc 1320 & 13.67 & 2698 & 2 & n & 0.32 & s2 + 03 25 04.9 & -12 18 07 & mcg -02 - 09 - 040 & 14.93 & 4400 & -2 & n & 0.33 & s2 + 03 25 25.3 & -06 08 39 & mrk 0609 & 15.06 & 10236 & -5 & n & 0.75 & s2 + 03 30 40.9 & -03 08 16 & mrk 0612 & 15.10 & 6195 & 3 & y & 0.62 & s2 + 03 33 39.7 & -05 05 22 & ngc 1358 & 13.30 & 4013 & 0 & y & 0.77 & s2 + 03 33 36.4 & -36 08 25 & ngc 1365 & 10.34 & 1659 & 3 & y & 0.71 & s1 + 03 36 46.4 & -35 59 58 & ngc 1386 & 12.64 & 924 & -2 & n & 0.38 & s2 + 03 37 03.2 & -25 14 57 & eso 482- g 014 & 15.39 & 12889 & 1 & n & 0.40 & s2 + 03 42 03.2 & -21 14 50 & eso 548- g 081 & 12.92 & 4341 & 1 & n & 0.87 & s1 + 03 43 26.5 & -31 44 38 & gsc 7028 00092 & 14.90 & 9574 & 3 & n & 0.64 & s2 + 03 45 12.5 & -39 34 30 & gsc 7569 01468 & 15.48 & 12884 & -5 & n & 0.96 & s1 + 03 55 09.3 & -17 28 10 & eso 549- g 036 & 14.64 & 8501 & 4 & y & 0.75 & s2 + 03 57 38.2 & -19 12 59 & ngc 1489 & 14.36 & 11421 & 3 & y & 0.44 & s2 + 04 02 25.7 & -18 02 51 & eso 549- g 049 & 14.36 & 7872 & 3 & n & 0.75 & s2 + 04 02 46.1 & -21 07 09 & eso 549- g 050 & 14.17 & 7532 & 3 & n & 0.67 & s2 + 04 04 27.5 & -10 10 07 & mcg -02 - 11 - 014 & 15.41 & 9267 & 2 & y & 0.62 & s2 + 04 13 49.7 & -32 00 25 & eso 420- g 013 & 13.52 & 3594 & 0 & n & 0.90 & s2 + 04 39 37.3 & -37 05 08 & gsc 7045 01444 & 15.22 & 12292 & 3 & n & 0.48 & s2 + 04 40 59.6 & -37 34 11 & eso 304- g 011 & 15.34 & 12392 & 3 & y & 0.45 & s2 + 09 42 33.3 & -03 41 55 & ngc 2974 & 12.30 & 1908 & 1 & n & 0.57 & s2 + 09 50 56.5 & -04 59 07 & mcg -01 - 25 - 049 & 14.46 & 6628 & 1 & y & 0.42 & s2 + 10 02 00.1 & -08 09 42 & gsc 5476 01170 & 15.22 & 4569 & 15 & n & 0.90 & s1 + 10 18 03.3 & -03 47 41 & gsc 4907 01728 & 15.36 & 11947 & 2 & n & 1.00 & s2 + 10 20 57.8 & -04 57 03 & gsc 4908 01685 & 15.25 & 11889 & 1 & y & 1.00 & s2 + 10 25 50.2 & -11 26 07 & mcg -02 - 27 - 003 & 15.00 & 11858 & 0 & n & 1.00 & s2 + 10 26 51.9 & -03 27 52 & ic 0614 & 15.39 & 10194 & 0 & n & 0.71 & s2 + 10 35 27.3 & -14 07 07 & mcg -02 - 27 - 009 & 14.33 & 4577 & 2 & n & 0.26 & s2 + 10 38 20.6 & -10 07 04 & gsc 5495 00478 & 15.41 & 8710 & 2 & y & 0.50 & s1 + 10 39 46.3 & -05 29 00 & mcg -01 - 27 - 031 & 14.07 & 6194 & 1 & y & 0.75 & s1 + 10 42 18.7 & -17 38 07 & mcg -03 - 27 - 026 & 14.80 & 6191 & -2 & n & 0.29 & s2 + 10 57 53.6 & -04 54 18 & ic 0657 & 15.44 & 9697 & 1 & y & 0.42 & s2 + 11 00 45.4 & -06 34 43 & gsc 4927 01023 & 15.36 & 8997 & 3 & n & 0.24 & s2 + 11 23 32.2 & -08 39 30 & ngc 3660 & 13.06 & 3678 & 4 & y & 0.87 & s2 + 11 24 56.3 & -03 48 41 & gsc 4926 01061 & 15.04 & 6788 & 2 & n & 0.40 & s2 + 11 25 02.1 & -05 04 07 & mcg -01 - 29 - 018 & 14.53 & 8518 & 2 & n & 0.90 & s2 + 11 34 23.3 & -09 40 07 & mcg -01 - 30 - 005 & 15.12 & 6419 & -3 & n & 1.00 & s2 + 11 43 18.6 & -12 52 42 & ngc 3831 & 13.60 & 5254 & 1 & n & 0.22 & s2 + 11 45 11.6 & -09 18 51 & ngc 3858 & 13.94 & 5729 & 0 & y & 0.67 & s2 + 11 45 40.5 & -18 27 16 & gsc 6096 00769 & 14.70 & 9877 & 2 & n & 0.55 & s1 + 11 52 38.2 & -05 12 07 & mcg -01 - 30 - 041 & 14.70 & 5749 & 0 & y & 0.50 & s2 + 12 00 43.3 & -20 50 01 & gsc 6100 00370 & 15.46 & 18639 & 15 & n & 0.76 & s1 + 12 16 60.0 & -26 12 36 & eso 505-ig 031 & 15.49 & 11840 & 15 & y & 0.88 & s2 + 12 39 39.4 & -05 20 39 & ngc 4593 & 12.21 & 2698 & 2 & y & 0.83 & s1 + 12 39 59.4 & -11 37 23 & ngc 4594 & 9.50 & 994 & 0 & n & 0.62 & s2 + 12 42 25.3 & -06 58 16 & ngc 4628 & 14.50 & 2828 & 1 & n & 0.23 & s2 + 12 51 32.4 & -14 13 17 & mcg -02 - 33 - 030 & 14.23 & 4298 & 1 & n & 0.43 & s1 + 12 52 12.4 & -13 24 54 & ngc 4748 & 14.27 & 4426 & 15 & n & 0.86 & s1 + 12 52 36.4 & -21 54 39 & eso 575-ig 016 ned02 & 14.87 & 6865 & 15 & n & 0.70 & s2 + 12 56 36.3 & -06 49 03 & ngc 4813 & 14.15 & 1373 & 1 & n & 0.40 & s2 + 13 00 52.9 & -13 26 59 & ngc 4897 & 13.23 & 2546 & 3 & y & 0.80 & s2 + 13 04 14.3 & -10 20 23 & ngc 4939 & 11.99 & 3117 & 3 & n & 0.38 & s2 + 13 04 13.1 & -05 33 06 & ngc 4941 & 12.43 & 1108 & 1 & n & 0.48 & s2 + 13 06 56.5 & -23 55 01 & ic 4180 & 13.88 & 2972 & 0 & y & 0.77 & s2 + 13 07 05.9 & -23 40 39 & ngc 4968 & 13.92 & 2930 & 0 & n & 0.47 & s2 + 13 10 17.3 & -07 27 07 & mcg -01 - 34 - 008 & 14.54 & 6713 & 3 & n & 0.86 & s2 + 13 17 03.4 & -02 15 41 & ic 4218 & 14.90 & 5808 & 1 & n & 0.27 & s1 + 13 19 31.7 & -12 39 26 & ngc 5077 & 12.41 & 2817 & -5 & n & 0.50 & s2 + 13 20 06.4 & -17 07 06 & mcg -03 - 34 - 049 & 15.14 & 6713 & 0 & n & 0.83 & s1 + 13 22 24.5 & -16 43 07 & mcg -03 - 34 - 064 & 14.64 & 6393 & -2 & n & 0.80 & s2 + 13 24 35.2 & -19 45 11 & gsc 6128 00106 & 14.97 & 5284 & 3 & n & 0.77 & s2 + 13 27 12.7 & -24 51 41 & gsc 6717 00254 & 15.42 & 12131 & 2 & n & 0.79 & s1 + 13 30 41.7 & -21 39 47 & gsc 6133 00078 & 15.49 & 7292 & 0 & y & 0.73 & s2 + 13 31 13.9 & -25 24 10 & eso 509- g 038 & 14.74 & 7786 & 15 & n & 0.42 & s1 + 13 32 39.1 & -10 28 53 & mcg -02 - 35 - 001 & 14.82 & 6638 & 2 & y & 0.67 & s1 + 13 33 08.3 & -23 32 37 & gsc 6714 00714 & 14.50 & 10342 & 3 & n & 0.81 & s2 + 13 34 39.6 & -23 26 48 & eso 509-ig 066 & 14.70 & 10031 & 15 & n & 0.77 & s2 + 13 37 35.1 & -21 12 58 & gsc 6134 00646 & 15.26 & 15703 & 3 & n & 0.49 & s2 + 13 37 50.0 & -23 59 41 & eso 509-ig 083 ned01 & 14.70 & 9044 & 15 & n & 0.63 & s2 + 13 40 19.7 & -23 51 29 & ngc 5260 & 13.42 & 6539 & 2 & y & 0.88 & s2 + 13 46 20.1 & -03 25 42 & mcg + 00 - 35 - 020 ned02 & 15.02 & 6948 & 2 & n & 0.38 & s2 + 13 58 59.7 & -20 02 44 & gsc 6144 00913 & 15.50 & 11797 & -5 & n & 1.00 & s2 + 14 03 26.1 & -06 01 51 & ngc 5427 & 12.50 & 2618 & 3 & n & 1.00 & s2 + 14 13 14.9 & -03 12 27 & ngc 5506 & 13.37 & 1753 & 3 & n & 0.57 & s2 + 14 26 12.3 & -11 54 07 & mcg -02 - 37 - 004 & 15.48 & 12480 & 3 & y & 0.80 & s2 + 14 33 43.5 & -14 37 11 & ngc 5664 & 14.52 & 4537 & 2 & n & 0.42 & s2 + 14 42 23.9 & -17 15 10 & ngc 5728 & 12.81 & 2793 & 4 & y & 0.58 & s2 + 14 59 24.8 & -16 41 36 & ngc 5793 & 14.17 & 3522 & 2 & n & 0.35 & s2 + 15 33 20.7 & -08 42 07 & mcg -01 - 40 - 001 & 14.64 & 6888 & 4 & y & 0.38 & s2 + 21 07 59.9 & -29 50 10 & gsc 6936 00748 & 15.10 & 5750 & 1 & n & 0.90 & s2 + 21 27 02.8 & -22 59 32 & eso 530- g 047 & 14.51 & 9761 & 2 & n & 0.59 & s2 + 21 48 19.5 & -34 57 06 & ngc 7130 & 13.33 & 4829 & 15 & n & 0.94 & s2 + 21 56 56.6 & -11 39 31 & gsc 5797 01005 & 15.37 & 17650 & 1 & n & 0.54 & s1 + 22 00 21.6 & -13 08 49 & ic 1417 & 14.36 & 5446 & 3 & n & 0.29 & s2 + 22 01 17.0 & -37 46 07 & mcg -06 - 48 - 013 & 14.82 & 10005 & -2 & n & 0.75 & s2 + 22 02 01.7 & -31 52 18 & ngc 7172 & 12.95 & 2593 & 1 & n & 0.52 & s2 + 22 08 28.3 & -34 06 23 & eso 404- g 032 & 14.33 & 4431 & 4 & n & 0.20 & s2 + 22 09 07.7 & -27 48 34 & ngc 7214 & 13.05 & 6832 & 4 & n & 0.74 & s1 + 22 14 42.0 & -38 48 23 & eso 344- g 016 & 15.08 & 11906 & 2 & y & 1.00 & s1 + 22 34 49.8 & -25 40 37 & eso 533- g 050 & 14.40 & 7928 & 0 & y & 0.89 & s2 + 22 35 45.9 & -26 03 01 & ngc 7314 & 11.88 & 1430 & 4 & n & 0.42 & s2 + 22 36 46.5 & -12 32 43 & mcg -02 - 57 - 023 & 14.82 & 7169 & 4 & n & 0.30 & s1 + 22 36 55.9 & -22 13 15 & eso 602- g 031 & 14.88 & 10101 & 3 & n & 0.50 & s1 + 22 47 47.6 & -11 48 59 & ngc 7378 & 13.64 & 2580 & 2 & y & 0.64 & s2 + 22 49 37.1 & -19 16 07 & mcg -03 - 58 - 007 & 14.84 & 9488 & 3 & y & 0.80 & s2 + 22 54 15.7 & -37 04 59 & eso 406- g 018 & 15.29 & 17215 & 15 & n & 0.54 & s2 + 22 55 01.0 & -39 39 41 & ngc 7410 & 11.61 & 1751 & 1 & y & 0.29 & s2 + 22 59 01.4 & -25 31 42 & eso 535- g 001 & 13.84 & 9015 & 2 & n & 0.94 & s2 + 23 00 47.8 & -12 55 07 & ngc 7450 & 14.00 & 3134 & 1 & y & 1.00 & s1 + 23 03 11.1 & -08 59 21 & ic 1464 & 14.90 & 7247 & 15 & n & 0.67 & s2 + 23 04 43.5 & -08 41 09 & mcg -02 - 58 - 022 & 15.00 & 14185 & 1 & n & 1.00 & s1 + 23 16 37.4 & -02 19 50 & ngc 7566 & 14.27 & 7972 & 1 & y & 0.54 & s2 + 23 18 22.5 & -04 24 59 & ngc 7592 & 14.50 & 7328 & 15 & n & 0.80 & s2 + 23 25 24.2 & -38 26 51 & gsc 8013 01279 & 14.89 & 10761 & 5 & n & 1.00 & s1 + 23 30 32.3 & -02 27 45 & mcg -01 - 59 - 027 & 15.14 & 10006 & 0 & y & 0.68 & s1 + 23 30 47.7 & -13 29 08 & ic 1495 & 14.14 & 6391 & 2 & y & 0.77 & s2 + 23 42 05.2 & -39 13 00 & gsc 8014 00236 & 15.42 & 12828 & 3 & n & 0.71 & s1 + 23 53 19.8 & -30 09 03 & eso 471-ig 037 ned02 & 15.24 & 14660 & 15 & n & 0.65 & s2 + 23 57 28.0 & -30 27 41 & mcg -05 - 01 - 013 & 14.97 & 9080 & 3 & n & 0.56 & s1 + 23 59 10.7 & -04 07 37 & ic 1524 & 13.94 & 5662 & 2 & y & 0.62 & s1 +

the dynamic response of a coated inclusion is considered in the context of active cloaking . the active cloak is achieved for a coated inclusion in the presence of membrane and flexural waves . in this paper , we investigate the design of an active cloak for a coated inclusion in three frequency regimes : the very low frequency ( monopole dominated ) range , the intermediate range , and the higher frequency range in which scattering resonances occur . in the first of these ranges , we validate previous work , which resulted in a simple mass - compensation design for the monopole scatterer , while in the second and third ranges , a combination of the use of an appropriate coating and the appropriate choice of the amplitudes of the active cloaking sources is necessary . we show that such cloaking can indeed be effective in the region of strong scattering resonances . we give closed form analytic expressions for the required amplitudes of the active cloaking sources in the three frequency regions and provide asymptotic estimates and numerical illustrations . * active cloaking of resonant coated inclusions for waves in membranes and kirchhoff plates * j. oneill , . selsil , r.c . mcphedran , a.b . movchan , n.v . movchan and c. henderson moggach department of mathematical sciences , university of liverpool , liverpool l69 7zl , uk cudos arc centre of excellence , school of physics , university of sydney , sydney , + new south wales 2006 , australia

1505.06138

cloaking of objects from detection by waves governed by the helmholtz equation is nowadays regarded as an interesting and important problem ( see , for example , @xcite , @xcite ) . the cloaking methods , well - developed for the helmholtz operator , have also been extended to advanced structured media in electromagnetism , elasticity , as well as waves in fluids ( see @xcite @xcite ) . the cloaking system can be understood as a by - pass structure , which routes a probe wave around an object , reducing the scattering by the object to a minimum . we make a distinction between passive and active cloaking . passive cloaks do not change their properties and remain unchanged for all frequencies and all types of incident waves . the drawback of a passive cloak is that its practical implementation can only be effective over a limited range of frequencies ( see , for example , @xcite , @xcite ) . active cloaks on the other hand can easily overcome this limitation in their effective frequency range , and can also adapt to various types of incident waves . active cloaking relies on being able to radiate compensating fields which cancel out perturbations of the incident or probe field by the object to be cloaked @xcite . in a set of papers @xcite@xcite , a continuous model of an active cloak for the laplace and helmholtz equations was developed . the model was set as a variational problem and was analysed mathematically in depth , and the effectiveness of the active cloaking was demonstrated . finite cloaking devices in the continuous framework were introduced to shield any objects within the given `` shielded '' region . in a recent paper @xcite , an analytical model of an approximate cloak for flexural waves was developed by introducing a finite number of active sources , whose complex amplitudes were chosen to suppress the scattered field resulting from an interaction of an incident wave with a finite obstacle . the multipole representation was used for the solution , and the required number of multipoles in the scattered field were constrained to be zero . although the truncation of the multipole representation introduces an approximation error , this approximation was shown to be entirely adequate for a sufficiently wide range of frequencies . in this paper , we investigate passive scattering reduction and active cloaking in the context of both the helmhotz equation ( applying to waves in elastic membranes ) and the fourth - order kirchhoff plate theory ( applying to flexural waves in thin elastic plates ) . we study these phenomena in three frequency ranges : in the first the frequency is sufficiently low , so that wave scattering is dominated by the monopole term , for both membranes and kirchhoff plates . we show that the following mass - compensation formula @xmath3 is applicable to reduction of scattering by an appropriately chosen coating around the inclusion . here @xmath4 are mass densities of the inclusion , coating , and ambient matrix , respectively , and @xmath5 are the interior and the exterior radii of the inclusion coating . this classical formula @xcite has been recently used @xcite in the context of scattering reduction for coated inclusions at low frequencies for which the dominance of the monopole term guarantees the applicability of the criterion ( [ mass_balance ] ) for both the membrane and kirchhoff plate waves . we show here that , while this formula is effective for membrane waves in reducing the scattering from a second - order to a fourth - order effect , the same is not in general true for kirchhoff waves , where for only very specific choices of material parameters , does the formula ( [ mass_balance ] ) deliver the reduction of scattering to fourth - order . the failure of ( [ mass_balance ] ) to deliver the reduction to fourth order scattering occurs because in general , an appropriate balance between the helmholtz and modified helmholtz parts of the kirchhoff wave expansion requires suitable choices for flexural rigidities , not just densities . we also show that , because scattering in the low frequency range is very weak , active cloaking is not necessary . as the frequency increases dipole and higher order terms become significant . we show that the choice of coating parameters is not sufficient to generate effective scattering reduction in this intermediate frequency range . a small number of control sources then delivers effective cloaking . as the frequency increases further , we enter the resonant scattering region . these resonances result in strongly enhanced scattering amplitudes , which change rapidly with frequency . this causes difficulties to arise for the application of active cloaking techniques , as suggested by the following argument . one strategy by which active cloaking might be defeated would be for the probe beam to be swept in frequency , so that the compensating fields would also have to be frequency - swept in corresponding fashion . a circumstance under which this would be difficult occurs when the object to be cloaked has resonant scattering properties . in the presence of resonances , it is difficult to avoid time lags between the change of the probe frequency and the corresponding changes required in the compensating fields around the cloaked object . these time lags inevitably increase with the sharpness ( or q factor ) of the resonance . thus , for active cloaking to work well near resonant frequencies , one needs to be able to modify the configuration of the cloaked object , to avoid rapid changes in scattering properties with frequency . the configurational change we investigate here is the use of the coating around the inclusion to deliver a flat reflectance as a function of frequency , rather than a very low reflectance . ( the latter is inevitably narrow - band at any but the lowest frequency range . ) this we show enables the active cloaking of the coated inclusion to operate effectively over the frequency region of flat reflectance . up to twelve active sources may be required , but good quality cloaking is always possible by this method . the paper is structured as follows . section [ prob_form ] includes the governing equations and the multipole representations of flexural waves near a coated inclusion . an illustration of resonance regimes and active cloaking by point sources is given in section [ membrane ] for the case of membrane waves , as a simple precursor to the more involved discussion necessary for flexural waves . in section [ flex_res ] , it is shown that active cloaking is still effective in resonance regimes at higher frequencies , as a result of the strategy described above of using an appropriate coating to move the inclusion resonances into the desired frequency range . a closed - form expression is given controlling the choice of inclusion and coating parameters , which delivers a zero monopole scattering coefficient . this generalises the classical result of konenkov @xcite to coated inclusions and active cloaking . the supplementary material gives a full analytic description of our cloaking method for both membranes and kirchhoff plates . the out - of - plane elastic displacement @xmath6 in a kirchhoff plate satisfies the equation of motion @xmath7 where @xmath8 represents active point sources placed in the exterior of the coated inclusion ; the explicit representation for @xmath9 and positioning of the point sources are discussed in section [ active_section ] . we note that the super / sub - script @xmath10 is to be replaced by @xmath11 or @xmath12 , depending on whether we are considering the inclusion , coating or exterior , respectively . in equation ( [ gov_eqn ] ) , @xmath13 is the biharmonic operator , a dot on the variable denotes the derivative with respect to time @xmath14 , @xmath15 is the mass density , @xmath16 is the thickness of the plate , @xmath17 is the corresponding region , @xmath18 $ ] is the flexural rigidity , with @xmath19 the young s modulus and @xmath20 the poisson s ratio of the corresponding elastic material . assuming time - harmonic vibrations , that is @xmath21 , outside the support of the function @xmath8 the governing equation ( [ gov_eqn ] ) can be reduced to @xmath22 here @xmath23 is the spectral parameter . consequently , @xmath24 can be written as the sum of solutions of the helmholtz and modified helmholtz equations @xmath25 where @xmath26 in the case of a circular geometry , the inclusion , the coating and the exterior are defined as follows : @xmath27 as shown in fig . [ setting ] . is incident on the coated inclusion along the @xmath28-axis . the inclusion , coating and exterior are associated with the subscripts @xmath29 , @xmath30 and @xmath12 , respectively . ] on the interface boundaries @xmath31 where the inclusion and coating meet , and @xmath32 where the coating and exterior meet , the conditions for perfect bonding , that is the continuity of the displacement , its normal derivative , the moment and the transverse force , lead to @xmath33 and , with spatial dependence suppressed , @xmath34 & = d_l \left [ \frac{\partial^2 w^{(l)}}{\partial r^2 } + \frac{\nu_l}{r } \left ( \frac{\partial w^{(l)}}{\partial r } + \frac{1}{r } \frac{\partial^2 w^{(l)}}{\partial \theta^2}\right ) \right ] , \label{moment_cont}\\ \nonumber \\ d_c \left[\frac{\partial}{\partial r } \delta_{r\theta } w^{(c ) } + \frac{1-\nu_c}{r^2 } \frac{\partial^2}{\partial \theta^2 } \left(\frac{\partial w^{(c)}}{\partial r } - \frac{w^{(c)}}{r } \right ) \right ] & = d_l \left[\frac{\partial}{\partial r } \delta_{r\theta } w^{(l ) } + \frac{1-\nu_l}{r^2 } \frac{\partial^2}{\partial \theta^2 } \left(\frac{\partial w^{(l)}}{\partial r } - \frac{w^{(l)}}{r } \right ) \right ] . \label{transverse_force__cont}\end{aligned}\ ] ] on @xmath31 we have @xmath35 and on @xmath32 we have @xmath36 . excluding the positions of point sources , the fields @xmath37 and @xmath38 can be written @xmath39 e^{in\theta } , \label{wice_rep } \\\ ] ] where @xmath10 is either @xmath11 or @xmath12 , depending on the respective region . we note that when @xmath40 , the coefficients @xmath41 and @xmath42 are identically zero to ensure that there are no forces or moments at the origin . using the representations ( [ wice_rep ] ) , equations ( [ disp_and_normal_der_disp_cont])-([transverse_force__cont ] ) give us two matrix relations @xmath43 where @xmath44 and @xmath45 are @xmath46 and @xmath47 matrices , respectively ; and @xmath48 where @xmath49 and @xmath50 are both @xmath46 matrices . the representations of these four matrices are given in the supplementary material . at this stage , it is important to emphasise that reducing the matrices @xmath44 and @xmath49 to block - diagonal structure is essential to pursue the derivation of an analytical formula for the scattering matrix @xmath51 . we note that in what follows @xmath52 denote the required matrices in block - diagonal form and @xmath53 are the associated matrices , with all these matrices being given in the supplementary material . we use the four matrices @xmath54 , @xmath55 , @xmath56 and @xmath57 to construct two further matrices @xmath58 and @xmath59 as follows : @xmath60 where the expressions which would represent the results of the matrix inversions in the previous equation are too cumbersome to be given explicitly . the final step in the construction is to cascade via a matrix product the progression from the interface between the inclusion and the inner coating to the interface between the outer coating and the matrix , giving the explicit transfer matrix @xmath61 , which incorporates the interface conditions at both interfaces : @xmath62 the transfer matrix @xmath61 can then be used to give the wave coefficients @xmath63 , @xmath64 , @xmath65 and @xmath66 in the exterior region in terms of the wave coefficients @xmath67 , @xmath68 in the inclusion region : @xmath69 the final step in this procedure is to eliminate the wave coefficients in the inclusion region to determine the @xmath70 scattering matrix @xmath71 which relates outgoing wave coefficients in the exterior region to incoming wave coefficients in the exterior region . the relevant equations resulting from the procedure are described in the supplementary material . and lead to the following matrix equation @xmath72 where on the left - hand side we have outgoing wave coefficients , and on the right - hand side incoming wave coefficients . if we now assume that the scattering occurs from a plane incident wave travelling along the @xmath28-axis , the coefficients @xmath73 associated with the inclusion , @xmath74 associated with the coating and @xmath75 associated with the exterior , respectively , can be readily found from ( [ main_matrix_eqn ] ) ( see supplementary material ) . note that in this particular case @xmath76 since @xmath77 and the @xmath64 are identically zero . it can be shown that the general expression for the scattered field is ( see , for example , @xcite , @xcite , @xcite ) @xmath78 \cos \left(n \theta\right ) , \label{wsc_monopole_dipole_general_simplified}\ ] ] where @xmath79 , denote the entries of the scattering matrix . at this point , we emphasise that the coefficient @xmath80 in the right - hand side of the expression ( [ wsc_monopole_dipole_general_simplified ] ) is in fact equal to the scattering coefficient @xmath81 ( using @xmath82 and @xmath83 ) . as has been claimed previously @xcite , with a suitably designed coating @xmath84 . ( in fact , @xmath81 can be made equal to zero exactly , with an appropriate choice of the material parameters of the coating - see ( [ cienanrel ] ) , ( [ scattering_coeff ] ) below . ) now , we assume that a finite number of active sources are applied in the exterior of the coated inclusion in order to reduce the scattering of an incident plane wave by the inclusion . here we derive formally an algebraic system for the active source amplitudes , which are computed to make the required number of multipole coefficients vanish in the representation of the outgoing scattered field . in order to find the total wave scattered from the coated inclusion , we identify two model problems illustrated in fig . [ modpr ] . these problems deal with scattering from the coated circular inclusion , with the centre at the origin . in the first model problem , the incident field is a plane wave advancing along the @xmath28-axis , whereas in the second problem , the incident field is generated by a point source of unit amplitude placed outside the coated inclusion . ( -280,-10 ) ( a ) ( -100,-10 ) ( b ) * the plane wave problem . * the first model problem involves scattering of a plane wave from the coated inclusion . the displacement field @xmath85 satisfies the equation @xmath86 supplemented by interface conditions ( [ disp_and_normal_der_disp_cont])([transverse_force__cont ] ) on the interior and exterior interfaces of the coated inclusion . here @xmath87 is the region occupied by the coated inclusion . the total field @xmath85 can be written as the sum of the incident field @xmath88 and the scattered field @xmath89 , that is @xmath90 where @xmath88 is the same as @xmath91 defined in ( [ incident_plane_wave ] ) . outside the coated inclusion , the outgoing scattered field @xmath89 has the representation @xmath92 \ , e^{i n \theta}.\ ] ] for @xmath93 , the evanescent term in the scattered field can be neglected since @xmath94 ( see ( 9.7.2 ) in @xcite ) , and hence @xmath95 * the single source model problem . * the second model problem involves scattering of a radial wave from the coated inclusion ( see fig . [ modpr](b ) ) . the displacement field , solving this model problem , is denoted by @xmath96 and satisfies the equation @xmath97 where @xmath98 denotes the dirac delta function , centred at @xmath99 . it is supplemented by the transmission conditions ( [ disp_and_normal_der_disp_cont])([transverse_force__cont ] ) . the solution @xmath96admits the representation @xmath100 and hence , @xmath101 where exponentially small terms satisfying the modified helmholtz equation , are not shown . here @xmath102 denotes the green s function for the biharmonic operator , @xmath103 is the scattered field due to the unit source at @xmath99 , and @xmath104 are constant coefficients . assuming that the coefficients @xmath105 and @xmath104 are given , active cloaking is achieved by introducing a set of @xmath106 control sources of complex amplitudes @xmath107 placed at the points @xmath99 in the exterior of the scatterer @xmath87 . after the truncation to order @xmath108 in the expansions ( [ plane_wave_scattered_field ] ) , ( [ arbitrary_source_field ] ) , we choose @xmath109 , so that the total displacement field @xmath110 is approximately equal to the incident field @xmath88 , that is @xmath111 to find @xmath112 , we substitute ( [ plane_wave_scattered_field ] ) and ( [ arbitrary_source_field ] ) into ( [ total_field ] ) , and obtain the following system of linear algebraic equations @xmath113 this algebraic system holds not only for the flexural waves in kirchhoff plates , but it is also valid for the membrane waves governed by the helmholtz equation . of course , in the latter case the coefficients @xmath114 and @xmath115 will change accordingly . it is assumed that the positions of active sources are chosen so that the system ( [ fourier_coeff_eqn ] ) is non - degenerate . in the text below , this algebraic system is implemented for the case of 4 , 8 and 12 sources positioned on a circular contour of a sufficiently large radius @xmath116 , and with its centre coinciding with the centre of the coated inclusion . this leads to an approximate active cloaking , which is efficient in the required frequency range . the numerical examples , based on our semi - analytic method , are also presented and demonstrate clearly the effectiveness of the method even in the resonant regimes . the ultimate aim of this paper is to cloak an inclusion with a suitably designed coating from flexural waves in resonant regimes using active control sources . in such regimes , a small change in frequency results in abrupt variations in scattering and thus it becomes important to eliminate frequency sensitivity of active sources . with this in mind , as a preliminary problem , it is instructive to study the interaction of flexural waves in a membrane containing a coated inclusion . this is because solutions to the biharmonic equation are superpositions of solutions to the helmholtz ( a mixture of propagating and evanescent waves ) and modified helmholtz equations ( all of which are evanescent waves ) . in this section , we give an illustrative example for active cloaking in resonant regimes in a membrane , where waves are governed by the helmholtz equation . the out - of - plane displacement of the membrane satisfies the following equations and transmission conditions in three regions , with @xmath117 , @xmath118 and @xmath119 , denoting the displacements in the exterior , coating and inclusion respectively : @xmath120 where @xmath121 is defined as @xmath122 , @xmath10 runs over @xmath123 @xmath30 or @xmath29 , and the domains @xmath124 are as in ( [ domains ] ) ; @xmath125 are the shear moduli in respective regions , and @xmath126 is the force term representing the active sources . is incident on the coated inclusion along the @xmath28-axis . the inclusion , coating and exterior are associated with the subscripts @xmath29 , @xmath30 and @xmath12 , respectively . ] we denote , as before , the radii of the inclusion and the exterior boundary of the coating by @xmath127 and @xmath128 , respectively ( see fig . [ setting_membrane ] ) . thus , the interface conditions can be written as @xmath129 and @xmath130 excluding the positions of point sources , the associated solutions of equations ( [ cidisp ] ) are @xmath131e^{in\theta } } , \quad k = i , c , e,\label{ksol}\end{aligned}\ ] ] where @xmath41 are identically zero . using the representations ( [ ksol ] ) with the interface conditions ( [ interface_cond_ic ] ) and ( [ interface_cond_ce ] ) , it is straightforward to derive the outgoing wave coefficient @xmath65 in terms of the coefficient @xmath63 ( characterising the incoming wave ) as @xmath132 where @xmath133 here @xmath134 stands for the scattering coefficient , and the representations for @xmath135 ( @xmath136 ) and @xmath137 are given in the supplementary material . we note that once the relation ( [ cienanrel ] ) is established , the coefficients @xmath138 associated with the coating , as well as the coefficient @xmath67 associated with the inclusion can be readily found ( see supplementary material ) . also note that the scattered field in this particular case can be represented as follows ( compare with ( [ wsc_monopole_dipole_general_simplified ] ) ) @xmath139 in the case of an inclusion with no coating it can be shown that formula ( [ cienanrel ] ) reduces to @xmath140 in particular , for the monopole term @xmath141 @xmath142 where @xmath143 are non - dimensional quantities ( considering a plane wave travelling in the @xmath28 direction as before , @xmath144 ) . we also note that the monopole term for static problems corresponds to a net force produced by an elastic inclusion , and is consequently zero ; this differs significantly from the dynamic case , when the inclusion possesses a non - zero mass density . in the dynamic case , the coefficient @xmath81 of the monopole term is defined by the formula ( [ efroma_no_coating ] ) , and plays an important role in scattering since it determines the scattering cross - section ( see chapter 13 of born and wolf @xcite ) . for low frequencies , as evident from equation ( [ e0e_asy ] ) , the coefficient @xmath145 is small , but , in general , it is a rapidly varying function of frequency , and may indeed show strongly resonant behaviour for high - contrast inclusions ( here caused by the ratio of @xmath146 , see fig . [ acoust_scat_uncoated_inclusion ] ) . ( figleft ) at ( 0.0,0.0 ) , @xmath147 , @xmath148 , @xmath149 , @xmath150 : ( a ) @xmath145 as a function of @xmath151 , ( b ) detail from ( a ) ( solid blue rectangle ) showing @xmath145 ( dashed green ) and the empirical fit @xmath152 ( solid grey).,title="fig : " ] ; at ( 4.5,-2.1 ) @xmath151 ; at ( -4.0,3.0 ) @xmath145 ; ( -3.63,-2.0 ) rectangle ( -3.5,-2.4 ) ; at ( 1.0,-3.0 ) ( a ) ; ( figright ) at ( 8.0,0.0 ) , @xmath147 , @xmath148 , @xmath149 , @xmath150 : ( a ) @xmath145 as a function of @xmath151 , ( b ) detail from ( a ) ( solid blue rectangle ) showing @xmath145 ( dashed green ) and the empirical fit @xmath152 ( solid grey).,title="fig : " ] ; at ( 11.7,-1.55 ) @xmath151 ; at ( 5.0,2.5 ) @xmath145 ; at ( 9.0,-3.0 ) ( b ) ; figure [ acoust_scat_uncoated_inclusion](a ) shows the monopole coefficient @xmath145 as a function of @xmath151 for a specific choice of high - contrast inclusion ; whilst the densities of the exterior and the inclusion are comparable ( @xmath153 , @xmath147 ) , the shear modulus of the exterior is tenfold that of the inclusion ( @xmath148 , @xmath149 ) . a prominent feature of the low frequency region ( enclosed by a solid blue rectangle ) , where @xmath145 is small , is that both the function @xmath145 itself and its derivative tend to zero with @xmath151 ( for which a diagram of the scattered field is given in fig . [ acoust_scat_uncoated_inclusion](b ) ) . sufficiently away from the low frequency region , the major features of fig . [ acoust_scat_uncoated_inclusion ] include a strong peak near @xmath154 , a narrow zero near @xmath155 and a rapid rise to a maximum near @xmath156 . the region of rapid variation of @xmath145 with @xmath151 between 2.62 and 3.0 would give rise to difficulties for active cloaking of a frequency - swept incident wave , and thus will be discussed in detail below . the influence of @xmath145 on the scattered field is illustrated in figs . [ omega_030_no_coating ] , [ omega_262_no_coating ] and [ omega_300_no_coating ] : fig . [ omega_030_no_coating ] shows the total displacement for a small frequency value of @xmath157 , which clearly shows a low scattering pattern [ omega_262_no_coating ] ( with @xmath155 ) and [ omega_300_no_coating ] ( with @xmath156 ) contrast the total displacement field patterns around the inclusion for the frequencies corresponding respectively to the first non - trivial zero of @xmath145 and its subsequent maximum . in fig . [ omega_262_no_coating ] scattering is primarily due to terms of dipole order , but the wave fronts behind the inclusion are practically straight . in fig . [ omega_300_no_coating ] , both monopole and dipole terms are contributing to the distortion of the wavefronts behind the inclusion . the scattering patterns change dramatically even though the frequency difference is small . + + [ membrane_nocoating_contours ] we now discuss two ways of scattering control by an elastic coating placed around an inclusion . first , we consider the very low frequency regime dominated by the monopole term , leading to the mass - compensation criterion for the choice of the parameters of the coating around the inclusion , as mentioned in the introduction ( see formula ( [ mass_balance ] ) ) . second , we analyse the higher frequency regime , which may include resonance modes . formula ( [ cienanrel ] ) leads to @xmath158 this suggests the design ( [ mass_balance ] ) for a coating which compensates the density difference between the inclusion and exterior , so that the scattering cross - section at low frequencies is minimised accordingly . in particular @xmath159 in the low frequency regime implies that ( [ mass_balance ] ) holds , that is , the average mass density of the inclusion combined with the coating equals the mass density of the exterior material . when @xmath160 is chosen according to the formula ( [ mass_balance ] ) , whilst the shear moduli of the coating @xmath161 and the exterior @xmath162 are chosen to be equal to each other , the graph of @xmath145 versus @xmath151 is given in fig . [ e0_coated_inclusion1](a ) . comparing this with fig . [ acoust_scat_uncoated_inclusion](a ) and their respective detailed views in figs [ acoust_scat_uncoated_inclusion](b ) and [ e0_coated_inclusion1](b ) , it is evident that in the low frequency regime , the scattering cross - section is substantially reduced . however , it is also shown in fig . [ e0_coated_inclusion1](a ) ( see for example , the region enclosed by the red rectangle ) that for higher frequencies we observe resonance scattering associated with rapid variation and increase of @xmath145 . consequently , the formula ( [ mass_balance ] ) is of no use in such regimes . ( figleft ) at ( 0.0,0.0 ) , @xmath147 , @xmath163 , @xmath148 , @xmath164 , @xmath149 , @xmath150 , @xmath165 : @xmath166 as a function of @xmath151 . note , the curve flattens approximately for @xmath167 $ ] . ( b ) detail from ( a ) ( green rectangle ) showing @xmath166 ( orange / dot - dashed ) and the empirical fit @xmath168 ( grey / solid).,title="fig : " ] ; at ( 4.4,-2.1 ) @xmath151 ; at ( -4.0,3.0 ) @xmath145 ; ( -3.63,-2.0 ) rectangle ( -3.5,-2.4 ) ; at ( 0.5,-3.0 ) ( a ) ; ( figright ) at ( 8.0,0.0 ) , @xmath147 , @xmath163 , @xmath148 , @xmath164 , @xmath149 , @xmath150 , @xmath165 : @xmath166 as a function of @xmath151 . note , the curve flattens approximately for @xmath167 $ ] . ( b ) detail from ( a ) ( green rectangle ) showing @xmath166 ( orange / dot - dashed ) and the empirical fit @xmath168 ( grey / solid).,title="fig : " ] ; at ( 11.5,-1.45 ) @xmath151 ; at ( 5.3,2.2 ) @xmath145 ; at ( 9.0,-3.0 ) ( b ) ; ( -0.0,-2.3 ) rectangle ( 0.8,2.5 ) ; the method of active cloaking that will be discussed in the next section , has the advantage of being efficient even at higher frequency regimes . we illustrate here the limitations of passive design , where the parameters of the coating are chosen to control the monopole term in the scattered field . at a frequency when the monopole coefficient @xmath145 is small but rapidly varying , active cloaking requires a rapid adjustment of the source amplitudes ( see the red box enclosing @xmath169 in fig . [ e0_coated_inclusion1](a ) ) . in these particular regions the scattered field increases significantly and thus we use the term resonant scattering " . to overcome this difficulty , the design of the coating can be revised to reduce the variation of the coefficient @xmath145 for this frequency range . in fig . [ e0_coated_inclusion2](a ) , we combine data from figs [ acoust_scat_uncoated_inclusion](a ) , [ e0_coated_inclusion1](a ) together with @xmath145 resulting from the revised design . the revised choice of parameters gives a flattened " curve near the resonant scattering frequency @xmath169 ( see the region enclosed by the red box ) . a magnified view is given in fig . [ e0_coated_inclusion2](b ) . figure [ e0_coated_inclusion2](c ) shows the behaviour of @xmath145 in the low frequency regime . it is noted that the mass balance criterion for the coating of the inclusion clearly reduces the monopole coefficient from being _ _ o__@xmath170 to _ _ o__@xmath171 . however , this is of little import since the monopole amplitude is already very small even for an uncoated inclusion . ( figleft ) at ( 0.0,0.0 ) , @xmath147 , @xmath148 , @xmath149 , @xmath150 ( green / dashed ) , ( ii ) a coated inclusion with parameter values : @xmath153 , @xmath147 , @xmath148 , @xmath164 , @xmath149 , @xmath150 , @xmath165 and @xmath163 ( orange / dot - dashed ) and ( iii ) a coated inclusion with parameter values : @xmath153 , @xmath172 , @xmath147 , @xmath148 , @xmath173 , @xmath149 , @xmath174 , @xmath165 ( blue / solid ) . note , the red box highlights where the blue / solid curve flattens approximately for @xmath175 $ ] , a lower frequency range than that seen in the green / dashed and orange / dot - dashed curves . ( b ) monopole amplitude for membrane wave scattering for the parameters of ( i)-(iii ) in ( a ) , but for a range of @xmath151 highlighted by the red box in ( a ) . we see strong variation in @xmath145 for an uncoated inclusion in this frequency range , but it is clear that with an appropriate choice of parameter values for the coating , we can create a region where @xmath145 remains fairly flat . ( c ) monopole amplitude for membrane wave scattering for the parameters of ( i)-(iii ) in ( a ) where @xmath151 is small . the upper two curves ( green / dashed and blue / solid ) in ( c ) correspond to second order variation of @xmath145 with @xmath151 while the third ( orange / dot - dashed ) corresponds to a fourth order variation.,title="fig : " ] ; at ( 4.4,-2.1 ) @xmath151 ; at ( -4.0,3.0 ) @xmath145 ; at ( 4.0,2.2 ) ( i ) ; at ( 4.0,0.8 ) ( ii ) ; at ( 4.0,-1.7 ) ( iii ) ; at ( 0.5,-3.0 ) ( a ) ; ( -2.58,-2.2 ) rectangle ( -2.58,-0.70 ) ; at ( -2.58,-2.5 ) @xmath176 ; ( -2.44,-2.2 ) rectangle ( -2.44,1.8 ) ; at ( -2.44,-2.5 ) @xmath177 ; ( -1.83,-2.2 ) rectangle ( -1.83,0.47 ) ; at ( -1.87,-2.5 ) @xmath176 ; at ( -1.79,-2.5 ) @xmath176 ; ( figright ) at ( 8.0,0.0 ) , @xmath147 , @xmath148 , @xmath149 , @xmath150 ( green / dashed ) , ( ii ) a coated inclusion with parameter values : @xmath153 , @xmath147 , @xmath148 , @xmath164 , @xmath149 , @xmath150 , @xmath165 and @xmath163 ( orange / dot - dashed ) and ( iii ) a coated inclusion with parameter values : @xmath153 , @xmath172 , @xmath147 , @xmath148 , @xmath173 , @xmath149 , @xmath174 , @xmath165 ( blue / solid ) . note , the red box highlights where the blue / solid curve flattens approximately for @xmath175 $ ] , a lower frequency range than that seen in the green / dashed and orange / dot - dashed curves . ( b ) monopole amplitude for membrane wave scattering for the parameters of ( i)-(iii ) in ( a ) , but for a range of @xmath151 highlighted by the red box in ( a ) . we see strong variation in @xmath145 for an uncoated inclusion in this frequency range , but it is clear that with an appropriate choice of parameter values for the coating , we can create a region where @xmath145 remains fairly flat . ( c ) monopole amplitude for membrane wave scattering for the parameters of ( i)-(iii ) in ( a ) where @xmath151 is small . the upper two curves ( green / dashed and blue / solid ) in ( c ) correspond to second order variation of @xmath145 with @xmath151 while the third ( orange / dot - dashed ) corresponds to a fourth order variation.,title="fig : " ] ; at ( 11.60,-1.80 ) @xmath151 ; at ( 5.0,2.55 ) @xmath145 ; at ( 9.0,-3.0 ) ( b ) ; ( -0.0,-2.3 ) rectangle ( 0.8,2.5 ) ; ( figdown ) at ( 3.5,-6.5 ) , @xmath147 , @xmath148 , @xmath149 , @xmath150 ( green / dashed ) , ( ii ) a coated inclusion with parameter values : @xmath153 , @xmath147 , @xmath148 , @xmath164 , @xmath149 , @xmath150 , @xmath165 and @xmath163 ( orange / dot - dashed ) and ( iii ) a coated inclusion with parameter values : @xmath153 , @xmath172 , @xmath147 , @xmath148 , @xmath173 , @xmath149 , @xmath174 , @xmath165 ( blue / solid ) . note , the red box highlights where the blue / solid curve flattens approximately for @xmath175 $ ] , a lower frequency range than that seen in the green / dashed and orange / dot - dashed curves . ( b ) monopole amplitude for membrane wave scattering for the parameters of ( i)-(iii ) in ( a ) , but for a range of @xmath151 highlighted by the red box in ( a ) . we see strong variation in @xmath145 for an uncoated inclusion in this frequency range , but it is clear that with an appropriate choice of parameter values for the coating , we can create a region where @xmath145 remains fairly flat . ( c ) monopole amplitude for membrane wave scattering for the parameters of ( i)-(iii ) in ( a ) where @xmath151 is small . the upper two curves ( green / dashed and blue / solid ) in ( c ) correspond to second order variation of @xmath145 with @xmath151 while the third ( orange / dot - dashed ) corresponds to a fourth order variation.,title="fig : " ] ; at ( 7.05,-8.05 ) @xmath151 ; at ( 0.6,-4.1 ) @xmath145 ; at ( 4.40,-9.0 ) ( c ) ; ( figleft ) at ( 0.0,0.0 ) , @xmath178 , of the displacement field , where @xmath116 is the radius of circle on which the control sources are placed , as a function of @xmath179 for @xmath155 . parameter values as for ( blue / solid ) curve in fig . [ e0_coated_inclusion2 ] : @xmath153 , @xmath172 , @xmath147 , @xmath148 , @xmath173 , @xmath149 , @xmath174 , @xmath165 , @xmath180.,title="fig : " ] ; at ( 4.4,-2.2 ) @xmath179 ; at ( -4.0,3.0 ) @xmath181 ; next , in fig . [ useless_coating ] , we present contour plots of the membrane wave amplitude for a small frequency value ; the inclusion has a coating whose density is chosen according to the formula ( [ mass_balance ] ) . the parameter values , with the exception of the mass density and elastic modulus of the coating , are chosen exactly as in fig . [ omega_030_no_coating ] ; we choose the elastic modulus of the coating equal to that of the exterior and the frequency value as @xmath182 . as in fig . [ e0_coated_inclusion2](c ) , it is apparent that the scattering has been reduced compared with fig . [ omega_030_no_coating ] , however , the benefit of such a reduction is very low , as even without the coating there is no significant scattering . for the higher frequency near @xmath183 , we have chosen the coating to flatten " the monopole term @xmath81 curve over the resonant range of @xmath151 . we then observe in fig . [ eneversusnplate_a ] that the multipole field components of higher order become non - negligible , and hence passive reduction of scattering would not be appropriate any longer . active cloaking by point sources , as described in @xcite , enables us to make the multipole coefficients vanish to the required order . as in @xcite , we add active sources of membrane waves to annul a number of multipole coefficients of the propagating scattered field . our study in @xcite confirmed that for an effective cloaking of flexural waves in a kirchhoff plate at least four sources were required . thus we first attempt to create a cloak for the coated inclusion with four active sources . these sources are symmetrically located on the @xmath28- and @xmath184-axes , all at a distance @xmath185 away from the origin . note that the symmetry of the inclusion dictates that we set the amplitudes of the sources on the @xmath184-axis to be the same . since the green s function for the helmholtz operator is given by @xmath186 where @xmath187 is the position of the point source , using graf s addition theorem for the sources located at @xmath188 and @xmath189 we obtain @xmath190 here @xmath191 are the amplitudes of the two sources located on the @xmath28-axis at @xmath192 , whereas @xmath193 is the amplitude of both sources located on the @xmath184-axis at @xmath194 ; note that these amplitudes are to be found . the @xmath179-th order coefficient for the total wave incident on the coated inclusion is thus @xmath195 as in @xcite we set the @xmath179-th order coefficient of the @xmath196 term equal to zero when @xmath197 and obtain @xmath198,\\ 4 { \cal t}_1 & = & ( q_+-q_- ) [ { \cal t}_1 h_1^{(1)}(\beta_e a_s ) + j_1 ( \beta_e a_s)],\\ 4i { \cal t}_2 & = & ( q_+ + q_- + 2p ) [ { \cal t}_2 h_2^{(1 ) } ( \beta_e a_s ) + j_2(\beta_e a_s)],\end{aligned}\ ] ] which we solve for @xmath191 and @xmath193 for given @xmath199 . finally , we write the displacement outside the coated inclusion as @xmath200 where @xmath106 needs to be chosen sufficiently large to ensure wave amplitude accuracy ( @xmath201 for all the examples considered in this paper ) . in fig . [ e0eflatmembranewithsources](a ) , we present the membrane wave amplitude for the frequency @xmath155 for an inclusion with a suitably designed coating chosen to give the flattened variation of @xmath145 in fig . [ e0_coated_inclusion2](a ) ; this is the frequency value at which a small deviation creates large fluctuations in the monopole term . comparing fig . [ e0eflatmembranewithsources](a ) with fig . [ omega_262_no_coating ] ( left ) or even with fig . [ omega_300_no_coating ] ( left ) , it is clear that the increase in strength of the monopole term results in higher scattering . figures [ e0eflatmembranewithsources](b ) and ( c ) illustrate two attempts , with four and eight active sources located at 1.57 units away from the origin , to cloak the coated inclusion . the equations to be solved to obtain the amplitudes of eight and twelve sources can be found in the supplementary material . we also remark that employing four active sources is perfectly adequate for smaller frequencies . to illustrate this , we present cloaking attempts at frequency values @xmath202 denoted by @xmath203 in fig . [ e0_coated_inclusion2](a ) , respectively . these frequency values respectively correspond to small , large and intermediate values of the monopole term , as shown in the same figure using dotted lines . the contour plots for the membrane wave amplitudes when the coated inclusion is surrounded by four active control sources are presented in fig . [ sourcesfreq0p70_0p80_1p20membrane ] : ( a ) corresponds to @xmath204 , ( b ) to @xmath205 and ( c ) to @xmath154 . in each of these plots , that on the left is with no sources , whereas the middle and the right ones have four sources , the last showing the membrane wave amplitude over a wider range of @xmath206 values . these computations confirm that the values of frequencies in the neighbourhood of the points of maximum for @xmath145 ( see fig . [ e0_coated_inclusion2 ] ) correspond to highly localised vibration modes , as seen in fig . [ e0_flat_cloak_0p80 ] . for the higher frequencies of fig . [ e0eflatmembranewithsources ] , it is necessary to use far more than four sources to achieve good quality cloaking . this is demonstrated in fig . [ e0eflatwithsources ] , where cloaking of very good accuracy is achieved using twelve sources . this observation is in good accord with the plot given in fig . [ eneversusnplate_a ] of the magnitudes of multipole field components . bearing in mind that the number of active sources is twice the value of the order showing on the horizontal axis of fig . [ eneversusnplate_a ] , it is evident that good quality cloaking requires ten or more active sources . + table [ owg_coeffs_membrane ] shows the total outgoing wave coefficients propagating away from the configuration of active sources and coated inclusion . for the case where there are no active sources , we see that there is a core set of multipole orders in @xmath207 $ ] with magnitudes of order unity . when active sources are introduced , it is clear that the method reduces to zero the targeted coefficients ( see shaded cells ) . we see an increase in magnitude for non - targeted coefficents which is generally significant for the two orders adjacent to the core . noting for example , the magnitude @xmath208 for @xmath209 with eight sources we see why the cloaking in fig . [ e0_flat_tot_ex_zoom_witheightsources157 ] is still not adequate , but with twelve sources , we see from table [ owg_coeffs_membrane ] that it is clearly adequate . as discussed in section [ prob_form ] , the fields @xmath210 of ( [ gov_eqn_simp ] ) are represented as a sum of solutions of the helmholtz and modified helmholtz equations , @xmath211 and @xmath212 respectively ( see ( [ w_eq_whh_wmhh ] ) , ( [ hh_mhh ] ) ) . in particular , when mass densities differ from the matrix to the inclusion , but elastic properties , flexural rigidities and poisson s ratio , of the inclusion , coating and ambient matrix are the same , the four transmission conditions ( [ disp_and_normal_der_disp_cont])([transverse_force__cont ] ) on each circular interface reduce to conditions of continuity of the displacement @xmath213 and its three derivatives with respect to @xmath214 , i.e. in terms of jump conditions @xmath215 = 0 , ~[\frac{\partial^k}{\partial r^k } w({\bf x } ) ] = 0 , ~\mbox{with } ~ k=1,2,3 , ~\mbox{and } { \bf x } ~\mbox{being a point on a circular interface}.\ ] ] in this case and at large wavelengths ( compared with the inclusion size ) the displacement @xmath38 in the matrix is dominated by the helmholtz wave , which satisfies the first two interface conditions ( [ disp_and_normal_der_disp_cont ] ) for the displacement and its normal derivative , with the discrepancy in two remaining interface conditions being asymptotically small ( see supplementary material ) . the mass balance formula ( [ mass_balance ] ) , applied for membrane waves to make the monopole term in the scattered field vanish , can then be used as an approximation . however , in the intermediate regime when the frequency of the incident wave becomes sufficiently large , the coupling between @xmath211 and @xmath212 becomes essential through the interface conditions ( [ moment_cont ] ) , ( [ transverse_force__cont ] ) . in the intermediate regime , the formula ( [ mass_balance ] ) is no longer valid , just as it is invalid when the elastic properties of the inclusion , coating and ambient matrix are different . as shown in the previous section for membrane waves , the coating around an inclusion can be used to control the monopole term in the expansion of the outgoing wave . note also that the desired control is not necessarily aimed at making the monopole term vanish . near the resonant frequencies of membranes , we have shown that the active cloaking can not be achieved using the simple mass balance argument of @xcite . on the other hand , the use of the coating enables one to shift away from a resonance mode , and hence to implement high - precision active cloaking by multiple sources . here , we show that kirchhoff plates allow a similar treatment of resonant regimes . however , the coated inclusion argument requires a significant rethinking , as the mass balance argument , working very well for membrane waves , proves to be inadequate for the case of kirchhoff plates . as expected , the additional features of flexural kirchhoff waves , scattered by an inclusion , are related to the interaction between the helmholtz and the modified helmholtz waves @xmath216 and @xmath217 defined in ( [ hh_mhh ] ) . the scattering of a monochromatic plane flexural wave by a circular obstacle with various boundary conditions was intensively studied by konenkov @xcite . in section 4 of @xcite , the author gives a formula for the scattered field for small values of @xmath218 ( in original notations @xmath10 stands for the wave number and @xmath219 for the radius of the inclusion ; this corresponds to @xmath220 in our notations ) at a large distance away from the obstacle . this formula translates to @xmath221 + 2 \left[\frac{d_e}{d_i}\nu_e-(1+\nu_i)\right]}{\frac{d_e}{d_i } \left[(1+\nu_i)+\frac{d_e}{d_i}(1-\nu_e)\right ] } ( \beta_e a_i)^2 + \dots\right\ } ~~ \text{as } ~\beta_e a_i \to 0.\ ] ] with the aim of making the monopole coefficient @xmath81 asymptotically zero to the leading - order as @xmath222 , we require @xmath223 + 2 \left[\frac{d_e}{d_i}\nu_e-(1+\nu_i)\right ] = 0 . \label{e0eplatenocoatingasympzero}\ ] ] this immediately highlights the difference between membrane and flexural waves in our investigation : all material properties of the inclusion and the exterior are present in equation ( [ e0eplatenocoatingasympzero ] ) for flexural waves , whereas for membrane waves the vanishing of @xmath81 to the leading - order merely involves the densities ( see formula ( [ e0e_asy ] ) ) . as in the case of an uncoated inclusion in the presence of membrane waves ( section [ nocoatingmembrane ] ) , a strong resonant behaviour of the scattering coefficient @xmath81 for a high - contrast inclusion is observed for flexural waves . figure [ flexural_scat_uncoated_inclusion](a ) shows @xmath224 for a specific choice of high - contrast material properties @xmath225 , @xmath147 , @xmath226 , @xmath227 and @xmath228 . in fig . [ flexural_scat_uncoated_inclusion](b ) we illustrate the behaviour of @xmath224 for small @xmath151 ( solid blue rectangle in fig . [ flexural_scat_uncoated_inclusion](a ) ) . we see the quadratic behaviour of the monopole term for small @xmath229 ; this is expected according to konenkov s asymptotic formula ( see the top right formula on p.156 of @xcite ) in which he states this behaviour is @xmath230 ( in the original notations @xmath231 , @xmath232 ) . to clarify , it is important to note that a comparison between fig . [ flexural_scat_uncoated_inclusion ] and fig . [ acoust_scat_uncoated_inclusion ] requires @xmath145 to be plotted as a function of @xmath229 for flexural waves ; this is because the spectral parameters are different for membrane wave and flexural wave problems , that is @xmath122 and @xmath23 , respectively . in other words , @xmath145 versus @xmath151 for membrane waves corresponds to @xmath145 versus @xmath229 ( up to a multiplicative constant @xmath233 ) , which is directly comparable to fig . [ flexural_scat_uncoated_inclusion ] . at ( 0.0,0.0 ) , @xmath147 , @xmath234 , @xmath227 , @xmath150 : ( a ) @xmath145 as a function of @xmath229 ( lower horizontal axis ) and of @xmath151 ( upper horizontal axis ) , ( b ) detail from ( a ) showing @xmath145 ( green / dashed ) and the empirical fit @xmath235 ( grey / solid ) . , title="fig : " ] ; at ( 4.2,-2.1 ) @xmath229 ; at ( 4.2,2.2 ) @xmath151 ; at ( -4.5,0.0 ) @xmath145 ; ( -3.28,-1.8 ) rectangle ( -3.15,-2.2 ) ; at ( 0.2,-3.0 ) ( a ) ; ( figright ) at ( 8.0,0.0 ) , @xmath147 , @xmath234 , @xmath227 , @xmath150 : ( a ) @xmath145 as a function of @xmath229 ( lower horizontal axis ) and of @xmath151 ( upper horizontal axis ) , ( b ) detail from ( a ) showing @xmath145 ( green / dashed ) and the empirical fit @xmath235 ( grey / solid ) . , title="fig : " ] ; at ( 11.2,-1.50 ) @xmath229 ; at ( 4.5,0.0 ) @xmath145 ; at ( 11.2,1.6 ) @xmath151 ; at ( 8.5,-3.0 ) ( b ) ; for three different frequencies we give flexural wave spatial distributions in fig . [ e0e_plates_small_min_max ] . for the low frequency ( @xmath157 ) the scattering amplitudes are small . the second frequency considered is @xmath236 , the first non - trivial zero of @xmath145 where again vibrational amplitudes are small . going from @xmath237 to @xmath238 , causes dramatic changes in both the spatial distribution around the inclusion and in its amplitude variation . + + to obtain a uniform asymptotic formula for @xmath81 in the case of a coated inclusion in a thin plate is a daunting task . nonetheless , our aim in this section is to make the monopole coefficient @xmath81 asymptotically zero to the leading order , which compared to the previous is somewhat easier . the equation which must to be satisfied in order to make the scattering coefficient zero ( to the leadingorder ) can be derived as @xmath239 \,a_c^4 + [ 2d_c \beta_c^4 - 2d_i \frak{a}_2 \frak{b}_2 \beta_i^4 ] \,a_i^4 \nonumber \\ & & + [ 4 d_e \frak{a}_2 ( \frak{b}_2 + d_e ) \beta_e^4 + \frak{a}_2 \frak{b}_2 ^ 2 \beta_c^2 \beta_e^2 -2d_i \frak{a}_1 \frak{b}_1 \beta_i^4 + 2d_c ( \frak{a}_1 \frak{b}_1 - \frak{a}_2 \frak{b}_2 ) \beta_c^4 ] \ , a_i^2 a_c^2 = 0 , \label{e0eplateasympleadorder}\end{aligned}\ ] ] where @xmath240 this fairly unpleasant formula should of course lead us to the result obtained by konenkov @xcite under the assumption of identical material parameter values , for example of the exterior and the coating , hence reducing the problem to the case when there is no coating around the inclusion . assuming @xmath241 and thus @xmath242 , the left - hand side of equation ( [ e0eplateasympleadorder ] ) indeed simplifies to @xmath243 in ( [ e0eplatenocoatingasympzero ] ) . it is also important to note that in the very special case of @xmath244 , @xmath245 but @xmath246 , equation ( [ e0eplateasympleadorder ] ) reduces to equation ( [ mass_balance ] ) which is exactly what is expected . the mass - compensation equation ( [ mass_balance ] ) is , in general , inadequate to annul the leading - order part of the monopole term @xmath81 . if one uses the mass - compensation formula unthinkingly , the result may be beneficial or detrimental but in either case the result will , in general , still be of second order in @xmath229 . finally , we draw attention to equation ( [ e0eplateasympleadorder ] ) noting that several parameters associated with the coating may be fine tuned to satisfy the equation . this is in stark contrast to the case of membrane waves , where density of the coating was the only parameter available for fine - tuning to minimise the monopole term . ( figleft ) at ( 0.0,0.0 ) , @xmath147 , @xmath247 , @xmath227 , @xmath248 , @xmath249 , @xmath150 ( green / dashed ) , ( ii ) a coated inclusion with parameter values as in ( i ) but with material properties of the coating reading : @xmath250 , @xmath251 , @xmath252 and @xmath165 ( blue / solid ) . ( b ) flexural wave scattering for the parameters of ( i ) and ( ii ) in ( a ) , but for a range of @xmath229 highlighted by the red box in ( a ) . we see strong variation in @xmath145 for an uncoated inclusion in this range for @xmath229 , but it is clear that with an appropriate choice of parameter values , we can create a region where @xmath145 remains fairly flat.,title="fig : " ] ; at ( 4.1,-2.0 ) @xmath229 ; at ( 4.05,2.2 ) @xmath151 ; at ( -3.8,3.0 ) @xmath145 ; at ( 4.0,0.15 ) ( ii ) ; at ( 4.0,1.0 ) ( i ) ; at ( 0.5,-3.0 ) ( a ) ; ( figright ) at ( 8.4,0.0 ) , @xmath147 , @xmath247 , @xmath227 , @xmath248 , @xmath249 , @xmath150 ( green / dashed ) , ( ii ) a coated inclusion with parameter values as in ( i ) but with material properties of the coating reading : @xmath250 , @xmath251 , @xmath252 and @xmath165 ( blue / solid ) . ( b ) flexural wave scattering for the parameters of ( i ) and ( ii ) in ( a ) , but for a range of @xmath229 highlighted by the red box in ( a ) . we see strong variation in @xmath145 for an uncoated inclusion in this range for @xmath229 , but it is clear that with an appropriate choice of parameter values , we can create a region where @xmath145 remains fairly flat.,title="fig : " ] ; at ( 12.00,-1.80 ) @xmath229 ; at ( 12.00,1.90 ) @xmath151 ; at ( 4.9,2.7 ) @xmath145 ; at ( 9.0,-3.0 ) ( b ) ; ( 0.08,-2.2 ) rectangle ( 1.4,2.1 ) ; ( figleft ) at ( 0.0,0.0 ) against the multipole order @xmath179 for @xmath236 . parameter values as for ( blue / solid ) curve in fig . [ e0_coated_inclusion2_plates ] : @xmath225 , @xmath250 , @xmath147 , @xmath253 , @xmath248 , @xmath252 , @xmath227 , @xmath249 , @xmath174 , @xmath165 , @xmath180.,title="fig : " ] ; at ( 4.4,-2.1 ) @xmath179 ; at ( -3.85,3.0 ) @xmath254 ; here we implement the algorithm of section [ algebsys ] for active cloaking in regimes , where the frequency is sufficiently high , so that the monopole coefficient may vary rapidly as a function of frequency . in this case , the coating is introduced to provide a smooth behaviour of @xmath255 in the required frequency range . we also comment on the low - frequency regime already discussed above for the case of membrane waves . explicit analytical representations for amplitudes of active sources are included in the supplementary material . as noted in the text above , the scattering produced by the elastic inclusion , in the limit when the wavelength of the incident plane wave tends to infinity compared to the size of the inclusion , is very weak . in such a regime , the scattering pattern is also dominated by the helmholtz type waves discussed earlier . however , as the frequency of the incident wave continues to increase , the role played by the modified helmholtz waves becomes more important , and we show in this section that a small number of active sources is capable of producing a substantial reduction of the scattered field . in fig . [ foursources_platea ] , we present the field plot , where @xmath256 , and hence @xmath257 , the wavelength of the incident plane wave is evidently comparable with the diameter of the elastic inclusion . the geometrical and elastic parameters of the inclusion and the coating are the same as in fig . [ e0eflatplate ] . figure [ foursources_platea ] shows a non - negligible perturbation of the incident wave by the inclusion . as described in section [ algebsys ] , four active sources can be introduced to compensate for the action produced by the coated inclusion on the incident field . the symmetric configuration of sources , shown in fig . [ foursources_plateb ] , is used in the simulation , and the amplitudes of the active sources ( as outlined in section [ algebsys ] ) are derived in an explicit closed form ( see supplementary material ) . the cloaking with four sources , as in fig . [ foursources_plateb ] , is quite successful in compensating for the scattering by the coated inclusion for @xmath256 ; beyond this frequency four sources will no longer be adequate . in the following section , we introduce larger number of sources and demonstrate how the larger - order multipole terms can be suppressed in higher - frequency regimes . the multipole representation ( [ total_field ] ) for the total field is used to produce the numerical data discussed below . the sources are placed on the circle of radius @xmath258 , outside the coated inclusion , whose external radius is @xmath259 . the simulations are presented for the cases of @xmath260 , @xmath261 and @xmath262 sources , together with tabular data on the values of high - order multipole coefficients . in fig . [ e0eflatwithsourcesplate ] , we present three cases corresponding to an incident plane wave scattered by a coated inclusion . the parameters of the coating are chosen to be the same as in the simulation depicted in figure [ e0eflatplate ] . the case when active sources are absent ( fig . [ e0eflatwithsourcesplate]a ) is compared with the two cases incorporating four and eight active sources , respectively . although the scattering pattern has been reduced , in this higher - frequency regime it is clear from fig . [ eneversusnplate ] that around twelve sources are required to make the significant multipole coefficients vanish . the required combination of twelve sources , with the amplitudes computed according to ( [ fourier_coeff_eqn ] ) is given in fig . [ e0eflatwithsourcesplatea ] , which shows exemplary quality of cloaking , with the scattered field being suppressed . it is also noted that figs . [ e0eflatwithsourcesplatea]a and [ e0eflatwithsourcesplatea]b use a different colour code scheme , so that the diagram ( a ) plots the field in the exterior of the ambient matrix only , while the diagrams ( b ) and ( c ) show the field in the inclusion , which also indicates the field localisation . the data in table [ ogw_coeff_plates ] for multipole coefficients demonstrates that the core region of multipole orders runs from @xmath263 $ ] . our procedure reduces all the amplitudes of targeted orders to zero , while increasing the moduli of adjacent orders . by inspecting the amplitudes of orders @xmath264 for four sources and @xmath265 for eight sources , we can see that in both cases we have an insufficient number of sources for successful cloaking . for twelve sources , both the entries in table [ ogw_coeff_plates ] and the displacement field components shown in fig . [ e0eflatwithsourcesplatea ] confirm good quality cloaking . this paper has presented a novel idea of active cloaking for coated inclusions subjected to incident waves of frequencies ranging from zero up to higher frequency regimes , where the monopole term in the outgoing scattered field can be a rapidly varying function of frequency . the technique involves choosing the coating to diminish the frequency sensitivity of the scattering problem , and then applying active control sources . a comparison between the membrane waves and flexural waves in kirchhoff plates has revealed interesting connections for low frequencies . in particular , we have identified the cases when the flexural wave scattering is driven by the helmholtz waves . on the other hand , we have also identified important differences where the evanescent waves which are present in the expressions for the solutions to the fourth - order problem contribute significantly . the utility of the proposed algorithm is high , as the amplitudes of active sources are given in the explicit analytical form , and the accuracy of the cloaking of coated inclusions in resonant regimes has been well demonstrated . j. oneill would like to greatly acknowledge the support from the epsrc through the grant ep / l50518/1 . . selsil , r.c . mcphedran and n.v . movchan acknowledge the financial support of the european commission s seventh framework programme under the contract number piapp - ga-284544-parm-2 . movchan acknowledges the support from epsrc programme grant ep / l024926/1 . d. j. colquitt , i. s. jones , n. v. movchan , a. b. movchan , m. brun and r. c. mcphedran , making waves round a structured cloak : lattices , negative refraction and fringes , _ proc . a _ , * 469 * , ( 2013 ) , 20130218 . f. guevara vasquez , g.w . milton , d. onofrei and p. seppecher , transformation elastodynamics and active exterior acoustic cloaking , _ acoustic metamaterials , springer series in materials science _ , * 166 * , ( 2013 ) , 289 - 318 . j. oneill , . selsil , r.c . mcphedran , a.b . movchan and n.v . movchan , active cloaking of inclusions for flexural waves in thin elastic plates , _ quart . j. mech . _ , to appear . preprint available as active cloaking of finite defects for flexural waves in elastic plates " , _

supersymmetry is the most natural framework for physics above the tev scale , and the corresponding framework for early - universe cosmology , including inflation , is supergravity . no - scale supergravity emerges from generic string compactifications and yields a non - negative potential , and is therefore a plausible framework for constructing models of inflation . no - scale inflation yields naturally predictions similar to those of the starobinsky model based on gravity , with a tilted spectrum of scalar perturbations : , and small values of the tensor - to - scalar perturbation ratio , as favoured by planck and other data on the cosmic microwave background ( cmb ) . detailed measurements of the cmb may provide insights into the embedding of inflation within string theory as well as its links to collider physics . july 2015 kcl - ph - th/2015 - 28 , lcts/2015 - 20 , cern - ph - th/2015 - 144 + act-05 - 15 , mi - th-1521 , umn - th-3442/15 , ftpi - minn-15/32

1507.02308

data on the cosmic microwave background ( cmb ) from the planck satellite @xcite and other experiments are qualitatively consistent with generic expectations from cosmological inflation @xcite . in particular , they are consistent with the prototypical starobinsky model based an @xmath0 extension of minimal einstein gravity @xcite , and are constraining or excluding many other models . as discussed elsewhere in this volume , some still consider alternatives to inflation @xcite , but we are encouraged by the progressive and impressive improvement in experimental precision to explore specific planck - compatible models in more detail . as we discuss , detailed measurements of the cmb provide a window on fundamental physics that is complementary to laboratory experiments , casting light on particle physics at energies far beyond the reach of colliders and possibly giving us insight into string compactifications . run 1 of the lhc revealed the higgs boson @xcite , which is an existence proof for an apparently elementary scalar boson . as such , it may serve as a prototype for the inflaton , and it has been proposed that the higgs could even be the inflaton itself @xcite . this requires a rather large non - minimal gravitational coupling of the higgs field , and seems impossible unless the standard model of particle physics is supplemented , since naive extrapolation of the standard model leads to a negative higgs potential at high scales . at the time of writing the lhc has yet to reveal any new physics beyond the standard model , but there are many reasons to expect new physics , and we consider supersymmetry to be the best - motivated possibility @xcite . the appearance of the higgs boson with a mass @xmath3 gev @xcite sharpens the problem of the naturalness of the electroweak scale , which low - energy supersymmetry could mitigate . moreover , simple supersymmetric models actually predicted correctly the mass of the higgs boson @xcite , and also that its couplings would resemble those in the standard model @xcite - which they do , so far . these are new motivations for supersymmetry provided by run 1 of the lhc , in addition to the rles that supersymmetry could play in grand unified theories and string theory . it is therefore natural to consider supersymmetric models of inflation . supersymmetric versions of inflation were originally proposed in the context of a growing set of problems @xcite besetting the new inflationary theory @xcite based on the one - loop ( coleman - weinberg ) potential for breaking su(5 ) . for example , the vacuum tended to evolve to a minimum different from that containing the standard model @xcite , and quantum fluctuations destabilized the inflationary vacuum @xcite unless the higgs effective quartic self - coupling was small , @xmath4 , whereas its value was fixed by the su(5 ) gauge coupling to be @xmath5 . but the biggest problem for new inflation was the magnitude of density fluctuations @xcite . the new inflationary model based on su(5 ) predicted that they should be @xmath6 , whereas experimentally they are @xmath7 . several of these problems are tied to the magnitude of the effective quartic potential coupling , which must be tuned to @xmath8 to insure acceptable density fluctuations . indeed , in any model of inflation based on an elementary scalar field , its effective potential must have some parameter that is small in natural units where the reduced planck mass @xmath9 gev is set to unity . in a supersymmetric theory , such parameters are renormalized multiplicatively , so the quantum corrections to small values are under control . for this reason , it was suggested that inflation cries out for supersymmetry @xcite . in this framework the magnitude of the self - coupling could be linked to the ratio of supersymmetry breaking to the gut scale rather than to the gut gauge coupling alone @xcite . tension due to fine tuning and the duration of inflation could be further relieved if the inflationary field value were separated from the gut scale and pushed to the planck scale , the scenario of primordial inflation @xcite using a gauge singlet field @xcite that was baptized the inflaton @xcite . primordial supersymmetric inflation made it easy to render natural the fact that the magnitude of the observed scalar density perturbations is @xmath7 @xcite . however , it is clear that any discussion of early - universe cosmology , including inflation , should also incorporate gravity in an essential way , and hence be set in the framework of supergravity @xcite . in general , a simple supergravity theory is characterized by a hermitian function of the matter scalar fields @xmath10 , called the khler potential @xmath11 , that captures its geometry , a holomorphic function of the scalar fields , called the superpotential @xmath12 , that describes their interactions , and another holomorphic function @xmath13 that characterizes their couplings to gauge fields @xmath14 @xcite . in minimal @xmath15 supergravity , the khler metric is flat : k = ^i _ i^ * , where the sum is over all scalar components in the theory . the simplest inflationary theory in minimal supergravity is defined by the superpotential @xcite w = m^2 ( 1 - ) ^2 , [ hrrw ] where @xmath16 is the inflaton and @xmath17 in planck units . however , because inflation in this model is effectively driven by a cubic term in the scalar potential , it leads to a prediction for a scalar perturbation spectrum with tilt , @xmath18 , which is now in serious disagreement with the determination by planck @xcite : @xmath19 . moreover , a generic supergravity theory coupled to matter is not suitable for cosmology , because its effective scalar potential is proportial to @xmath20 , scalars typically pick up masses proportional to @xmath21 , where @xmath22 is the hubble parameter @xcite . though the theory defined by ( [ hrrw ] ) is constructed to avoid this @xmath23 problem , a generic inflationary model is in general plagued by this problem of large masses . in addition , the spontaneous breaking of local supersymmetry introduces additional challenges for constructing a successful supergravity inflationary scenario @xcite , stemming from the introduction of a chiral superfield whose scalar components have weak - scale masses but planck - scale vacuum expectation values ( vevs ) @xcite . the question then arises , which type of supergravity theory to choose for formulating models of inflation ? an attractive way to avoid the @xmath23 problem is provided @xcite by no - scale supergravity @xcite . in the minimal two - field case @xcite useful for inflation , its khler potential can be written in the logarithmic form @xmath24 where @xmath25 and @xmath16 are complex scalar fields and the @xmath26 represent possible additional matter fields problem may also be avoided in theories with a shift symmetry in which the khler potential is a function @xmath27 rather than @xmath28 @xcite . in this case the danergous term arising from @xmath20 does not depend on the combination @xmath29 , which may then be chosen as the inflaton . ] . moreover , no - scale supergravity emerges as the effective four - dimensional low - energy field theory in generic compactifications of string theory @xcite , with @xmath25 being identified as the compactification volume modulus . we therefore consider it to be the best - motivated framework for constructing field - theoretical models of inflation @xcite . a simple version of a no - scale inflationary model is defined by the superpotential @xmath30 @xcite . however , in this model too , inflation is effectively driven by a cubic term in the scalar potential , so that it yields the same prediction @xmath18 as in the minimal model @xcite . the planck 2013 data @xcite , with their confirmation of a tilt in the spectrum of scalar perturbations with @xmath1 and their strengthening of previous upper bounds on the tensor - to - scalar ratio @xmath31 triggered to re - examine no - scale inflation @xcite . as already mentioned , the planck data are highly consistent with the predictions of the starobinsky @xmath0 model @xcite . we were therefore very impressed to discover that the simplest possible wess - zumino superpotential @xcite @xmath32 in conjunction with the simplest no - scale khler potential ( [ nsk ] ) reproduced the starobinsky predictions for suitable choices of @xmath33 and @xmath34 @xcite . subsequently , many other examples of no - scale inflationary models yielding predictions compatible with the planck data have been discovered and studied @xcite . in parallel , the second data release from planck @xcite has sharpened the observational constraints on no - scale models of inflation , and interest in the observability of tensor perturbations in the cmb has been kindled by results from bicep2 @xcite and the prospects for other experiments searching for primordial @xmath35-mode polarization in the cmb . in this article we review these and other recent developments in no - scale inflation , with particular emphasis on the rle of planck data in motivating and constraining no - scale models @xcite . we also address the prospects for tying no - scale models of inflation more closely to string theory and particle physics at accessible collider energies . an @xmath15 supergravity theory is characterized by its hermitian khler function @xmath11 and holomorphic superpotential @xmath12 via the combination @xmath36 @xcite . the kinetic terms for scalar fields are then given in terms of the khler metric @xmath37 by @xmath38 , and ( discarding @xmath39-terms associated with gauge interactions ) the effective potential is @xmath40 \ , , \label{effpot}\ ] ] where @xmath41 is the inverse of @xmath42 . equation ( [ effpot ] ) displays the major problem for models of cosmology based on @xmath15 supergravity , namely that the effective potential is generically not flat , and has holes " with a depth that is @xmath6 in natural ( planckian ) units @xcite . it is easy to verify that the no - scale model ( [ nsk ] ) avoids this problem @xcite , since the effective potential becomes @xmath43 which is clearly positive semidefinite with no holes " . the kinetic terms for the scalar fields @xmath25 and @xmath16 in ( [ nsk ] ) are @xmath44 assuming that the @xmath25 field has a vev @xmath45 and @xmath46 ( we return later to a discussion how this may come about ) , we may neglect the kinetic mixing between the @xmath25 and @xmath16 fields in ( [ no - scalel ] ) , and are left with the following effective lagrangian for the inflaton field @xmath16 : @xmath47 which is the starting - point for our discussion of no - scale inflation . in order to study the inflaton potential in the model ( [ nsk ] , [ wz ] ) given by ( [ phil ] ) , we first make the field transformation @xcite @xmath48 and introduce a rescaled parameter @xmath49 , in terms of which the effective potential becomes @xmath50 writing @xmath51 in terms of its real and imaginary parts : @xmath52 and choosing the specific case @xmath53 ( in planck units ) , we have @xmath54 at @xmath55 ( fixed by the potential at large @xmath56 ) , the potential for the real part of the inflaton takes the form @xmath57 this potential is displayed as the black line in fig . [ fig : eno6 ] , which also shows as coloured lines the potential for values of @xmath34 slightly different from the reference value @xmath53 in planck units . the value of @xmath58 ( and hence @xmath34 ) is fixed by the magnitude of density perturbations a_s = = ^4 ( x/ ) , where @xmath59 is one of the slow - roll parameters , @xmath60 . using @xmath61 @xcite and @xmath62 to insure 50 - 60 efolds of inflation , we obtain @xmath63 . [ b ] remarkably , the potential for @xmath53 is _ identical _ with that in the starobinsky model . we recall that , starting from the modified einstein - hilbert action @xcite @xmath64 making the conformal transformation @xmath65 along with the field redefinition @xmath66 , one obtains the action @xcite @xmath67 \ , .\ ] ] this has the standard einstein - hilbert form with a canonically - normalized scalar field @xmath68 whose potential is identical with ( [ nswzpot ] ) . cosmological perturbations in this model were first calculated in @xcite and , as already commented , are in perfect agreement with the planck data @xcite . for @xmath69 , the spectral tilt is @xmath70 , where @xmath71 , and the tensor to scalar perturbation ratio is @xmath72 . however , there is a price to be paid for the success of the starobinsky model , namely that the measured magnitude of the scalar density perturbations corresponds to a value of @xmath73 in natural units , i.e. , to an unnaturally ( ? ) large coefficient for the @xmath74 term in ( [ starob ] ) . in contrast , in the no - scale wess - zumino model the magnitude of the perturbations requires @xmath75 , which is technically natural thanks to supersymmetry . the wess - zumino superpotential is not the only way to derive the starobinsky model from no - scale supergravity . indeed , it was not even the first scenario for deriving the effective action ( [ starob ] ) from no - scale supergravity @xcite , although the connection with inflation was not made previously to @xcite . ref . @xcite exhibited a no - scale supergravity example with the superpotential @xmath76 and argued that the no - scale khler potential ( [ nsk ] ) was the only consistent supergravity extension of @xmath0 gravity . in the example ( [ cecotti ] ) , the canonically - normalized real part @xmath77 of the field @xmath25 : re@xmath78 plays the rle of the inflaton , and has a starobinsky potential when @xmath16 is fixed at zero . many other starobinsky - like avatars of no - scale supergravity were derived and discussed in @xcite . in some of these the inflaton was identified with the matter field @xmath16 in ( [ nsk ] ) , and in others it was identified with the @xmath25 field , which could be interpreted as a modulus of compactification @xcite . it is possible to represent ( [ nsk ] ) in a more symmetric form @xcite : @xmath79 where the complex fields @xmath80 and @xmath81 are related to the fields @xmath82 in ( [ nsk ] ) by @xmath83 it is important to note that the effective superpotential is modified when the coordinates are transformed as in ( [ tphiwrite ] ) : @xmath84 the transformation ( [ tphiwrite ] ) shows that , at the level of the khler potential ( [ k21symm ] ) that determines the geometry of the khler manifold , there is no real distinction between the modulus " field @xmath25 and the matter " field @xmath16 . however , the distinction becomes important when one considers the accompanying superpotential , which is an essential step in constructing a starobinsky avatar of no - scale supergravity . as an example @xcite , if one considers the superpotential @xmath85 \ , , \label{w1}\ ] ] which is not obviously of wess - zumino form , and assumes that @xmath86 , one finds an effective potential @xmath87 which yields exactly the starobinsky potential when rewritten in terms of the canonically - normalized inflaton field @xmath88 . moreover , transforming back to the @xmath89 basis using the inverse of ( [ tphiwrite ] ) , we find that the the khler potential and the superpotential have exactly the forms ( [ nsk ] , [ wz ] ) . on the other hand , if one interchanges @xmath90 and makes the same transformation ( [ tphiwrite ] ) , one finds the same khler potential ( [ nsk ] ) but the following superpotential : @xmath91 this yields the asymptotically dilatation - invariant effective potential @xmath92 and , making the transformation @xmath93 , we see that this example also reproduces the starobinsky potential , but with the inflaton identified as the `` modulus '' field and with @xmath16 fixed at 0 . many more examples of starobinsky - like models with the inflaton identified as either a modulus field or a matter field have been constructed and discussed in @xcite . a generic issue in these and other models of inflation inspired by string theory is how to fix the vacuum moduli , specifically in the no - scale inflationary models discussed above the @xmath25 field , which may be identified with the overall compactification volume modulus . the phenomenological approach to this problem taken in @xcite was inspired by @xcite , namely adding quartic terms inside the logarithm in the khler potential ( [ nsk ] ) : k = - 3 ( t + t^ * - + ) , [ ekn ] where @xmath94 is a mass scale assumed to be smaller than the planck scale , and @xmath95 is a parameter that breaks the invariance under the imaginary translations of the khler potential . for simplicity , we can choose @xmath96 , in which case the masses of the real and imaginary parts of @xmath25 are equal . a non - zero mass for @xmath25 is most easily obtained in this context by simply adding a constant term @xmath97 to the superpotential @xcite , thus breaking supersymmetry . this constant term induces a gravitino mass @xmath98 and a modulus mass @xmath99 , which is hierarchically larger than the gravitino mass . for @xmath100 , the potential for @xmath16 ( or canonical @xmath51 as in ( [ chipot ] ) ) is indistinguishable from the starobinsky potential . however , stabilization terms with similar forms have not yet been derived in string theory , whereas the corrections to ( [ nsk ] ) that have been motivated by string theory , see for example @xcite , do not stabilize the volume modulus while maintaining the starobinsky form of the potential necessary for inflation @xcite . solving the string modulus stabilization problem lies beyond the scope of our discussion of inflation , so we just flag it here as an important open problem . although starobinsky - like models are certainly highly consistent with the cmb data from planck and other experiments , there is scope for models with larger values of @xmath31 , so it is interesting to explore the scope for such a possibility within the no - scale supergravity framework @xcite . as an example , we consider here a khler potential of the form @xcite : @xmath101 ^ 2}{\lambda^2 } \right)+\frac{|\phi|^2}{(t+\bar{t})^3 } \ , , \label{ourk}\ ] ] where the second term is typical of how a matter field with modular weight @xmath102 would appear in an orbifold compactification of string @xcite , and we postulate a superpotential of the form @xmath103 where @xmath104 is some coefficient @xmath105 . in this case , we have a model where the modulus , @xmath25 plays the role of the inflaton , and one linear combination of the real and imaginary parts of @xmath25 , defined by the angle @xmath106 , is stabilized for @xmath107 . it is easy to check that @xmath16 is driven to zero , for which value the effective potential takes the simple form @xmath108 we decompose @xmath25 into its real and imaginary parts @xmath109 , where @xmath110 is normalized canonically and @xmath111 is canonical at the minimum when @xmath112 : @xmath113 the potential is minimized when @xmath114 , in which case the effective lagrangian is given by @xmath115 and it is easy to see that the minimum of this effective potential is at @xmath116 when @xmath110 is at its minimum , the effective lagrangian for @xmath111 is @xmath117 and we recover the minimal chaotic inflationary model with a quadratic potential @xcite . conversely , when @xmath111 is at its minimum , the effective lagrangian for @xmath110 is @xmath118 which yields the familiar starobinsky potential @xcite . the potential ( [ effl ] ) for the choice @xmath119 is shown in fig . [ fig : egno2 - 1 ] , where we see the starobinsky form in the @xmath110 direction ( [ lrho ] ) and the quadratic form in the @xmath111 direction ( [ lsigma ] ) , is given in @xcite . ] . this model has two dynamical fields coupled through the kinetic term of the largrangian ( [ effl ] ) , and a full discussion of their behaviour during inflation requires a complete two - field analysis @xcite , which we summarize in the next section . for now , we just note that for small @xmath94 the @xmath106-dependent stabilization terms in ( [ ourk ] ) reduce the model to a family of nearly single - field models characterized by an angle @xmath106 in the @xmath120 plane . if the coefficient @xmath121 of the quartic stabilization term is large enough , the inflaton trajectory is confined to a narrow valley in field space , like a bobsleigh running down a narrow track . it is clear that @xmath122 and @xmath31 must depend on the initial condition for the complex inflaton field @xmath25 , and particularly the value of @xmath106 . as an example , we consider initial conditions in the @xmath123 plane that lead to @xmath124 e - foldings of inflation , for @xmath69 , assuming @xmath125 and setting @xmath126 . the resulting @xmath106 dependences of @xmath31 and @xmath122 are shown in fig . [ fig : egno2 - 2 ] . here we see clearly how the model ( [ effl ] ) interpolates between the limits of quadratic and starobinsky - like inflation as @xmath106 increases from @xmath127 @xcite . the planck 2015 data @xcite disfavour the small values of @xmath106 that yield @xmath128 . [ b ] [ b ] values of @xmath31 that are smaller than in the starobinsky model are possible in other models related to no - scale supergravity @xcite . the starobinsky potential can be expressed parametrically @xmath129 where @xmath56 is canonically normalized , the value of @xmath130 is determined by the magnitude of the scalar density perturbations , and @xmath131 in the starobinsky model . the inflationary predictions are derived at large @xmath56 where the potential is dominated by the constant and leading term @xmath132 in ( [ staroab ] ) . one can consider phenomenological generalizations of ( [ staroab ] ) where @xmath133 and @xmath134 and @xmath35 treated as free parameters that may differ from the starobinsky values @xmath135 and @xmath131 . at leading order in @xmath136 one finds @xmath137 implying @xmath138 these predictions are independent of @xmath134 , and the prediction for @xmath122 is independent of @xmath35 . the only model - dependence is that @xmath31 depends on @xmath35 . within the no - scale framework , different values of @xmath35 could be obtained in models with multiple moduli @xmath139 that share the no - scale property @xcite @xmath140 if one identifies the inflaton with one of the moduli @xmath139 , one finds that has a potential of the form ( [ staroab ] ) with @xmath141 and hence @xmath142 the leading alternative to the single - modulus case with @xmath143 may be that with three moduli @xmath139 , each with @xmath144 , one of which is identified with the inflaton . in this case @xmath31 would be a factor of 3 smaller than in the starobinsky model . this example shows that , within the class of no - scale models discussed here , an eventual measurement of @xmath31 might provide some observational clues to the form of string compactification . since the building - blocks of supersymmetric models are complex scalar fields , in general , supersymmetric models of inflation must take multi - field effects into account @xcite . for example , the original no - scale model ( [ nsk ] ) has four field components in general , as does the model ( [ ourk ] ) discussed in the previous section . early studies of no - scale models took the ( over-)simplified approach of fixing some ( combination ) of the field components , as was done in ( [ ourk ] ) . what happens if one relaxes this assumption , and considers the full multi - field dynamics of the inflaton field ? in general , it is known that the inflaton field trajectory will be curved and that , as a result , isocurvature fluctuations perpendicular to the direction of inflaton motion motion source adiabatic perturbations as the field trajectory evolves towards the global minimum . this extra source of adiabatic scalar perturbations tends to suppress the tensor - to - scalar ratio @xmath31 , and may in addition source non - gaussian effects such as @xmath145 @xcite . we have studied such effects in the specific two - field no - scale model introduced in the previous section @xcite , and some representative results are shown in fig . [ fig : egno3 ] in the case where the stabilization term in ( [ ourk ] ) has the value @xmath126 , i.e. , with strong stabilization of the inflaton trajectory , although they were unimportant in this case . ] . the left panel of fig . [ fig : egno3 ] shows predictions for the tensor - to - scalar ratio @xmath31 as a function of the starting - point in the @xmath109 plane for @xmath146 , i.e. , with the stabilization term in ( [ ourk ] ) switched off ( almost ) . we see in the left panel that , as expected , @xmath31 is reduced compared to what might have been expected from a naive single - field analysis . only very close to the vertical axis ( @xmath147 ) does @xmath31 approach the value predicted by the chaotic inflationary model with a quadratic potential . in most of the plane of possible initial conditions , @xmath31 takes values comparable to those in the starobinsky model . the right panel shows that these planck - compatible values of @xmath31 are not accompanied by large non - gaussianity @xcite : in all the plane , @xmath148 , well within the planck bound on this measure of non - gaussianity : @xmath149 @xcite . as is well - known , the predictions of any specific model of inflation depend on @xmath150 , the number of e - folds at some reference scale @xmath151 . for example , in starobinsky - like models one has @xmath152 the small values of @xmath31 in ( [ n*dependences ] ) are unlikely to be probed in the near future , but the planck measurement of @xmath122 is already imposing interesting constraints on @xmath150 @xcite . this is related to inflationary model parameters via @xcite @xmath153 where we have ignored the possibility of entropy generation after reheating @xcite , @xmath154 and @xmath155 are the present hubble expansion rate and cosmological scale factor , respectively , @xmath156 is the inflationary energy density at the reference scale , @xmath157 and @xmath158 are the energy densities at the end of inflation and after reheating , @xmath159 is the _ e - fold _ average of the equation - of - state parameter @xmath160 during the thermalization epoch , and @xmath161 is the number of equivalent bosonic degrees of freedom after reheating : @xmath162 . the values of the reheating energy density @xmath158 and temperature @xmath163 depend , in turn , on the rate of inflaton decay , @xmath164 , and we find @xcite that @xmath165 where @xmath134 parametrizes the approach to complete thermalization : @xmath166 and @xmath167 is the _ time _ average of the equation - of - state parameter @xmath160 during the thermalization epoch , which we find to be @xmath168 for @xmath169 . as a result , @xmath150 depends on the inflaton decay rate as follows : @xmath170 \ , + \ , \dots \ , , \,\ ] ] where the @xmath26 represent other terms in the full expression for @xmath150 given in @xcite , where it is also shown that @xmath171 in starobinsky - like models . fig . [ fig : egno5 - 1 ] shows the corresponding numerical relation between @xmath164 , @xmath150 and @xmath122 in starobinsky - like models including those based on no - scale supergravity . the diagonal red and blue lines are numerical results and analytic approximations , respectively , which agree quite well , and the vertical lines represent specific models of inflaton decay discussed in @xcite . the upper axis shows values of the two - body decay coupling @xmath172 corresponding to @xmath164 via the relation @xmath173 , for values of the coupling @xmath172 ranging from @xmath174 ( vertical red line ) to the value @xmath175 ( vertical purple line ) , which would correspond to a reheating temperature @xmath176 mev , below which the successful conventional big bang nucleosynthesis calculations would need to be modified substantially . in order to avoid the overproduction of gravitinos , whose decays could also aversely affect big bang nucleosynthesis and overpopulate the universe with dark matter particles @xcite , one may require @xmath177 , corresponding to the vertical green line in fig . [ fig : egno5 - 1 ] . the vertical magenta line corresponds to the rate for three - body decays involving top quarks that dominate in some no - scale models @xcite , and the vertical yellow and magenta lines bracket the expected range for decays into gauginos that may be found in models with non - trivial gauge kinetic functions @xmath178 . [ fig : egno5 - 2 ] confronts the planck 2015 constraints @xcite - the yellow shaded regions are favoured at the 68% cl and the blue shaded regions are allowed at the 95% cl - with the predictions of the simplest wess - zumino no - scale model @xcite for different values of @xmath150 and @xmath179 near the value 1/3 that reproduces starobinsky - like predictions @xcite . we see in the upper panel that the planck 2015 constraints are almost independent of @xmath31 in the displayed range of @xmath31 , and that the contours of fixed @xmath150 ( coloured lines ) intersect the contours of fixed @xmath179 ( black lines ) at acute angles . consequently , as seen in the lower panel , the planck 2015 constraints on @xmath150 depend quite strongly on the value of @xmath179 . for the starobinsky value @xmath180 we find lower bounds @xmath181 which are shown as horizontal yellow and blue lines , respectively , in fig . [ fig : egno5 - 1 ] . their intersections with the diagonal blue line give the corresponding lower bounds on @xmath182 and @xmath172 . we see that the lower bounds are already stronger than those imposed by successful big bang nucleosynthesis ( vertical purple line ) and the 68% cl lower limit approaches the upper limit on @xmath183 suggested by gravitino decays ( vertical green line ) . thus , the planck data @xcite already impose interesting constraints on inflaton decays in no - scale models : other examples are discussed in @xcite . one of the biggest challenges in constructing models of inflation is to relate them to particle physics at accessible ( collider ) energies . one connection may be provided in principle by the discussion of inflaton decays in the previous section @xcite . for example , in models where the inflaton is identified with the supersymmetric partner of a singlet ( right - handed ) neutrino in a see - saw model of neutrino masses , the constraints on the two - body decay coupling @xmath172 have potential implications for low - energy observables such as flavour - changing lepton transitions as well as scenarios for neutrino mixing . another possible connection arises via supersymmetry breaking . for example , if the gauge kinetic function @xmath13 depends same non - trivially on the inflaton field @xmath16 : @xmath184 the inflaton decays into standard model gauge bosons and gauginos are given by @xcite @xmath185 and the gaugino masses are given by @xmath186 as already commented , the vertical yellow and magenta lines in fig . [ fig : egno5 - 1 ] bracket the range of @xmath172 and hence @xmath182 found in starobinsky - like no - scale models , which correspond to @xmath187 , with the upper end of this range corresponding to @xmath188 and hence @xmath189 . additionally , different assignments for the inflaton and matter fields lead to different possibilities for the pattern of supersymmetric particle masses , via their dependences on the model - dependent soft supersymmetry - breaking scalar masses @xmath190 , gaugino masses @xmath191 and bilinear and trilinear scalar couplings @xmath192 and @xmath193 @xcite . among the possibilities are the original _ no - scale _ boundary conditions @xmath194 @xcite , _ cmssm_-like boundary conditions in which @xmath195 and @xmath193 are non - zero and universal for different scalar species and determined by the gravitino mass @xcite , _ msugra_-like boundary conditions in which @xmath196 @xcite , etc . thus , if supersymmetry is discovered and sparticle masses measured at accelerators , it may be possible constrain models of inflation , and vice versa if models of inflation can be constrained . an example is given in fig . [ fig : eno8 ] , which displays results for a no - scale scenario with @xmath197 and @xmath198 at some input renormalization scale @xmath199 in an su(5 ) gut model with superpotential terms @xmath200 , where @xmath201 and @xmath202 are @xmath203 and @xmath204 higgs representations , respectively , for the representative values @xmath205 discussed in @xcite . we see that , within this particular model , only a restricted range of @xmath206 gev is consistent with the lhc data . the relations ( [ m12m32 ] , [ gaugedecay ] , [ dgphi ] ) show how this type of constraint can then be applied to models of inflation and string compactification . -0.9 in plane in the su(5 ) no - scale model with @xmath207 at @xmath199 , for @xmath208 and @xmath209 @xcite . the brown shaded regions are excluded because the lsp is charged , the green shaded regions are excluded by @xmath210 , the renormalization - group equations are unstable in the ochre shaded region of the upper panel . the relic cold dark matter density lies within the cosmological range along the narrow dark blue strips . the near - vertical continuous purple line is the atlas 95% cl lower limit on the gaugino mass @xmath191 , the dash - dotted ( red ) lines are contours of @xmath211 as calculated using feynhiggs 2.10.0 @xcite , and the solid diagonal green lines marks a 95% cl upper limit on br(@xmath212 ) . the continuous ( black ) lines in the upper panel are contours of @xmath213 , the ratio of supersymmetric higgs vevs @xcite . _ , title="fig : " ] as we have reviewed in this article , no - scale supergravity is an attractive framework for constructing models of inflation , since supersymmetry accommodates naturally the required hierarchy between the scale of inflation and the planck scale @xcite , gravity must be incorporated in any discussion of cosmology , and no - scale supergravity @xcite emerges from generic string compactifications @xcite and yields an effective potential that is positive semi - definite @xcite . we have discussed various no - scale inflationary scenarios , reviewing how they naturally yield an effective potential , and hence predictions for @xmath122 and @xmath31 , that are coincident @xcite with the planck - friendly starobinsky model based on @xmath0 gravity @xcite . moreover , no - scale models achieve this in a technically natural way , via small superpotential couplings rather than surprisingly large non - minimal gravitational couplings as in the starobinsky and higgs inflation models . the no - scale framework is , moreover , more flexible , being able to accommodate models intermediate between the starobinsky model and a quadratic potential suitable for chaotic inflation @xcite , which could yield larger values of @xmath31 that are still compatible with the constraints from planck and other experiments . as in other supersymmetric models of inflation , it is necessary to take into account two - field effects : in the no - scale models we study @xcite , we find that these tend to reduce @xmath31 without generating large non - gaussianity . in principle , no - scale inflation could provide a phenomenological bridge between string theory and collider physics @xcite . in addition to the above - mentioned model - dependence of @xmath31 , the value of @xmath122 is related directly in starobinsky - like models to the number of e - folds during inflation @xcite , which is in turn sensitive to the rate of inflaton decay and thereby the assignment of the inflaton as a modulus or matter field . the pattern of soft supersymmetry breaking is also sensitive to this assignment , and would be measurable at the lhc or in other collider experiments . as we have mentioned , open issues in no - scale inflation include the mechanism for stabilization of the various moduli , including the volume modulus that plays a prominent rle in building models . the inflationary observables are sensitive to the mechanism of modulus stabilization , and therefore may be able to cast some light on this basic issue in string phenomenology . more generally , for the foreseeable future measurements of inflationary observables are likely to take us closer to the string scale than any other experiments , and no - scale inflationary models may be the best platform for exploiting this scientific opportunity . the work of j.e . was supported in part by the london centre for terauniverse studies ( lcts ) , using funding from the european research council via the advanced investigator grant 267352 and from the uk stfc via the research grant st / l000326/1 . the work of d.v.n . was supported in part by the doe grant de - fg03 - 95-er-40917 and in part by the alexander s. onassis public benefit foundation . the work of m.a.g.g . and k.a.o . was supported in part by doe grant de - sc0011842 at the university of minnesota . 99 p. a. r. ade _ et al . _ [ planck collaboration ] , arxiv:1502.01589 [ astro-ph.co ] ; arxiv:1502.02114 [ astro-ph.co ] . j. martin , c. ringeval and v. vennin , phys . dark univ . * 5 - 6 * , 75 - 235 ( 2014 ) [ arxiv:1303.3787 [ astro-ph.co ] ] ; j. martin , c. ringeval , r. trotta and v. vennin , jcap * 1403 * ( 2014 ) 039 [ arxiv:1312.3529 [ astro-ph.co ] ] ; j. martin , arxiv:1502.05733 [ astro-ph.co ] . a. a. starobinsky , phys . b * 91 * , 99 ( 1980 ) . a. ijjas , p. j. steinhardt and a. loeb , phys . b * 723 * ( 2013 ) 261 [ arxiv:1304.2785 [ astro-ph.co ] ] ; j. l. lehners and p. j. steinhardt , phys . d * 87 * ( 2013 ) 12 , 123533 [ arxiv:1304.3122 [ astro-ph.co ] ] ; a. ijjas , p. j. steinhardt and a. loeb , phys . b * 736 * ( 2014 ) 142 [ arxiv:1402.6980 [ astro-ph.co ] ] . g. aad _ et al . _ [ atlas collaboration ] , phys . b * 716 * ( 2012 ) 1 [ arxiv:1207.7214 [ hep - ex ] ] ; s. chatrchyan _ et al . _ [ cms collaboration ] , phys . b * 716 * ( 2012 ) 30 [ arxiv:1207.7235 [ hep - ex ] ] . f. l. bezrukov and m. shaposhnikov , phys . b * 659 * , 703 ( 2008 ) [ arxiv:0710.3755 [ hep - th ] ] . h. p. nilles , phys . * 110 * ( 1984 ) 1 ; h. e. haber and g. l. kane , phys . * 117 * ( 1985 ) 75 . g. aad _ et al . _ [ atlas and cms collaborations ] , phys . * 114 * ( 2015 ) 191803 [ arxiv:1503.07589 [ hep - ex ] ] . j. r. ellis , g. ridolfi and f. zwirner , phys . b * 257 * ( 1991 ) 83 ; h. e. haber and r. hempfling , phys . * 66 * ( 1991 ) 1815 ; y. okada , m. yamaguchi and t. yanagida , prog . * 85 * ( 1991 ) 1 . j. r. ellis , s. heinemeyer , k. a. olive and g. weiglein , phys . b * 515 * ( 2001 ) 348 [ hep - ph/0105061 ] . k. a. olive , phys . rept . * 190 * , 307 ( 1990 ) . a. d. linde , phys . b * 108 * , 389 ( 1982 ) ; a. albrecht and p. j. steinhardt , phys . lett . * 48 * , 1220 ( 1982 ) . a. billoire and k. tamvakis , nucl . b * 200 * , 329 ( 1982 ) ; j. d. breit , s. gupta and a. zaks , phys . lett . * 51 * , 1007 ( 1983 ) . a. d. linde , phys . b * 116 * , 335 ( 1982 ) . press , _ phys . * 21 * ( 1980 ) 702 ; v.f . mukhanov and g.v . chibisov , _ jetp lett . _ * 33 * ( 1981 ) 532 ; s.w . hawking , _ phys . lett . _ * 115b * ( 1982 ) 295 ; a.a . starobinsky , _ phys . lett . _ * 117b * ( 1982 ) 175 ; a.h . guth and s.y . pi , _ phys . lett . _ * 49 * ( 1982 ) 1110 ; j.m . bardeen , p.j . steinhardt and m.s . turner , _ phys . rev . _ * d28 * ( 1983 ) 679 . j. r. ellis , d. v. nanopoulos , k. a. olive and k. tamvakis , phys . b * 118 * ( 1982 ) 335 . j. r. ellis , d. v. nanopoulos , k. a. olive and k. tamvakis , nucl . b * 221 * ( 1983 ) 52 . j. r. ellis , d. v. nanopoulos , k. a. olive and k. tamvakis , phys . b * 120 * ( 1983 ) 331 . k. nakayama and f. takahashi , jcap * 1110 * , 033 ( 2011 ) [ arxiv:1108.0070 [ hep - ph ] ] . g. d. coughlan and g. g. ross , phys . b * 157 * , 151 ( 1985 ) ; a. albrecht and r. h. brandenberger , phys . rev . d * 31 * , 1225 ( 1985 ) ; l. g. jensen and k. a. olive , phys . b * 159 * , 99 ( 1985 ) . d. v. nanopoulos , k. a. olive and m. srednicki , phys . b * 127 * , 30 ( 1983 ) . d. v. nanopoulos , k. a. olive , m. srednicki and k. tamvakis , phys . b * 123 * , 41 ( 1983 ) . a. b. goncharov and a. d. linde , phys . b * 139 * , 27 ( 1984 ) . e. cremmer , s. ferrara , l. girardello and a. van proeyen , nucl . b * 212 * ( 1983 ) 413 . r. holman , p. ramond and g. g. ross , phys . b * 137 * , 343 ( 1984 ) . e. j. copeland , a. r. liddle , d. h. lyth , e. d. stewart and d. wands , phys . d * 49 * , 6410 ( 1994 ) [ astro - ph/9401011 ] ; e. d. stewart , phys . d * 51 * , 6847 ( 1995 ) [ hep - ph/9405389 ] . g. d. coughlan , w. fischler , e. w. kolb , s. raby and g. g. ross , phys . b * 131 * , 59 ( 1983 ) ; a. s. goncharov , a. d. linde and m. i. vysotsky , phys . b * 147 * , 279 ( 1984 ) ; t. banks , d. b. kaplan and a. e. nelson , phys . rev . d * 49 * , 779 ( 1994 ) [ hep - ph/9308292 ] ; b. de carlos , j. a. casas , f. quevedo , e. roulet , phys . lett . b * 318 * , 447 ( 1993 ) [ hep - ph/9308325 ] ; m. kawasaki , t. moroi and t. yanagida phys . b * 370 * , 52 ( 1996 ) [ hep - ph/9509399 ] . i. joichi and m. yamaguchi , phys . b * 342 * , 111 ( 1995 ) [ hep - ph/9409266 ] ; m. dine , r. kitano , a. morisse and y. shirman , phys . rev . d * 73 * , 123518 ( 2006 ) [ hep - ph/0604140 ] ; r. kitano , phys . b * 641 * , 203 ( 2006 ) [ hep - ph/0607090 ] ; r. kallosh and a. d. linde , jhep * 0702 * , 002 ( 2007 ) [ hep - th/0611183 ] ; h. abe , t. higaki and t. kobayashi , phys . d * 76 * ( 2007 ) 105003 [ arxiv:0707.2671 [ hep - th ] ] ; j. fan , m. reece and l .- t . wang , jhep * 1109 * , 126 ( 2011 ) [ arxiv:1106.6044 [ hep - ph ] ] . e. dudas , c. papineau and s. pokorski , jhep * 0702 * , 028 ( 2007 ) [ hep - th/0610297 ] ; h. abe , t. higaki , t. kobayashi and y. omura , phys . rev . d * 75 * , 025019 ( 2007 ) [ hep - th/0611024 ] . r. kallosh , a. linde , k. a. olive and t. rube , phys . d * 84 * , 083519 ( 2011 ) [ arxiv:1106.6025 [ hep - th ] ] ; a. linde , y. mambrini and k. a. olive , phys . d * 85 * , 066005 ( 2012 ) [ arxiv:1111.1465 [ hep - th ] ] ; e. dudas , a. linde , y. mambrini , a. mustafayev and k. a. olive , arxiv:1209.0499 [ hep - ph ] . j. l. evans , m. a. g. garcia and k. a. olive , jcap * 1403 * , 022 ( 2014 ) [ arxiv:1311.0052 [ hep - ph ] ] . m. k. gaillard , h. murayama and k. a. olive , phys . b * 355 * ( 1995 ) 71 [ hep - ph/9504307 ] . e. cremmer , s. ferrara , c. kounnas and d. v. nanopoulos , phys . b * 133 * ( 1983 ) 61 ; j. r. ellis , a. b. lahanas , d. v. nanopoulos and k. tamvakis , phys . lett . b * 134 * ( 1984 ) 429 . e. witten , phys . b * 155 * ( 1985 ) 151 ; see also s. b. giddings , s. kachru and j. polchinski , phys . d * 66 * , 106006 ( 2002 ) [ hep - th/0105097 ] ; v. balasubramanian , p. berglund , j. p. conlon and f. quevedo , jhep * 0503 * , 007 ( 2005 ) [ hep - th/0502058 ] . j. r. ellis , k. enqvist , d. v. nanopoulos , k. a. olive and m. srednicki , phys . b * 152 * ( 1985 ) 175 [ erratum - ibid . * 156b * ( 1985 ) 452 ] . k. enqvist , d. v. nanopoulos and m. quiros , phys . b * 159 * , 249 ( 1985 ) ; p. bintruy and m. k. gaillard , phys . d * 34 * , 3069 ( 1986 ) ; h. murayama , h. suzuki , t. yanagida and j. yokoyama , phys . rev . d * 50 * , 2356 ( 1994 ) [ arxiv : hep - ph/9311326 ] ; s. c. davis and m. postma , jcap * 0803 * , 015 ( 2008 ) [ arxiv:0801.4696 [ hep - ph ] ] ; s. antusch , m. bastero - gil , k. dutta , s. f. king and p. m. kostka , jcap * 0901 * , 040 ( 2009 ) [ arxiv:0808.2425 [ hep - ph ] ] ; s. antusch , m. bastero - gil , k. dutta , s. f. king and p. m. kostka , phys . lett . b * 679 * , 428 ( 2009 ) [ arxiv:0905.0905 [ hep - th ] ] ; s. antusch , k. dutta , j. erdmenger and s. halter , jhep * 1104 * ( 2011 ) 065 [ arxiv:1102.0093 [ hep - th ] ] ; r. kallosh and a. linde , jcap * 1011 * , 011 ( 2010 ) [ arxiv:1008.3375 [ hep - th ] ] ; r. kallosh , a. linde and t. rube , phys . d * 83 * , 043507 ( 2011 ) [ arxiv:1011.5945 [ hep - th ] ] ; t. li , z. li and d. v. nanopoulos , jcap * 1402 * , 028 ( 2014 ) [ arxiv:1311.6770 [ hep - ph ] ] ; w. buchmuller , c. wieck and m. w. winkler , phys . b * 736 * , 237 ( 2014 ) [ arxiv:1404.2275 [ hep - th ] ] . j. ellis , d. v. nanopoulos and k. a. olive , phys . * 111 * ( 2013 ) 111301 [ phys . rev . lett . * 111 * ( 2013 ) 12 , 129902 ] [ arxiv:1305.1247 [ hep - th ] ] . p. a. r. ade _ et al . _ [ planck collaboration ] , astron . astrophys . * 571 * ( 2014 ) a16 [ arxiv:1303.5076 [ astro-ph.co ] ] ; astron . astrophys . * 571 * ( 2014 ) a22 [ arxiv:1303.5082 [ astro-ph.co ] ] . d. croon , j. ellis and n. e. mavromatos , phys . b * 724 * ( 2013 ) 165 [ arxiv:1303.6253 [ astro-ph.co ] ] . j. ellis , d. v. nanopoulos and k. a. olive , jcap * 1310 * ( 2013 ) 009 [ arxiv:1307.3537 ] . j. ellis , h. j. he and z. z. xianyu , phys . d * 91 * ( 2015 ) 2 , 021302 [ arxiv:1411.5537 [ hep - ph ] ] . p. a. r. ade _ et al . _ [ bicep2 and planck collaborations ] , phys . rev . lett . * 114 * , no . 10 , 101301 ( 2015 ) [ arxiv:1502.00612 [ astro-ph.co ] ] . s. ferrara , a. kehagias and a. riotto , fortsch . phys . * 62 * , 573 ( 2014 ) [ arxiv:1403.5531 [ hep - th ] ] ; s. ferrara , a. kehagias and a. riotto , fortsch . phys . * 63 * , 2 ( 2015 ) [ arxiv:1405.2353 [ hep - th ] ] ; r. kallosh , a. linde , b. vercnocke and w. chemissany , jcap * 1407 * , 053 ( 2014 ) [ arxiv:1403.7189 [ hep - th ] ] ; k. hamaguchi , t. moroi and t. terada , phys . b * 733 * , 305 ( 2014 ) [ arxiv:1403.7521 [ hep - ph ] ] . unfortunately , we do not have space in this summary to do justice to the extensive recent literature following @xcite . see , for example : r. kallosh and a. linde , jcap * 1306 * ( 2013 ) 028 [ arxiv:1306.3214 [ hep - th ] ] ; t. li , z. li and d. v. nanopoulos , jcap * 1404 * , 018 ( 2014 ) [ arxiv:1310.3331 [ hep - ph ] ] ; t. li , z. li and d. v. nanopoulos , eur . j. c * 75 * , no . 2 , 55 ( 2015 ) [ arxiv:1405.0197 [ hep - th ] ] ; c. p. burgess , m. cicoli and f. quevedo , jcap * 1311 * ( 2013 ) 003 [ arxiv:1306.3512 [ hep - th ] ] ; f. farakos , a. kehagias and a. riotto , nucl . b * 876 * , 187 ( 2013 ) [ arxiv:1307.1137 [ hep - th ] ] ; s. ferrara , r. kallosh , a. linde and m. porrati , phys . d * 88 * ( 2013 ) 8 , 085038 [ arxiv:1307.7696 [ hep - th ] ] ; w. buchmller , v. domcke and c. wieck , phys . lett . b * 730 * , 155 ( 2014 ) [ arxiv:1309.3122 [ hep - th ] ] ; c. pallis , jcap * 1404 * , 024 ( 2014 ) [ arxiv:1312.3623 [ hep - ph ] ] ; c. pallis , jcap * 1408 * , 057 ( 2014 ) [ arxiv:1403.5486 [ hep - ph ] ] ; i. antoniadis , e. dudas , s. ferrara and a. sagnotti , phys . b * 733 * , 32 ( 2014 ) [ arxiv:1403.3269 [ hep - th ] ] ; w. buchmuller , e. dudas , l. heurtier and c. wieck , jhep * 1409 * , 053 ( 2014 ) [ arxiv:1407.0253 [ hep - th ] ] ; c. kounnas , d. lst and n. toumbas , fortsch . phys . * 63 * , 12 ( 2015 ) [ arxiv:1409.7076 [ hep - th ] ] ; t. terada , y. watanabe , y. yamada and j. yokoyama , jhep * 1502 * , 105 ( 2015 ) [ arxiv:1411.6746 [ hep - ph ] ] ; a. b. lahanas and k. tamvakis , phys . d * 91 * , no . 8 , 085001 ( 2015 ) [ arxiv:1501.06547 [ hep - th ] ] ; i. dalianis and f. farakos , arxiv:1502.01246 [ gr - qc ] ; d. roest and m. scalisi , arxiv:1503.07909 [ hep - th ] ; m. scalisi , arxiv:1506.01368 [ hep - th ] . b. whitt , phys . lett . b * 145 * ( 1984 ) 176 . v. f. mukhanov and g. v. chibisov , jetp lett . * 33 * , 532 ( 1981 ) [ pisma zh . fiz . * 33 * , 549 ( 1981 ) ] ; a. a. starobinsky , sov . * 9 * , 302 ( 1983 ) . s. cecotti , phys . b * 190 * ( 1987 ) 86 . j. ellis , m. a. g. garcia , d. v. nanopoulos and k. a. olive , arxiv:1503.08867 [ hep - ph ] . j. ellis , m. a. g. garca , d. v. nanopoulos and k. a. olive , jcap * 1405 * ( 2014 ) 037 [ arxiv:1403.7518 [ hep - ph ] ] . j. r. ellis , c. kounnas and d. v. nanopoulos , phys . lett . b * 143 * , 410 ( 1984 ) . s. kachru , r. kallosh , a. d. linde and s. p. trivedi , phys . d * 68 * , 046005 ( 2003 ) [ arxiv : hep - th/0301240 ] ; r. kallosh and a. d. linde , jhep * 0412 * , 004 ( 2004 ) [ arxiv : hep - th/0411011 ] ; r. kallosh and a. d. linde , jcap * 0704 * , 017 ( 2007 ) [ arxiv:0704.0647 [ hep - th ] ] . e. dudas and c. wieck , arxiv:1506.01253 [ hep - th ] . j. ellis , m. a. g. garcia , d. v. nanopoulos and k. a. olive , jcap * 1408 * ( 2014 ) 044 [ arxiv:1405.0271 [ hep - ph ] ] . j. ellis , m. a. g. garca , d. v. nanopoulos and k. a. olive , jcap * 1501 * ( 2015 ) 01 , 010 [ arxiv:1409.8197 [ hep - ph ] ] . l. j. dixon , v. kaplunovsky and j. louis , nucl . b * 329 * ( 1990 ) 27 ; j. a. casas , in _ proceedings , international europhysics conference on high - energy physics _ , jerusalem , israel , august 19 - 25 , 1997 , eds . d. lellouch , g. mikenberg and e. rabinovici , pp 914 - 917 [ hep - ph/9802210 ] . c. gordon , d. wands , b. a. bassett and r. maartens , phys . d * 63 * , 023506 ( 2001 ) [ astro - ph/0009131 ] ; c. t. byrnes and d. wands , phys . d * 74 * , 043529 ( 2006 ) [ astro - ph/0605679 ] ; r. easther , j. frazer , h. v. peiris and l. c. price , phys . * 112 * ( 2014 ) 161302 [ arxiv:1312.4035 [ astro-ph.co ] ] . s. renaux - petel and k. turzynski , jcap * 1506 * ( 2015 ) 06 , 010 [ arxiv:1405.6195 [ astro-ph.co ] ] . k. dutta and a. maharana , phys . d * 91 * ( 2015 ) 4 , 043503 [ arxiv:1409.7037 [ hep - ph ] ] ; k. das , k. dutta and a. maharana , arxiv:1506.05745 [ hep - ph ] . m. bolz , a. brandenburg and w. buchmuller , nucl . b * 606 * , 518 ( 2001 ) [ erratum - ibid . b * 790 * , 336 ( 2008 ) ] [ hep - ph/0012052 ] ; r. h. cyburt , j. ellis , b. d. fields and k. a. olive , phys . d * 67 * , 103521 ( 2003 ) [ astro - ph/0211258 ] ; f. d. steffen , jcap * 0609 * , 001 ( 2006 ) [ hep - ph/0605306 ] ; m. kawasaki , k. kohri , t moroi and a.yotsuyanagi , phys . rev . d * 78 * , 065011 ( 2008 ) [ arxiv:0804.3745 [ hep - ph ] ] ; r. h. cyburt , j. ellis , b. d. fields , f. luo , k. a. olive and v. c. spanos , jcap * 0910 * , 021 ( 2009 ) [ arxiv:0907.5003 [ astro-ph.co ] ] ; r. h. cyburt , j. ellis , b. d. fields , f. luo , k. a. olive and v. c. spanos , jcap * 1305 * , 014 ( 2013 ) [ arxiv:1303.0574 [ astro-ph.co ] ] . j. r. ellis , k. a. olive , y. santoso and v. c. spanos , phys . b * 573 * ( 2003 ) 162 [ arxiv : hep - ph/0305212 ] , and phys . d * 70 * ( 2004 ) 055005 [ arxiv : hep - ph/0405110 ] . j. ellis , d. v. nanopoulos and k. a. olive , phys . d * 89 * ( 2014 ) 4 , 043502 [ arxiv:1310.4770 [ hep - ph ] ] . t. hahn , s. heinemeyer , w. hollik , h. rzehak and g. weiglein , phys . lett . * 112 * ( 2014 ) 14 , 141801 [ arxiv:1312.4937 [ hep - ph ] ] .

we study static , spherically symmetric solutions with an electric field in an extension of general relativity ( gr ) containing a ricci - squared term and formulated in the palatini formalism . we find that all the solutions present a central core whose area is proportional to the planck area times the number of charges . far from the core , curvature invariants quickly tend to those of the usual reissner - nordstrm solution , though the structure of horizons may be different . in fact , besides the structures found in the reissner - nordstrm solution of gr , we find black hole solutions with just one nondegenerate horizon ( schwarzschild - like ) , and nonsingular black holes and naked cores . the charge - to - mass ratio of the nonsingular solutions implies that the core matter density is independent of the specific amounts of charge and mass and of order the planck density . we discuss the physical implications of these results for astrophysical and microscopic black holes , construct the penrose diagrams of some illustrative cases , and show that the maximal analytical extension of the nonsingular solutions implies a bounce of the radial coordinate .

1207.6004

according to general relativity ( gr ) , the fate of any sufficiently massive star is to end up its lifetime forming a black hole , a spacetime region that contains a zero - volume singularity of infinite density cloaked by an event horizon . singularity and uniqueness theorems , together with the cosmic censorship conjecture @xcite , tell us that black holes are described by solely three parameters : mass , charge and angular momentum , a result known as the no - hair theorem that yields the kerr - newman family @xcite . for any black hole , the event horizon acts as a sort of no - way - out layer that separates the physics outside the horizon , which is the one that we can directly explore and find it in excellent agreement with the gr predictions @xcite , from the physics of its interior , where much less is known . when quantum phenomena come into play , the mere existence of an event horizon induces the emission of thermal particles via hawking radiation @xcite . quantum effects may also affect the details of gravitational collapse , as recently studied in @xcite , and are likely to play a very important role when the spacetime curvature reaches the planck scale . in fact , it is generally accepted that classical black hole singularities should be removed by quantum gravitational effects . however , our current understanding of quantum gravity is not mature enough to provide a clear and detailed description of how this could occur . it seems thus justified to explore phenomenological approaches to this problem hoping to gain some new insights that help us to better understand how black hole structure could be modified by quantum gravity effects . as the singularity theorems only state some precise physical conditions under which the appearance of singularities is unavoidable , several approaches have been carried out to find conditions that avoid or ameliorate the formation of black hole singularities . a well known example is bardeen s black hole @xcite , in which _ exotic _ matter sources get rid of the singularity while keeping the horizons and the asymptotically flat character . this singularity avoidance is realized through the formation of a central matter ( de sitter ) core , such that the corresponding spacetime is interpreted as the gravitational field of a nonlinear magnetic monopole and can be derived from a nonlinear electrodynamics model @xcite . that approach became a prototype for most developments on regular black holes within classical gr . in this sense , there has been much activity aimed at finding alternative matter sources for the interior region , such as introducing nonlinearities as in the case of nonlinear theories of electrodynamics @xcite , implementing a de sitter core that matches the exterior field in some junction " region ( see e.g. @xcite and references therein ) , or using new ideas inspired by noncommutative geometry @xcite , in such a way that singularities are removed . also other regular magnetically charged solutions within gr have been found @xcite . a different approach comes from the idea that extensions of gr with high - curvature corrections could be able to capture some essential features useful or required to find an effective description of the quantum gravity dynamics . this point is supported by the study of quantum fields in curved spacetimes and by approaches to quantum gravity based on string theory @xcite . in this sense , gravity theories containing higher - order curvature invariants naturally appear as modifications of gr in these quantum gravity approaches . such theories generically lead to higher - order partial derivative equations , which is a manifestation of the fact that new high - energy degrees of freedom are being taken into account . the resulting solutions are thus expected to depend on a larger number of parameters ( integration constants ) , which should provide more freedom and/or new mechanisms to avoid the singularities . however , nonsingular black holes of this kind are rather scarce ( see e.g. @xcite ) and usually require the addition of exotic sources of matter again . this suggests that the addition of new degrees of freedom in the problem is unable by itself to resolve the problem of singularities . moreover , these theories suffer from ghosts and other perturbative instabilities @xcite . these problems , however , can be avoided if the curvature invariants appear in appropriate combinations , because then the equations of motion may remain second - order like in gr . these are known as lovelock gravities @xcite ( see , for instance , @xcite for a pedagogical introduction ) . for example , the simplest extension of einstein gravity via higher - curvature terms in this context corresponds to gauss - bonnet gravity and picks up three new terms , @xmath0 , which supplement the einstein - hilbert lagrangian . it should be noted that in @xmath1 dimensions these new terms are topological invariants not contributing to the equations of motion , which means that lovelock gravities only provide modified dynamics in the context of higher dimensions . exact static spherically symmetric solutions to gauss - bonnet theory in vacuum @xcite and with electrostatic fields @xcite are known , but they still contain singularities or are ill defined . an alternative strategy to obtain modified gravitational dynamics beyond gr is to assume that the metric and affine structures of the theory are independent @xcite . when the connection is not constrained _ a priori _ to be given by the christoffel symbols of the metric , one finds that even @xmath2 extensions yield second - order field equations , which contrasts with the usual ( metric ) formulation of those theories @xcite and the general belief that only lovelock gravities have second - order equations . this approach , known as palatini formalism , has been recently used to obtain a covariant action for the effective hamiltonian dynamics of loop quantum cosmology @xcite , an approach to quantum cosmology based on the nonperturbative quantization techniques of loop quantum gravity @xcite . in the palatini approach @xcite , metric and connection are regarded as independent entities and the field equations are obtained by independent variation of the action with respect to both of them . though this does not affect the dynamics of gr @xcite , it does have important consequences for extensions of it @xcite . in general , one finds that in palatini theories the matter plays an active role in the construction of the independent connection , which ends up producing modified dynamics . when there is no matter , the field equations boil down to those of gr with an effective cosmological constant , which depends on the form of the particular lagrangian chosen . this property has made these theories very attractive for cosmological applications . the unusual role played by the matter in the construction of the geometry in palatini theories together with the second - order character of their field equations makes them specially interesting to explore the effects of new gravitational physics on the structure of black holes . in this sense , in ref.@xcite we considered palatini @xmath2 modifications of gr in interaction with modified matter sources , through nonlinear electrodynamics ( neds ) . in palatini @xmath2 theories , the modified dynamics is due to a number of new terms on the right - hand side of the equations that depend on the trace @xmath3 of the energy - momentum tensor of the matter . unlike maxwell s electrodynamics , the stress - energy tensor of neds possesses , in general , a nonvanishing trace , which makes them specially suitable to excite the palatini modified dynamics in electrovacuum scenarios . in the particular case of born - infeld ned coupled to the gravity theory @xmath4 , where @xmath5 is the planck curvature , we found that the degree of divergence of the kretschmann scalar near the singularity can be much weaker than in gr . in this work we go beyond ref.@xcite and explore how black hole structure is affected by new physics at the planck scale by considering a palatini theory of the form @xmath6 , where @xmath7 , @xmath8 is the ( symmetric ) ricci tensor , and @xmath9 and @xmath10 are constants . terms of this kind have been considered in the metric approach in an attempt to find ghost and singularity free theories of gravity @xcite , and also in the study of black holes coupled to neds @xcite . in the palatini framework , @xmath11 theories yield second - order field equations that exactly boil down to the usual einstein - de sitter equations in vacuum ( see details in @xcite and below ) , which guarantees the absence of ghosts and other dynamical instabilities . the presence of a ricci - squared term in the action is very important because it leads to modified dynamics even for traceless matter sources , such as radiation and the usual maxwell electromagnetic field , which contrasts with @xmath2 theories . as a result , this quadratic @xmath11 model provides the simplest extension beyond gr of the usual ( nonrotating ) reissner - nordstrm black hole . the quadratic @xmath11 palatini model proposed here has already been studied in cosmological scenarios , where it was found that the big bang singularity is replaced by a cosmic bounce in isotropic and anisotropic ( bianchi i ) universes filled with standard sources of matter and radiation @xcite . in refs.@xcite we reported on several key aspects of nonsingular black hole solutions found in this model . in this paper we shall go further on this subject extending those results and providing a comprehensive account of all the derivations and technical details . the paper is organized as follows . in section [ section : ii ] we recall some basic aspects of palatini @xmath11 theories and write the associated field equations . the particular actions for matter ( maxwell ) and gravity sectors of our theory are introduced in section [ section : iii ] , where we construct all the relevant geometric objects . in section [ section : iv ] we comment on the choice of ansatz for the line element , and proceed to solve the metric field equations . section [ section : v ] is devoted to the analysis of the interior region and to the characterization of the event horizons . in particular we construct the penrose diagrams and their maximal analytical extensions associated to the different black holes and _ naked cores _ found . the physical aspects of these solutions , including singular and nonsingular black holes , and microscopic _ naked cores _ , are discussed in section [ section : vi ] . we conclude in section [ section : vii ] with a summary and some future perspectives . we define palatini @xmath11 theories as follows @xmath12=\frac{1}{2\kappa^2}\int d^4x \sqrt{-g}f(r , q ) + s_m[g,\psi_m],\ ] ] where @xmath13 , @xmath14 $ ] represents the matter action , @xmath15 is the space - time metric , @xmath16 , @xmath17 , @xmath18 , and @xmath19 variation of ( [ eq : action ] ) with respect to metric and connection leads to the following equations @xcite @xmath20&=&0 \ , \label{eq : con - varx}\end{aligned}\ ] ] were we have used the short - hand notation @xmath21 . in the above derivation we have assumed a symmetric ricci tensor , @xmath22 , and vanishing torsion . the condition on the ricci tensor , @xmath23}=0 $ ] , forces the connection components @xmath24 to be the gradient of a scalar function , @xmath25 . in the usual formulation of gr , where the connection is given by the christoffel symbols of the metric , one finds that @xmath26 . in our theory ( [ eq : action ] ) , the condition @xmath23}=0 $ ] is equivalent to assuming that @xmath27 can be solved as the levi - civita connection of an auxiliary metric @xmath28 , which leads to @xmath29 . the explicit relation between @xmath28 and @xmath30 follows from the field equations and will be discussed later . a reason to set @xmath23}=0 $ ] is that then the field equations of ( [ eq : action ] ) in vacuum boil down exactly to those of gr ( with possibly a cosmological constant , depending on the function @xmath11 chosen ) . this guarantees that there are no new propagating degrees of freedom and , therefore , the resulting theory is not affected by ghosts or other dynamical instabilities . in regions containing sources , the equations of ( [ eq : action ] ) differ from those of gr due to the presence of new matter / energy - dependent terms induced by the mismatch between @xmath28 and @xmath30 , which leads to modified gravitational dynamics without introducing new dynamical degrees of freedom or higher - order derivatives of the metric . if the condition @xmath23}=0 $ ] is relaxed , then @xmath31 must also have a purely vectorial component , @xmath32 , which adds new dynamical degrees of freedom to the theory . in that case , the dynamics of ( [ eq : action ] ) differs from that of gr even in the absence of matter / energy sources ( see @xcite for related results in this direction ) . we now focus on working out a solution for ( [ eq : con - varx ] ) . at first sight , since @xmath33 and @xmath8 are functions of the connection and its first derivatives , ( [ eq : con - varx ] ) can be seen as a nonlinear , second - order partial differential equation for the unknown connection . however , there exist algebraic relations between @xmath33 , @xmath8 and the energy - momentum tensor of the matter that make the problem easier to handle . to proceed , we first define the matrix @xmath34 , whose components are @xmath35 , which allows us to express ( [ eq : met - varx ] ) as @xmath36 in matrix notation , this equation reads @xmath37 where @xmath38 is the matrix representation of @xmath39 . note that @xmath33 and @xmath40 are the trace of @xmath34 and @xmath41 , respectively . the solution of ( [ eq : met - varrq2 ] ) implies that @xmath34 can be expressed as a function of the components of the energy - momentum tensor , i.e. , @xmath42 . assuming that for a given @xmath11 theory such a solution exists , eq.([eq : con - varx ] ) can now be seen as an algebraic equation for the connection in which , besides the metric @xmath30 , there is an explicit dependence on the energy - momentum tensor of the matter . to solve it , we look for a metric @xmath43 such that ( [ eq : con - varx ] ) becomes @xmath44=0 $ ] . this guarantees that the independent connection can be expressed as the levi - civita connection of @xmath43 . using matrix notation , we have @xmath45 where we have defined @xmath46 . computing the determinant of this expression , we find @xmath47 . with this result , we have @xmath48 this shows that the connection of @xmath11 theories can be explicitly solved in terms of the physical metric @xmath30 and the matter sources . with the above results , the metric field equations can be rewritten in a more compact and transparent form . expressing ( [ eq : met - varrq2 ] ) as @xmath49 and using ( [ eq : h - g ] ) , we can rewrite @xmath50 as @xmath51 . this allows to express ( [ eq : met - varrq3 ] ) as @xmath52 where @xmath53 . the schwarzschild black hole is the most general spherically symmetric , nonrotating vacuum solution of gr and also of ( [ eq : met - varx ] ) . however , that solution assumes that all the matter is concentrated on a point of infinite density , which is not consistent with the dynamics of ( [ eq : met - varx ] ) . in fact , if one considers the collapsing object as described by a perfect fluid that behaves as radiation during the last stages of the collapse , explicit computation of the scalar @xmath54 shows that the energy density @xmath55 is bounded from above by @xmath56 , where @xmath57 g/@xmath58 is planck s density @xcite . therefore , one should study the complicated process of collapse of a spherical nonrotating object to determine how the schwarzschild metric is modified in our theory . for this reason we study instead vacuum space - times with an electric field , which possess a nonzero stress - energy tensor able to excite the palatini dynamics even in static settings . the resulting solutions should therefore be seen as planck - scale modifications of the usual reissner - nordstrm solution . let us thus consider as the matter source in action ( [ eq : action ] ) the maxwell lagrangian @xmath59 = -\frac{1}{2\kappa^2}\int d^4x \sqrt{-g}f_{\alpha\beta}f^{\alpha\beta}\ ] ] whose associated stress - energy tensor is written as @xmath60 \ , \ ] ] where @xmath61 is the field strength tensor of the vector potential @xmath62 . for a purely electrostatic configuration and taking a spherically symmetry line element of the form @xmath63 , with @xmath64 , one finds that the only nonvanishing component is @xmath65 . it is then easy to see that the ( sourceless ) field equations @xmath66 lead to @xmath67 where @xmath68 is an integration constant that represents the charge generating the electric field . with this result , the product @xmath69 in ( [ eq : em ] ) becomes @xmath70 where @xmath71 and @xmath72 represent the @xmath73 identity and zero matrices , respectively . using this , we find that @xmath74 in order to write the field equations in the form ( [ eq : met - varrq4 ] ) , we first need to find the explicit form of @xmath34 for this matter source , which will allow us to construct @xmath75 and compute its determinant . to do this , we use ( [ eq : tmn - em ] ) and write ( [ eq : met - varrq2 ] ) as @xmath76 where @xmath77 and we have defined @xmath78 . it is easy to see that there are @xmath79 square roots that satisfy this equation , namely , @xmath80 where @xmath81 . agreement with gr in the low curvature regime ( where @xmath82 and @xmath83 ) requires @xmath84 . for this reason , we simplify the notation and take @xmath85 from this it follows that the matrix @xmath75 is given by @xmath86 where @xmath87 . from this expression it is easy to see that @xmath88 and , therefore , the field equations ( [ eq : met - varrq4 ] ) become @xmath89 to work out the explicit form of the metric we must specify an @xmath11 model . it is very useful to consider the family @xmath90 , where @xmath91 is planck s length , because tracing ( [ eq : met - varx ] ) with the metric @xmath92 leads to the well known @xmath2 relation @xmath93 , which implies that @xmath94 . since for maxwell theory @xmath95 , it follows that the @xmath96 part of the @xmath11 theory does not play a very relevant role in the dynamics . we will just assume that the @xmath96 part is close to gr , @xmath97 , and that @xmath98 for simplicity ( and for consistency with the choice @xmath84 above ) . this is true , in particular , for the model @xmath99 whose cosmological dynamics has been carefully studied in the literature @xcite and that we set as the model to be discussed from now on . for this model we have that when @xmath100 is given by ( [ eq : tmn - em ] ) then @xmath101 , @xmath102 , and @xmath103 . using this in ( [ eq : m_ab ] ) and taking the trace we find @xmath104 from which we obtain @xmath105 which coincides with the expression of gr . from this result it follows that @xmath106 and @xmath107 . noting also that @xmath108 , ( [ eq : ricci - h3 ] ) becomes @xmath109 which fully determines the dynamics of our @xmath11 theory coupled to maxwell electrodynamics . these equations exactly recover gr in the limit @xmath110 . in order to solve for the metric @xmath30 , it is more convenient to solve first for @xmath28 using ( [ eq : ricci - h4 ] ) and then transform back to @xmath30 using the relation @xmath111 that follows from ( [ eq : h - g ] ) . to do this , it is convenient to define a line element associated to the metric @xmath28 using a set of schwarzschild - like coordinates as follows @xmath112 this line element is formally identical to that corresponding to the physical metric @xmath30 , @xmath113 but their relation is not trivial due to the nonconformal relation between the two metrics and the different choice of radial coordinate @xmath114 . this can be seen from the relation @xmath115 , which leads to @xmath116 from the line element ( [ eq : ds2h ] ) it is easy to see that @xmath117 implies that @xmath118 . it is also easy to see that @xmath119 . since the time coordinate is the same in the two line elements , for the @xmath120 component we just have @xmath121 . it should be noted that , in general , the line elements can be written without using @xmath122 and @xmath123 as coordinates @xcite . in that case , we would have @xmath124 and @xmath125 , with the relations @xmath126 and @xmath127 , being the latter clearly independent of the choice of @xmath128 . at this point , it is important to note that the radial function @xmath129 vanishes when @xmath130 . this means that the @xmath131spheres of the @xmath132geometry ( the geometry associated to the independent connection ) can only be put into correspondence with the @xmath131spheres of the @xmath133geometry up to @xmath134 . since the @xmath132geometry deviates from the @xmath133geometry by the effects of a matter - induced deformation ( represented by the matrix @xmath135 ) , the impossibility of mapping any portion of the @xmath132geometry into the @xmath136 sector of the @xmath133 geometry suggests that the matter ( the electromagnetic field in this case ) can not penetrate in that region of the physical spacetime . the description and analysis of the properties of the hypersurface @xmath137 will be one of the main goals of this paper . working with the schwarzschild ( canonical ) coordinates of above and introducing the ansatz @xmath138 and @xmath139 , the components of the tensor @xmath140 become @xmath141 \\ { r_{\tilde{r}}}^{\tilde{r}}&=&-\frac{1}{2h_{\tilde{r}\tilde{r}}}\bigg[\frac{a_{\tilde{r}\tilde{r}}}{a}- \left(\frac{a_{\tilde{r}}}{a}\right)^2 + 2\psi_{\tilde{r}\tilde{r}}+ \nonumber \\ & + & \left(\frac{a_{\tilde{r}}}{a}+ 2\psi_{\tilde{r}}\right)\left(\frac{a_{\tilde{r}}}{a}+\psi_{\tilde{r}}\right)+ \frac{2}{\tilde{r}}\frac{a_{\tilde{r}}}{a}\bigg ] \label{eq : rrr}\\ { r_\theta}^\theta&=&\frac{1}{\tilde{r}^2}\left[1-a(1+\tilde{r}\psi_{\tilde{r}})-\tilde{r}a_{\tilde{r}}\right ] \ . \label{eq : rzz}\end{aligned}\ ] ] from ( [ eq : ricci - h4 ] ) it is easy to see that the combination @xmath142 . using ( [ eq : rtt ] ) and ( [ eq : rrr ] ) , this combination implies that @xmath143 , like in gr and @xmath2 theories . as usual , this constant can be eliminated by a redefinition of the time coordinate , leaving a single function ( @xmath144 ) to be determined . from ( [ eq : rzz ] ) we find that @xmath144 satisfies the following equation @xmath145=\frac{\tilde{\kappa}^2q^2}{2r^2}\frac{1}{\left(r^2-\frac{\tilde{\kappa}^2q^2l_p^2}{r^2}\right)}\ ] ] using the ansatz @xmath146 and the relation @xmath147 , ( [ eq : a(r ) ] ) turns into @xmath148 taking into account that @xmath149 , the above expression becomes @xmath150 which reduces the problem to solving a first - order differential equation in the variable @xmath122 . the integration constant of this equation can be identified with the schwarzschild mass @xmath151 of the vacuum problem ( @xmath152 ) . we can thus write @xmath153 to emphasize that it is the function @xmath154 which encodes the electrostatic contribution to the mass function . in order to obtain @xmath154 , it is useful to introduce some definitions to work with dimensionless variables . we thus define a length ( squared ) associated to the charge , @xmath155 , and a dimensionless radial variable @xmath156 . with this notation , the metric @xmath30 can be expressed as @xmath157}{\delta_2 z \sigma_-^{1/2 } } \ .\ ] ] where @xmath158 , we used the relation @xmath159 and defined @xmath160 , together with @xmath161 the ( dimensionless ) function @xmath162 has a purely electrostatic origin and satisfies the following equation @xmath163 we stress that the two scales of the problem , namely , the integration constants @xmath68 and @xmath164 , have been replaced by the dimensionless ratios @xmath165 and @xmath166 given in ( [ eq : d1d2 ] ) . the integration of ( [ eq : gz ] ) to obtain the function @xmath162 can be carried out straightforwardly using power series expansions in two regions of interest . one is the far limit @xmath167 and the other is @xmath168 . to study this limit , it is useful to express @xmath169 as follows @xmath170 using the binomial expansion @xmath171 , the above expression becomes @xmath172 the integration of this expression is immediate and gives @xmath173 from the first terms of this expansion , @xmath174 , one can verify that when @xmath175 the gr limit ( i.e. the standard reissner - nordstrm solution ) is recovered @xmath176 where the first - order corrections come from the @xmath177 functions rather than from the expansion of @xmath162 . as it follows from ( [ eq : gtt - far ] ) and ( [ eq : grr - far ] ) , as long as @xmath178 , the location of the external horizon in these black holes is essentially the same as in gr . how this picture changes for microscopic black holes will be discussed later on . to study this limit , we first consider the change of variable @xmath180 , which leads to @xmath181 \ .\ ] ] using again the binomial expansion , we find the following solution @xmath182 \nonumber \\ & \approx & 2 \sqrt{z-1}-\frac{11}{6 } ( z-1)^{3/2}+o[z-1]^{5/2}.\end{aligned}\ ] ] to have agreement with the solution ( [ eq : far ] ) found before , we need to add an integration constant @xmath183 on the right hand side of this expansion , which leads to @xmath184 \ , \ ] ] where @xmath185 $ ] is a hypergeometric function , and @xmath186 . the mass function can then be written as @xmath187\right ) \ .\ ] ] the series expansion provided here can be used to perform analytical studies of the geometry near @xmath188 . for that purpose , the computer algebra mathematica package xact @xcite will be very useful . to study the internal geometry of black holes , it is convenient to use coordinates in which the metric is well - defined even at the event horizons . for line elements of the form we are using , @xmath189 , it is particularly useful to replace the @xmath190 coordinates by the so - called eddington - finkelstein coordinates @xmath191 @xcite that turn the line element into @xmath192 where @xmath193 with @xmath194 , and @xmath195 is such that @xmath196 . from our definitions in ( [ eq : g ] ) , we have @xmath197 , @xmath198 , @xmath199 , where @xmath200 , and we can also define @xmath201 ( for notational convenience , from now on we use the dimensionless variables @xmath202 and @xmath203 instead of @xmath122 and @xmath204 ) . the line element ( [ eq : ef ] ) puts forward that the geometry is fully characterized by the functions @xmath205 and @xmath206 . the relation between @xmath202 and @xmath203 can be found by direct integration and is given by @xmath207 z$ ] . for @xmath167 , we have @xmath208 , whereas for @xmath179 we find @xmath209 , where @xmath210}/{\gamma\left[\frac{1}{4}\right]}\approx 0.59907 $ ] . the relation between @xmath202 and @xmath203 is monotonic and invertible in the region @xmath211 . from ( [ eq : gz ] ) we also see that @xmath205 is only defined for @xmath211 . as we pointed out above , this is a dynamical consequence of the theory , not a coordinate problem . to learn more about this point , we need to study the properties of the geometry as the region @xmath179 is approached . the general expansion of the metric function @xmath205 of ( [ eq : ef ] ) around @xmath212 leads to @xmath213 this expansion shows that the metric component @xmath214 is in general divergent as @xmath179 . however , it also points out the existence of a particular mass - to - charge ratio for which the divergence disappears . in fact , if we take @xmath215 , we get @xmath216 where , according to ( [ eq : d1d2 ] ) , @xmath217 for macroscopic black holes . this result implies that when @xmath218 the line element and the metric components are finite everywhere . it is worth noting that this feature is quite similar to what is found in some neds in gr , for which a particular combination between matter and black hole parameters give rise to metrics which are finite everywhere @xcite . in our case , it is the nonlinearity in the gravitational action ( [ eq : action ] ) , as opposed to the nonlinearity in the matter sector of neds , which gives rise to this effect . it should be noted that even though in gr with some neds the metric may be finite everywhere , some of the curvature invariants always blow up at the center of the solutions , according to a well known theorem holding for electrostatic fields with maxwell weak field limit @xcite . to see what happens in our theory , we consider next the behaviour of the ricci scalar , ricci squared , and kretschmann scalar of the metric @xmath30 in the regions @xmath179 and also @xmath167 . using the exact solution for the metric found above , we can compute some relevant curvature invariants to extract useful information about the geometry . the simplest such objects are the ricci scalar @xmath219 , the ricci squared scalar @xmath220 , and the kretschmann scalar @xmath221 , where @xmath222 represents the riemann tensor of the metric @xmath15 . in gr , these objects have the following exact values @xmath223 though the ricci scalar in gr is zero , because the stress - energy tensor of the electromagnetic field is traceless , the other two scalars are nonzero and , in fact , diverge as @xmath224 , which signals the existence of a strong singularity at @xmath225 . since the solution that we found for the metric in our theory is given in terms of infinite series expansions , we find it useful to compute the above scalars in the two natural regimes in which those solutions were found , namely , in the region @xmath167 and in the limit @xmath179 . when @xmath167 , we find the following expansions ( @xmath226 and @xmath227 ) @xmath228 it is clear that these results recover @xmath229 when @xmath230 ( or , equivalently , @xmath167 ) . however , in the region @xmath179 the behaviour of those curvature scalars is completely different from the @xmath167 expansions . when @xmath179 , we find @xmath231 \nonumber,\end{aligned}\ ] ] @xmath232 \nonumber \\ & + & \left(1-\frac{\delta _ 1^*}{\delta _ 1}\right)^2\left[\frac{1}{8 \delta _ 2 ^ 2 ( z-1)^3}-o\left(\frac{1}{({z-1})^2}\right)\right ] \nonumber , \end{aligned}\ ] ] @xmath233+\nonumber \\ & + & \left(1-\frac{\delta _ 1^*}{\delta _ 1}\right){}^2\left[\frac{1}{4 \delta _ 2 ^ 2 ( z-1)^3}+o\left(\frac{1}{({z-1})^2}\right)\right ] \nonumber.\end{aligned}\ ] ] from these expansions we see that the curvature scalars diverge at @xmath188 except for those configurations whose charge - to - mass ratio satisfies the condition @xmath218 , since in that case all of them become finite . this is an important result whose physical consequences will be explored in more detail later . for now , it should be noted that the avoidance of the singularity is a local , nonperturbative effect that has no impact on the structure of the black hole at distances @xmath167 , which quickly tends to that of gr regardless of the particular value of @xmath165 , as can be seen from the expansions in ( [ eq : rgr ] ) , ( [ eq : qgr ] ) , and ( [ eq : kgr ] ) . consider the normal to a hypersurface @xmath234constant , namely , @xmath235 . if @xmath236 , then @xmath237 , @xmath238 , and @xmath239 . this result shows that the vector @xmath240 is spacelike outside the external horizon , null at the horizon , and timelike inside the horizon except at @xmath188 , where it becomes null again ( regardless of the value of @xmath165 ) due to the presence of @xmath241 . this implies that the singularity found for @xmath242 lies on a null hypersurface , which contrasts with the schwarzschild ( spacelike ) and reissner - nordstrm ( timelike ) singularities of gr . in the context of gr , null singularities have been found in the interior of reissner - nordstrm black holes perturbed by neutral matter @xcite . further exploration of the connection between these two results shall be done elsewhere . to learn more about this null hypersurface , we compute now its surface gravity using the killing vector @xmath243 . in the coordinates ( [ eq : ef ] ) , which are regular across the external horizon , the components of this vector are @xmath244 and @xmath245 . since the surface gravity @xmath246 is defined by @xmath247 ( evaluated at @xmath188 ) , it follows that @xmath248 , which leads to @xmath249 note that , strictly speaking , the surface gravity only makes sense when evaluated on a horizon . since in the cases @xmath242 the null surface @xmath188 is singular , we believe that the divergence of @xmath246 in those cases is just a manifestation of the breakdown of the geometric description on that surface . on the contrary , the vanishing of @xmath246 at the horizon @xmath188 when @xmath250 puts forward the smoothness of the geometry at that location . we refrain ourselves from interpreting these results in a thermodynamic context because this aspect of palatini theories of gravity has not been considered in the literature with sufficient detail yet . horizons are located at the points where the metric function @xmath251 vanishes . from the definitions given above , this happens when the curves @xmath252 and @xmath253 meet . we have already seen analytically that for large black holes , @xmath254 , the external horizon lies almost at the same location as in gr . however , since the internal structure of our black holes is different from that of gr the very existence of an inner horizon is not guaranteed a priori . moreover , the location of the external horizon for microscopic black holes may also significantly differ from the prediction of gr . for these reasons , in this section we focus on these points to complete our analysis of the internal structure of these black holes . we note that due to the character of the solutions , given as infinite power series , the best way to determine whether inner horizons exist or not is by using a graphical representation of the functions @xmath255 and @xmath256 . from these representations , see figs . [ fig:1 ] , [ fig:2 ] , and [ fig:3 ] , one realizes that the first terms of the expansions may provide useful information to understand the main features of the various cases of interest . in this sense , it is worth noting that both @xmath255 and @xmath256 are monotonic functions whose asymptotes are @xmath257 and @xmath258 , respectively . this implies that if for some @xmath259 we have @xmath260 then at some @xmath261 we will have @xmath262 , which implies the existence of an horizon ( as can be verified graphically ) . following this reasoning , we expand @xmath255 and @xmath256 around @xmath188 and identify @xmath263 as the condition for the existence of an inner horizon in that region . using the expansions @xmath264 and @xmath265 , we find the following cases : * * if @xmath218 then @xmath266 . * + this relation translates into the condition @xmath267 . expressing the charge as @xmath268 , where @xmath269 is the electron charge and @xmath270 the number of charges , we can write @xmath271 , where @xmath272 is the electromagnetic fine structure constant . with this notation , the above condition becomes @xmath273 . this result means that in order to have an horizon when @xmath274 , the number of charges must be at least equal or greater than @xmath275 . for smaller values of the charge one can verify graphically that there are neither inner nor outer horizons , which represents _ naked core _ solutions . when the charge is greater than @xmath275 then we have only an external horizon and no inner horizon . all of these solutions are free of curvature singularities , as results from the analysis of section [ sec : v.b ] . + ( solid green ) with @xmath276 ( solid red curves labeled by @xmath270 ) . the dashed ( orange ) straight lines that meet at the origin correspond to @xmath277 . the dashed black curve is @xmath278 . note that @xmath279 and @xmath276 quickly recover the gr behaviour for @xmath167 . as a result , the location of the external horizon almost coincides with the gr prediction for @xmath280 . [ fig:1],scaledwidth=45.0% ] * * if @xmath281 then @xmath282 . * + this condition indicates that once the charge - to - mass ratio @xmath283 and the number of charges @xmath270 are specified , one can always find some @xmath284 that verifies the inequality , which implies the existence of an horizon . therefore , regardless of the value of @xmath270 , when @xmath281 we always have an ( external ) horizon ( see fig.[fig:2 ] ) . + . when @xmath281 , we have @xmath285 , which forces all the curves @xmath256 to cut @xmath255 in a single point . except for very small values of @xmath270 , the location of the ( external ) horizon almost coincides with the gr prediction . in this plot @xmath286 . [ fig:2],scaledwidth=45.0% ] * * if @xmath287 then @xmath288 . * + from this it follows that for some combinations of @xmath165 and @xmath270 there may or may not exist a @xmath284 that satisfies the inequality . this means that in some cases there is no horizon ( @xmath289 always ) , which implies a naked singularity , while in other cases there may be up to two horizons . this is verified graphically in fig . [ fig:3 ] , where we can appreciate solutions without horizon , solutions with two horizons , and solutions with one ( degenerate ) horizon ( extreme black hole ) , a situation analogous to that of the standard reissner - nordstrm solution of gr . when @xmath287 , some configurations have two horizons ( see the curves with @xmath290 ) while others have no horizons ( like @xmath291 ) . the limiting case @xmath292 ( not shown here ) has only one ( degenerate ) horizon and represents an extreme black hole . the set @xmath293 represents naked singularities . in this plot @xmath294 . [ fig:3],scaledwidth=45.0% ] the previous discussion about horizons provides useful information for the construction of the penrose diagrams corresponding to the solutions found . for instance , we have seen that when @xmath281 , all solutions represent black holes with a single horizon . behind the horizon we find a singularity located at @xmath295 . this configuration is essentially the same as that found in gr for schwarzschild black holes , except for the fact that the singularity is now null instead of spacelike . as a result , the corresponding penrose conformal diagram is that represented in fig . [ fig : penrose_i ] . . unlike in the schwarzschild black hole , where the @xmath225 singularity is spacelike , the singularity here is null and appears at @xmath295 . [ fig : penrose_i],scaledwidth=45.0% ] when @xmath287 , we may find the same subcases as in gr , namely , solutions with two horizons , with one double ( degenerate ) horizon , and naked singularities . like in the previous example , the main difference is that the singularity is null rather timelike . to illustrate how the penrose diagram is modified , we plot the case with two horizons in fig . [ fig : penrose_ii ] . with two horizons ( @xmath296 and @xmath297 ) . the only difference with respect to the gr diagram is that the singularity lies on a null surface at @xmath295 . [ fig : penrose_ii],scaledwidth=45.0% ] the case @xmath218 deserves special attention because the null hypersurface @xmath295 is nonsingular . unlike in the other cases with @xmath298 , the absence of a singularity suggests that the geometry may admit some analytical extension beyond that point . this extension is naturally obtained from the relation @xmath299 that defines the function @xmath300 in ( [ eq : ef ] ) . in our analysis following ( [ eq : ef ] ) , we implicitly assumed that @xmath301 , and omitted the possibility of having a branch with the negative sign , @xmath302 . in the singular cases @xmath298 , the omission of the branch with @xmath303 is justified because there is no natural way to extend the geometry beyond the singularity at @xmath295 . however , if there is no singularity at @xmath295 , the divergence of @xmath304 at this point simply states that the function @xmath305 has reached a minimum at the point @xmath306 ( see the discussion following eq.([eq : ef ] ) and fig . [ fig : zzstar ] ) . for values of @xmath307 , the branch with @xmath303 describes a new region in which the area of the @xmath131spheres grows as @xmath308 . the relation between @xmath122 and @xmath204 can thus be written explicitly as follows ( see fig . [ fig : zzstar ] ) @xmath309 r & \text { if } r^*\ge r^*_c \\ 2r^*_c-{_{2}f}_1\left[-\frac{1}{4},\frac{1}{2},\frac{3}{4},\frac{r_c^4}{r^4}\right ] r & \text { if } r^*\le r^*_c \end{array}\right.\ ] ] in terms of @xmath204 , we also have @xmath310 , which gives continuity to the metric across the bounce . , @xmath311 and @xmath312 . the minimum of @xmath206 occurs at @xmath313 , where @xmath311 vanishes and @xmath312 reaches its maximum value . recall that @xmath201 . [ fig : zzstar],scaledwidth=45.0% ] using @xmath314 and @xmath315 to parameterize the @xmath131spheres when @xmath316 and @xmath317 , respectively , the line element ( [ eq : ef ] ) can be written as @xmath318 this representation is useful to understand that for future directed ( @xmath319 ) worldlines , @xmath320 on timelike or null worldlines if @xmath316 , but @xmath321 if @xmath307 . from a physical point of view , this means that a spherical shell of matter that collapses and crosses the external horizon will unavoidably shrink to a sphere of area @xmath322 before bouncing off into an outgoing shell of increasing area . the outgoing shell , obviously , can not return to the region from which it was sent initially because it crossed an event horizon . as can be seen from the penrose diagram of this spacetime in fig . [ fig : penrose_iii ] , the outgoing shell can reach several different final regions ( different @xmath323 regions for light rays ) . . the null surfaces labeled as @xmath324 represent the region where the area of the @xmath131spheres reaches its minimum ( nonzero ) value . an incoming null geodesic ( light ray ) propagating from @xmath325 towards the horizon @xmath296 , will reach @xmath324 and bounce off as an outgoing null geodesic into another region , eventually reaching @xmath326 ( if no scattering takes place ) . the surface @xmath324 , therefore , plays the role of ( attractive ) black hole and ( repulsive ) white hole at the same time . [ fig : penrose_iii],scaledwidth=45.0% ] we now discuss several points useful to understand some physical aspects of the solutions described in previous sections . we first note that using @xmath327 we find that the area of the @xmath295 surface is given by @xmath328 , where @xmath329 is planck s area . this admits a nice physical interpretation since it suggests that each charge sourcing the electric field has associated an elementary quantum of area of magnitude @xmath330 . from this it follows that the ratio of the total charge @xmath68 by the area of this surface gives a universal constant , @xmath331 , which up to a factor @xmath332 coincides with the planck surface charge density . this result is independent of the mass of the black hole and , therefore , applies both to singular as well as to regular solutions . if we focus now on the regular solutions only , we find that the regularity condition @xmath218 sets the following mass - to - charge relation @xmath333 which can be interpreted in the sense that the matter density inside a sphere of radius @xmath334 becomes another universal constant @xmath335 from the definition of @xmath165 and @xmath336 , comparing any two configurations one can verify that @xmath337 this relation is very useful to classify the different black hole configurations , because the relation @xmath338 that appears in ( [ eq : gtt_series ] ) can be written as @xmath339 , where @xmath340 is the density of the regular core . this representation allows for a more physical interpretation and classification of the solutions given in sec . [ section : v ] . in particular , the nonsingular configuration @xmath341 is the only case in which the core density is a magnitude independent of @xmath164 and @xmath68 and , in fact , turns out to be given in terms of fundamental constants only . this fact puts forward the special quantum - gravitational nature of the nonsingular solutions . in this sense , we believe that the regularity condition @xmath341 , rather than as a fine - tuning issue , should be interpreted in the spirit of a quantum constraint relating the mass and charge ( or charges , in general ) of the solutions to avoid the singularity . it should also be noted that adiabatic changes of the charge and mass of the black hole @xcite do not allow transitions between nonsingular configurations , which indicates that such configurations have different quantum numbers ( different @xmath68 and @xmath164 but the same @xmath340 ) . this must have important consequences for hawking radiation , because if the emission of quanta is to be compatible with the regularity of the core , then the spectrum must necessarily have a discrete structure . we shall leave the exploration of this issue for future works . the above results picture a black hole interior with an ultracompact core of radius @xmath342 which contains all the mass in its interior and all the charge on its surface . this view is physically very appealing but should be compatible with the mathematical solution represented in fig . [ fig : penrose_iii ] . in fact , fig . [ fig : penrose_iii ] represents an exact mathematical solution of a physically incomplete problem , because it does not take into account the necessary existence of the massive charged particles that generate the electrostatic field . the fact that the surface @xmath295 is null suggests that if massive charged particles were explicitly included in the problem , then the killing vector field @xmath343 could become again timelike in the region hidden by the @xmath295 horizon , where the sources should be located , thus allowing for the existence of static interior solutions of the type suggested by the constraint ( [ eq : rq - rs ] ) . therefore , for the description of the geometry behind the @xmath295 horizon , one should specify the @xmath39 of the sources that carry the mass of the core and the charge that generates the external electric field , which would allow to define a new auxiliary metric @xmath344 able to parameterize the internal geometry of the core ( assuming that suitable matching conditions can be found at @xmath295 ) . since the regularity condition ( [ eq : rq - rs ] ) supports that the core matter density is bounded , we expect the existence of completely regular solutions in agreement with the results found for this same theory of gravity in cosmological scenarios @xcite . it is worth pointing out that if the @xmath295 null surface could be smoothly matched to an interior region filled with matter , where @xmath343 were timelike , then the analytical extension of the exterior spacetime beyond @xmath295 would admit two possible and different branches , which would appear on different sheets of a larger conformal diagram . one sheet would contain the matter - filled region and another the analytical extension shown in fig . [ fig : penrose_iii ] , where the radial coordinate _ bounces_. this could give complete physical reality to the spacetime of fig . [ fig : penrose_iii ] in such a way that particles approaching @xmath295 would be scattered into the _ white hole _ region instead of falling into the matter - filled sector . this and related phenomenological issues will be explored elsewhere . from the large @xmath202 expansion in ( [ eq : gtt - far ] ) and ( [ eq : grr - far ] ) , we saw that the gr solution @xmath345 is an excellent approximation for any @xmath175 . this implies that the location of the external horizon of these charged black holes is essentially the same as in gr . we thus find @xmath346 where ( [ eq : rq - rs ] ) has been used . for a solar mass black hole , where the number of protons is around @xmath347 , eq.([eq : rq - rs ] ) implies that the number of charges needed to avoid the @xmath188 singularity is just @xmath348 ( or @xmath349 moles ) , which is a very tiny amount on astrophysical terms . in fact , this amount of charge is comparatively so small , @xmath350 , that it seems reasonable to expect that a quantum gravitational violation of electric charge conservation could naturally act to avoid black hole singularities in stellar collapse processes . in general , @xmath351 implies that in astrophysical scenarios @xmath352 ( see figs . [ fig:1 ] , [ fig:2 ] and [ fig:3 ] ) . we have already seen that as @xmath270 drops below the critical value @xmath353 , the external horizon disappears and the core undresses becoming directly _ observable _ ( see fig . [ fig:1 ] ) . the penrose diagram corresponding to the maximal analytical extension of these naked regular cores , @xmath354 , is depicted in fig . [ fig : penrose_iv ] . without event horizon , @xmath354 . if the asymptotic region @xmath326 is identified with @xmath355 , then the surface @xmath324 should behave as a perfectly reflecting boundary . in the general case , however , light rays reaching @xmath324 would go through into a new universe . [ fig : penrose_iv],scaledwidth=45.0% ] from ( [ eq : rq - rs ] ) , it follows that the mass of these objects is @xmath356 for the particular case @xmath357 , we find @xmath358 ^ 2)\approx 1.23605 m_p$ ] . it is worth mentioning that this very particular number also appears in the computation of the total energy associated to the electric field in born - infeld nonlinear electrodynamics , which indeed is found to be @xmath359 ^ 2 ) \beta^{1/2}q^{1/2}$ ] with @xmath183 the born - infeld parameter @xcite . its presence here puts forward the regularizing role played by gravitation . moreover , as the electromagnetic field itself is expected to receive corrections at high energies ( see @xcite for some string theory results in this regard ) , it also points out that an improved description of the electromagnetic field through nonlinear electrodynamics in these scenarios could provide interesting new insights on the interaction between the nonlinearities of matter and gravitation . all these facts suggest that the theory ( [ eq : action ] ) may shed new light on the problem of sources in electrodynamics coupled to gravitation @xcite , which already arises in the reissner - nordstrm and schwarzschild solutions of gr @xcite . finally it is worth noting that if @xmath270 is seen as an integer number , besides the core area , the mass of these nonsingular naked cores and black holes is quantized , as has been recently claimed in @xcite on very general grounds . as pointed out before , one thus expects a discrete spectrum of hawking radiation , because physically allowed transitions should occur only between regular configurations . this illustrates how planck - scale physics may affect the perturbative predictions of the semiclassical approach . new investigations on all these issues are currently underway and shall be published elsewhere . from the analysis of the geometry in the @xmath179 region , we have found that the solutions of our theory can be classified into three types according to the value of @xmath165 or , equivalently , according to the value of the core density @xmath336 . when @xmath360 , then the conformal diagram of fig . [ fig : penrose_i ] shows a structure very similar to that found for the schwarzschild black holes of gr . this admits a nice physical interpretation because when the core contains more mass than in the nonsingular case , the black hole looks more like an uncharged object , such as a schwarzschild black hole . on the other hand , when the mass - to - charge ratio is smaller than the expected value of a regular configuration , @xmath361 , then the resulting structure is closer to a typical reissner - nordstrm black hole of gr ( see fig . [ fig : penrose_ii ] ) . our black holes , therefore , are somehow smoothly interpolating between the abrupt structures found in gr . the nonsingular solutions , @xmath341 , lie in the middle of these two disconnected branches ( schwarzschild - vs - reissner - nordstrm ) , and represent a kind of object not found in gr but that emerges when planck scale effects are incorporated in the problem . the smooth interpolation found here between the schwarzschild and reissner - nordstrm solutions of gr shares some resemblance with the behaviour of the contracting and expanding branches of the nonsingular cosmological solutions found in this palatini @xmath11 model @xcite . in gr , one can classify cosmological solutions in two groups , one corresponding to an expanding branch that starts in a ( big bang ) singularity , and its time reversal , which corresponds to a contracting branch that ends in that ( big crunch ) singularity . in the quadratic palatini model studied here , the two singular branches of gr are smoothly connected through a bouncing solution that avoids the singularity . in this work we have studied the structure of electrically charged black holes in a palatini extension of gr characterized by a ricci - squared term . theories of this type could be naturally motivated by quantum effects in curved spacetimes . these theories provide modified dynamics without introducing new dynamical degrees of freedom . this implies , in particular , that the resulting solutions can be completely classified using the same parameters as one finds in gr , namely , the total charge @xmath68 and total mass @xmath164 . the absence of new dynamical degrees of freedom also guarantees that these theories are free from ghosts and other potential dynamical instabilities . we have obtained exact analytical solutions expressed as infinite power series expansions . these solutions show that the structure of these black holes coincides with that of the well known reissner - nordstrm black holes of gr for values of @xmath230 , where @xmath362 , with @xmath270 representing the number of charges and @xmath363 . important modifications arise as we approach the minimum of the radial coordinate , the region @xmath364 . at this radius , the gauge invariant quantity @xmath365 that represents the energy density of the electromagnetic field reaches its maximum value @xmath366 , and other quantities of interest such as @xmath39 also take ( finite ) planck scale values . this means that the modified dynamics of our palatini @xmath11 model has the effect of setting upper bounds at the planck scale on the energy density of the matter fields involved , a property already observed in cosmological settings . we have found that @xmath295 is a singular null hypersurface if @xmath242 ( see ( [ eq : d1d2 ] ) for the definition of @xmath165 ) . when the mass - to - charge ratio @xmath165 is set to the particular value @xmath367 ^ 2/\sqrt{2\pi^3}\approx 0.572 $ ] , then this null surface becomes nonsingular and the geometry can be analytically extended by means of a _ bounce _ of the radial coordinate ( see fig . [ fig : zzstar ] ) . we pointed out that the area of the null hypersurface @xmath295 grows linearly with the number of charges , @xmath328 . as already mentioned , this behaviour suggests that each charge sourcing the electric field has associated an elementary quantum of area of magnitude @xmath330 . using this result , direct computation of the surface charge density , @xmath368 , gives @xmath369 , which up to a factor @xmath332 coincides with the planck surface charge density . for the nonsingular solution , the condition @xmath250 can be seen as indicating that the mass density of the core is @xmath370 , i.e. , it is of order the planck mass density . it must be noted that @xmath371 and @xmath340 are given in terms of the fundamental constants @xmath372 and the speed of light @xmath373 , and are independent of @xmath68 and @xmath164 ( in the singular cases , @xmath336 does depend on @xmath68 and @xmath164 ) . in our opinion , this is a clear manifestation of the quantum gravitational nature of the nonsingular solutions . rather than as a fine - tuning problem , this constraint on the core density should be seen as a quantization condition that selects a discrete set among all the classically allowed solutions . in a sense , this is analogous to bohr s atomic model , where the stability of hydrogen under electromagnetic emission of radiation was postulated assuming the existence of certain privileged orbits that had to satisfy specific quantization conditions . on the other hand , we have found the mass spectrum given in ( [ eq : spectrum ] ) , which is valid for all ( positive ) values of @xmath270 . this mass spectrum has important implications for the emission of hawking quanta . this is so because if physically allowed transitions occur between nonsingular configurations only , then the resulting spectrum must be discrete . our analysis also puts forward the existence of a new kind of nonsingular objects which are not hidden by an external horizon . these _ naked cores _ exist for values of the charge comprised within the interval @xmath374 . an important lesson that follows from our analysis is that the boundedness of the energy density does not necessarily imply that the spacetime is nonsingular . in order to find nonsingular solutions , it is necessary that charge and mass satisfy a particular relation . in this sense , it seems fair to say that the mass spectrum and sizes of the nonsingular objects described here need not be in correspondence with actual physical particles . from the discussion of section [ section : iv ] on the relation between @xmath375 and @xmath129 , it follows that including new scales in the problem ( such as the masses and different gauge charges of the particles making up the system ) should have an impact on the resulting value of @xmath324 , defined as the value of @xmath122 at which the energy density reaches its maximum . the analogous of the regularity condition @xmath218 could also set more complicated constraints between the total mass and total charges of the system , thus providing a richer structure to the set of nonsingular solutions . these aspects , together with the process of formation and the stability under perturbations of the nonsingular solutions ( with and without external horizon ) studied here is currently underway . the work of g. j. o. has been supported by the spanish grant fis2008 - 06078-c03 - 02 , fis2011 - 29813-c02 - 02 , the consolider program cpan ( csd2007 - 00042 ) , and the jae - doc program of the spanish research council ( csic ) . d. r. -g . thanks the hospitality of the theoretical physics group at valencia u. , where part of this work was carried out . we are very grateful to a. fabbri and j. navarro - salas for their useful comments and illuminating discussions . r. penrose , phys . * 14 * , 57 ( 1965 ) ; s. hawking and r. penrose , proc . london a * 314 * , 529 ( 1970 ) ; b. carter , phys . * 26 * , 331 ( 1971 ) ; s. w. hawking and g. f. r. ellis , _ the large scale structure of space and time _ ( cambridge university press , cambridge , 1973 ) . l. parker and d. j. toms , _ quantum field theory in curved spacetime : quantized fields and gravity _ ( cambridge university press , 2009 ) ; n. d. birrell and p. c. w. davies , _ quantum fields in curved space _ ( cambridge university press , 1982 ) . d. l. wiltshire , phys . d * 38 * , 2445 ( 1988 ) ; m. aiello , r. ferraro , and g. giribet , phys . d * 70 * , 104014 ( 2004 ) ; class . * 22 * , 2579 ( 2005 ) ; h. maeda , m. hassaine , and c. martinez , phys . d * 79 * , 044012 ( 2009 ) . s. capozziello and m. de laurentis , phys . rept . * 509 * , 167 ( 2011 ) ; a. de felice , and s. tsujikawa , living rev . * 13 * , 3 ( 2010 ) ; t. p. sotiriou and v. faraoni , rev . * 82 * , 451 ( 2010 ) ; s. capozziello and m. francaviglia , gen . grav . * 40 * , 357 ( 2008 ) ; s. nojiri and s. d. odintsov , int . j. geom . methods mod . * 4 * , 115 ( 2007 ) . m. bojowald , living rev . rel . * 8 * , 11 ( 2005 ) ; a. ashtekar , nuovo cim . * 122 b * , 135 ( 2007 ) ; a. ashtekar , t. pawlowski and p. singh , phys . * 96 * , 141301 ( 2006 ) ; phys . rev . d * 73 * , 124038 ( 2006 ) ; * 74 * , 084003 ( 2006 ) ; l. szulc , w. kaminski , and j. lewandowski , class . quant * 24 * , 2621 ( 2007 ) ; a. ashtekar , t. pawlowski , p. singh , and k. vandersloot , phys . rev . d * 75 * , 024035 ( 2007 ) ; k. vandersloot , phys . d * 75 * , 023523 ( 2007 ) ; e. bentivegna and t. pawlowski , phys . d * 77 * , 124025 ( 2008 ) . t. thiemann , _ modern canonical quantum general relativity _ , ( cambridge university press , 2007 ) ; c. rovelli , _ quantum gravity _ , ( cambridge university press , 2004 ) ; a. ashtekar and j. lewandowski , class . * 21 * , r53 ( 2004 ) . h. a. buchdahl , j. phys . a : math . gen . * 12 * , 1235 ( 1979 ) ; v. tapia and m. ujevic , class . 15 * , 3719 ( 1998 ) ; v. vitagliano , t. p. sotiriou , and s. liberati , phys . rev . d * 82 * , 084007 ( 2010 ) . h. p. de oliveira , class . grav . * 11 * , 1469 ( 1994 ) ; g. w. gibbons and d. a. rasheed , nucl . b * 454 * , 185 ( 1995 ) ; s. fernando , phys . d * 74 * , 104032 ( 2006 ) ; j. diaz - alonso and d. rubiera - garcia , phys . d * 81 * , 064021 ( 2010 ) . e. s. fradkin and a. a. tseytlin , phys . b * 163 * , 123 ( 1985 ) ; r. g. leigh , mod . a * 4 * , 2767 ( 1989 ) ; a. a. tseytlin , nucl . b * 501 * , 41 ( 1997 ) ; d. brecher , phys . b * 442 * , 117 ( 1998 ) .

for the quantum ising chain , the self - dual block renormalization procedure of fernandez - pacheco [ phys . rev . d 19 , 3173 ( 1979 ) ] is known to reproduce exactly the location of the zero - temperature critical point and the correlation length exponent . recently , miyazaki and nishimori [ phys . rev . e 87 , 032154 ( 2013 ) ] have proposed to study the disordered quantum ising model in dimensions by applying the fernandez - pacheco procedure successively in each direction . to avoid the inequivalence of directions of their approach , we propose here an alternative procedure where the directions are treated on the same footing . for the pure model , this leads to the correlation length exponents in ( to be compared with the 3d classical ising model exponent ) and ( to be compared with the 4d classical ising model mean - field exponent ) . for the disordered model in dimension , either ferromagnetic or spin - glass , the numerical application of the renormalization rules to samples of linear size yields that the transition is governed by an infinite disorder fixed point , with the activated exponent , the typical correlation exponent and the finite - size correlation exponent . we discuss the similarities and differences with the strong disorder renormalization results . # 1#2 # 1#2

1409.8459

the quantum ising model defined in terms of pauli matrices @xmath12 @xmath13 is the basic model to study quantum phase transitions at zero - temperature @xcite . on a hypercubic lattice in dimension @xmath2 , the pure model with the same transverse field @xmath14 on all sites , and the same ferromagnetic coupling @xmath15 between nearest - neighbors is well understood via the equivalence with the classical ising model in dimension @xmath16 , i.e. the time plays the role of an extra space - dimension @xcite , and the dynamical exponent is @xmath17 . in particular , the quantum model in @xmath18 corresponds to the exactly solved 2d classical ising model , the quantum model in @xmath4 corresponds to the 3d classical ising model , and the quantum model in dimension @xmath19 is characterized by the standard mean - field exponents . from the point of view of block - renormalization for quantum models , there exists a special self - dual procedure introduced by fernandez - pacheco @xcite , which is able to reproduce the exact critical point @xmath20 and the exact correlation length exponent @xmath21 . various generalizations of this procedure to higher dimensions @xmath1 have been studied @xcite , as well as extensions to other quantum models like the potts and the ashkin - teller models @xcite . in the disordered case , where the transverse fields @xmath22 and the couplings @xmath23 are random variables , many exact results have been obtained in @xmath18 by daniel fisher @xcite via the asymptotically exact strong disorder renormalization procedure ( for a review , see @xcite ) . in particular , the transition is governed by an infinite disorder fixed point and presents unconventional scaling laws with respect to the pure case . in dimension @xmath1 , the strong disorder renormalization procedure has been studied numerically with the conclusion that the transition is also governed by an infinite - disorder fixed point in dimensions @xmath24 @xcite . these numerical renormalization results are in agreement with the results of independent quantum monte - carlo in @xmath4 @xcite . the strong disorder renormalization is thus a very powerful method , but it leads to a complicated renormalized topology for the surviving clusters as soon as @xmath1 . in particular , a large number of very weak bonds are a priori generated during the rg , that will eventually not be important for the forthcoming rg steps . this is why recent numerical implementations of strong disorder rg rules are based on algorithms avoiding this proliferation of weak generated bonds @xcite . a natural question is whether infinite disorder fixed points can be also reproduced by more standard block - renormalization . recently , miyazaki and nishimori @xcite have proposed to generalize the block - renormalization of fernandez - pacheco @xcite concerning the pure model in @xmath18 , and their extension for the pure model in @xmath25 @xcite to the random case : their results are in agreement with the infinite disorder fixed point scalings . however in dimension @xmath1 , their procedure has the drawback that the various directions are treated inequivalently , so that they need to re - symmetrize afterward @xcite . in addition , in the random case , they have used some pool method that does not keep all the generated correlations between renormalized parameters , and the only critical exponent that they measure is the finite - size correlation length exponent @xmath26 , whereas it is interesting to measure also the activated exponent @xmath27 and the typical correlation length exponent @xmath28 in the present paper , we thus propose another generalization in dimensions @xmath4 and @xmath29 of the self - dual one - dimensional block - renormalization of fernandez - pacheco , that we study analytically in the pure case and numerically in the disordered case . the paper is organized as follows . in section [ sec_rgrule ] , we describe the elementary renormalization rule that will be applied throughout the paper . the application to the one - dimensional quantum ising model of eq . [ hqising ] , both pure and random , is recalled in section [ sec_d1 ] . in section [ sec_d2 ] , we derive the block renormalization rules for the two - dimensional case . the applications of these two - dimensional rules are discussed respectively in section [ sec_d2pure ] for the pure ferromagnetic model , in section [ sec_d2randomferro ] for the random ferromagnetic model , and in section [ sec_d2spinglass ] for the spin - glass model . our conclusions are summarized in section [ sec_conclusion ] . the extension to dimension @xmath29 is presented in appendix [ sec_d3 ] . in this section , we consider the following elementary hamiltonian involving @xmath30 spins @xmath31 with @xmath32 @xmath33 where @xmath34 plays the special role of the master , and the @xmath35 spins @xmath31 with @xmath36 are the slaves. for each eigenvalue @xmath38 of @xmath39 , one has to diagonalize independently the @xmath35 hamiltonians concerning the single quantum spin @xmath31 in the transverse field @xmath22 and in the effective magnetic field @xmath40 @xmath41 the two eigenvalues read @xmath42 with the following corresponding eigenvectors @xmath43 where @xmath44 in summary , the hamiltonian @xmath37 of eq . [ hintra ] has two degenerate ground states labeled by the two values @xmath38 . the ground - state energy is simply obtained by adding the @xmath35 contributions @xmath45 @xmath46 the two corresponding ground - states are obtained by the tensor products @xmath47 the projector onto the two ground - states reads @xmath48 it is thus convenient to define the renormalized spin @xmath49 from these two ground - states @xmath50 with the corresponding operators @xmath51 to evaluate @xmath53 , we have to consider the action of the operator @xmath54 on each ground state @xmath55 this yields the following trivial projection rule in terms of the renormalized operator @xmath56 of eq . [ sigmar ] @xmath57 to evaluate @xmath59 , we need to compute the action of the operator @xmath60 on each ground state @xmath61 using @xmath62 one obtains the following projection rule in terms of the renormalized operator @xmath56 of eq . [ sigmar ] @xmath63 to compute @xmath65 , we have to consider the action of the operator @xmath66 on each ground state @xmath67 using @xmath68 one obtains the following projection rule in terms of the renormalized operator @xmath69 of eq . [ sigmar ] @xmath70 \sigma_{r0}^x \label{projsigmax0}\end{aligned}\ ] ] the physical meaning of the procedure derived above is thus very simple : the renormalized spin @xmath71 represents the master spin @xmath72 dressed by its b slave spins @xmath73 . for each slave @xmath36 , we may consider the two limiting cases : \(a ) if @xmath74 , then the slave spin @xmath73 is ferromagnetically locked to its master @xmath72 ( eq . [ projsigmazi ] becomes @xmath75 to be compared with eq . [ projsigmaz0 ] ) , so that the flipping of the master spin is affected by the small ratio @xmath76 ( eq . [ projsigmax0 ] ) . \(b ) if @xmath77 , then the slave spin @xmath73 is mostly disordered and only weakly polarized by its master @xmath72 ( eq . [ projsigmazi ] becomes @xmath78 ) , so that the flipping of the master spin is unchanged ( eq . [ projsigmax0 ] ) . the projection rules above are thus compatible with the strong disorder rg rules in the two limits @xmath74 and @xmath77 , but can also apply to cases where @xmath79 . so they can be used to analyze both pure and random quantum ising models , as recalled in the following section for @xmath18 . the hamiltonian of the quantum ising chain reads @xmath80 in a block renormalization rule , one wishes to replace each block of two spins @xmath81 by a single renormalized spin @xmath82 . it is thus convenient to rewrite eq . [ h1d ] as @xmath83 \label{h1di}\end{aligned}\ ] ] the idea of fernandez - pacheco @xcite , written here for the random case @xcite , is the following choice of the intra - block hamiltonian @xmath84 since @xmath85 has the form the hamiltonian of eq . [ hintra ] analyzed in section [ sec_rgrule ] , the two spins @xmath86 can be renormalized via a single renormalized spin @xmath87 . the application of the projection rules of eqs [ projsigmaz0 ] , [ projsigmazi ] and [ projsigmax0 ] read for the present case @xmath88 as a consequence , the projection of the remaining part of the hamiltonian @xmath89 p_{intra}^{(1 ) } = \sum_{i } h_{i}^{r } \nonumber \\ h_{i}^{r } = & & - h^r ( 2i ) \sigma_{r(2i)}^x - j_{2 \vec x}^r ( 2i-2 ) \sigma^z_{r(2i-2 ) } \sigma^z_{r(2i ) } \label{h1drfin}\end{aligned}\ ] ] has the same form as the initial hamiltonian of eq . [ h1d ] , in terms of \(i ) the renormalized transverse fields on the remaining even sites @xmath90 \(ii ) the renormalized couplings between the remaining even sites @xmath91 if the initial parameters are @xmath92 on the whole chain , one obtains after one rg step the renormalized parameters ( eqs [ rgh1d ] and [ rgj1d2x ] ) @xmath93 so that the ratio @xmath94 evolves according to the simple rule @xmath95 the disordered attractive fixed point @xmath96 and the ferromagnetic attractive fixed point @xmath97 are separated by the unstable fixed point @xmath98 characterized by the correlation length exponent @xmath0 obtained by @xmath99 . the fact that both @xmath100 and @xmath26 are in agreement with the exact solution @xcite shows that the fernandez - pacheco choice @xcite of the intra - block hamiltonian of eq . [ h1intra1d ] is better than other choices @xcite . since the quantum ising chain represents the anisotropic limit of the two - dimensional classical ising model and is the the same universality class , it is interesting to compare with all the real - space renormalization procedures concerning classical spin models , from the early migdal - kadanoff schemes to the more recent tensor networks formulations ( see the recent review @xcite and references therein ) . for the two - dimensional ising model , whenever the two directions are treated on the same footing , the various real - space rg procedures that have been proposed are able to produce very good approximations of the exponent @xmath26 ( see for instance the table i of the review @xcite ) but never yield exactly @xmath0 , in contrast to the fernandez - pacheco quantum procedure described above . so it seems presently that the only way to obtain exactly @xmath0 for the 2d ising model is by defining a renormalization procedure _ for the transfer matrix _ @xcite in order to inherit the exactness of the exponent @xmath26 of the fernandez - pacheco quantum procedure . in terms of the ratios @xmath101 the rg rules of eq . [ rgh1d ] and [ rgj1d2x ] reads @xmath102 and thus corresponds to a simple addition in log - variables @xmath103 so after @xmath104 rg steps corresponding to a length @xmath105 , the renormalized ratio @xmath106 reads in terms of the initial variables @xmath107 \label{inffp}\end{aligned}\ ] ] the central limit theorem yields the asymptotic behavior @xmath108 + l^{1/2 } \sqrt { \left [ var[\ln j_{\vec x } ] + var[\ln h ] \right ] } u \label{inffpclt}\end{aligned}\ ] ] where @xmath109 is a gaussian random variable . the first term yields that the critical point corresponds to the condition @xmath110 and that the typical correlation length exponent is @xmath111 outside criticality , the competition between the first and the second term shows the finite - size correlation exponent is @xmath112 at criticality where the first term vanishes , the second random term of order @xmath113 corresponds to an infinite disorder fixed point of exponent @xmath114 all these conclusions of eqs [ criti1d ] , [ ntyp1d ] , [ nav1d ] , [ psi1d ] obtained via the application of the fernandez - pacheco renormalization to the random quantum ising chain @xcite , are in agreement with the fisher strong disorder renormalization exact results @xcite . it is thus interesting to look for an appropriate generalization of the fernandez - pacheco renormalization in higher dimensions , and first of all in dimension @xmath4 . the initial quantum ising hamiltonian defined on the square lattice of unit vectors @xmath115 reads @xmath116 \label{h2d}\end{aligned}\ ] ] various generalizations of the one - dimensional fernandez - pacheco renormalization procedure have been already proposed in dimension @xmath4 , both for the pure case @xcite and for the disordered case @xcite , with the drawbacks recalled in the introduction . in this section , we thus introduce another procedure where the two directions are considered on the same footing . we wish to define a block renormalization rule , where each block of four spins @xmath117 will be replaced by a single renormalized spin @xmath118 , after two elementary renormalization steps . it is convenient to start by rewriting eq . [ h2d ] as @xmath119 \nonumber \\ & & - \sigma_{(2i,2j-1)}^z [ j_{\vec x}(2i,2j-1 ) \sigma_{(2i+1,2j-1 ) } ^z + j_{\vec y}(2i,2j-1 ) \sigma_{2i,2j}^z \nonumber \\ & & \ \ \ \ \ \ \ \ \ \ \ \ \ \ + j_{\vec x}(2i-1,2j-1)\sigma_{(2i-1,2j-1)}^z + j_{\vec y}(2i,2j-2 ) \sigma_{(2i,2j-2)}^z ] \nonumber\end{aligned}\ ] ] for the first renormalization step , we choose the following intra - block hamiltonian @xmath120 since @xmath121 has the form the hamiltonian of eq . [ hintra ] analyzed in section [ sec_rgrule ] , the three spins @xmath122 can be renormalized via a single renormalized spin @xmath123 , whereas the spin @xmath124 that is not involved in @xmath121 remains unchanged . the application of the projection rules of eqs [ projsigmaz0 ] , [ projsigmazi ] and [ projsigmax0 ] read for the present case @xmath125 as a consequence , the projection of the remaining part of the hamiltonian reads @xmath126 p_{intra}^{(1 ) } = \sum_{(i , j ) } h_{i , j}^{r } \nonumber \\ h_{i , j}^{r } = & & - h ( 2i-1,2j-1 ) \sigma_{(2i-1,2j-1)}^x - h^r ( 2i,2j ) \sigma_{r(2i,2j)}^x \nonumber \\ & & - j^r_{2 \vec x } ( 2i-2,2j ) \sigma^z_{r(2i-2,2j ) } \sigma^z_{r(2i,2j ) } - j^r_{2\vec y}(2i,2j-2 ) \sigma^z_{r(2i,2j-2 ) } \sigma^z_{r(2i,2j ) } \nonumber \\ & & - j^r _ { \vec x- \vec y } ( 2i,2j ) \sigma^z_{r(2i,2j ) } \sigma_{(2i+1,2j-1 ) } ^z - j^r _ { -\vec x+ \vec y } ( 2i,2j ) \sigma^z_{r(2i,2j ) } \sigma_{2i-1,2j+1}^z \nonumber \\ & & - j^r _ { \vec x+ \vec y } ( 2i-1,2j-1 ) \sigma_{(2i-1,2j-1)}^z \sigma^z_{r(2i,2j ) } \label{h2drfin}\end{aligned}\ ] ] in terms of \(i ) the renormalized transverse fields of even - even sites @xmath127 \(ii ) the renormalized couplings along the horizontal directions at distance two @xmath128 \(iii ) the renormalized couplings along the vertical directions at distance two @xmath129 \(iv ) the renormalized couplings along the diagonal directions @xmath130 and @xmath131 @xmath132 \(v ) the renormalized couplings along the diagonal direction @xmath133 within each block @xmath134 for the hamiltonian @xmath135 of eq . [ h2drfin ] , we choose the following intra - block hamiltonian @xmath136 it has the form the hamiltonian of eq . [ hintra ] analyzed in section [ sec_rgrule ] , so that the two spins @xmath137 can be replaced by a single renormalized spin @xmath118 . the application of the projection rules of eqs [ projsigmaz0 ] , [ projsigmazi ] and [ projsigmax0 ] read for the present case @xmath138 ^ 2 } } \sigma^z_{rr(2i,2j ) } \nonumber \\ p_{intra}^{(2 ) } \sigma^x_{r(2i,2j ) } p_{intra}^{(2 ) } & & = \frac{h ( 2i-1,2j-1)}{\sqrt{h^2(2i-1,2j-1)+[j^r_{\vec x+\vec y } ( 2i-1,2j-1)]^2 } } \sigma_{rr(2i,2j)}^x \label{proj2dp2}\end{aligned}\ ] ] as a consequence , the projection of the remaining part of the hamiltonian reads @xmath139 p_{intra}^{(2 ) } = \sum_{(i , j ) } h_{i , j}^{rr } \nonumber \\ h_{i , j}^{rr } = & & - h^{rr } ( 2i,2j ) \sigma_{rr(2i,2j)}^x % \nonumber \\ & & - j^r_{2 \vec x } ( 2i-2,2j ) \sigma^z_{rr(2i-2,2j ) } \sigma^z_{rr(2i,2j ) } - j^r_{2\vec y}(2i,2j-2 ) \sigma^z_{rr(2i,2j-2 ) } \sigma^z_{rr(2i,2j ) } \label{h2drrfin}\end{aligned}\ ] ] i.e. it has the same form as the initial hamiltonian on the square lattice , in terms of \(i ) the renormalized transverse fields @xmath140 ^ 2+[j^r_{\vec x+\vec y}(2i-1,2j-1)]^2 } } \label{rghrr}\end{aligned}\ ] ] \(ii ) the renormalized couplings along the horizontal directions at distance two @xmath141 ^ 2 + [ j^r _ { \vec x+ \vec y}(2i-1,2j-1)]^2 } } \label{rgj2ddrrh}\end{aligned}\ ] ] \(ii ) the renormalized couplings along the vertical directions at distance two @xmath142 ^ 2 } } \label{rgj2ddrrv}\end{aligned}\ ] ] in the following sections , we discuss the application of these renormalization rules to the pure case , to the random ferromagnetic case , and to the spin - glass case . if we start from the pure model of parameters @xmath92 , the renormalization rules of eqs [ rghrr ] [ rgj2ddrrh ] [ rgj2ddrrv ] reduce to @xmath143 ^ 2 } } = \frac{h^4 } { \sqrt{h^2+j^2 } \sqrt{h^2(h^2+j^2)+4 j^4 } } \nonumber \\ j^{rr } & & = \frac{j^2}{\sqrt{h^2+j^2 } } \left [ 1 + \frac { 2 j^2 } { \sqrt{h^2(h^2+j^2)+ 4j^4 } } \right ] \label{purrg2d}\end{aligned}\ ] ] in terms of the ratio @xmath94 , the renormalization rule reads @xmath144 \equiv \phi(k ) \label{xevol2d}\end{aligned}\ ] ] we are now interested into the critical point satisfying the fixed point equation @xmath145 ( between the ferromagnetic fixed point @xmath146 and the disordered fixed point @xmath147 ) : eliminating the square - root yields the fourth degree equation @xmath148 the only positive root reads ( the other three roots can be disregarded since one is negative , and the two others are imaginary ) @xmath149 the correlation length exponent @xmath26 can be obtained from @xmath150 and the corresponding numerical value @xmath151 is very close to the numerical estimate @xmath5 for the 3d classical ising model . the two - dimensional procedure that we have proposed is thus much simpler that the miyazaki - nishimori - ortiz procedure that needs re - symmetrization between the two directions @xcite and yields a very good approximation for the correlation length exponent @xmath26 , better than other real - space renormalization procedures @xcite . the extension to the pure model in @xmath29 is described in the appendix a , and we now turn to the random models in dimension @xmath4 . in order to compare with the strong disorder renormalization numerical results @xcite , we have adopted the standard choice of a flat distribution of couplings between @xmath152 and @xmath153 @xmath154 and a flat distribution of transverse fields between @xmath155 and @xmath156 @xmath157 so that the control parameter of the zero - temperature transition is @xmath158 we have applied numerically the renormalization rules derived above to @xmath159 disordered two - dimensional samples of linear size @xmath160 ( containing @xmath161 spins ) with periodic boundary conditions . the renormalization procedure is stopped at the scale @xmath162 , where there remains only four sites and eight links in each sample . as a consequence for this largest length @xmath162 , the statistics is over @xmath163 random fields and over @xmath164 random couplings . at each renormalization step corresponding to the lengths @xmath165 with @xmath166 we have analyzed the statistical properties of the renormalized transverse fields and of the renormalized couplings . more precisely , we have measured the rg flows of the typical values defined by @xmath167 and of the widths of the probability distributions @xmath168 as a function of the length @xmath169 for 28 values of the control parameter @xmath170 of eq . [ control ] . the linear size @xmath171 and the statistics over @xmath159 are thus of the same order of those used in recent strong disorder renormalization studies in @xmath4 @xcite , but of course the implementation is much simpler here since the spatial structure remains a square lattice upon rg instead of an evolving non - trivial topology . another difference is that we analyze the statistics over samples at fixed control parameter @xmath170 and fixed size @xmath169 , whereas strong disorder renormalization studies of ref . @xcite are based on the determination of the pseudo - critical parameter for each sample . on fig . [ figflowh ] , we show in log - log scale the rg flows of the typical renormalized transverse field @xmath172 of eq . [ deftyp ] @xmath173 and of the width @xmath174 of eq . [ defwidth ] @xmath175 as a function of @xmath176 : + ( i ) ordered phase @xmath177 : the asymptotic slope is @xmath4 , as shown here for @xmath178 ( squares ) and @xmath179 ( diamond ) + ( ii ) disordered phase @xmath180 : the asymptotic slope is @xmath181 , as shown here for @xmath182 ( triangles up ) and @xmath183 ( triangles right ) + ( iii ) critical point @xmath184 ( circles ) : the asymptotic slope is @xmath185 . + ( b ) @xmath186 as a function of @xmath187 : + ( i ) ordered phase @xmath177 : the asymptotic slope is @xmath188 , as shown here for @xmath178 ( squares ) and @xmath179 ( diamond ) + ( ii ) disordered phase @xmath180 : the asymptotic slope is @xmath181 , as shown here for @xmath182 ( triangles up ) and @xmath183 ( triangles right ) + ( iii ) critical point @xmath184 ( circles ) : the asymptotic slope is @xmath185 . , title="fig:",height=226 ] as a function of @xmath176 : + ( i ) ordered phase @xmath177 : the asymptotic slope is @xmath4 , as shown here for @xmath178 ( squares ) and @xmath179 ( diamond ) + ( ii ) disordered phase @xmath180 : the asymptotic slope is @xmath181 , as shown here for @xmath182 ( triangles up ) and @xmath183 ( triangles right ) + ( iii ) critical point @xmath184 ( circles ) : the asymptotic slope is @xmath185 . + ( b ) @xmath186 as a function of @xmath187 : + ( i ) ordered phase @xmath177 : the asymptotic slope is @xmath188 , as shown here for @xmath178 ( squares ) and @xmath179 ( diamond ) + ( ii ) disordered phase @xmath180 : the asymptotic slope is @xmath181 , as shown here for @xmath182 ( triangles up ) and @xmath183 ( triangles right ) + ( iii ) critical point @xmath184 ( circles ) : the asymptotic slope is @xmath185 . , title="fig:",height=226 ] at the critical point @xmath189 , the rg flows of the typical value and of the width display the same activated scaling of an infinite disorder fixed point @xmath190 where @xmath191 is some @xmath192 random variable . the values obtained here for the location of the critical point @xmath193 and the activated exponent @xmath185 turn out to be different from the estimations @xmath194 and @xmath195 obtained via the strong disorder renormalization ( see @xcite and references therein ) : the origin of these differences is not clear to us . in the ordered phase , the logarithm of the typical renormalized transverse field grows extensively with respect to the volume @xmath196 ( with @xmath4 here ) @xmath197 where the length scale @xmath198 represents the characteristic size of finite disordered clusters within this ordered phase . from our numerical data concerning the ordered phase , the asymptotic behavior of eq . [ hlorder ] allows to measure @xmath198 and its divergence near criticality @xmath199 in the disordered phase , the asymptotic typical value @xmath200 diverges with an essential singularity as a function of the control parameter @xmath201 in the critical region , the finite - size scaling form is governed by some finite - size scaling correlation length exponent @xmath202 @xmath203 the matching with the behavior in the ordered phase ( eqs [ hlorder ] and [ nuh ] ) and in the disordered phase ( eq . [ defkappa ] ) yields the relations @xmath204 and @xmath205 the previous numerical measures of @xmath27 , @xmath206 and @xmath207 yield the estimate @xmath208 as shown on fig . [ figfss ] , this value of @xmath202 gives satisfactory finite - size scaling plots of the numerical data of fig . [ figflowh ] . corresponding to the sizes @xmath209 with the values @xmath184 for the critical point and @xmath210 for the finite - size correlation length exponent : + ( a ) @xmath211 as a function of @xmath212 . + ( b ) @xmath213 as a function of @xmath212 . , title="fig:",height=226 ] corresponding to the sizes @xmath209 with the values @xmath184 for the critical point and @xmath210 for the finite - size correlation length exponent : + ( a ) @xmath211 as a function of @xmath212 . + ( b ) @xmath213 as a function of @xmath212 . , title="fig:",height=226 ] this value for @xmath202 ( eq . [ nufss ] ) agrees with the estimations obtained via the strong disorder renormalization ( see @xcite and references therein ) , and with the asymmetric block renormalization of ref @xcite . as a function of @xmath176 : + ( i ) ordered phase @xmath177 : the flow is non - monotonic , as shown here for @xmath178 ( squares ) and @xmath179 ( diamond ) : the asymptotic behavior ( not visible on this plot appropriate to the critical region ) corresponds to the classical random ferromagnet growth @xmath214 + ( ii ) disordered phase @xmath180 : the asymptotic slope is @xmath188 , as shown here for @xmath182 ( triangles up ) and @xmath183 ( triangles right ) + ( iii ) critical point @xmath184 ( circles ) : the asymptotic slope is @xmath185 . + ( b ) @xmath215 as a function of @xmath187 : + ( i ) ordered phase @xmath177 : the asymptotic slope is @xmath216 , as shown here for @xmath178 ( squares ) + ( ii ) disordered phase @xmath180 : the asymptotic slope is @xmath217 , as shown here for @xmath182 ( triangles up ) + ( iii ) critical point @xmath184 ( circles ) : the asymptotic slope is @xmath185 . , title="fig:",height=226 ] as a function of @xmath176 : + ( i ) ordered phase @xmath177 : the flow is non - monotonic , as shown here for @xmath178 ( squares ) and @xmath179 ( diamond ) : the asymptotic behavior ( not visible on this plot appropriate to the critical region ) corresponds to the classical random ferromagnet growth @xmath214 + ( ii ) disordered phase @xmath180 : the asymptotic slope is @xmath188 , as shown here for @xmath182 ( triangles up ) and @xmath183 ( triangles right ) + ( iii ) critical point @xmath184 ( circles ) : the asymptotic slope is @xmath185 . + ( b ) @xmath215 as a function of @xmath187 : + ( i ) ordered phase @xmath177 : the asymptotic slope is @xmath216 , as shown here for @xmath178 ( squares ) + ( ii ) disordered phase @xmath180 : the asymptotic slope is @xmath217 , as shown here for @xmath182 ( triangles up ) + ( iii ) critical point @xmath184 ( circles ) : the asymptotic slope is @xmath185 . , title="fig:",height=226 ] on fig . [ figflowj ] , we show in log - log scale the rg flows of the typical renormalized coupling @xmath218 of eq . [ deftyp ] @xmath219 and of the width @xmath220 of eq . [ defwidth ] @xmath221 at the critical point @xmath193 , the rg flows of the typical value and of the width display the same activated scaling of infinite disorder fixed point as in eq . [ hlcriti ] @xmath222 where @xmath223 is some @xmath192 random variable . in the disordered phase , the typical renormalized coupling @xmath224 decays exponentially with the size @xmath169 @xmath225 where @xmath226 represents the typical correlation length . from our numerical data concerning the disordered phase , the asymptotic behavior of eq . [ jldisordertyp ] allows to measure @xmath227 and its divergence near criticality @xmath228 the compatibility with the finite - size scaling form analogous to eq . [ fsshtyp ] @xmath229 implies the relation @xmath230 it is satisfied by the previously quoted estimates of @xmath231 , @xmath232 and @xmath27 . to study the two - dimensional spin - glass quantum ising model , we have replaced the probability distribution of eq . [ pjflat ] concerning the random ferromagnetic case by the flat distribution of zero - mean @xmath233 we have then repeated exactly the same numerical analysis as in the previous section [ sec_d2randomferro ] : the critical point is now found at @xmath234 , and the critical exponents @xmath9 , @xmath235 , @xmath10 , @xmath11 are the same as for the random ferromagnet presented in the previous section , as expected for infinite disorder fixed points @xcite , and as found within the miyasaki - nishimori asymmetric scheme @xcite . in this paper , we have proposed a simple generalization in @xmath1 of the self - dual block renormalization procedure of fernandez - pacheco @xcite , that we have tested for pure and random quantum ising models , with the following conclusions . for the pure models , where the fernandez - pacheco procedure is known to reproduce the exact correlation length exponent @xmath236 @xcite , we have obtained @xmath237 ( to be compared with the 3d classical ising model exponent @xmath5 ) and @xmath238 ( to be compared with the 4d classical ising model mean - field exponent @xmath7 ) . for the random models , where the fernandez - pacheco procedure is known to reproduce exactly the location of the critical point and the critical exponents @xmath239 , @xmath240 and @xmath241 of the infinite disorder fixed point @xcite , we have applied numerically the renormalization rules to two - dimensional samples of linear size @xmath242 , with either random ferromagnetic disorder or spin - glass disorder , both types of disorder leading to the same infinite disorder fixed point : the finite - size correlation exponent @xmath11 coincides with strong disorder renormalization result ( see @xcite and references therein ) , and with the asymmetric block renormalization of ref @xcite , but the activated exponent @xmath9 turns out to be somewhat higher than strong disorder renormalization estimate @xmath243 ( see @xcite and references therein ) . the origin of this difference remains to be clarified . we have also analyzed the rg flows in the disordered and ordered phases , in order to extract the typical correlation length exponent @xmath10 and the analog @xmath235 , and tested the finite - size scaling . in summary , the generalization in @xmath1 of the self - dual block renormalization procedure of fernandez - pacheco @xcite is able to reproduce both the conventional scaling of pure critical points and the activated scaling of infinite disorder fixed points . it would be thus interesting to develop such methods in models governed by strong ( not infinite ) disorder fixed points like the quantum ising model with long - ranged interactions @xcite or the superfluid - insulator transition @xcite , as well as in models where the transition at weak disorder could be in another universality class @xcite . the initial quantum hamiltonian on the cubic lattice reads @xmath244 \label{h3d}\end{aligned}\ ] ] we wish to define a block renormalization rule , where each block of @xmath245 spins ( @xmath246 ; @xmath247 ; @xmath248 ; @xmath249 ; @xmath250 ; @xmath251 ; @xmath252 ; @xmath253 ) will be replaced by a single renormalized spin @xmath254 , after three elementary renormalization steps . it is thus convenient to start by rewriting eq . [ h3d ] as @xmath255 \nonumber \\ & & - \sigma_{(2i,2j-1,2k)}^z [ j_{\vec y}(2i,2j-1,2k ) \sigma_{(2i,2j,2k ) } ^z + j_{\vec x}(2i,2j-1,2k ) \sigma_{(2i+1,2j-1,2k)}^z + j_{\vec z}(2i,2j-1,2k ) \sigma_{(2i,2j-1,2k+1)}^z \nonumber \\ & & + j_{\vec y}(2i,2j-2,2k ) \sigma_{(2i,2j-2,2k)}^z + j_{\vec x}(2i-1,2j-1,2k ) \sigma_{(2i-1,2j-1,2k)}^z + j_{\vec z}(2i,2j-1,2k-1 ) \sigma_{(2i,2j-1,2k-1)}^z ] \nonumber \\ & & - \sigma_{(2i,2j,2k-1)}^z [ j_{\vec z}(2i,2j,2k-1 ) \sigma_{(2i,2j,2k ) } ^z + j_{\vec x}(2i,2j,2k-1 ) \sigma_{(2i+1,2j,2k-1)}^z + j_{\vec y}(2i,2j,2k-1 ) \sigma_{(2i,2j+1,2k-1)}^z \nonumber \\ & & + j_{\vec z}(2i,2j,2k-2 ) \sigma_{(2i,2j,2k-2)}^z + j_{\vec x}(2i-1,2j,2k-1 ) \sigma_{(2i-1,2j,2k-1)}^z + j_{\vec y}(2i,2j-1,2k-1 ) \sigma_{(2i,2j-1,2k-1)}^z ] \nonumber \\ & & - \sigma_{(2i-1,2j-1,2k-1)}^z [ j_{\vec x}(2i-1,2j-1,2k-1 ) \sigma_{(2i-1,2j-1,2k ) } ^z + j_{\vec y}(2i-1,2j-1,2k-1 ) \sigma_{(2i-1,2j,2k-1)}^z \nonumber \\ & & + j_{\vec z}(2i-1,2j-1,2k-1 ) \sigma_{(2i-1,2j-1,2k)}^z + j_{\vec x}(2i-2,2j-1,2k-1 ) \sigma_{(2i-2,2j-1,2k-1)}^z \nonumber \\ & & + j_{\vec y}(2i-1,2j-2,2k-1 ) \sigma_{(2i-1,2j-2,2k-1)}^z + j_{\vec z}(2i-1,2j-1,2k-2 ) \sigma_{(2i-1,2j-1,2k-2)}^z ] \label{h3dbox}\end{aligned}\ ] ] in the first renormalization step , we choose the following intra - block hamiltonian @xmath256 it has the form the hamiltonian of eq . [ hintra ] analyzed in section [ sec_rgrule ] , so that the four spins ( @xmath246 ; @xmath257 ; @xmath248 ; @xmath258 ) are replaced by a single renormalized spin @xmath259 , whereas the four other spins @xmath260 that are not involved in @xmath261 remain unchanged . as a consequence , the projection of the remaining part of the hamiltonian reads @xmath263 p_{intra}^{(1 ) } = \sum_{(i , j ) } h_{i , j , k}^{r } \nonumber \\ & & h_{i , j , k}^{r } = - h^r ( 2i,2j,2k ) \sigma_{r(2i,2j,2k)}^x - h ( 2i-1,2j-1,2k-1 ) \sigma_{(2i-1,2j-1,2k-1)}^x \nonumber \\ & & - h ( 2i-1,2j-1,2k ) \sigma_{(2i-1,2j-1,2k)}^x - h ( 2i,2j-1,2k-1 ) \sigma_{(2i,2j-1,2k-1)}^x - h ( 2i-1,2j,2k-1 ) \sigma_{(2i-1,2j,2k-1)}^x \nonumber \\ & & - \sigma^z_{r(2i,2j,2k ) } [ j_{2\vec x}^r(2i-2,2j,2k ) \sigma_{r(2i-2,2j,2k)}^z + j_{2 \vec y}^r(2i,2j-2,2k ) \sigma_{r(2i,2j-2,2k)}^z + j_{2\vec z}^r(2i,2j,2k-2 ) \sigma_{r(2i,2j,2k-2)}^z \nonumber \\ & & + j^r _ { \vec x- \vec y } ( 2i,2j,2k)\sigma_{2i+1,2j-1,2k}^z + j^r _ { -\vec x+ \vec y } ( 2i,2j,2k ) \sigma_{2i-1,2j+1,2k}^z \nonumber \\ & & + j^r _ { \vec x- \vec z } ( 2i,2j,2k ) \sigma_{2i+1,2j,2k-1}^z + j^r _ { -\vec x+ \vec z } ( 2i,2j,2k ) \sigma_{2i-1,2j,2k+1}^z \nonumber \\ & & + j^r _ { \vec y- \vec z } ( 2i,2j,2k)\sigma_{2i,2j+1,2k-1}^z + j^r _ { -\vec y+ \vec z } ( 2i,2j,2k)\sigma_{2i,2j-1,2k+1}^z \nonumber \\ & & + j^r _ { \vec x+ \vec y } ( 2i-1,2j-1,2k ) \sigma_{2i-1,2j-1,2k}^z + j^r _ { \vec x+ \vec z } ( 2i-1,2j,2k-1 ) \sigma_{2i-1,2j,2k-1}^z \nonumber \\ & & + j^r _ { \vec y+ \vec z } ( 2i,2j-1,2k-1 ) \sigma_{2i,2j-1,2k-1}^z ] \nonumber \\ & & - \sigma_{(2i-1,2j-1,2k-1)}^z [ j_{\vec x}(2i-1,2j-1,2k-1 ) \sigma_{(2i-1,2j-1,2k ) } ^z + j_{\vec y}(2i-1,2j-1,2k-1 ) \sigma_{2i-1,2j,2k-1}^z \nonumber \\ & & + j_{\vec z}(2i-1,2j-1,2k-1 ) \sigma_{2i-1,2j-1,2k}^z + j_{\vec x}(2i-2,2j-1,2k-1 ) \sigma_{(2i-2,2j-1,2k-1)}^z \nonumber \\ & & + j_{\vec y}(2i-1,2j-2,2k-1 ) \sigma_{(2i-1,2j-2,2k-1)}^z + j_{\vec z}(2i-1,2j-1,2k-2 ) \sigma_{(2i-1,2j-1,2k-2)}^z ] \label{h3drfin}\end{aligned}\ ] ] with for the second renormalization step , we choose the following intra - block hamiltonian @xmath274 it has the form the hamiltonian of eq . [ hintra ] analyzed in section [ sec_rgrule ] , so that the four spins ( @xmath275 ; @xmath276 ; @xmath277 ; @xmath278 ) are replaced by a single renormalized spin @xmath279 . the application of the projection rules of eqs [ projsigmaz0 ] , [ projsigmazi ] and [ projsigmax0 ] read for the present case @xmath280 ^ 2 } } \sigma^z_{rr(2i,2j,2k ) } \nonumber \\ & & p_{intra}^{(2 ) } \sigma^z_{2i-1,2j,2k-1 } p_{intra}^{(2 ) } = \frac { j^r_{\vec x+\vec z } ( 2i-1,2j,2k-1 ) } { \sqrt { h^2(2i-1,2j,2k-1)+[j^r_{\vec x+\vec z } ( 2i-1,2j,2k-1)]^2 } } \sigma^z_{rr(2i,2j,2k ) } \nonumber \\ & & p_{intra}^{(2 ) } \sigma^z_{2i,2j-1,2k-1 } p_{intra}^{(2 ) } = \frac { j^r_{\vec y+\vec z } ( 2i,2j-1,2k-1 ) } { \sqrt { h^2(2i,2j-1,2k-1)+[j^r_{\vec y+\vec z } ( 2i,2j-1,2k-1)]^2 } } \sigma^z_{rr(2i,2j,2k ) } \nonumber \\ & & p_{intra}^{(2 ) } \sigma^x_{r(2i,2j,2k ) } p_{intra}^{(2 ) } = \frac{h ( 2i-1,2j-1,2k)}{\sqrt{h^2(2i-1,2j-1,2k)+[j^r_{\vec x+\vec y } ( 2i-1,2j-1,2k)]^2 } } \nonumber \\ & & \frac{h ( 2i-1,2j,2k-1)}{\sqrt{h^2(2i-1,2j,2k-1)+[j^r_{\vec x+\vec z } ( 2i-1,2j,2k-1)]^2 } } \frac{h ( 2i,2j-1,2k-1)}{\sqrt{h^2(2i,2j-1,2k-1)+[j^r_{\vec y+\vec z } ( 2i,2j-1,2k-1)]^2 } } \sigma_{rr(2i,2j)}^x \nonumber\end{aligned}\ ] ] as a consequence , the projection of the remaining part of the hamiltonian reads @xmath281 p_{intra}^{(2 ) } = \sum_{(i , j ) } h_{i , j , k}^{rr } \nonumber \\ & & h_{i , j , k}^{rr } = - h^{rr } ( 2i,2j,2k ) \sigma_{rr(2i,2j,2k)}^x - h ( 2i-1,2j-1,2k-1 ) \sigma_{(2i-1,2j-1,2k-1)}^x \nonumber \\ & & - j_{2\vec x}^{rr}(2i-2,2j,2k ) \sigma_{rr(2i-2,2j,2k)}^z \sigma^z_{rr(2i,2j,2k ) } - j_{2 \vec y}^{rr}(2i,2j-2,2k ) \sigma_{rr(2i,2j-2,2k)}^z \sigma^z_{rr(2i,2j,2k ) } \nonumber \\ & & - j_{2\vec z}^{rr}(2i,2j,2k-2 ) \sigma_{rr(2i,2j,2k-2)}^z \sigma^z_{rr(2i,2j,2k ) } \nonumber \\ & & - j^{rr}_{\vec x-\vec y-\vec z } ( 2i,2j,2k ) \sigma^z_{rr(2i,2j,2k ) } \sigma_{(2i+1,2j-1,2k-1)}^z -j^{rr}_{-\vec x+\vec y-\vec z } ( 2i,2j,2k ) \sigma^z_{rr(2i,2j,2k ) } \sigma_{(2i-1,2j+1,2k-1)}^z \nonumber \\ & & - j^{rr}_{-\vec x-\vec y+\vec z } ( 2i,2j,2k ) \sigma^z_{rr(2i,2j,2k ) } \sigma_{(2i-1,2j-1,2k+1)}^z \nonumber \\ & & - j^{rr}_{\vec x+\vec y+\vec z } ( 2i-1,2j-1,2k-1 ) \sigma_{(2i-1,2j-1,2k-1)}^z \sigma^z_{rr(2i,2j,2k ) } \label{h3drrfin}\end{aligned}\ ] ] in terms of \(i ) the renormalized transverse fields @xmath282 ^ 2 } } \nonumber \\ & & \frac{h ( 2i-1,2j,2k-1)}{\sqrt{h^2(2i-1,2j,2k-1)+[j^r_{\vec x+\vec z}(2i-1,2j,2k-1)]^2 } } \frac{h ( 2i,2j-1,2k-1)}{\sqrt{h^2(2i,2j-1,2k-1)+[j^r_{\vec y+\vec z}(2i,2j-1,2k-1)]^2 } } \label{rghrr3d}\end{aligned}\ ] ] \(ii ) the renormalized couplings along the lattice directions at distance two @xmath283 ^ 2 + [ j^r _ { \vec x+ \vec y}(2i-1,2j-1,2k)]^2 } } \nonumber \\ & & + j^r _ { \vec x- \vec z } ( 2i-2,2j,2k ) \frac{j^r _ { \vec x+ \vec z } ( 2i-1,2j,2k-1)}{\sqrt{[h^r(2i-1,2j,2k-1)]^2 + [ j^r _ { \vec x+ \vec z}(2i-1,2j,2k-1)]^2 } } \nonumber \\ j^{rr}_{2 \vec y } ( 2i,2j-2,2k ) & & = j^r_{2 \vec y}(2i,2j-2,2k ) \nonumber \\ & & + j^r _ { -\vec x+ \vec y } ( 2i,2j-2,2k ) \frac{j^r _ { \vec x+ \vec y } ( 2i-1,2j-1,2k ) } { \sqrt{[h^r(2i-1,2j-1,2k)]^2+[j^r _ { \vec x+ \vec y}(2i-1,2j-1,2k)]^2 } } \nonumber \\ & & + j^r _ { \vec y- \vec z } ( 2i,2j-2,2k ) \frac{j^r _ { \vec y+ \vec z } ( 2i,2j-1,2k-1 ) } { \sqrt{[h^r(2i,2j-1,2k-1)]^2+[j^r _ { \vec y+ \vec z}(2i,2j-1,2k-1)]^2 } } \nonumber \\ j^{rr}_{2 \vec z } ( 2i,2j,2k-2 ) & & = j^r_{2 \vec z}(2i,2j,2k-2 ) \nonumber \\ & & + j^r _ { -\vec x+ \vec z } ( 2i,2j,2k-2 ) \frac{j^r _ { \vec x+ \vec z } ( 2i-1,2j,2k-1 ) } { \sqrt{[h^r(2i-1,2j,2k-1)]^2+[j^r _ { \vec x+ \vec z}(2i-1,2j,2k-1)]^2 } } \nonumber \\ & & + j^r _ { -\vec y+ \vec z } ( 2i,2j,2k-2 ) \frac{j^r _ { \vec y+ \vec z } ( 2i,2j-1,2k-1 ) } { \sqrt{[h^r(2i,2j-1,2k-1)]^2+[j^r _ { \vec y+ \vec z}(2i,2j-1,2k-1)]^2 } } \label{rgj3ddrrdiagneg}\end{aligned}\ ] ] \(iii ) the renormalized couplings along @xmath284 ; @xmath285 ; @xmath285 @xmath286 ^ 2 } } \nonumber \\ j^{rr}_{-\vec x+\vec y-\vec z } ( 2i,2j,2k ) & & = j_{\vec y}(2i-1,2j,2k-1 ) \frac { j^r_{\vec x+\vec z } ( 2i-1,2j,2k-1 ) } { \sqrt { h^2(2i-1,2j,2k-1)+[j^r_{\vec x+\vec z } ( 2i-1,2j,2k-1)]^2 } } \nonumber \\ j^{rr}_{-\vec x-\vec y+\vec z } ( 2i,2j,2k ) & & = j_{\vec z}(2i-1,2j-1,2k ) \frac { j^r_{\vec x+\vec y } ( 2i-1,2j-1,2k ) } { \sqrt { h^2(2i-1,2j-1,2k)+[j^r_{\vec x+\vec y } ( 2i-1,2j-1,2k)]^2 } } \label{proj3djrrdiagneg}\end{aligned}\ ] ] \(iv ) the renormalized couplings along @xmath287 @xmath288 ^ 2 } } \nonumber \\ & & + j_{\vec y}(2i-1,2j-1,2k-1 ) \frac { j^r_{\vec x+\vec z } ( 2i-1,2j,2k-1 ) } { \sqrt { h^2(2i-1,2j,2k-1)+[j^r_{\vec x+\vec z } ( 2i-1,2j,2k-1)]^2 } } \nonumber \\ & & + j_{\vec z}(2i-1,2j-1,2k-1)\frac { j^r_{\vec x+\vec y } ( 2i-1,2j-1,2k ) } { \sqrt { h^2(2i-1,2j-1,2k)+[j^r_{\vec x+\vec y } ( 2i-1,2j-1,2k)]^2 } } \label{proj3drrdiagpos}\end{aligned}\ ] ] for the third renormalization step , we choose the following intra - block hamiltonian @xmath289 it has the form the hamiltonian of eq . [ hintra ] analyzed in section [ sec_rgrule ] , so that the two spins @xmath290 can be replaced by a single renormalized spin @xmath291 . the application of the projection rules of eqs [ projsigmaz0 ] , [ projsigmazi ] and [ projsigmax0 ] read for the present case @xmath292 ^ 2 } } \sigma^z_{rrr(2i,2j,2k ) } \nonumber \\ p_{intra}^{(3 ) } \sigma^x_{rr(2i,2j,2k ) } p_{intra}^{(2 ) } & & = \frac { h ( 2i-1,2j-1,2k-1 ) } { \sqrt { h^2(2i-1,2j-1,2k-1)+[j^{rr}_{\vec x+\vec y+\vec z } ( 2i-1,2j-1,2k-1)]^2 } } \sigma_{rrr(2i,2j,2k)}^x \label{proj3dp3}\end{aligned}\ ] ] the projection of the remaining part of the hamiltonian reads @xmath293 p_{intra}^{(3 ) } = \sum_{(i , j , k ) } h_{i , j , k}^{rrr } \nonumber \\ h_{i , j , k}^{rrr } = & & - h^{rrr } ( 2i,2j,2k ) \sigma_{rrr(2i,2j,2k)}^x \nonumber \\ & & - j_{2\vec x}^{rrr}(2i-2,2j,2k ) \sigma_{rrr(2i-2,2j,2k)}^z \sigma^z_{rrr(2i,2j,2k ) } - j_{2 \vec y}^{rrr}(2i,2j-2,2k ) \sigma_{rrr(2i,2j-2,2k)}^z \sigma^z_{rrr(2i,2j,2k ) } \nonumber \\ & & - j_{2\vec z}^{rrr}(2i,2j,2k-2 ) \sigma_{rrr(2i,2j,2k-2)}^z \sigma^z_{rrr(2i,2j,2k ) } \label{h3drrrfin}\end{aligned}\ ] ] i.e. it has the same form as the initial hamiltonian , in terms of @xmath295 ^ 2 } } \nonumber \\ j^{rrr}_{2 \vec y } ( 2i,2j-2,2k ) & & = j^{rr}_{2 \vec y}(2i,2j-2,2k ) \nonumber \\ & & + j^{rr}_{-\vec x+\vec y-\vec z } ( 2i,2j-2,2k ) \frac { j^{rr}_{\vec x+\vec y+\vec z } ( 2i-1,2j-1,2k-1 ) } { \sqrt { h^2(2i-1,2j-1,2k-1)+[j^{rr}_{\vec x+\vec y+\vec z } ( 2i-1,2j-1,2k-1)]^2 } } \nonumber \\ j^{rrr}_{2 \vec z } ( 2i,2j,2k-2 ) & & = j^{rr}_{2 \vec z}(2i,2j,2k-2 ) \nonumber \\ & & + j^{rr}_{-\vec x-\vec y+\vec z } ( 2i,2j,2k-2 ) \frac { j^{rr}_{\vec x+\vec y+\vec z } ( 2i-1,2j-1,2k-1 ) } { \sqrt { h^2(2i-1,2j-1,2k-1)+[j^{rr}_{\vec x+\vec y+\vec z } ( 2i-1,2j-1,2k-1)]^2 } } \label{rgj3ddrrrv}\end{aligned}\ ] ] if we start from the pure model of parameters @xmath92 , the renormalization rules of eqs [ rghrrr ] and [ rgj3ddrrrv ] reads for the ratio @xmath94 to @xmath296 \equiv \phi(k ) \label{evolk3d}\end{aligned}\ ] ] the critical point satisfying @xmath145 is found to be @xmath297 the correlation length exponent @xmath26 given by @xmath298 @xmath299 is very close to the mean - field value @xmath300 of the 4d classical ising model . e. altman , y. kafri , a. polkovnikov and g. refael , phys . 100 , 170402 ( 2008 ) ; + e. altman , y. kafri , a. polkovnikov and g. refael , phys . rev . b 81 , 175528 ( 2010 ) ; + s. iyer , d. pekker and g. refael , phys . rev . b 85 , 094202 ; + g. refael and e. altman , comptes rendus physique 14 , 725 ( 2013 ) .

we present an analysis of over 2200 images taken on 14 nights at the mt . laguna 1 m telescope of the open cluster m67 . our observations overlap but extend beyond the field analyzed by gilliland et al . ( 1991 ) , and complement data recently published by van den berg et al . ( 2002 ) and stassun et al . ( 2002 ) . we show variability in the light curves of all 4 of the known w uma variables on timescales ranging from a day to decades ( for ah cnc ) . we have modeled the light curve of ah cnc , and the total eclipses allow us to determine and . the position of this system near the turnoff of m67 makes it useful for constraining the turnoff mass for the cluster . we have also detected two unusual features in the light curve of ah cnc that may be caused by prominences . we have also monitored cluster blue stragglers for variability , and we present evidence hinting at low level variations in the stragglers s752 , s968 , and s1263 , and we place limits on the variability of a number of other cluster blue stragglers . finally , we provide photometry of the sub - subgiant branch star s1063 showing variability on timescales similar to the orbital period , while the `` red straggler '' s1040 shows evidence of an unexplained drop in brightness at phases corresponding to the passage of the white dwarf in front of the giant .

astro-ph0301278

thanks to high stellar densities and small velocity dispersions , stellar clusters stand out as environments with high frequencies of strong gravitational interactions between stars . in the cores of the densest globular clusters , collisions of single stars may occur relatively frequently ( hills & day 1976 ) . however , even for less active environments like open clusters , strong interactions between binary stars probably play an important role in modifying the orbital parameters of the binaries and facilitating stellar collisions ( hurley et al . 2001 ; portegies zwart et al . current thinking identifies blue stragglers as one of the likely products of stellar interactions , so that an in - depth understanding of these stars could lead to a better understanding of the overall importance of environmental effects within stellar clusters ( see bailyn 1995 for a review ) . as a byproduct of our project to monitor the partially - eclipsing blue straggler s1082 ( designation from sanders 1977 ) , we made extensive observations of the fields around the core of the open cluster m67 . in order to get complete coverage of the light curve of s1082 ( period 1.0677978 d ) observations over several nights needed to be made . after it was discovered that the light curve was variable on timescales of a month or less , our observational dataset was expanded further . the observations of that system are presented and analyzed in a separate paper ( sandquist et al . however , in combination with the large archive of radial velocity measurements for m67 stars ( e.g. milone & latham 1994 ) , photometric studies of blue stragglers can provide us with important clues to their current states , and may also lead to an understanding of the evolutionary route they followed before becoming identifiable as blue stragglers . w uma variables are one class of binary star that can produce blue stragglers after angular momentum losses cause the two stars to coalesce . m67 is a somewhat difficult target for comprehensive variability studies because of its large angular size , but several groups have presented results on the cluster . for example , gilliland et al . ( 1991 ) combined relatively deep observations from several different observatories in a study of the core of the cluster . they serendipitously discovered two w uma variables ( s1036 and iii-79 ) , two blue stragglers with low amplitude @xmath3 scuti pulsations ( s1280 and s1284 ) , and evidence of longer period variations in other stars . their field was relatively small , however , so that it was nearly certain that other variables would be found . recently van den berg et al . ( 2002 , hereafter vsvm ) and stassun et al . ( 2002 , hereafter svmv ) presented photometric studies of variables in a larger field in m67 . their studies were initiated to look for variability among the x - ray sources within the cluster . several of the known x - ray sources were shown to be low - amplitude variables , and photometry was presented for an additional w uma variable ( s757 ) . the present study complements these photometric studies in several ways . in a number of instances we are able to present better delineated light curves than vsvm and svmv because we took shorter exposures and focused on observations in a single filter band . in addition , by combining our results with those of the other studies we are able to build a better picture of the variability of the light curves themselves on timescales ranging from days to decades for some of the better known variables . in 2 we briefly describe the observations , in 3 we discuss the reduction of the photometry , and in 4 we discuss the variables . all of the photometry for this study was taken at the 1 m telescope at the mt . laguna observatory using a @xmath4 ccd on 19 nights between december 2000 and march 2002 . the nights of observations are given in table [ obs ] . the photometry was primarily in @xmath0 band with typical exposure times of 20 s ( ranging between 15 and 60 s depending partly on atmospheric transparency ) to optimize the counts for the variable s1082 . exposures were usually separated by about 2.5 minutes due to a relatively long readout time for the ccd . guiding jitter and poor air flow at the site of the telescope typically restricts image quality to greater than 4 pixels ( 1.6 arcsec ) in the best conditions . observing conditions for the nights varied greatly , reaching fwhm of 13 pixels for some of the worst frames . the relative sparseness of the cluster worked in our favor though because decent photometry could still be done for many of the stars in the field . several hundred columns of the chip developed charge transfer problems during the december 2000 run . because of this , measurements of stars that fell near these columns were eliminated . the remainder of the chip was not noticeably affected by this problem . in the january and february 2001 runs this problem was almost entirely corrected . the ccd was replaced for the march 2001 run and a two - amplifier readout was employed , which doubled the duty cycle for our observations . for the remainder of the observations the replacement ccd was used with a one - amplifier readout . the object frames were reduced in usual fashion , using overscan subtraction , bias frames , and flat fields ( usually twilight flats , with the exception of dec . 11/12 , 2000 , feb . 17/18 and 18/19 , 2001 , for which dome flats had to be used ) . we chose to rely on aperture photometry for this study . in the analysis we used the iraf tasks daofind and phot from the apphot package . curve - of - growth analysis was conducted using the iraf tool digiphot.photcal.mkapfile in order to bring all photometric measurements to the same total aperture size . the general procedure is discussed in stetson ( 1990 ) . individual nights were run separately through the curve - of - growth analysis . in order to improve the accuracy of the relative photometry for the light curves , we used an ensemble photometry method similar to that described by honeycutt ( 1992 ) in order to get a simultaneous solution for median magnitudes of all stars and relative zero - points for all image frames . our implementation is described in more detail in sandquist et al . the solution was improved iteratively until none of the frame zero - points or median star magnitudes changed by more than 0.0005 mag between iteration steps . we allowed for the possibility that magnitude residuals could be a function of star position on the ccd for each frame by fitting second - order polynomials to the residuals in the @xmath5 and @xmath6 directions , and subtracting these fits during the solution iteration . occasionally these corrections amounted to a few centimag . in reducing the data , we paid close attention to possible correlations of residuals with external variables like seeing , airmass , sky intensity , exposure time , and the photometric zeropoint of the frame ( see gilliland et al . 1991 for a detailed discussion of the reasons ) . during the photometric reductions , we discovered that the photometry of a handful of the bluest stars in the sample showed significant variations of approximately 0.02 mag from night to night . on further investigation , we found that for several of these stars the photometric variations around the median values were definitely correlated . although we have been unable to determine the exact cause of this problem , we were able to verify that the correlations in the variation were only detectable in stars with @xmath7 . in our sample this affected the blue stragglers s752 , 968 , 977 , 1066 , 1263 , and 1267 . none of these stars had strong variations , but in order to try to look for low amplitude changes we computed a running weighted average of the residuals derived from all of the stars observed on a frame and on up to 6 other frames closest in time . the correction was subtracted from the brightness measurement for each star . the results of this analysis will be discussed in more detail in [ otherv ] . to judge the significance of observed photometric variations , we used two different measures : the rms variations about the median magnitude @xmath8 and the welch - stetson index @xmath9 ( welch & stetson 1993 ) . @xmath9 measures the degree of correlation of pairs of brightness measurements , and is generally the more sensitive method of the two for detecting variability . this index gives high scores for previously known variables , with the @xmath3 scuti star s1280 ( a variable of low - amplitude ) having the lowest score ( @xmath10 ) . in fig . [ var ] , we plot both measures as a function of magnitude for the stars in our samples to provide the reader with a means of judging the significance of the variations discussed below . this is complicated by the fact that seeing variations could result in correlated residuals for a star if a nearby star of comparable brightness contributed light to the aperture used to photometer the star . the variability indices plotted in fig . [ var ] were calculated for measurements with the best seeing ( fwhm @xmath11 ) in order to reduce this confusion , although some of the stars with high scores are still affected by this effect . however , for the objects discussed in this paper , we verified that there is no correlation between the photometric residuals ( defined as observed minus median magnitude ) and seeing . we used two different techniques to determine periods for the variables we observed : the lafler - kinman statistic ( * ? ? ? * hereafter l - k ) , and the lomb - scargle ( * ? ? ? * hereafter l - s ) periodogram . these two statistical methods measure slightly different characteristics of trial light curves . the l - k statistic measures the quality of a light curve for a given trial period using the sum of the differences in magnitude between observations made at adjacent phases . variations in the overall brightness of the system on different orbits can cause problems with this test , however . the l - s periodogram is basically a harmonic decomposition of the observations . for the systems with the most stable light curves , we modeled the light curve using the program nightfallrwichman / nightfall.html for the program and a user manual ( wichmann 1998 ) ] , which includes model atmospheres ( hauschildt et al . 1999 ) and physical effects such as detailed reflection ( which is important for close binaries ) . detailed fitting of the eclipsing binaries requires some care in choosing the prescription for limb - darkening . significant systematic differences in parameters such as mass ratio and inclination can result from this choice . we have chosen to use a two - parameter square - root law primarily because the use of linear or quadratic limb darkening laws resulted in significantly higher @xmath12 values due to poor fits to the eclipses for our variables . our choice is supported by comparisons between predictions from model atmospheres and best - fit limb - darkening laws @xcite that indicate that square - root or logarithmic laws are to be preferred for stars with surface temperatures near those of our w uma variables . in this section , we discuss the important features of the light curves of the variable stars we detected . unless otherwise listed , the i d numbers are from sanders ( 1977 ) . a color - magnitude diagram for the systems discussed below is shown in fig . [ fig1 ] . four w uma binaries are currently known to be members of m67 . we present observations of all of these systems below , and we list their properties in table [ wuma ] . * s757 : * this star was slightly outside the field analyzed by gilliland et al . ( 1991 ) , but was identified as a probable variable star by both nissen et al . ( 1986 ) and rajamohan et al . svms discuss their identification of this system as a w uma variable based on a likely period of 0.3600 d and color @xmath13 . they presented a partial light curve generated from two runs in 1998 and 2000 that indicate maxima and minima of roughly equal brightness . we independently identified the system , and are able to refine the period to @xmath14 d from two seasons of data . in this case we found that the best period from the l - k method was in a smaller peak in the l - s periodogram . the amplitude of the light curve in @xmath0 is approximately 0.09 mag , although this does vary from month to month . regardless , the small amplitude indicates that the system has a fairly low inclination ( @xmath15 ) . our data phased to this best period are shown in fig . [ s757 ] . figure [ s757.2 ] shows our most complete single - day light curves . the observations from january 2001 in particular show significant changes in the light curve from night to night . the relative heights of the maxima clearly changed between dec . 7 , 2000 and jan . 30 , 2001 with phase 0.25 slightly brighter than phase 0.75 in december , but about 0.05 mag fainter in january . the magnitude level of the secondary eclipse is the portion of the light curve that seems to remain the most constant . the month timescale of these variations leads us to believe that the cause is starspots . spots have been hypothesized on many of w uma systems in connection with unequal brightness of the two maxima ( generally called the oconnell effect ; oconnell 1951 ) . the case of s757 adds an additional wrinkle because the maximum following the primary eclipse did change from being brighter than the other maximum to being fainter ( generally called a variable oconnell effect ) . this could indicate that either there are significant spots present on _ both _ components or that hot _ and _ cool spots appeared on one component of the binary . we tend to believe the former explanation because the levels of the two maxima do not appear to remain at constant brightness . because we focused on @xmath0-band observations , we do not have enough information to conduct a more detailed analysis of the temperatures of hypothesized spots . in any case , the fact that we are unable to establish the true brightness of the maximum of the light curve makes a derivation of reliable system parameters nearly impossible . in spite of the apparent spot activity , s757 was not detected as an x - ray source in the cluster by belloni , verbunt , & mathieu ( 1998 ) . both of the other w uma systems of comparable brightness ( s1036 and s1282 ) were detected . @xmath16 increases with increasing color for field stars , which implies that s 757 should be brighter than s1282 in x - rays . the periods of the two systems are very nearly the same , which means that differences in the rate of rotation are not likely to account for differences in x - ray activity . in addition , s757 falls in the region where the two x - ray fields of belloni et al . overlap , meaning that there was a longer total integration on s757 . there does not appear to be any unidentified x - ray sources in the vicinity of s757 , so we are unable to explain the non - detection . * s1036 ( ev cnc ) : * s1036 is classified as a blue straggler because accurate photometry shows it to be about 0.1 mag ( in @xmath17 ) to the blue of the cluster turnoff . gilliland et al . ( 1991 ) determined the period to be 10.59 h. vsvm redetermined the period to be @xmath18 d , and they note that this does not appear to be consistent with the gilliland et al . value , although they could not judge how significant this was because gilliland et al . did not quote an error . using the vsvm data in combination with our own , we are able to improve the determination of the period very slightly . both the l - k and l - s methods agree on slightly lower but different periods . the l - s result appears to phase the data slightly better , and the measured light curve parameters are presented in table [ wuma ] . the asymmetry in the light curve of s1036 makes it somewhat unusual : for a contact binary with uniform surface temperature , the light - curve maxima should have equal brightness . for s1036 , the two maxima differ in brightness by approximately 0.02 mag . the two light - curve minima also differ in brightness by approximately 0.09 mag , as can seen in fig . [ s1036 ] . for eclipsing binaries a difference in brightness of the light curve minima normally implies a difference in temperature for the two stars , something which is hard to understand if the stars are in contact . in table [ wuma ] , we include measurements of eclipse depths @xmath19 and @xmath20 , where @xmath21 and @xmath22 refer to primary and secondary minima , and @xmath23 refers to the global maximum of the light curve . we conducted limited modeling of s1036 using the program nightfall ( see [ lcanal ] for details of the code ) . two factors conspire to prevent definitive determination of system parameters from the light curve : low inclination and the possibility of spots . for warnings about deriving system parameters from low - inclination systems , see rucinski ( 2001 ) and the discussion related to the system s1282 below . the asymmetries in the light curves of w uma variables have led many researchers to model these systems with spots , usually with no more than two spots . however , there are potentially many combinations of spot parameters ( spot latitude , longitude , size , and dimming factor ) that can model a single light curve . this is true of s1036 as well . however , to check on the possibility that the determination of a subset of the system parameters might be robustly determined from these models , we first determined a set of system and spot parameters that fits the system well ( @xmath24 ) , and then proceeded to compute grids of models varying two parameters at a time . this baseline solution had @xmath25 , @xmath26 , @xmath27 , and two spots ( spot 1 : facing us at phase 0.52 , radius @xmath28 , dimming factor 0.72 ; spot 2 : facing us at phase 0.23 , radius @xmath29 , dimming factor 0.80 ) . the filling factor @xmath30 , where c is the jacobi potential of the common envelope , and @xmath31 and @xmath32 are the corresponding potentials of the @xmath33 and @xmath34 points we provide @xmath35 contour maps for two of these experiments in fig . [ s1036chi ] . as can be seen in the figure , there is very little correlation between @xmath36 and mass ratio @xmath37 ( which is very poorly constrained ) . we checked for correlations between inclination and spot parameters , and found that they were minor . the primary degeneracy is between @xmath36 and the filling factor @xmath38 . the light curve clearly allows for solutions covering the range of filling factors observed for most contact binaries ( @xmath39 ; rucinski 1997 ) . the reader should keep in mind that the right panel of the figure was computed for a constant set of spot parameters , and some adjustment of these could reduce the @xmath35 value for a given combination of @xmath36 and @xmath38 . thus , solutions with the component stars detached should very much be considered to be viable . if the system is in contact , however , the inclination can be constrained to be @xmath40 . at least one spot appears to be necessary in models of the light curve because the shallower minimum is displaced away from the expected position at phase 0.5 and the two maxima still have different brightness . vsvm saw no evidence for variation in color during the orbit , although their quoted upper limit on the color variation was 0.05 in @xmath17 . the limits on color variations placed by the vsvm data do not distinguish between the possibilities that i ) the surface temperatures of the stars are equal and the differences in maximum and minimum brightness are entirely due to cold spots , or ii ) there is one strong cold spot that explains the differences in maximum brightness and a large temperature difference between the two stars that explains the differences in minimum brightness . however , neither hypothesis predicts color variations of more than about 0.01 mag , so much more accurate color information is needed to constrain the models strongly . as a result , we did not attempt to do a more systematic study of the binary parameters using our photometry . during our observations , we did detect slight variations in the shape of the light curve from month to month , although they are not as noticeable as those seen for s 757 . given that the timescales for variations in the light curves of the other w uma variables discussed in this paper are in the range of months or years , studies of the light curve of s1036 over a longer time baseline could help determine whether starspots are the primary factor in determining the light curve shape . * s1282 ( ah cnc ) : * with data from the studies of whelan et al . ( 1979 ) and gilliland et al . ( 1991 ) , there are currently good datasets covering the entire light curve of this w uma system for three epochs separated by over a decade . gilliland et al . noted that both minima in the light curve appeared to have changed from curved and continuously varying ( possibly indicating partial eclipses ) in the whelan et al . dataset to flatter ( indicating total eclipses ) in their dataset . our light curves closely resemble those of gilliland et al . : one flat - bottom eclipse covering approximately 0.1 in orbital phase , and a slightly deeper and more curved primary eclipse . figure [ s1282 ] shows our observations along with those of vsvm and @xcite . we have phased our observations using the ephemeris of kurochkin ( 1979 ) , which includes a second - order time term , although we have added a phase shift of 0.5 to bring the deeper minimum to phase 0 . the deeper minimum in the whelan et al . data is _ not _ the deeper minimum in later data according to this ephemeris . as can be seen in the observations of whelan et al . , the minima were at approximately the same brightness level in 1973 the kurochkin ephemeris does a good job of phasing the observations of vsvm with ours , clearly illustrating the necessity of using the second - order term . we had to include an additional phase shift of 0.04 to our data and those of vsvm in order to bring the primary minimum to phase 0 , which indicates that the ephemeris needs to be revised . [ we should note that the zero - point of the vsvm data appears to be different from that of our data . shifting their data according to the difference between our median @xmath0 magnitude ( 13.44 ) and their average @xmath0 magnitude ( 13.54 ; svmv ) we find that the light curves differ by 0.05 mag . the difference is partly due to the difference between the median and average for this light curve . for figure [ s1282 ] , we simply added a magnitude correction to their data to make comparison of the light curves easier . ] based on the appearance of the eclipses in our data , the system is clearly an a - type w uma binary ( the deeper eclipse being a transit of the larger star by the smaller star , as seen from the curvature of the light curve near phase 0 ) rather than w - type as identified by whelan et al . given the position of s1282 near the turnoff of m67 in the color - magnitude diagram , it might have been predicted that it should be an a - type system since they are generally believed to have evolved components ( mochnacki 1981 ) . the relationship between type and evolutionary state may be misleading , however , since s1282 appears to have effectively changed type in less than a decade ( between the observations of whelan et al . and those of gilliland et al . ) . one should also look at the data for the systems s757 ( discussed above ) and iii-79 ( discussed below ) , which show that one eclipse minimum can go from being the fainter of the two to the brighter in a matter of months or years . because this is much too short to correspond to any reasonable evolutionary timescale ( nuclear , thermal , or even dynamical ) for the stars themselves , we believe that this shows that the evolutionary status of a w uma variable can not be determined without an extended study of the light curve and its potential for variation . for s1282 , we do have additional information that is not always available for a w uma system : the binary is found at the turnoff of the cluster , and so the primary star is likely to have depleted much of its core hydrogen supply . because s1282 was close to our primary target ( s1082 ) in the cluster , we have photometry from most nights of observations , and on some nights we were able to cover almost an entire orbital period . previous modeling based on the whelan et al . data ( maceroni et al 1984 ) indicated that the system significantly overfills the roche lobes of the two stars . however , the presence and duration of the total eclipse invalidates the low inclination ( @xmath41 ) that maceroni et al . derived . given that the whelan et al . observations of light curve minima occurred over a relatively short period of 2.5 months , it is possible that there was a short - term change in the light curve shape . we have observed shorter timescale changes to the light curve shape ( for example , the deeper primary eclipse observed on january 25 , 2001 ) , and the variability seen by vsvm in their secondary eclipse observations during two runs seems to support this . however , the amplitude of the light curve from whelan et al . seems to be somewhat smaller than the amplitude in our data , which would tend to rule out cool spots as the cause . while whelan et al . did present radial velocity measurements in order to measure the spectroscopic mass ratio , they were not able to constrain the velocity semi - amplitude of the secondary star precisely . the large differences between their light curve and later ones might also indicate that their radial velocities may have been systematically affected by spots ( to give just one possibility ) . as a result , we will not take their spectroscopic mass ratio to be a primary constraint on light curve models . the radial velocity measurements were good enough to show , however , that the less massive star is totally eclipsed as it passes behind the more massive star , as would be expected if the more massive star is also larger in size . in most cases , w uma light curves by themselves provide very little information about the system parameters , and attempts to derive parameters using @xmath35 fits are extremely dangerous ( see rucinski 2001 for a discussion ) . however , when a contact system shows total eclipses , the duration of totality strongly constrains the combination of mass ratio and inclination , and is insensitive to the degree of contact ( mochnacki & doughty 1972 ) . because s1282 clearly has total eclipses , a fairly stable light curve during the period of our observations , and has not been modelled previously , light - curve models can provide valuable constraints . we modeled the light curve of s 1282 using nightfall ( see [ lcanal ] for details ) . we focus on data from december 2000 and january 2001 because nearly the entire light curve was observed in relatively short periods of time . we decided not to attempt to model data taken in 2002 because the maximum following primary eclipse was found to be significantly fainter ( @xmath42 mag ) than the maximum following secondary eclipse , probably indicating that spots were a significant influence . the large width of both eclipses requires an inclination near @xmath43 and a small mass ratio . our best - fit values for the december 2000 and january 2001 data separately are presented in table [ tfits ] . in order to estimate the possible errors , we ran grids of models varying @xmath37 and @xmath36 to determine changes in @xmath35 . the resulting contour maps are shown in fig . the contours correspond to levels 1.0 and 4.0 above the minimum @xmath35 fit for that dataset , which should roughly delineate the 1 and @xmath44 confidence regions @xmath37 and @xmath36 taken individiually . as seen in the contour maps , there is little correlation between @xmath37 and @xmath36 . the best fits contours for the two datasets agree well , although there is a small shift in @xmath37 between the results for the two datasets . using the information from the @xmath45 maps and the systematic shifts between the best fit models for the two datasets , we quote best fit values of @xmath1 and @xmath2 . the variations in the best - fit parameters due to month - to - month changes in the light curve appear to be comparable in importance to the random errors in the fitted value of @xmath37 . in fig . [ fits ] , we show a selection of theoretical models against the observational data . the best fit model is almost lost among the observational points , with the poorest fit near @xmath46 . the upper panel of the figure shows two models with the best - fit inclination but with mass ratios @xmath47 away from the best model . the mass ratio is primarily constrained by the eclipse depths . the lower panel shows light curves with the best - fit mass ratio , but with different inclinations . the @xmath48 curve is almost indistinguishable from the best fit model , but the @xmath49 curve has eclipses of noticeably shorter duration . although the derived mass ratio for the system falls near the low end of the distribution for w uma systems , w uma systems tend to be found preferentially at low @xmath37 values ( rucinski 2001 ) , so this fact is not unusual . the filling factor for the system is constrained by the shape of the light curve before and after primary eclipse , and although it is correlated with the inclination ( in the sense that lower inclinations require larger filling factors ) , there is negligible degeneracy between mass ratio and filling factor . for both of our best - fit models we find it necessary to use a filling factor ( @xmath50 ) that is probably larger than the majority of observed systems ( rucinski 1997 ) . a lower limit on the filling factor is @xmath51 based on @xmath45 values measured for models with @xmath48 . the results of the light curve analysis contradict the earlier spectroscopic analysis of @xcite . although their derived radial velocities were very uncertain ( @xmath52 @xmath53 km s@xmath54 , 138 @xmath53 km s@xmath54@xmath55 240 @xmath53 km s@xmath54 ) , the implied mass ratios are inconsistent with our photometric value . in addition , if it is assumed that the maximum @xmath56 value is approximately correct ( giving a spectroscopic @xmath37 value closest to our photometric value ) , the derived total system mass is substantially less that that predicted for a turnoff star in m67 from stellar evolution models ( @xmath57 ) in spite of the system s position near the turnoff in the color - magnitude diagram . we are forced to conclude that the radial velocity data for this system are not currently of high enough quality to derive trustworthy masses . the importance of a good radial velocity curve for this system should be emphasized , since it would provide a valuable constraint on the cluster turnoff mass , and would thereby help more accurately age - date m67 . our extensive observations allowed us to discover some unusual transient features in the light curves ( fig . [ weird ] ) . on january 23/24 and 25/26 , 2001 , we observed short ( @xmath58 min ) brightness increases ( @xmath59 mag for both , compared to eclipse depths of 0.33 and 0.39 mag ) . unusual features of these two brightenings were : they were observed at almost identical phase positions shortly before maximum , the following maximum did not show the feature , and on the two days the brightness increase was seen on _ different _ maxima . in addition , the primary minimum preceding the brightness increase on jan . 25/26 was significantly deeper ( by @xmath60 mag ) than any other minimum observed . these features disappeared within two days . these transient features bear some resemblance to variations seen in the light curve of the nearby w uma variable 44i boo ( e.g. duerbeck 1978 ) , which appears to have active periods of a few years duration interspersed with quieter periods with undisturbed light curves . a similar cycle ( probably related to the magnetic field ) might help to explain the difference between the light curves of @xcite and other observers . the timing of the features shortly before maximum seems to argue against stellar flares . the short duration of the features argues against hot spots , which should persist for longer times . one hypothesis is based on magnetically - confined gas ( prominences ) on one of the stars on the portion of the surface facing away from the other star . if the gas has sufficient optical depth , it could potentially increase the effective surface area of the binary when a large slab is aligned edge - on , but could cause either modest or negligible dimming of the system when viewed face - on through the thin dimension of the slab . the increased depth of the first half of the primary eclipse on january 25 indicates that the feature was probably on the side of the secondary star facing almost directly away from the primary . because our light curve was taken entirely in @xmath0 , we have no direct information on temperature . ( we wish to thank m. blake for the suggestion that initiated this line of thought . ) * iii-79 ( et cnc ) : * this system was discovered by gilliland et al . ( 1991 ) at @xmath61 from the cluster center , and was also observed by svmv . this binary is considerably fainter than the other three contact systems known . this system was in the field of our march 2001 observations , with the best data coming from night 13 . we averaged observations to increase the s / n ratio in the data . our data are presented in fig . [ etcnc ] along with those of svmv from january 1998 , phased to the same linear ephemeris using a period ( @xmath62 d ) that is consistent with the value quoted by svmv to within the errors . our measured median magnitude ( @xmath63 ) agrees very well with the average magnitude of svmv ( 15.89 ) , so no shifting in magnitude was done . although individual data points have considerably larger errors than those for the other w uma systems discussed , a comparison of the light curves indicates a noticeable change in the light curve shape . while the brighter of the two light curve maxima follows the brighter minimum in the svmv and gilliland et al datasets , the brighter maximum appears to follow the fainter minimum in our data . although we do not have data completely covering the minimum plotted at phase @xmath64 , several points with good errors indicate that it has dimmed noticeably . independent of this , the @xmath64 minimum does not reach the same depth as the @xmath65 minimum in the light curves of gilliland et al . and so the indication is that this system shows fairly extreme variations in the shape of the light curve . without more information about the nature of this variation , it would be reckless to try to model this system . during the course of the observations we observed most of the cluster s blue stragglers during the majority of at least one night in order to look for variability on various timescales . the results of this search are given in table [ straggler ] . our list of stragglers comes primarily from the list of ahumada & lapasset ( 1995 ) , although stars have been excluded if the accurate photometry of @xcite indicated that the star was close to the cluster turnoff . stars s2223 and s2226 were added to the list based on the photometry of @xcite and proper motion membership @xcite . we have monitored a number of stragglers for the first time in this study , but we do not find any of these to be noticeably variable . we provide the values for our variability indicators @xmath8 and @xmath9 ( discussed in [ vari ] ) in columns 4 and 5 respectively . typically nonvariable stars in the magnitude range of the stragglers had @xmath66 and @xmath67 . we discuss most of the variability candidates in [ otherv ] below . we note that both indicators can give erroneously high scores if observations were taken during poor seeing conditions and a contaminating star was present nearby , as was the case for the straggler s975 . we have provided information on the nights the stars were observed in order to lay the groundwork for constraining variations on longer timescales . we do , however , rule out the possibility that any of the other blue stragglers observed are w uma variables of amplitude more than about @xmath68 mag . the blue stragglers in the cmd cover the region where the instability strip occurs . we have roughly translated the instability strip of pamyatnykh ( 2000 ) into our observational cmd , as shown in fig [ fig1 ] . we will discuss our oscillation mode analyses of the two known blue straggler @xmath3 scuti stars s1280 and s1284 in a separate paper in preparation . however , we find several other stragglers with photometry that places them inside the instability strip : s1263 , s1267 , s968 , s1066 and s1434 to the blue of the known pulsators , and s752 , s2226 , s1082 and s975 to the red . we did not observe s1434 or s2226 . s752 , s975 , and s1267 are known to be long period variables ( @xmath69 d , 1221 d , and 846 d , respectively ; latham & milone 1996 ) , while s 1082 contains a close binary that may be part of a triple @xcite . therefore , these systems are less likely to show photometric variations due to pulsation ( although it is possible that components of the long period variables will still fall in the instability strip : in particular , the brightest component of s 1082 shows some evidence of being a @xmath3 scuti star ; sandquist et al . 2003 ) . as a result , s 968 , s 1066 , and s 1263 are the best candidates to search for pulsation . @xcite observed s 1263 extensively , and also found no sign of variation . we do not find convincing evidence of pulsation in any of the three stars , although this may be due to low pulsation amplitudes or higher frequency ( overtone ) oscillations . we discuss all three stars in [ otherv ] with regards to longer period variations . as discussed earlier , to tightly constrain the photometric variations present in some of the bluest blue stragglers , we were forced to subtract off a systematic error ( probably related to color response of the filter / ccd combination ) that appeared as correlated slow low - amplitude variations among several stars . once this correction was made several of the stars no longer showed any signs of significant variation ( s 977 , s 1066 , s 1267 ) . the remaining three ( s 752 , s 968 , s 1263 ) still showed trends , and will be discussed below . none of these stars tripped the variability condition @xmath70 , although in the case of s 752 at least there is fairly clear evidence of variability in one interval of time . before presenting the results , we note that we looked for correlations of the magnitude residuals with seeing and airmass for each star below , and found no evidence that such effects were responsible . * s 752 : * as noted earlier , s 752 is known to be a binary system ( @xmath69 d ; @xmath71 ; latham & milone 1996 ) containing an am star . this star was outside the field observed by @xcite , and although it was observed by svvm , it was not reported as variable . simoda ( 1991 ) observed the star and found it to be non - variable . we present our observations in fig . while the vast majority of our measurements show no indication of variability , our december 2000 observations give the indication of a decrease in brightness of 0.03 mag over the course of a week , punctuated with something resembling a flare ( 0.05 mag increase in brightness lasting over 3 hours ) . however , see gilliland et al . 1991 for a discussion of false flaring . * s 968 : * this blue straggler was not in the field observed by @xcite , but was observed by svmv although they did not report it as being variable . our observations are shown in fig . [ s968 ] . for most of the nights this star was observed , it was fairly consistent with constant brightness . on other nights there appeared to be noticeable trends during a night ( most notably february 18/19 , 2001 ) . this behavior could explain the non - detection by svmv . * s 1263 : * svmv indicate that their observations of this star have significant scatter of up to 0.03 mag , and kim et al . ( 1996 ) find large dispersion in their measurements . there does appear to be a long term drift in the brightness of the star ( of approximately 0.02 mag ) as seen in fig . [ s1263 ] , even after removing the effects of correlated residuals . some of our nights of observations were compromised by very poor seeing and contamination by mmj 5951 @xcite , which is about @xmath72 away . based on trends seen in the photometric residuals in measurements during the poorest seeing , we eliminated measurements with seeing greater than @xmath73 fwhm . we attempted to look for periodic signals in the data , although we did not find convincing evidence of any ( the best indications were for a period of 17.5 d ) . this system is not known to be part of a binary , so it should be monitored further to determine the manner of its variation . * s 1040 : * this system was determined by mathieu , latham , & griffin ( 1990 ) to be a single - lined spectroscopic binary having a circular orbit and a period @xmath74 d. the position of s 1040 to the blue side of the red giant branch in the color - magnitude diagram makes it a `` red straggler '' . using ultraviolet imaging and spectra , @xcite showed that the secondary companion to the red giant primary is likely to be a helium white dwarf . @xcite also laid out a hypothetical history for the system in which a main sequence binary began mass transfer at a period of approximately 2 d when the more massive star had a helium core mass of @xmath75 . mass transfer continued until the envelope of the initially more massive star was depleted , creating a system composed of a helium white dwarf and a blue straggler . once the blue straggler evolved to the red giant branch , a system like s 1040 would be created . in previous variability studies , @xcite found evidence for low amplitude ( 0.012 mag ) variation with a period of 7.97 d , while vsvm found no evidence of variation . we plot our data ( averaged in 0.1 d bins to improve the errors ) in fig . [ s1040 ] , and tabulate it in table [ s1040tab ] . we have observed several nights for which the system brightness decreased by as much as 0.06 mag compared to the peak level . when the data is phased to the ephemeris of mathieu et al . ( 1990 ) , there is an indication of a systematic decrease in brightness at a phase corresponding to the passage of the white dwarf in front of the giant . because our phase coverage is incomplete , we are unable to determine the exact depth of the feature . however , a feature of this depth and shape can not be explained as an eclipse of the red giant by the white dwarf . observations on two nights do not match up well with the light curve formed from the rest of the observations . we probed a variety of periods from 1 to 70 d , but did not find satisfactory light curves from any of the best trial periods from the l - k or l - s methods . additional observations should be made to better determine the shape of the light curve , and the nature of the variation . * s 1063 : * this system is a binary with orbital period of @xmath76 d and eccentricity @xmath77 ( latham et al . 1992 ; mathieu et al . 2002 ) , although an orbit has not yet been published for the system . s 1063 is known to be unusual in its x - ray emission @xcite and position in the color - magnitude diagram ( fainter than the subgiant branch ) . there is good evidence of proper motion membership though ( * ? ? ? * ; * ? ? ? * @xmath78 probability ) . this system fell just outside the gilliland et al . field , but was observed by vsvm in their examination of x - ray sources . vsvm identified the system as as a non - periodic variable based on an examination of periods up to the length of their longest interval of continuous observations ( 18 d ) . their figure 5 makes it clear that at the very least there is a great deal of variability in the light curve , and that the variation is not obviously related to the orbital characteristics of the system . we averaged observations from a given night in 0.1 d windows to improve the error in our measurements , and they are presented in table [ s1063tab ] . in our data , the system shows slight photometric trends during the course of a night , but systematic offsets in the star s photometry are clearly detected from night to night . using an l - s periodogram , we do not find any evidence of variation on the orbital timescale , but there is slight evidence of variation with a period of approximately 23 days . this period seems to bring the maxima and minima of our data into approximate alignment . as shown in fig . [ s1063 ] , a comparison of data from different observing runs indicates that the brightness level of the minimum of the light curve probably varies a few centimag from cycle to cycle . during the course of our observations the light curve appeared to maintain a roughly similar shape , but one that appears to differ from the portions observed by vsvm . although the system does not show periodicity on the orbital period , the similarity of the timescale may indicate that the variability is related via tides and/or magnetic activity . further progress on this system will require observations over much longer periods and a more detailed examination of the relationship with the orbit . [ we have three nights of observations of another sub - subgiant branch system ( s 1113 ) , but our phase coverage for this system is fairly poor , so we do not discuss it . ] we have presented @xmath0 observations for the open cluster m67 , and have discussed the light curves for the known w uma contact binaries and the monitoring of the majority of the blue stragglers for variability . we find that all of the known w uma binaries show light curve variations that occur on timescales of days to months . two systems ( s757 and iii-79 ) show large changes in the shapes of their light curves . the relative brightnesses of both the two maxima and the two minima in the light curve of s757 have been observed to change on timescales of less than a month . the faint system iii-79 shows a substantial change in the shape of the light curve between the january 1998 of svmv and our observations in march 2001 . the other two systems ( s1036 and s1282 ) show smaller variations . s1282 changed between a w - type and an a - type configuration between the 1973 observations of whelan et al . ( 1979 ) and the 1988 observations of gilliland et al . the existence of two systems which appear to change between these subtypes indicates that the classification is perhaps not a robust indicator of the evolutionary state of the stellar components . the blue straggler system s 1036 shows smaller light curve variations , but a stronger and more stable oconnell effect . we have verified that s 1282 is a highly - inclined totally - eclipsing system with @xmath79 and @xmath80 . because this system falls right at the cluster turnoff , we strongly encourage further spectroscopic work to provide a constraint on the cluster turnoff mass , and thus on the cluster age . this system has also shown unusual short disturbances in its light curve that may relate to magnetic activity , and which should be followed up . among the blue stragglers , in addition to the two known @xmath3 scuti pulsating variables , we find possible evidence of longer period variations in the stars s752 , s968 , and s1263 . while these stars did not satisfy our criteria for a definite claim of variability , we see trends in the photometry and apparent changes in the mean brightness level that should be investigated . finally , we present a series of observations of two long - period binary systems . for the poorly understood sub - subgiant branch system s 1063 , there are indications of quasi - periodicity on timescales similar to the orbital period , as well as variations in the mean brightness level of the light curve . for the giant - white dwarf system s 1040 , we find evidence of a drop in brightness at phases corresponding to the passage of the white dwarf in front of the giant using the ephemeris derived by mathieu et al . although this drop in brightness can not be due to an eclipse of the giant by the white dwarf itself , there may be associated material within the system that could account for the variability . this information can probably be used to contrain the inclination of the system . we would especially like to thank m. van den berg and k. stassun for the loan of data from their photometry papers . e.l.s . would like to thank m. blake , s. rucinski , w. welsh , r. taam , and p. etzel for helpful conversations during the course of this work . we would also like to thank the director of mount laguna observatory ( p. etzel ) for generous allocations of telescope time . this work has been partly supported by nsf grant ast-0098696 to e.l.s . ahumada , j. , & lapasset , e. 1995 , , 109 , 375 bailyn , c. d. 1995 , , 22 , 133 belloni , t. , verbunt , f. , & mathieu , r. d. 1998 , , 339 , 431 daz - cordovs , j. , claret , a. , & gimnez , a. 1995 , , 110 , 329 dinescu , d. i. , demarque , p. , guenther , d. b. , & pinsonneault , m. h. 1995 , , 109 , 2090 duerbeck , h. w. 1978 , , 32 , 361 fan , x. et al . 1996 , , 112 , 628 gilliland , r. l. & brown , t. m. 1992 , , 103 , 1945 gilliland , r. l. , brown , t. m. , duncan , d. k. , suntzeff , n. b. , lockwood , g. w. , thompson , d. t. , schild , r. e. , jeffrey , w. a. , & penprase , b. e. , 1991 , , 101 , 541 girard , t. m. , grundy , w. m. , lopez , c. e. , & van altena , w. f. 1989 , , 98 , 227 hauschildt , p. h. , allard , f. , & baron , e. 1999 , , 512 , 377 hills , j. g. & day , c. a. 1976 , astrophys . lett . , 17 , 87 honeycutt , r. k. 1992 , , 104 , 435 hurley , j. r. , tout , c. a. , aarseth , s. j. , & pols , o. r. 2001 , , 323 , 630 kim , s .- l . , chun , m. y. , park , b .- , & lee , s .- w . 1996 , jkas , 29 , 43 kurochkin , n. e. 1979 , astron . circular ussr , 1076 , 2 lafler , j. & kinman , t. d. 1965 , , 11 , 216 landsman , w. , aparicio , j. , bergeron , p. , di stefano , r. , & stecher , t. p. 1997 , , 481 , l93 latham , d. , mathieu , r. , milone , a. , & davis , r. 1992 , in binaries as tracers of stellar formation , ed . a. duquennoy and m. mayor ( cambridge : cambridge university press ) , 132 latham , d. w. , & milone , a. a. e. 1996 , in asp conf . 90 , the origins , evolution , and destinies of binary stars in clusters , ed . e. f. milone , and j .- c . mermilliod ( san francisco : asp ) , 385 maceroni , c. , milano , l. , & russo , g. 1984 , , 58 , 405 mathieu , r. d. , latham , d. w. , & griffin , r. f. 1990 , , 100 , 1859 mathieu , r. d. , van den berg , m. , torres , g. , latham , d. w. , verbunt , f. , & stassun , k. 2002 , aj , submitted milone , a. a. e & latham , d. w. 1994 , , 108 , 1828 mochnacki , s. w. 1981 , , 245 , 650 mochnacki , s. w. & doughty , n. a. 1972 , , 156 , 51 montgomery , k. a. , marschall , l. a. , & janes , k. a. 1993 , , 106 , 181 nissen , p. e. , twarog , b. a. , & crawford , d. l. 1986 , , 92 , 1100 . oconnell , d. j. k. 1951 , riverview coll . publ . , 2 , 85 pamyatnykh , a. a. 2000 , in delta scuti and related stars , ed . m. breger and m. h. montgomery ( san francisco : asp ) , 215 peterson , r. c. , carney , b. w. , & latham , d. w. 1984 , , 279 , 237 portegies zwart , s. f. , mcmillan , s. l. w. , hut , p. & makino , j. 2001 , , 321 , 199 rajamohan , r. , bhattacharyya , j. c. , subramanian , v. , & kuppuswamy , k. 1988 , bull . india , 16 , 139 rasio , f. a. 1996 , , 444 , l41 rucinski , s. m. 1997 , , 113 , 1112 rucinski , s. m. 1998 , , 116 , 2998 rucinski , s. m. 2001 , , 122 , 1007 rucinski , s. m. & duerbeck , h. w. 1997 , , 109 , 1340 sanders , w. l. 1977 , , 27 , 89 sandquist , e. l. , latham , d. w. , shetrone , m. d. , & milone , a. a. e. 2003 , , in press scargle , j. d. 1982 , , 263 , 835 simoda , m. 1991 , ibvs , 3675 , 1 stassun , k. g. , van den berg , m. , mathieu , r. d. , & verbunt , f. 2002 , , 382 , 899 ( svmv ) stetson , p. b. 1990 , , 102 , 932 van den berg , m. , orosz , j. , verbunt , f. , & stassun , k. 2001 , , 375 , 375 van den berg , m. , stassun , k. g. , verbunt , f. , & mathieu , r. d. 2002 , , 382 , 888 ( vsvm ) van hamme , w. 1993 , , 106 , 2096 welch , d. l. & stetson , p. b. 1993 , , 105 , 1813 whelan , j. a. j. , worden , s. p. , ruciski , s. m. , & romanishin , w. 1979 , , 186 , 729 cccc 1 & dec . 5/6 , 2000 & 1884.878 & 60 + 2 & dec . 7/8 , 2000 & 1886.806 & 111 + 3 & dec . 11/12 , 2000 & 1890.817 & 57 + 4 & dec . 12/13 , 2000 & 1891.820 & 116 + 5 & jan . 23/24 , 2001 & 1933.675 & 206 + 6 & jan . 25/26 , 2001 & 1935.673 & 176 + 7 & jan . 29/30 , 2001 & 1939.676 & 44 + 8 & jan . 30/31 , 2001 & 1940.657 & 130 + 9 & jan . 31/feb . 1 , 2001 & 1941.663 & 161 + 10 & feb . 17/18 , 2001 & 1958.655 & 63 + 11 & feb . 18/19 , 2001 & 1959.790 & 75 + 12 & mar . 1/2 , 2001 & 1970.619 & 205 + 13 & mar . 3/4 , 2001 & 1972.631 & 205 + 14 & mar . 5/6 , 2001 & 1974.721 & 32 + 15 & jan . 21/22 , 2002 & 2296.687 & 89 + 16 & jan . 24/25 , 2002 & 2299.690 & 124 + 17 & feb . 5/6 , 2002 & 2311.649 & 119 + 18 & feb . 10/11 , 2002 & 2316.638 & 167 + 19 & mar . 18/19 , 2002 & 2352.625 & 131 + [ obs ] lcccc @xmath81 & 2451800.129 & 2450500.047 & & 2450000.036 + @xmath82 ( d ) & @xmath83 & @xmath84 & 0.360452 & 0.2704 + @xmath85 & 0.08 & 0.13 & 0.39 & @xmath86 + @xmath87 & 0.06 & 0.06 & 0.33 & 0.11 + @xmath88 & 13.51 & 12.81 & 13.44 & 15.90 + @xmath89 & 0.61 & 0.50 & 0.52 & 1.11 + @xmath36 & & @xmath90 & @xmath91 & + @xmath37 & & & @xmath92 & + [ wuma ] ccrcrlrl 145 & & & & & none & & + 277 & & & & & none & & + 751 & 12.70 & & 0.008 & 0.53 & @xmath95 & 777 & no variation + 752 & 11.32 & 0.30 & 0.018 & 0.75 & @xmath96 & 1418 & am star ; possible flare detected + 792 & 11.99 & 0.60 & 0.007 & 1.01 & @xmath97 & 1610 & no variation + 968 & 11.25 & 0.13 & 0.010 & 0.74 & @xmath98 & 1261 & am star ; possible low amplitude , + & & & & & & & timescale @xmath99 days + 975 & 11.05 & 0.43 & 0.077 & 7.26 & @xmath100 & 1659 & photometry affected by faint companion + 977 & 10.02 & @xmath101 & 0.006 & 0.69 & @xmath102 & 504 & no variation + 984 & 12.26 & 0.57 & 0.014 & 1.64 & @xmath100 & 1732 & no variation + 997 & 12.13 & 0.46 & 0.006 & 0.92 & @xmath100 & 1655 & no variation + 1005 & 12.68 & 0.52 & 0.010 & 1.57 & @xmath100 & 1654 & no variation + 1031 & 13.29 & 0.46 & 0.007 & 0.32 & @xmath100 & 1644 & no variation + 1036 & 12.81 & 0.49 & 0.046 & 90.59 & @xmath100 & 1651 & w uma variable ( ev cnc ) + 1066 & 10.95 & 0.11 & 0.003 & 0.24 & @xmath103 & 2061 & no variation + 1072 & 11.31 & 0.61 & 0.006 & 0.97 & @xmath103 & 2073 & no variation + 1082 & 11.20 & 0.42 & 0.024 & 33.59 & @xmath103 & 2097 & rs cvn variable ( es cnc ) + 1165 & & & & & none & & + 1183 & 12.66 & & 0.011 & 0.07 & 15 & 66 & no variation + 1195 & 12.29 & & 0.003 & 0.40 & 15 & 78 & no variation + 1263 & 11.06 & 0.19 & 0.007 & 0.98 & @xmath104 & 1275 & possible low amplitude , + & & & & & & & timescale @xmath105 10 days + 1267 & 10.90 & & 0.005 & 0.45 & @xmath106 & 484 & no variation + 1273 & 12.25 & 0.57 & 0.006 & 0.63 & @xmath103 & 1451 & no variation + 1280 & 12.23 & 0.26 & 0.009 & 2.55 & @xmath103 & 1685 & @xmath3 scu variable ( ew cnc ) + 1282 & 13.44 & 0.52 & 0.133 & 108.08 & @xmath107 & 1864 & w uma variable ( ah cnc ) + 1284 & 10.93 & 0.22 & 0.012 & 9.41 & @xmath103 & 2015 & @xmath3 scu variable ( ex cnc ) + 1434 & 10.70 & 0.11 & & & none & & + 1440 & & & & & none & & + 1947 & & & & & none & & + 2204 & 12.89 & 0.45 & 0.013 & 1.03 & @xmath100 & 1717 & no variation + 2223 & 13.30 & 0.50 & 0.012 & 0.92 & @xmath103 & 2084 & no variation + 2226 & & & & & none & & + [ straggler ] cccr 1884.9294 & 11.5231 & 0.0006 & 33 + 1885.0391 & 11.5243 & 0.0008 & 27 + 1886.8586 & 11.5184 & 0.0007 & 36 + 1886.9615 & 11.5157 & 0.0006 & 43 + 1887.0468 & 11.5177 & 0.0007 & 32 + 1890.8688 & 11.5076 & 0.0007 & 42 + 1890.9386 & 11.4998 & 0.0014 & 14 + 1891.8734 & 11.5064 & 0.0008 & 37 + 1891.9872 & 11.5092 & 0.0006 & 51 + 1892.0607 & 11.5101 & 0.0008 & 28 + 1930.7983 & 11.5395 & 0.0058 & 1 + 1933.7305 & 11.5209 & 0.0005 & 55 + 1933.8308 & 11.5210 & 0.0005 & 63 + 1933.9312 & 11.5191 & 0.0005 & 57 + 1934.0118 & 11.5181 & 0.0007 & 31 + 1935.7288 & 11.5128 & 0.0007 & 45 + 1935.8314 & 11.5152 & 0.0005 & 60 + 1935.9333 & 11.5138 & 0.0006 & 54 + 1936.0000 & 11.5114 & 0.0010 & 17 + 1939.8739 & 11.4991 & 0.0008 & 36 + 1939.9386 & 11.4952 & 0.0018 & 8 + 1940.7053 & 11.5028 & 0.0007 & 37 + 1940.8386 & 11.4972 & 0.0007 & 46 + 1940.9371 & 11.5010 & 0.0007 & 47 + 1941.7175 & 11.5034 & 0.0005 & 49 + 1941.8186 & 11.5074 & 0.0005 & 48 + 1941.9199 & 11.5052 & 0.0005 & 50 + 1941.9883 & 11.5018 & 0.0010 & 14 + 1958.6849 & 11.4962 & 0.0036 & 7 + 1958.8989 & 11.4972 & 0.0006 & 50 + 1958.9531 & 11.4961 & 0.0017 & 5 + 1959.8451 & 11.5077 & 0.0006 & 47 + 1959.9218 & 11.5017 & 0.0009 & 28 + 2296.7175 & 11.5428 & 0.0008 & 21 + 2300.0020 & 11.5423 & 0.0008 & 20 + 2311.6980 & 11.5368 & 0.0009 & 23 + 2311.7991 & 11.5394 & 0.0005 & 44 + 2311.8994 & 11.5383 & 0.0005 & 46 + 2311.9536 & 11.5345 & 0.0019 & 5 + 2316.6882 & 11.5028 & 0.0009 & 46 + 2316.7888 & 11.5104 & 0.0006 & 53 + 2316.8887 & 11.5079 & 0.0007 & 45 + 2316.9590 & 11.5038 & 0.0011 & 23 + 2352.6743 & 11.5638 & 0.0006 & 47 + 2352.7751 & 11.5613 & 0.0005 & 50 + 2352.8530 & 11.5622 & 0.0006 & 34 + [ s1040tab ] cccr 1884.9297 & 13.4520 & 0.0010 & 33 + 1885.0391 & 13.4550 & 0.0014 & 27 + 1886.8594 & 13.4854 & 0.0011 & 36 + 1886.9609 & 13.4899 & 0.0008 & 43 + 1887.0469 & 13.4912 & 0.0010 & 32 + 1890.8711 & 13.5676 & 0.0011 & 43 + 1890.9375 & 13.5754 & 0.0025 & 13 + 1891.8750 & 13.5556 & 0.0012 & 37 + 1891.9883 & 13.5590 & 0.0010 & 52 + 1892.0625 & 13.5538 & 0.0015 & 27 + 1933.7305 & 13.5066 & 0.0007 & 55 + 1933.8320 & 13.5129 & 0.0006 & 64 + 1933.9336 & 13.5140 & 0.0007 & 58 + 1934.0117 & 13.5138 & 0.0010 & 29 + 1935.7305 & 13.5668 & 0.0008 & 45 + 1935.8320 & 13.5704 & 0.0007 & 60 + 1935.9336 & 13.5697 & 0.0007 & 54 + 1936.0000 & 13.5757 & 0.0013 & 17 + 1939.8750 & 13.6060 & 0.0010 & 36 + 1939.9375 & 13.6033 & 0.0027 & 7 + 1940.7070 & 13.5926 & 0.0009 & 37 + 1940.8398 & 13.5944 & 0.0010 & 47 + 1940.9375 & 13.5924 & 0.0011 & 46 + 1941.7188 & 13.5715 & 0.0008 & 50 + 1941.8242 & 13.5696 & 0.0007 & 49 + 1941.9258 & 13.5693 & 0.0007 & 49 + 1941.9922 & 13.5667 & 0.0016 & 12 + 1958.6797 & 13.5500 & 0.0104 & 4 + 1958.8984 & 13.5577 & 0.0008 & 49 + 1958.9531 & 13.5658 & 0.0025 & 5 + 1959.8984 & 13.5886 & 0.0010 & 51 + 1970.6875 & 13.5418 & 0.0008 & 63 + 1970.7930 & 13.5450 & 0.0008 & 75 + 1970.8828 & 13.5474 & 0.0009 & 67 + 1972.6914 & 13.5598 & 0.0007 & 97 + 1972.7930 & 13.5630 & 0.0010 & 89 + 1972.8594 & 13.5639 & 0.0022 & 17 + 1974.6758 & 13.5433 & 0.0024 & 30 + 2296.7188 & 13.5082 & 0.0008 & 21 + 2300.0000 & 13.5386 & 0.0009 & 20 + 2311.6992 & 13.6108 & 0.0010 & 24 + 2311.8008 & 13.6122 & 0.0006 & 44 + 2311.9023 & 13.6116 & 0.0006 & 48 + 2311.9531 & 13.6154 & 0.0027 & 3 + 2316.6875 & 13.5574 & 0.0016 & 42 + 2316.7891 & 13.5538 & 0.0008 & 53 + 2316.8906 & 13.5541 & 0.0009 & 46 + 2316.9609 & 13.5599 & 0.0019 & 22 + 2352.6758 & 13.5144 & 0.0007 & 48 + 2352.7773 & 13.5164 & 0.0006 & 51 + 2352.8555 & 13.5183 & 0.0008 & 32 + [ s1063tab ]

in this paper we study simplicial complexes as higher dimensional graphs in order to produce algebraic statements about their facet ideals . we introduce a large class of square - free monomial ideals with cohen - macaulay quotients , and a criterion for the cohen - macaulayness of facet ideals of simplicial trees . along the way , we generalize several concepts from graph theory to simplicial complexes .

math0307403

from the point of view of commutative algebra , the focus of this paper is on finding square - free monomial ideals that have cohen - macaulay quotients . in @xcite villarreal proved a criterion for the cohen - macaulayness of edge ideals of graphs that are trees . edge ideals are square - free monomial ideals where each generator is a product of two distinct variables of a polynomial ring . these ideals have been studied extensively by villarreal , vasconcelos and simis among others . in @xcite we studied a generalization of this concept ; namely the _ facet ideal _ of a simplicial complex . by generalizing the definition of a `` tree '' to simplicial complexes , we extended the results of @xcite from the class of edge ideals to all square - free monomial ideals . below we investigate the structure of simplicial complexes in order to show that villarreal s cohen - macaulay criterion for graph - trees extends to simplicial trees ( corollary [ cm - criterion ] ) . this is of algebraic and computational significance , as it provides an effective criterion for cohen - macaulayness that works for a large class of square - free monomial ideals . we introduce a condition on a simplicial complex that ensures the cohen - macaulayness of its facet ideal , and a method to build a cohen - macaulay ideal from any given square - free monomial ideal . along the road to the algebraic goal , this study sheds light on the beautiful combinatorial structure of simplicial complexes . the paper is organized as follows : sections [ basic - setup ] to [ basic - trees - section ] review the basic definitions and cover the elementary properties of trees . in section [ n - partite - section ] we draw comparisons between graph theory and simplicial complex theory , and prove a generalized version of knig s theorem in graph theory for simplicial complexes . we then go on to prove a structure theorem for unmixed trees in section [ structure - theorem - section ] . we introduce the notion of a _ grafted _ simplicial complex in section [ grafting - section ] , and show that for simplicial trees , being grafted and being unmixed are equivalent conditions . the notion of grafting brings us to section [ cm - section ] , where we prove that grafted simplicial complexes are cohen - macaulay , from which it follows that a simplicial tree is unmixed if and only if it is cohen - macaulay . in this section we define the basic notions that we will use later in the paper . some of the proofs that appeared earlier in @xcite have been omitted here ; we refer the reader to the relevant sections of @xcite when that is the case . a _ simplicial complex _ @xmath0 over a set of vertices @xmath1 is a collection of subsets of @xmath2 , with the property that @xmath3 for all @xmath4 , and if @xmath5 then all subsets of @xmath6 are also in @xmath0 ( including the empty set ) . an element of @xmath0 is called a _ face _ of @xmath0 , and the _ dimension _ of a face @xmath6 of @xmath0 is defined as @xmath7 , where @xmath8 is the number of vertices of @xmath6 . the faces of dimensions 0 and 1 are called _ vertices _ and _ edges _ , respectively , and @xmath9 . the maximal faces of @xmath0 under inclusion are called _ facets _ of @xmath0 . the dimension of the simplicial complex @xmath0 is the maximal dimension of its facets ; in other words @xmath10 we denote the simplicial complex @xmath0 with facets @xmath11 by @xmath12 and we call @xmath13 the _ facet set _ of @xmath0 . a simplicial complex with only one facet is called a _ simplex_. by a _ subcollection _ of a simplicial complex @xmath0 we mean a simplicial complex whose facet set is a subset of the facet set of @xmath0 . a simplicial complex @xmath14 is _ connected _ if for every pair @xmath15 , @xmath16 , there exists a sequence of facets @xmath17 of @xmath18 such that @xmath19 , @xmath20 and @xmath21 for @xmath22 . an equivalent definition is stated on page 222 of @xcite : @xmath18 as above is _ disconnected _ if its vertex set @xmath2 can be partitioned as @xmath23 , where @xmath24 and @xmath25 are nonempty subsets of @xmath2 , such that no facet of @xmath18 has vertices in both @xmath24 and @xmath25 . otherwise @xmath18 is _ connected_. let @xmath0 be a simplicial complex over @xmath26 vertices labeled @xmath27 . let @xmath28 be a field , @xmath29 be indeterminates , and @xmath30 be the polynomial ring @xmath31 $ ] . \(a ) we define @xmath32 to be the ideal of @xmath30 generated by all the square - free monomials @xmath33 , where @xmath34 is a facet of @xmath0 . we call @xmath32 the _ facet ideal _ of @xmath0 . \(b ) we define @xmath35 to be the ideal of @xmath30 generated by all the square - free monomials @xmath33 , where @xmath34 is not a face of @xmath0 . we call @xmath35 the _ non - face ideal _ or the _ stanley - reisner ideal _ of @xmath0 . we refer the reader to @xcite and @xcite for an extensive coverage of the theory of stanley - reisner ideals . throughout this paper we often use @xmath29 to denote both the vertices of @xmath18 and the variables appearing in @xmath36 . let @xmath37 be an ideal in a polynomial ring @xmath31 $ ] , where @xmath28 is a field and @xmath38 are square - free monomials in @xmath29 that form a minimal set of generators for @xmath39 . \(a ) we define @xmath40 to be the simplicial complex over a set of vertices @xmath27 with facets @xmath11 , where for each @xmath4 , @xmath41 . we call @xmath40 the _ facet complex _ of @xmath39 . \(b ) we define @xmath42 to be the simplicial complex over a set of vertices @xmath27 , where @xmath34 is a face of @xmath42 if and only if @xmath43 . we call @xmath42 the _ non - face complex _ or the _ stanley - reisner complex _ of @xmath39 . facet ideals give a one - to - one correspondence between simplicial complexes and square - free monomial ideals . notice that given a square - free monomial ideal @xmath39 in a polynomial ring @xmath31 $ ] , the vertices of @xmath40 are those variables that divide a monomial in the generating set of @xmath39 ; this set may not necessarily include all elements of @xmath44 . the fact that some extra variables may appear in the polynomial ring does not affect the algebraic or combinatorial structure of @xmath40 . on the other hand , if @xmath18 is a simplicial complex , being able to consider the facet ideals of its subcomplexes as ideals in the same ring simplifies many of our discussions . [ example1 ] let @xmath0 be the simplicial complex below . @xmath45 here @xmath46 , @xmath47 are ideals in the polynomial ring @xmath48 $ ] . [ example2 ] if @xmath49 $ ] , then @xmath42 is the 1-dimensional simplicial complex : @xmath50 and @xmath40 is the simple graph @xmath51 in this special case @xmath39 is also called the _ edge ideal _ of the graph @xmath40 ( this terminology is due to villarreal ; see @xcite ) . we now generalize some notions from graph theory to simplicial complexes . let @xmath0 be a simplicial complex with vertex set @xmath2 and facets @xmath11 . vertex cover _ for @xmath0 is a subset @xmath52 of @xmath2 , with the property that for every facet @xmath53 there is a vertex @xmath54 such that @xmath55 . minimal vertex cover _ of @xmath0 is a subset @xmath52 of @xmath2 such that @xmath52 is a vertex cover , and no proper subset of @xmath52 is a vertex cover for @xmath0 . the smallest cardinality of a vertex cover of @xmath0 is called the _ vertex covering number _ of @xmath0 and is denoted by @xmath56 . a simplicial complex @xmath0 is _ unmixed _ if all of its minimal vertex covers have the same cardinality . note that a simplicial complex may have several minimal vertex covers . let @xmath18 be a simplicial complex . a set @xmath57 of facets of @xmath18 is called an _ independent set _ if @xmath58 whenever @xmath59 . the maximum possible cardinality of an independent set of facets in @xmath18 , denoted by @xmath60 , is called the _ independence number _ of @xmath18 . an independent set of facets which is not a proper subset of any other independent set is called a _ maximal independent set _ of facets . [ example11 ] if @xmath18 is the simplicial complex @xmath61{silly - example.eps}\ ] ] then @xmath62 . also , @xmath18 is unmixed as its minimal vertex covers , listed below , all have cardinality equal to two : @xmath63 this , by the way , is an example of a `` grafted '' tree ( see definitions [ tree ] and [ grafting ] ) . we show later in the paper that all grafted trees are unmixed . the graph @xmath40 in example [ example2 ] however is not unmixed . this is because @xmath64 and @xmath65 are both minimal vertex covers for @xmath40 of different cardinalities . in this case the same argument shows that the simplicial complex in example [ example1 ] is not unmixed . the following is an easy but very useful observation ; see proposition 1 in @xcite for a proof . [ minprime ] let @xmath0 be a simplicial complex over @xmath26 vertices labeled @xmath29 . consider the ideal @xmath67 in the polynomial ring @xmath68 $ ] over a field @xmath28 . then an ideal @xmath69 of @xmath30 is a minimal prime of @xmath39 if and only if @xmath70 is a minimal vertex cover for @xmath0 . we say that a simplicial complex @xmath18 over a set of vertices @xmath29 is cohen - macaulay if for a given field @xmath28 , the quotient ring @xmath71/{{\mathcal{f}}}({\delta})\ ] ] is cohen - macaulay . it follows directly from proposition [ minprime ] , or from an elementary duality with stanley - reisner theory discussed in corollary 2 of @xcite , that in order for @xmath18 to be cohen - macaulay , it has to be unmixed . [ cm - is - unmixed ] suppose that @xmath0 is a simplicial complex with vertex set @xmath29 . if @xmath31/{{\mathcal{f}}}(\delta)$ ] is cohen - macaulay , then @xmath0 is unmixed . [ dimension ] it is worth observing that for a square - free monomial ideal @xmath39 , there is a natural way to construct @xmath42 and @xmath40 from each other using the structure of the minimal primes of @xmath39 . to do this , consider the vertex set @xmath2 consisting of all variables that divide a monomial in the generating set of @xmath39 . the following correspondence holds : @xmath72 facet of @xmath42 @xmath73 @xmath74 minimal vertex cover of @xmath40 also @xmath75 where the intersection is taken over all prime ideals @xmath76 of @xmath77 $ ] that are generated by a minimal vertex cover of @xmath40 ( or equivalently , primes @xmath76 that are generated by @xmath78 , where @xmath6 is a facet of @xmath42 ; see @xcite theorem 5.1.4 ) . regarding the dimension and codimension of @xmath39 , note that by theorem 5.1.4 of @xcite and the discussion above , setting @xmath79 $ ] as above , we have @xmath80 and @xmath81 we illustrate all this through an example . for @xmath82 , where @xmath40 and @xmath42 are drawn in example [ example2 ] , we have : [ cols="^,^ " , ] notice that in the case that @xmath0 is a graph , definition [ tree ] agrees with the definition of a tree in graph theory , with the difference that now the term `` leaf '' refers to an edge , rather than a vertex . [ two leaves ] let @xmath0 be a tree of two or more facets . then @xmath0 has at least two leaves . suppose that @xmath0 has @xmath83 facets @xmath11 where @xmath84 . we prove the lemma by induction on @xmath83 . the case @xmath85 follows from the definition of a leaf . to prove the general case suppose that @xmath86 is a leaf of @xmath0 and @xmath87 . consider the subcomplex @xmath88 of @xmath0 . by induction hypothesis @xmath89 has two distinct leaves ; say @xmath90 and @xmath91 are those leaves . at least one of @xmath90 and @xmath91 must be different from @xmath92 ; say @xmath93 . we show that @xmath90 is a leaf for @xmath0 . let @xmath94 . given any facet @xmath53 with @xmath95 , we already know by the fact that @xmath90 is a leaf of @xmath89 @xmath96 we need to verify this for @xmath97 . since @xmath86 is a leaf for @xmath0 and @xmath98 , @xmath99 intersecting both sides of this inclusion with @xmath90 , we obtain @xmath100 where the last inclusion holds because @xmath101 and @xmath90 is a leaf of @xmath89 . it follows that @xmath90 , as well as @xmath86 , is a leaf for @xmath0 . a promising property of trees from an algebraic point of view is that they behave well under localization , i.e. the localization of a tree is a forest . this property is in particular useful when making inductive arguments on trees , as localization usually corresponds to reducing the number of vertices of a simplicial complex . before proving this , we first determine what the localization of a simplicial complex precisely looks like . [ local - discussion ] suppose that @xmath102 is a simplicial complex over the vertex set @xmath103 . let @xmath76 be a prime ideal of @xmath31 $ ] generated by a subset of @xmath44 that contains @xmath104 ( we show later in the proof of lemma [ localization ] that this is the main case that we need to study ) . we would like to see what the simplicial complex associated to @xmath105 looks like . so @xmath106 now suppose @xmath107 where each @xmath108 is the monomial corresponding to the facet @xmath53 . it follows that @xmath109 where each @xmath110 is obtained by dividing @xmath108 by the product of all the variables in @xmath111 that appear in @xmath108 . some of the monomials in the generating set of @xmath105 are redundant after this elimination , so without loss of generality we can write : @xmath112 where @xmath113 are the redundant monomials . we use the notation @xmath114 to indicate the simplicial complex associated to the monomial ideal with the same generating set as the one described for @xmath105 in ( [ e : local ] ) , in the polynomial ring @xmath115 $ ] . it follows that : @xmath116 where for each @xmath4 , @xmath117 and @xmath118 each contain at least one of @xmath119 . this simplicial complex is called the _ localization _ of @xmath18 at the prime ideal @xmath76 . note that every minimal vertex cover @xmath52 of @xmath18 that is contained in @xmath120 remains a minimal vertex cover of @xmath114 , as the minimal prime over @xmath39 generated by the elements of @xmath52 remains a minimal prime of @xmath105 . moreover if @xmath18 is unmixed then @xmath114 is also unmixed . algebraically , this is easy to see , as the height of the minimal primes of @xmath105 remain the same . one can also see it from a combinatorial argument : if @xmath121 is a minimal vertex cover for @xmath114 , then @xmath122 covers all facets @xmath119 , and therefore @xmath118 , as well . therefore @xmath122 covers all of @xmath123 , and so has a subset @xmath124 of cardinality @xmath56 that is a minimal vertex cover for @xmath18 , and so @xmath124 must cover @xmath114 as well . therefore @xmath125 . we have thus shown that : [ unmixed - localizes ] let @xmath0 be an unmixed simplicial complex with vertices @xmath29 , and let @xmath126 be the facet ideal of @xmath0 in the polynomial ring @xmath68 $ ] where @xmath28 is a field . then for any prime ideal @xmath76 of @xmath30 , @xmath114 is unmixed with @xmath127 . we examine a specific case : [ local - example ] let @xmath18 be the simplicial complex below with @xmath128 its facet ideal in the polynomial ring @xmath129 $ ] . @xmath45 let @xmath130 be a prime ideal of @xmath30 . then @xmath131 . the tree @xmath114 , shown below , has minimal vertex covers @xmath132 and @xmath133 , which are the generating sets for the minimal primes of @xmath105 . @xmath134 if @xmath135 then @xmath136 which corresponds to the forest @xmath137 drawn below with minimal vertex covers @xmath65 and @xmath138 . @xmath139 example [ local - example ] above also demonstrates the following lemma . [ localization ] let @xmath0 be a tree with vertices @xmath29 , and let @xmath126 be the facet ideal of @xmath0 in the polynomial ring @xmath68 $ ] where @xmath28 is a field . then for any prime ideal @xmath76 of @xmath30 , @xmath114 is a forest . the first step is to show that it is enough to prove this for prime ideals of @xmath30 generated by a subset of @xmath140 . to see this , assume that @xmath76 is a prime ideal of @xmath30 and that @xmath141 is another prime of @xmath30 generated by all @xmath142 such that @xmath143 ( recall that the minimal primes of @xmath39 are generated by subsets of @xmath140 ) . so @xmath144 . if @xmath37 , then @xmath145 where for each @xmath4 , @xmath146 is the image of @xmath108 in @xmath147 . in other words , @xmath146 is obtained by dividing @xmath108 by the product of all the @xmath148 such that @xmath149 and @xmath150 . but @xmath150 implies that @xmath151 , and so it follows that @xmath152 . therefore @xmath153 . on the other hand since @xmath144 , @xmath154 , which implies that @xmath155 ( the equality and inclusions of the ideals here mean equality and inclusion of their generating sets ) . we now prove the theorem for @xmath156 . following the setup in discussion [ local - discussion ] , we let @xmath157 @xmath158 @xmath159 for some @xmath160 . to show that @xmath161 is a forest , we need to show that every nonempty subcollection of @xmath161 has a leaf . let @xmath162 be a subcollection of @xmath89 where @xmath163 are distinct facets . if @xmath164 , @xmath165 is obviously a leaf and so we are done ; so suppose @xmath166 . consider the corresponding subcollection @xmath167 of @xmath0 , which has a leaf , say @xmath168 . so there exists @xmath169 , such that @xmath170 for every facet @xmath171 . now since each of the @xmath172 is a nonempty facet of @xmath173 and @xmath174 , the same statement holds in @xmath173 ; so @xmath175 for every facet @xmath176 . this implies that @xmath177 is a leaf for @xmath173 . in this section we study simplicial complexes as graphs with higher dimension , drawing results that will help us later in inductive arguments on unmixed trees . [ unmixed - induction ] if @xmath18 is a simplicial complex that has a leaf @xmath6 with joint @xmath178 , then @xmath179 . suppose @xmath180 . let @xmath181 and let @xmath52 be a vertex cover of minimal cardinality for @xmath161 , which implies that @xmath182 , as any vertex cover of @xmath18 has a subset that is a vertex cover of @xmath161 . since @xmath6 is a facet of @xmath161 , there exists a vertex @xmath183 that belongs to @xmath6 . if @xmath184 is a free vertex of @xmath6 , we may replace it by a non - free vertex of @xmath6 to get a vertex cover @xmath185 of @xmath161 , with a subset @xmath186 that is a minimal vertex cover of @xmath161 , and so @xmath187 . but now @xmath186 is a minimal vertex cover for all of @xmath18 , and so @xmath188 which implies that @xmath189 . [ unmixed - corollary ] if the simplicial complex @xmath18 is a tree and @xmath190 is a joint , then @xmath191 . this means that in a tree with more than one facet , it is always possible to remove a facet without reducing the vertex covering number . moreover we show in proposition [ joint - removal - unmixed ] that if @xmath18 is an unmixed tree with a joint @xmath178 , then @xmath192 is also unmixed . as a result , one can use induction on the number of facets of an unmixed tree . note that all these arguments remain valid for a forest . we are now ready to extend knig s theorem from graph theory . [ konig ] if @xmath18 is a simplicial complex that is a tree ( forest ) and @xmath180 , then @xmath18 has @xmath193 independent facets , and therefore @xmath194 . we use induction on the number of facets @xmath83 of @xmath18 . if @xmath195 , then there is nothing to prove since @xmath196 . suppose that the theorem holds for forests with less than @xmath83 facets and let @xmath18 be a forest with @xmath83 facets . if every connected component of @xmath18 has only one facet , our claim follows immediately . otherwise , by corollary [ unmixed - corollary ] one can remove a joint of @xmath18 to get a forest @xmath161 with @xmath197 , and so by induction hypothesis @xmath161 has @xmath193 independent facets , which are also independent facets of @xmath18 ; so @xmath198 . on the other hand it is clear that @xmath199 , and so the assertion follows . this section is the combinatorial core of the paper . here we give a precise description of the structure of an unmixed tree . it turns out that a tree is unmixed if and only if it is `` grafted '' ( see definition [ grafting ] ) . the notion of grafting is what eventually builds a bridge between unmixed and cohen - macaulay trees . below @xmath200 stands for the vertex set of @xmath18 . [ basic - lemma ] let @xmath18 be an unmixed simplicial complex . suppose that @xmath201 , and @xmath202 is a maximal independent set of facets of @xmath18 . then every vertex of @xmath18 belongs to one of the @xmath53 . in other words , the vertex set of @xmath18 is the disjoint union of the vertex sets of the @xmath53 : @xmath203 let @xmath184 be an vertex of @xmath18 that does not belong to any of the @xmath53 . then one can find a minimal vertex cover @xmath52 of @xmath18 containing @xmath184 ( this is always possible ) . but then @xmath52 must contain one vertex of each of the @xmath53 as well , which implies that @xmath204 . since @xmath18 is unmixed , this is not possible . lemma [ basic - lemma ] does not hold in general for any unmixed simplicial complex . take , for example , the case of a complete graph @xmath178 over 5 vertices labeled @xmath205 ( every pair of vertices of @xmath178 are connected by an edge ) . this graph is unmixed with @xmath206 and @xmath207 . however , @xmath208 is a maximal independent set of facets and the fifth vertex @xmath209 of @xmath178 is missing from the vertex set of the graph @xmath210 , which contradicts the claim of lemma [ basic - lemma ] . lemma [ basic - lemma ] along with theorem [ konig ] provides us with the following property for unmixed trees . [ basic ] if @xmath18 is an unmixed tree with @xmath180 , and @xmath202 is a maximal independent set of facets of @xmath18 , then @xmath211 . [ leaf - independence ] if @xmath18 is an unmixed tree , then any maximal independent set of facets of cardinality @xmath56 of @xmath18 contains all the leaves . in particular , the leaves of an unmixed tree are independent . every leaf has a free vertex , and so it follows from above that a independent set of facets of cardinality @xmath56 must contain all the leaves . the claim then follows . [ joint - not - leaf ] if @xmath18 is an unmixed tree , then a maximal independent set of facets of cardinality @xmath56 of @xmath18 can not contain a joint . in particular , a joint of an unmixed tree can not be a leaf . if @xmath178 is a joint , it has to intersect a leaf @xmath6 by definition , and as @xmath6 is in every maximal independent set of facets of cardinality @xmath56 , @xmath178 can not be in any . but even more is true . for an unmixed tree @xmath18 , there is only one maximal independent set of facets with @xmath56 elements , and that is the set consisting of all the leaves . we prove this in theorem [ structuretheorem ] . the proposition below allows us to use induction on the number of facets of an unmixed tree . [ joint - removal - unmixed ] let @xmath18 be an unmixed tree with a leaf @xmath6 , and let @xmath178 be a joint of @xmath6 . then @xmath181 is also unmixed . we use induction on the number of vertices of @xmath18 . let @xmath102 and @xmath212 be the vertex set for @xmath18 . the case @xmath213 is clear . suppose that @xmath180 and @xmath52 is a minimal vertex cover for @xmath161 . by corollary [ unmixed - corollary ] @xmath214 as well . if @xmath52 contains any vertex of @xmath178 , then it is also a minimal vertex cover for @xmath18 and hence of cardinality @xmath193 . so suppose that @xmath215 * claim : * _ there is a vertex @xmath216 . _ * _ proof of claim : _ if not , then @xmath217 + we show that this is not possible . + notice that for any @xmath218 there is a facet @xmath219 such that @xmath220 ( if no such @xmath221 existed , then @xmath222 would also be a vertex cover ) . + from ( [ a = v - g ] ) it follows that @xmath223 + on the other hand using theorem [ konig ] we can assume @xmath202 is a maximal independent set of facets in @xmath18 . by corollary [ joint - not - leaf ] @xmath224 as @xmath225 , one of the @xmath53 , say @xmath226 , has to contain more than one element of @xmath52 , so suppose @xmath227 where @xmath166 and @xmath228 are distinct elements of @xmath52 . it follows from ( [ a = v - g ] ) that @xmath229 + from the discussion preceding ( [ hgh ] ) above , one can pick @xmath230 to be facets of @xmath161 such that @xmath231 for @xmath232 , and consider the tree @xmath233 which by lemma [ two leaves ] is supposed to have two leaves . but based on the descriptions of @xmath234 in ( [ frg ] ) and ( [ hig ] ) , only one facet of this tree , namely @xmath178 , could possibly have a free vertex , which is a contradiction . this proves the claim . we now proceed to showing that @xmath235 is not possible . let @xmath236 we localize at the prime ideal @xmath76 generated by @xmath237 , and use the induction hypothesis . let @xmath238 and let @xmath239 from discussion [ local - discussion ] we know that , without loss of generality , for some @xmath240 @xmath241 where @xmath242 , and each of @xmath243 contains at least one of @xmath244 . we also know by lemma [ localization ] that @xmath245 is a forest whose vertex set is a proper subset of @xmath2 . by lemma [ unmixed - localizes ] @xmath245 is unmixed with @xmath246 we now focus on @xmath247 . besides possibly @xmath248 , all other facets of @xmath245 and @xmath247 are the same . we show why this is true . let @xmath249 . then clearly @xmath250 on the other hand , as @xmath251 and @xmath252 , we have @xmath253 and so @xmath254 conversely , if @xmath255 , then @xmath256 which implies the same for all @xmath257 , and therefore @xmath249 . so there are two possible scenarios : 1 . if @xmath258 , then @xmath259 which implies that @xmath52 is also a minimal vertex cover of @xmath245 , which is unmixed , and hence @xmath260 ; a contradiction . 2 . if @xmath261 then @xmath262 if not , then for some facet @xmath221 of @xmath18 , we have @xmath263 , so @xmath264 and therefore , since @xmath178 is a joint of the leaf @xmath6 , @xmath265 which immediately results in @xmath266 which is not possible . + in fact , @xmath267 remains a leaf in @xmath245 , since if @xmath268 is a facet of @xmath245 such that @xmath269 , then @xmath270 and so @xmath248 is a joint of @xmath245 . + now by the induction hypothesis , @xmath271 is an unmixed forest . this again implies that @xmath260 ; a contradiction . although not obvious at a first glance , proposition [ joint - removal - unmixed ] does not necessarily hold if @xmath178 is not a tree . the following example of an unmixed graph @xmath178 with a leaf demonstrates this point . @xmath272 the graph @xmath178 above was taken from the table of unmixed graphs in @xcite . the minimal vertex covers of @xmath178 , all of cardinality 3 , are @xmath273 and @xmath274 . but once one removes the joint @xmath275 , @xmath276 has minimal vertex covers @xmath277 and @xmath278 of different cardinalities , and is therefore not unmixed . [ structuretheorem ] suppose that @xmath18 is an unmixed tree with more than one facet such that @xmath180 . then @xmath18 can be written as @xmath279 with the following properties : 1 . @xmath280 are all the leaves of @xmath18 ; 2 . @xmath281 ; 3 . for @xmath59 , @xmath282 ; 4 . if a facet @xmath283 is not a leaf , then it does not contain a free vertex . if we prove @xmath284 , then parts @xmath285 , @xmath286 and @xmath287 will follow from @xmath284 , corollary [ leaf - independence ] and corollary [ basic ] . we prove part @xmath284 by induction on the number of facets @xmath83 of @xmath18 . if @xmath288 , then @xmath289 ( if @xmath18 is a tree of two facets , both facets must be leaves by lemma [ two leaves ] , and since @xmath18 is connected , we can get minimal vertex covers of cardinalities one and two , which means that @xmath18 is not unmixed ) . so the base case for induction is when @xmath290 . in this case , let @xmath86 and @xmath90 be the two disjoint leaves of @xmath18 , and let @xmath92 be the third facet . since @xmath18 is connected and unmixed , @xmath92 can not be a leaf ( because the leaves are pairwise disjoint ) . so @xmath92 is a joint for both @xmath86 and @xmath90 and this settles the case @xmath290 . for the general case , suppose that @xmath178 is a joint of @xmath0 . by corollary [ joint - not - leaf ] , @xmath178 is not a leaf . by corollary [ unmixed - corollary ] and proposition [ joint - removal - unmixed ] , if we remove @xmath178 , the forest @xmath291 is unmixed and @xmath214 . by the induction hypothesis , @xmath292 where conditions @xmath284 to @xmath287 are satisfied . it is easy to see from condition @xmath287 that if @xmath6 is a leaf of @xmath18 , then it will still be a leaf of @xmath161 ( because it has a free vertex ) . our goal is to show that the converse is true , that is , to show that @xmath280 are all the leaves of @xmath18 . we have the following presentation for @xmath18 : @xmath293 there are two cases to consider . 1 . @xmath178 is the only joint of @xmath18 . + suppose , without loss of generality , that for some @xmath294 , @xmath295 are leaves of @xmath18 and @xmath296 are not leaves of @xmath18 . remove @xmath297 from @xmath18 to obtain the forest @xmath298 by lemma [ two leaves ] , @xmath299 has at least two leaves . neither one of @xmath300 could be a leaf , because neither one of them has a free vertex . to see this , note that by the induction hypothesis on @xmath161 and part ( @xmath301 ) of the theorem , @xmath300 do not have free vertices in @xmath161 , and hence they can not have free vertices in @xmath18 . as facets of @xmath299 , they still do not have free vertices , because as @xmath178 is the only joint of @xmath18 , @xmath302 since @xmath178 is a facet of @xmath299 the removal of @xmath297 does not free any vertices of @xmath303 . + this implies that at least one of @xmath296 is a leaf of @xmath299 . suppose that @xmath304 is a leaf . then there exists a facet @xmath305 such that @xmath306 since @xmath307 for @xmath308 , it follows that @xmath309 and so @xmath304 is a leaf of @xmath18 , which is a contradiction . @xmath18 has another joint @xmath276 distinct from @xmath178 . + consider the presentation of @xmath18 as in ( [ dd ] ) . as @xmath310 is a maximal independent set of facets in @xmath18 , it can not contain @xmath276 ( corollary [ joint - not - leaf ] ) . therefore @xmath311 + we show that , say , @xmath86 is a leaf of @xmath18 . + consider the two unmixed forests @xmath312 + we already know from before that @xmath86 is a leaf of @xmath161 . from the fact that @xmath310 is a maximal independent set of facets in @xmath299 and corollary [ leaf - independence ] and the induction hypothesis , it follows that @xmath86 is also a leaf of @xmath299 . + so , by the definition of a leaf , there is a facet , say @xmath92 , in @xmath161 , such that @xmath313 and a facet @xmath314 such that @xmath315 + the possible scenarios are the following . 1 . @xmath316 or @xmath317 . + suppose @xmath316 . in this case @xmath318 , and so because of ( [ label-2 ] ) @xmath319 which with ( [ label-1 ] ) and ( [ label-2 ] ) implies that @xmath320 hence @xmath86 is a leaf of @xmath18 . the case @xmath317 is identical . 2 . @xmath321 and @xmath322 . + in this case , statements ( [ label-1 ] ) and ( [ label-2 ] ) , respectively , translate into + @xmath323 + if @xmath86 is not a leaf of @xmath18 , it follows from ( [ case21 ] ) that @xmath324 + by ( [ case2 ] ) there exist @xmath325 + * claim : * _ there is a minimal vertex cover for @xmath326 that avoids all the vertices in @xmath327 , @xmath276 and @xmath86 . _ * _ proof of claim : _ we first show that there is no facet of @xmath326 that has all its vertices in @xmath328 . suppose that @xmath221 is such a facet : @xmath329 and consider the tree @xmath330 by lemma [ two leaves ] , @xmath331 must have two leaves . note that @xmath221 can not be a leaf , since because of ( [ h ] ) it has no free vertices . if @xmath86 is a leaf , then it can not have @xmath178 or @xmath276 as its joint , since that violates the first two conditions in ( [ case2 ] ) , and so @xmath221 must be its joint . but then it follows that @xmath332 this implies that @xmath333 ( where @xmath184 is defined in ( [ x - and - y ] ) ) , which along with the third part of ( [ case2 ] ) , results in @xmath334 , which is a contradiction . + so @xmath178 and @xmath276 are the two leaves of @xmath331 . consider @xmath178 first . if @xmath276 is a joint for @xmath178 , it follows that @xmath335 which contradicts ( [ case2 ] ) . + if @xmath221 is a joint of @xmath178 , then @xmath336 which implies that @xmath337 , but this again means @xmath334 ( because of ( [ case2 ] ) ) , which is a contradiction . so @xmath86 is the only possible joint for @xmath178 . + with an identical argument for @xmath276 , it follows that @xmath86 is a joint for both @xmath178 and @xmath276 in @xmath331 , and therefore @xmath338 which along with ( [ h ] ) implies that @xmath339 which is impossible since @xmath221 and @xmath86 are both facets of @xmath18 . + so we have shown that every facet of @xmath18 other than @xmath178 , @xmath276 and @xmath86 , has at least one vertex outside @xmath178 and @xmath276 ( and therefore by the third condition in ( [ case2 ] ) , outside @xmath86 ) . + for each facet @xmath221 of @xmath326 , pick a vertex @xmath340 that avoids all three facets @xmath178 , @xmath276 and @xmath86 . the set of these vertices is a vertex cover for @xmath341 , and so it has a subset that is a minimal vertex cover . this proves the claim . + now let @xmath52 be a minimal vertex cover for @xmath341 that avoids all the vertices in @xmath327 , @xmath276 and @xmath86 . since @xmath326 has @xmath342 independent facets , @xmath343 . now @xmath344 is a minimal vertex cover for @xmath18 with more than @xmath193 vertices , which contradicts the fact that @xmath18 is unmixed with vertex covering number equal to @xmath193 ( note that @xmath184 and @xmath345 do not belong to any facet of @xmath326 , as this would contradict the third condition in ( [ case2 ] ) ) . so both cases 1 and 2 lead to contradictions , therefore all of @xmath280 must be leaves of @xmath18 , which proves the theorem . [ structure - example ] the simplicial complex @xmath18 shown below is an unmixed tree , satisfying properties ( i ) to ( iv ) of theorem [ structuretheorem ] . @xmath346 it is important to notice that theorem [ structuretheorem ] does not suggest that every facet in an unmixed tree is either a leaf or a joint ( see example [ not - leaf - or - joint - example ] below ) . on the other hand two different leaves in an unmixed tree may share a joint , as is the case with the unmixed graph @xmath347 . for these reasons the two numbers @xmath193 and @xmath348 in the statement of theorem [ structuretheorem ] that count the number of leaves and non - leaves , respectively , do not seem to have any particular relationship to one another . [ not - leaf - or - joint - example ] the following simplicial complex , which is the facet complex of the ideal @xmath349 is an unmixed tree with a facet @xmath350 that is neither a leaf nor a joint . in fact , the two leaves @xmath351 and @xmath352 share a joint @xmath353 . @xmath354 above , for simplicity , an @xmath26-dimensional facet ( simplex ) is drawn as a shaded polygon with @xmath355 vertices . the picture in 3d is as follows : @xmath356 all that we proved in the previous section about unmixed trees can be put into one definition namely that of a grafted tree . in fact , the method of grafting works as an effective way to build an unmixed simplicial complex from any given simplicial complex by adding new leaves ( theorem [ grafted - is - unmixed ] ) . it turns out that a grafted simplicial complex is cohen - macaulay ( theorem [ cm ] ) . [ grafting ] a simplicial complex @xmath18 is a _ grafting _ of the simplicial complex @xmath357 with the simplices @xmath280 ( or we say that @xmath18 is _ grafted _ ) if @xmath358 with the following properties : 1 . @xmath359 ; 2 . @xmath280 are all the leaves of @xmath18 ; 3 . @xmath281 ; 4 . for @xmath59 , @xmath282 ; 5 . if @xmath360 is a joint of @xmath0 , then @xmath361 is also grafted . note that a simplicial complex that consists of only one facet or several pairwise disjoint facets is indeed grafted , as it could be considered as a grafting of the empty simplicial complex . it is easy to check that conditions ( i ) to ( v ) above are satisfied in this case . it is also clear that the union of two or more grafted simplicial complexes is itself grafted . [ embedded - intersections ] condition ( v ) above implies that if @xmath6 is a leaf of a grafted @xmath18 , then all the facets @xmath221 that intersect @xmath6 have embedded intersections ; in other words if @xmath362 and @xmath363 are both nonempty , then @xmath364 this implies that there is a chain of intersections @xmath365 where @xmath366 are all the facets of @xmath18 that intersect @xmath6 . [ remove - all - grafted ] condition ( v ) in definition [ grafting ] can be replaced by `` @xmath361 is grafted for all @xmath367 '' . this is because even if @xmath360 is not a joint of @xmath18 , @xmath361 satisfies properties ( i ) , ( iii ) and ( iv ) , and it satisfies ( ii ) and ( v ) because of remark [ embedded - intersections ] , and so @xmath361 is grafted . if @xmath161 in definition [ grafting ] is a tree , then @xmath18 is also a tree . to see this , consider any subcollection @xmath299 of @xmath18 . if @xmath299 contains @xmath53 for some @xmath4 , then by remarks [ embedded - intersections ] and [ remove - all - grafted ] @xmath53 is a leaf of @xmath299 . if @xmath299 contains neither of the @xmath53 , then it is a subcollection of the tree @xmath161 , which implies that @xmath299 has a leaf . the `` suspension '' of a graph , as defined in @xcite , is also a grafting of that graph . [ grafting - example ] the tree @xmath368 that appeared in example [ structure - example ] above is a grafting of the tree @xmath369 with the leaves @xmath86 and @xmath90 . in fact , there may be more than one way to graft a given simplicial complex . for example , some possible ways of grafting @xmath369 are shown below : ( 300,220 ) ( 0,180)@xmath370 ( 40,155 ) ( 150,180)@xmath371 ( 220,180)@xmath372 ( 250,150 ) ( 100,130)@xmath373 ( 110,105)_graft _ ( 0,30)@xmath374 ( 40,0 ) ( 180,130)@xmath375 ( 220,105)_graft _ ( 220,30)@xmath376 ( 250,10 ) [ grafted - is - unmixed ] let @xmath377 be a grafting of the simplicial complex @xmath378 with the simplices @xmath280 . then @xmath18 is unmixed , and @xmath379 . if @xmath378 is the empty simplicial complex , the claim is immediate , so we assume that it is nonempty . we argue by induction on the number of facets @xmath83 of @xmath18 . the first case to consider is @xmath290 . in this case , @xmath18 must have at least two leaves , as if there were only one leaf @xmath86 , i.e. if @xmath380 , then by condition ( i ) of definition [ grafting ] we would have @xmath381 and @xmath382 , which is impossible . so @xmath383 , where @xmath384 and @xmath385 . it is now easy to see that @xmath18 is unmixed with @xmath386 . suppose @xmath18 has @xmath387 facets , and let @xmath92 be a joint of the leaf @xmath86 . by part ( v ) of definition [ grafting ] @xmath388 is also grafted , and therefore by the induction hypothesis unmixed with @xmath214 . let @xmath52 be a minimal vertex cover of @xmath18 . we already know that @xmath389 as @xmath280 are @xmath193 independent facets of @xmath18 . now suppose that @xmath390 . since @xmath52 is also a vertex cover for @xmath161 , it has a subset @xmath186 that is a minimal vertex cover of @xmath161 with @xmath391 . since @xmath186 is a proper subset of @xmath52 , it is not a vertex cover for @xmath18 , and therefore @xmath186 can not contain a vertex of @xmath92 . so @xmath186 contains a free vertex @xmath184 of @xmath86 ( all non - free vertices of @xmath86 are shared with @xmath92 ) . now @xmath52 must contain a vertex @xmath345 of @xmath92 ; say @xmath392 ( @xmath393 , since in that case @xmath184 would be redundant ) . so @xmath394 on the other hand @xmath186 must also contain a vertex of @xmath90 , say @xmath209 . so @xmath90 contributes two vertices @xmath345 and @xmath209 to @xmath52 ; note that neither one of @xmath345 or @xmath209 could be a free vertex , as in that case the free one would be redundant . now suppose that @xmath395 is a joint of @xmath90 . remove @xmath395 from @xmath18 to get @xmath396 so @xmath52 has a subset @xmath185 , @xmath397 , that is a minimal vertex cover for @xmath299 . but as @xmath52 already has exactly one vertex in each of @xmath398 , the only way to get @xmath185 from @xmath52 is to remove one of @xmath345 or @xmath209 , this means that : @xmath399 in either case @xmath185 contains a vertex of @xmath395 , which implies that @xmath185 is a minimal vertex cover for @xmath18 ; a contradiction . example [ grafting - example ] demonstrates theorem [ grafted - is - unmixed ] : @xmath400 is a non - unmixed tree , which gets grafted with some leaves to make the unmixed trees @xmath161 , @xmath299 and @xmath401 . one could graft any simplicial complex , even a badly non - unmixed non - tree . let @xmath161 be the non - unmixed non - tree in example [ non - leaf - example ] . we could graft @xmath161 with three new leaves @xmath402 the resulting picture below is unmixed , and moreover , as we prove later , cohen - macaulay . @xmath61{not - tree - suspended.eps}\ ] ] in the case of a tree theorems [ structuretheorem ] and [ grafted - is - unmixed ] put together with corollary [ basic ] produce a much stronger statement : [ tree - unmixed - iff - grafted ] suppose the simplicial complex @xmath18 is a tree . then @xmath18 is unmixed if and only if @xmath18 is grafted . grafted simplicial complexes behave well under localization ; in other words , the localization of a grafted simplicial complex is also grafted . in the case of trees this follows directly from corollary [ tree - unmixed - iff - grafted ] , lemma [ unmixed - localizes ] and lemma [ localization ] . but the statement holds more generally . [ grafted - localization ] let @xmath104 where @xmath18 is a grafted simplicial complex with vertices labeled @xmath29 . suppose that @xmath28 is a field and @xmath76 is a prime ideal of the polynomial ring @xmath31 $ ] . then @xmath114 is a grafted simplicial complex . with notation as in definition [ grafting ] , let @xmath403 if @xmath18 has only one facet , the statement of the theorem follows immediately , so assume that @xmath18 has two or more facets . as in the proof of lemma [ localization ] , it is enough to assume that @xmath76 is generated by a subset of @xmath44 , so @xmath404 following discussion [ local - discussion ] , let @xmath405 where for @xmath406 and @xmath407 @xmath408 and @xmath409 each contain at least one of @xmath410 we now rename the facets of @xmath245 as follows . for @xmath411 , let @xmath412 for each @xmath413 , @xmath414 contains one of the facets appearing in ( [ the - extra - facets ] ) . but as by definition @xmath282 for all @xmath415 , there must be some @xmath416 for which @xmath417 . for this particular @xmath418 , set @xmath419 this choice of @xmath418 is well - defined : if there were some @xmath420 distinct from @xmath418 such that @xmath421 , then it would follow from remark [ embedded - intersections ] that either @xmath422 or @xmath423 , which contradicts the fact that both @xmath424 and @xmath425 are facets of @xmath245 . we now represent @xmath245 as @xmath426 where @xmath427 represent all the other facets of @xmath245 that were not labeled by some @xmath428 . our goal is to show that @xmath245 is a grafting of the simplicial complex @xmath429 with the simplices @xmath430 . it is clear by our construction that the facets @xmath430 are pairwise disjoint . to see this , notice that for each pair of distinct numbers @xmath431 , there is a pair of distinct numbers @xmath432 such that @xmath433 and as @xmath434 , @xmath435 so condition ( iv ) of definition [ grafting ] is satisfied . on the other hand , by theorem [ grafted - is - unmixed ] @xmath18 is unmixed , so by lemma [ unmixed - localizes ] @xmath245 is unmixed with @xmath436 . we now apply lemma [ basic - lemma ] to @xmath245 to deduce that @xmath437 which implies condition ( i ) in definition [ grafting ] . this also implies that @xmath427 can not have free vertices , and hence can not be leaves of @xmath245 . condition ( iii ) is satisfied by the construction of @xmath245 . we need to show that @xmath430 are all leaves of @xmath245 . if @xmath438 then @xmath245 is grafted by definition . so suppose that @xmath245 has a connected component @xmath161 with two or more facets . as @xmath161 is connected , it must contain some of the @xmath439 , and as @xmath440 , @xmath161 must also contain some of the @xmath441 . so we can without loss of generality assume that @xmath442 for some @xmath443 and @xmath444 . we now show that , for example , @xmath445 is a leaf for @xmath161 . there are two cases to consider : 1 . @xmath446 for some @xmath4 such that @xmath447 . + since @xmath161 is connected , it has some facets that intersect @xmath428 ; suppose that @xmath448 are all the facets of @xmath449 such that @xmath450 for @xmath451 . + for each @xmath451 suppose that @xmath452 + the above paragraph translates into @xmath453 and hence @xmath454 for @xmath451 . + from remark [ embedded - intersections ] it follows that there is some total order of inclusion on the nonempty sets @xmath455 ; we assume that @xmath456 which after intersecting each set with @xmath457 turns into @xmath458 which is equivalent to @xmath459 + it follows that @xmath445 is a leaf of @xmath161 , and in addition , condition ( v ) of definition [ grafting ] is satisfied . 2 . @xmath460 for some @xmath418 such that @xmath461 . + in this case for some @xmath4 , @xmath462 , @xmath463 + exactly as above , let @xmath464 be all the facets of @xmath449 such that @xmath465 , and let @xmath466 for @xmath451 . + as all the sets @xmath455 are nonempty , we follow the exact argument as above to obtain the chain @xmath467 + as @xmath468 , we can intersect all these sets with @xmath424 to obtain @xmath469 which is equivalent to @xmath470 + it follows that @xmath445 is a leaf of @xmath161 , and also condition ( v ) of definition [ grafting ] is satisfied . we are now ready to show that the facet ideal of a grafted simplicial complex has a cohen - macaulay quotient . besides revealing a wealth of square - free monomial ideals with cohen - macaulay quotients , this result implies that all unmixed trees are cohen - macaulay . it is interesting to observe how the previous sentence follows also from proposition [ grafted - localization ] : if @xmath477 is any other maximal ideal of @xmath474 , from the proof of lemma [ localization ] and proposition [ grafted - localization ] we see that if @xmath478 is the ideal generated by all of @xmath479 that belong to @xmath477 , then @xmath480 is the facet ideal of a grafted simplicial complex over the vertex set @xmath481 . so one can write @xmath482 where @xmath83 is a prime ideal of @xmath483 $ ] . it follows that @xmath484_p / i_p \otimes_k k[x_{e+1},\ldots , x_n]_q.\ ] ] as @xmath483_q$ ] is clearly cohen - macaulay , by theorem 5.5.5 of @xcite , it is enough to show that @xmath485_p / i_p$ ] is cohen - macaulay in order to conclude that @xmath486 is cohen - macaulay . but this is again the case of localizing at the irrelevant ideal . now suppose that for each @xmath487 , @xmath488 where @xmath489 is a free vertex of the leaf @xmath53 , and @xmath490 . we wish to show that @xmath491 is a regular sequence in @xmath474 . this follows from the process of `` polarization '' that we describe below . [ polarization ] let @xmath30 be the ring @xmath31/({m_1,\ldots , m_q})$ ] , where @xmath38 are monomials in the variables @xmath29 , and @xmath28 is a field . then there is an @xmath492 , and a set of square - free monomials @xmath493 in the polynomial ring @xmath494 $ ] , such that @xmath495 where @xmath496/(n_1 , \ldots , n_q)$ ] and @xmath497 is a regular sequence of forms of degree one in @xmath498 . it is then shown that @xmath509 is a non - zerodivisor in @xmath510 . if @xmath511 is square - free , we are done . otherwise one applies the same procedure to @xmath511 continually until the ideal becomes square - free . what we would like to show is that sequence ( [ the - sequence ] ) polarizes the ring @xmath512/(y_1^{u_1 + 1 } , \ldots , y_r^{u_r+1},e_1 , \ldots , e_s)\ ] ] into the ring @xmath474 , where @xmath513 are monomials corresponding to the facets @xmath514 , where each vertex belonging to @xmath53 has been replaced by the free vertex @xmath489 . in other words if @xmath515 we wish to show that @xmath516 it will then follow from the proof of proposition [ polarization ] ( as detailed in @xcite as well as in @xcite ) that sequence ( [ the - sequence ] ) is a regular sequence in @xmath474 . intuitively our claim is straightforward to see . the only problem that may arise is if after applying sequence ( [ the - sequence ] ) to @xmath517 , we end up with a permutation of the vertices of @xmath18 . to prevent this from happening , we use the subtle structure of a grafted simplicial complex ( remark [ embedded - intersections ] ) that the facets intersecting a leaf do so in an embedded ( and therefore ordered ) manner . in other words , suppose for the leaf @xmath53 , the facets @xmath518 are all the facets of @xmath519 that intersect @xmath53 , with the ordering @xmath520 so in sequence ( [ the - sequence ] ) , we order @xmath521 such that if for any @xmath294 and @xmath522 , @xmath523 then @xmath524 . we now use induction on the number of facets of @xmath18 . if we remove a joint , say @xmath525 , we obtain a grafted simplicial complex @xmath526 over the same set of vertices @xmath29 , with @xmath527 ( lemma [ unmixed - induction ] ) . therefore if @xmath528/{{\mathcal{f}}}({\delta}')\ ] ] then @xmath529 moreover , @xmath161 has @xmath280 as leaves . so by the induction hypothesis , sequence ( [ the - sequence ] ) polarizes the ring @xmath530/(y_1^{u_1 + 1 } , \ldots , y_r^{u_r+1},e_2 , \ldots , e_s)\ ] ] into @xmath531 , or in other words , @xmath532 the induction hypothesis has ensured that after applying sequence ( [ the - sequence ] ) to @xmath533 , all facets of @xmath161 are restored to their original positions and labeling . now it all reduces to showing that during this polarization process , @xmath534 turns into @xmath92 . this is clear , as for every @xmath4 , @xmath535 has its place in the ordered sequence ( [ ordered - sequence ] ) , and so if @xmath536 , then the first @xmath537 applications of sequence ( [ sub - polar ] ) restore @xmath535 before moving on to facets that have larger intersections with @xmath53 . as @xmath92 has disjoint intersections with @xmath280 , once sequence ( [ sub - polar ] ) has been applied for all @xmath4 , @xmath92 is restored to its proper position . [ cm - criterion ] let @xmath18 be a tree over a set of vertices @xmath29 , and let @xmath28 be a field . then the quotient ring @xmath31/{{\mathcal{f}}}({\delta})$ ] is cohen - macaulay if and only if @xmath18 is unmixed .